Maneuvering and Control of Marine Vehicles

MANEUVERING AND CONTROL
OF MARINE VEHICLES
Michael S. Triantafyllou
Franz S. Hover
Department of Ocean Engineering
Massachusetts Institute of Technology
Cambridge, Massachusetts USA
These notes were developed in the instruction of the MIT graduate subject
13.49: Maneuvering and Control of Surface and Underwater Vehicles. We plan
many enhancements; your comments are welcome!
Maneuvering and Control of Marine Vehicles

Latest Revision: January 11, 2002
Michael
c S. Triantafyllou and Franz S. Hover
Contents
1 MATH FACTS 1
1.1 Vectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.2 Vector Magnitude . . . . . . . . . . . . . . . . . . . . . 2
1.1.3 Vector Dot Product . . . . . . . . . . . . . . . . . . . . 2
1.1.4 Vector Cross Product . . . . . . . . . . . . . . . . . . . 2
1.2 Matrices . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Multiplying a Vector by a Matrix . . . . . . . . . . . . 3
1.2.3 Multiplying a Matrix by a Matrix . . . . . . . . . . . . 4
1.2.4 Common Matrices . . . . . . . . . . . . . . . . . . . . 4
1.2.5 Transpose . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.6 Determinant . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.7 Inverse . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2.8 Trace . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2.9 Eigenvalues and Eigenvectors . . . . . . . . . . . . . . 7
1.2.10 Modal Decomposition . . . . . . . . . . . . . . . . . . 9
1.2.11 Singular Value . . . . . . . . . . . . . . . . . . . . . . . 9
1.3 Laplace Transform . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.2 Convergence . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.3 Convolution Theorem . . . . . . . . . . . . . . . . . . . 11
1.3.4 Solution of Differential Equations by Laplace Transform 13
2 KINEMATICS OF MOVING FRAMES 14

2.1 Rotation of Reference Frames . . . . . . . . . . . . . . . . . . 14
2.2 Differential Rotations . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Rate of Change of Euler Angles . . . . . . . . . . . . . . . . . 18
2.4 Dead Reckoning . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3 VESSEL INERTIAL DYNAMICS 20

3.1 Momentum of a Particle . . . . . . . . . . . . . . . . . . . . . 20
3.2 Linear Momentum in a Moving Frame . . . . . . . . . . . . . 21
3.3 Example: Mass on a String . . . . . . . . . . . . . . . . . . . 22
3.3.1 Moving Frame Affixed to Mass . . . . . . . . . . . . . 23
3.3.2 Rotating Frame Attached to Pivot Point . . . . . . . . 23
i
3.3.3 Stationary Frame . . . . . . . . . . . . . . . . . . . . . 24
3.4 Angular Momentum . . . . . . . . . . . . . . . . . . . . . . . 24
3.5 Example: Spinning Book . . . . . . . . . . . . . . . . . . . . . 26
3.5.1 x-axis . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.5.2 y-axis . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.5.3 z-axis . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.6 Parallel Axis Theorem . . . . . . . . . . . . . . . . . . . . . . 28
3.7 Basis for Simulation . . . . . . . . . . . . . . . . . . . . . . . 28
4 HYDRODYNAMICS: INTRODUCTION 30
4.1 Taylor Series and Hydrodynamic Coefficients . . . . . . . . . . 30
4.2 Surface Vessel Linear Model . . . . . . . . . . . . . . . . . . . 31
4.3 Stability of the Sway/Yaw System . . . . . . . . . . . . . . . . 32
4.4 Basic Rudder Action in the Sway/Yaw Model . . . . . . . . . 34
4.4.1 Adding Yaw Damping through Feedback . . . . . . . . 36
4.4.2 Heading Control in the Sway/Yaw Model . . . . . . . . 36
4.5 Response of the Vessel to Step Rudder Input . . . . . . . . . . 37
4.5.1 Phase 1: Accelerations Dominate . . . . . . . . . . . . 37
4.5.2 Phase 3: Steady State . . . . . . . . . . . . . . . . . . 37
4.6 Summary of the Linear Maneuvering Model . . . . . . . . . . 38
4.7 Stability in the Vertical Plane . . . . . . . . . . . . . . . . . . 38
5 SIMILITUDE 40
5.1 Use of Nondimensional Groups . . . . . . . . . . . . . . . . . 40
5.2 Common Groups in Marine Engineering . . . . . . . . . . . . 42
5.3 Similitude in Maneuvering . . . . . . . . . . . . . . . . . . . . 44
5.4 Roll Equation Similitude . . . . . . . . . . . . . . . . . . . . . 46
6 CAPTIVE MEASUREMENTS 48
6.1 Towtank . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
6.2 Rotating Arm Device . . . . . . . . . . . . . . . . . . . . . . . 48
6.3 Planar-Motion Mechanism . . . . . . . . . . . . . . . . . . . . 49
7 STANDARD MANEUVERING TESTS 52

7.1 Dieudonné Spiral . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.2 Zig-Zag Maneuver . . . . . . . . . . . . . . . . . . . . . . . . . 52
7.3 Circle Maneuver . . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.3.1 Drift Angle . . . . . . . . . . . . . . . . . . . . . . . . 53
ii
7.3.2 Speed Loss . . . . . . . . . . . . . . . . . . . . . . . . 53
7.3.3 Heel Angle . . . . . . . . . . . . . . . . . . . . . . . . . 53
7.3.4 Heeling in Submarines with Sails . . . . . . . . . . . . 54
8 STREAMLINED BODIES 56
8.1 Nominal Drag Force . . . . . . . . . . . . . . . . . . . . . . . 56
8.2 Munk Moment . . . . . . . . . . . . . . . . . . . . . . . . . . 56
8.3 Separation Moment . . . . . . . . . . . . . . . . . . . . . . . . 57
8.4 Net Effects: Aerodynamic Center . . . . . . . . . . . . . . . . 57
8.5 Role of Fins in Moving the Aerodynamic Center . . . . . . . . 58
8.6 Aggregate Effects of Body and Fins . . . . . . . . . . . . . . . 60
8.7 Coefficients Zw , Mw , Zq , and Mq for a Slender Body . . . . . . 60
9 SLENDER-BODY THEORY 62
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
9.2 Kinematics Following the Fluid . . . . . . . . . . . . . . . . . 62
9.3 Derivative Following the Fluid . . . . . . . . . . . . . . . . . . 63
9.4 Differential Force on the Body . . . . . . . . . . . . . . . . . . 64
9.5 Total Force on a Vessel . . . . . . . . . . . . . . . . . . . . . . 64
9.6 Total Moment on a Vessel . . . . . . . . . . . . . . . . . . . . 65
9.7 Relation to Wing Lift . . . . . . . . . . . . . . . . . . . . . . . 67
9.8 Convention: Hydrodynamic Mass Matrix A . . . . . . . . . . 67
10 PRACTICAL LIFT CALCULATIONS 68

10.1 Characteristics of Lift-Producing Mechanisms . . . . . . . . . 68
10.2 Jorgensen’s Formulas . . . . . . . . . . . . . . . . . . . . . . . 68
10.3 Hoerner’s Data: Notation . . . . . . . . . . . . . . . . . . . . 70
10.4 Slender-Body Theory vs. Experiment . . . . . . . . . . . . . . 71
10.5 Slender-Body Approximation for Fin Lift . . . . . . . . . . . . 73
11 FINS AND LIFTING SURFACES 74

11.1 Origin of Lift . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
11.2 Three-Dimensional Effects: Finite Length . . . . . . . . . . . . 74
11.3 Ring Fins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
12 PROPELLERS AND PROPULSION 77

12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
12.2 Steady Propulsion of Vessels . . . . . . . . . . . . . . . . . . . 77
12.2.1 Basic Characteristics . . . . . . . . . . . . . . . . . . . 77
iii
12.2.2 Solution for Steady Conditions . . . . . . . . . . . . . 81
12.2.3 Engine/Motor Models . . . . . . . . . . . . . . . . . . 81
12.3 Unsteady Propulsion Models . . . . . . . . . . . . . . . . . . . 83
12.3.1 One-State Model: Yoerger et al. . . . . . . . . . . . . . 83
12.3.2 Two-State Model: Healey et al. . . . . . . . . . . . . . 84
13 TOWING OF VEHICLES 85
13.1 Statics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
13.1.1 Force Balance . . . . . . . . . . . . . . . . . . . . . . . 85
13.1.2 Critical Angle . . . . . . . . . . . . . . . . . . . . . . . 87
13.2 Linearized Dynamics . . . . . . . . . . . . . . . . . . . . . . . 89
13.2.1 Derivation . . . . . . . . . . . . . . . . . . . . . . . . . 89
13.2.2 Damped Axial Motion . . . . . . . . . . . . . . . . . . 91
13.3 Cable Strumming . . . . . . . . . . . . . . . . . . . . . . . . . 94
13.4 Vehicle Design . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
14 TRANSFER FUNCTIONS & STABILITY 96

14.1 Partial Fractions . . . . . . . . . . . . . . . . . . . . . . . . . 96
14.2 Partial Fractions: Unique Poles . . . . . . . . . . . . . . . . . 96
14.3 Example: Partial Fractions with Unique Real Poles . . . . . . 97
14.4 Partial Fractions: Complex-Conjugate Poles . . . . . . . . . . 98
14.5 Example: Partial Fractions with Complex Poles . . . . . . . . 98
14.6 Stability in Linear Systems . . . . . . . . . . . . . . . . . . . . 98
14.7 Stability ⇐⇒ Poles in LHP . . . . . . . . . . . . . . . . . . . 99
14.8 General Stability . . . . . . . . . . . . . . . . . . . . . . . . . 99
15 CONTROL FUNDAMENTALS 100

15.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
15.1.1 Plants, Inputs, and Outputs . . . . . . . . . . . . . . . 100
15.1.2 The Need for Modeling . . . . . . . . . . . . . . . . . . 100
15.1.3 Nonlinear Control . . . . . . . . . . . . . . . . . . . . . 101
15.2 Representing Linear Systems . . . . . . . . . . . . . . . . . . . 101
15.2.1 Standard State-Space Form . . . . . . . . . . . . . . . 101
15.2.2 Converting a State-Space Model into a Transfer Function 102
15.2.3 Converting a Transfer Function into a State-Space Model 102
15.3 PID Controllers . . . . . . . . . . . . . . . . . . . . . . . . . . 103
15.4 Example: PID Control . . . . . . . . . . . . . . . . . . . . . . 103
15.4.1 Proportional Only . . . . . . . . . . . . . . . . . . . . 104
iv
15.4.2 Proportional-Derivative Only . . . . . . . . . . . . . . 104
15.4.3 Proportional-Integral-Derivative . . . . . . . . . . . . . 105
15.5 Heuristic Tuning . . . . . . . . . . . . . . . . . . . . . . . . . 105
15.6 Block Diagrams of Systems . . . . . . . . . . . . . . . . . . . . 106
15.6.1 Fundamental Feedback Loop . . . . . . . . . . . . . . . 106
15.6.2 Block Diagrams: General Case . . . . . . . . . . . . . . 106
15.6.3 Primary Transfer Functions . . . . . . . . . . . . . . . 107
16 MODAL ANALYSIS 109

16.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
16.2 Matrix Exponential . . . . . . . . . . . . . . . . . . . . . . . . 109
16.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . 109
16.2.2 Modal Canonical Form . . . . . . . . . . . . . . . . . . 109
16.2.3 Modal Decomposition of Response . . . . . . . . . . . 110
16.3 Forced Response and Controllability . . . . . . . . . . . . . . 111
16.4 Plant Output and Observability . . . . . . . . . . . . . . . . . 112
17 CONTROL SYSTEMS – LOOPSHAPING 113

17.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
17.2 Roots of Stability – Nyquist Criterion . . . . . . . . . . . . . . 113
17.2.1 Mapping Theorem . . . . . . . . . . . . . . . . . . . . 114
17.2.2 Nyquist Criterion . . . . . . . . . . . . . . . . . . . . . 114
17.2.3 Robustness on the Nyquist Plot . . . . . . . . . . . . . 115
17.3 Design for Nominal Performance . . . . . . . . . . . . . . . . . 116
17.4 Design for Robustness . . . . . . . . . . . . . . . . . . . . . . 116
17.5 Robust Performance . . . . . . . . . . . . . . . . . . . . . . . 118
17.6 Implications of Bode’s Integral . . . . . . . . . . . . . . . . . . 118
17.7 The Recipe for Loopshaping . . . . . . . . . . . . . . . . . . . 119
18 LINEAR QUADRATIC REGULATOR 121

18.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
18.2 Full-State Feedback . . . . . . . . . . . . . . . . . . . . . . . . 121
18.3 Dynamic Programming . . . . . . . . . . . . . . . . . . . . . . 121
18.4 Dynamic Programming and Full-State Feedback . . . . . . . . 123
18.5 Properties and Use of the LQR . . . . . . . . . . . . . . . . . 125
v
19 KALMAN FILTER 128
19.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
19.2 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . 128
19.3 Step 1: An Equation for Σ̇ . . . . . . . . . . . . . . . . . . . . 129
19.4 Step 2: H as a Function of Σ . . . . . . . . . . . . . . . . . . 131
19.5 Properties of the Solution . . . . . . . . . . . . . . . . . . . . 132
19.6 Combination of LQR and KF . . . . . . . . . . . . . . . . . . 133
19.7 Proofs of the Intermediate Results . . . . . . . . . . . . . . . . 134
19.7.1 Proof that E(eT W e) = trace(ΣW ) . . . . . . . . . . . 134
∂
19.7.2 Proof that ∂H trace(−ΛHCΣ) = −ΛT ΣC T . . . . . . . 135
∂
19.7.3 Proof that ∂H trace(−ΛΣC T H T ) = −ΛΣC T . . . . . . 135
19.7.4 Proof of the Separation Principle . . . . . . . . . . . . 136
20 LOOP TRANSFER RECOVERY 137

20.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
20.2 A Special Property of the LQR Solution . . . . . . . . . . . . 138
20.3 The Loop Transfer Recovery Result . . . . . . . . . . . . . . . 139
20.4 Usage of the Loop Transfer Recovery . . . . . . . . . . . . . . 141
20.5 Three Lemmas . . . . . . . . . . . . . . . . . . . . . . . . . . 141
21 SYSTEM IDENTIFICATION 143

21.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
21.2 Visual Output from a Simple Input . . . . . . . . . . . . . . . 143
21.3 Transfer Function Estimation – Sinusoidal Input . . . . . . . . 145
21.4 Transfer Function Estimation – Broadband Input . . . . . . . 146
21.4.1 Fourier Transform of Sampled Data . . . . . . . . . . . 147
21.4.2 Estimating the Transfer Function . . . . . . . . . . . . 148
21.5 Time-Domain Simulation . . . . . . . . . . . . . . . . . . . . . 150
22 CARTESIAN NAVIGATION 153

22.1 Acoustic Navigation . . . . . . . . . . . . . . . . . . . . . . . 153
22.2 Global Positioning System (GPS) . . . . . . . . . . . . . . . . 155
23 REFERENCES 159
24 PROBLEMS 163
vi
1
1 MATH FACTS
1.1 Vectors
1.1.1 Definition
We use the overhead arrow to denote a column vector, i.e., a number with a
direction. For example, in three-space, we write
 
 2 
 
~a =  1  .

7 
The elements of a vector have a graphical interpretation, which is particularly

easy to see in two or three dimensions.
1
2
y
x
1. Vector addition is pointwise.
  
~a + ~b = ~

c 
 2 
   3 
   5 
 
1 + 3 = 4 .
 
    

7 
2  
9 
Graphically, addition is stringing the vectors together head to tail.

2 1 MATH FACTS
2. Scalar multiplication is pointwise.
   
 2 
  −4 
 

−2 ×  1  =  −2  .

7   −14 
1.1.2 Vector Magnitude

The total length of a vector of dimension m, its Euclidean norm, is given by
v
um
uX
||~x|| = t x2i ;
i=1
this scalar is commonly used to normalize a vector to length one.
1.1.3 Vector Dot Product

The dot product of two vectors is the sum of the products of the elements:
m
~x · ~y = ~xT ~y =
X
xi yi .
i=1
The dot product also satisfies
~x · ~y = ||~x||||~y || cos θ,
where θ is the angle between the vectors.
1.1.4 Vector Cross Product

The cross product of two three-dimensional vectors is another vector, ~x ×~y = ~z,
whose
1. direction is normal to the plane formed by the two vectors,
2. direction is given by the right-hand rule, rotating from ~x to ~y ,

1.2 Matrices 3
3. magnitude is the area of the parallelogram formed by the two vectors –

the cross product of two parallel vectors is zero – and
4. (signed) magnitude is equal to ||~x||||~y || sin θ, where θ is the angle between

the two vectors, measured from ~x to ~y .
The schoolbook formula is
 
 x2 y3 − x3 y2 
 
~x × ~y = x3 y1 − x1 y3 .
x1 y2 − x2 y1 

 
1.2 Matrices
1.2.1 Definition
A matrix, or array, is equivalent to a set of row vectors, arranged side by side,
say
 
2 3
~
A = [~a b] =  1 3  .
 
7 2
This matrix has three rows (m = 3) and two columns (n = 2); a vector is
a special case of a matrix with one column. Matrices, like vectors, permit
pointwise addition and scalar multiplication. We usually use an upper-case
symbol to denote a matrix.
1.2.2 Multiplying a Vector by a Matrix

If Aij denotes the element of matrix A in the i’th row and the j’th column,
then the multiplication ~c = A~v is constructed as:
n
X
ci = Ai1 v1 + Ai2 v2 + · · · + Ain vn = Aij vj ,
j=1
where n is the number of columns in A. ~c will have as many columns as

A has rows (m). Note that this multiplication is well-defined only if ~v has
4 1 MATH FACTS
as many rows as A has columns; they have consistent inner dimension n.

The product ~v A would be well-posed only if A had one row, and the proper
number of columns. There is another important interpretation of this vector
multiplication: Let the subscript : indicate all rows, so that each A:j is the
j’th column vector. Then
~c = A~v = A:1 v1 + A:2 v2 + · · · + A:n vn .
We are multiplying column vectors of A by the scalar elements of ~v .
1.2.3 Multiplying a Matrix by a Matrix

The multiplication C = AB is equivalent to a side-by-side arrangement of
column vectors C:j = AB:j , so that
C = AB = [AB:1 AB:2 · · · AB:k ],
where k is the number of columns in matrix B. The same inner dimension

condition applies as noted above: the number of columns in A must equal the
number of rows in B. Matrix multiplication is:
1. Associative. (AB)C = A(BC).
2. Distributive. A(B + C) = AB + AC, (B + C)A = BA + CA.
3. NOT Commutative. AB 6= BA, except in special cases.
1.2.4 Common Matrices

Identity . The identity matrix is usually denoted I, and comprises a square
matrix with ones on the diagonal, and zeros elsewhere, e.g.,
 
1 0 0
I3×3 = 0 1 0 

.
0 0 1
The identity always satisfies AIn×n = Im×m A = A.

1.2 Matrices 5
Diagonal Matrices . A diagonal matrix is square, and has all zeros off the
diagonal. For instance, the following is a diagonal matrix:
 
4 0 0
 0 −2 0  .
A= 
0 0 3
The product of a diagonal matrix with another diagonal matrix is diagonal,
and in this case the operation is commutative.
1.2.5 Transpose
The transpose of a vector or matrix, indicated by a T superscript results
from simply swapping the row-column indices of each entry; it is equivalent to
“flipping” the vector or matrix around the diagonal line. For example,
 
 1 
 
~a = 2 −→ ~a T = {1 2 3}
 

3 
 
1 2 " #
1 4 8
A =  4 5  −→ AT = .
 
2 5 9
8 9
A very useful property of the transpose is
(AB)T = B T AT .
1.2.6 Determinant
The determinant of a square matrix A is a scalar equal to the volume of the
parallelepiped enclosed by the constituent vectors. The two-dimensional case
is particularly easy to remember, and illustrates the principle of volume:
det(A) = A11 A22 − A21 A12

" #!
1 −1
det = 1 + 1 = 2.
1 1
6 1 MATH FACTS
y
Area = det(A) = 2
A:2 A:1
x
-1 1
In higher dimensions, the determinant is more complicated to compute. The

general formula allows one to pick a row k, perhaps the one containing the
most zeros, and apply
j=n
Akj (−1)k+j ∆kj ,
X
det(A) =
j=1
where ∆kj is the determinant of the sub-matrix formed by neglecting the k’th
row and the j’th column. The formula is symmetric, in the sense that one
could also target the k’th column:
j=n
Ajk (−1)k+j ∆jk .
X
det(A) =
j=1
If the determinant of a matrix is zero, then the matrix is said to be singular –

there is no volume, and this results from the fact that the constituent vectors
do not span the matrix dimension. For instance, in two dimensions, a singular
matrix has the vectors colinear; in three dimensions, a singular matrix has
all its vectors lying in a (two-dimensional) plane. Note also that det(A) =
det(AT ). If det(A) 6= 0, then the matrix is said to be nonsingular.
1.2.7 Inverse
The inverse of a square matrix A, denoted A−1 , satisfies AA−1 = A−1 A = I.
Its computation requires the determinant above, and the following definition
of the n × n adjoint matrix:
T
(−1)1+1 ∆11 · · · (−1)1+n ∆1n

adj(A) = 

··· ··· ···  .

n+1 n+n
(−1) ∆n1 · · · (−1) ∆nn .
1.2 Matrices 7
Once this computation is made, the inverse follows from
adj(A)
A−1 = .
det(A)
If A is singular, i.e., det(A) = 0, then the inverse does not exist. The inverse
finds common application in solving systems of linear equations such as
A~x = ~b −→ ~x = A−1~b.
1.2.8 Trace
The trace of a matrix is simply the sum of the diagonals:
n
X
tr(A) = Aii .
i=1
1.2.9 Eigenvalues and Eigenvectors

A typical eigenvalue problem is stated as
A~x = λ~x,
where A is an n × n matrix, ~x is a column vector with n elements, and λ is
a scalar. We ask for what nonzero vectors ~x (right eigenvectors), and scalars
λ (eigenvalues) will the equation be satisfied. Since the above is equivalent to
(A − λI)~x = ~0, it is clear that det(A − λI) = 0. This observation leads to the
solutions for λ; here is an example for the two-dimensional case:
" #
4 −5
A = −→
2 −3
" #
4−λ −5
A − λI = −→
2 −3 − λ
det(A − λI) = (4 − λ)(−3 − λ) + 10
= λ2 − λ − 2
= (λ + 1)(λ − 2).
8 1 MATH FACTS
Thus, A has two eigenvalues, λ1 = −1 and λ2 = 2. Each is associated with a

right eigenvector ~x. In this example,
(A − λ1 I)~x1 = ~0 −→
" #
5 −5
~x1 = ~0 −→
2 −2
n√ √ oT
~x1 = 2/2, 2/2 .
(A − λ2 I)~x2 = ~0 −→
" #
2 −5
~x2 = ~0 −→
2 −5
n √ √ oT
~x2 = 5 29/29, 2 29/29 .
Eigenvectors have arbitrary magnitude and sign; they are often normalized to
have unity magnitude, and positive first element (as above). A set of n eigen-
vectors is always linearly independent. The condition that rank(A − λi I) =
rank(A)−1 indicates that there is only one eigenvector for the eigenvalue λi . If
the left-hand side is less than this, then there are multiple unique eigenvectors
that go with λi .
The above discussion relates only the right eigenvectors, generated from the
equation A~x = λ~x. Left eigenvectors, also useful for many problems, pertain
to the transpose of A: AT ~y = λ~y . A and AT share the same eigenvalues λ,
since they share the same determinant. Example:
(AT − λ1 I)~y1 = ~0 −→
" #
5 2
~y1 = ~0 −→
−5 −2
n √ √ oT
~y1 = 2 29/29, −5 29/29 .
(AT − λ2 I)~y2 = ~0 −→
" #
2 2
~y2 = ~0 −→
−5 −5
n√ √ oT
~y2 = 2/2, − 2/2 .
1.2 Matrices 9
1.2.10 Modal Decomposition

The right and left eigenvectors of a particular eigenvalue have unity dot prod-
uct, that is ~xTi ~yi = 1, with the normalization noted above. The dot product
of a left eigenvector with the right eigenvector of a different eigenvalue is zero.
Thus, if
X = [~x1 · · · ~xn ] , and

Y = [~y1 · · · ~yn ] ,
then we have
Y T X = I, or
Y T = X −1 .
Next, construct a diagonal matrix of eigenvalues:
 
λ1 0
Λ=

· ;

0 λn
it follows that
AV = V Λ −→
A = V ΛW T
n
~ iT .
X
= λi~vi w
i=1
Hence A can be written as a sum of modal components.1
1.2.11 Singular Value

Let G(s) be an m × n, possibly complex matrix. The singular value decompo-
sition (SVD) computes three matrices satisfying
1
By carrying out successive multiplications, it can be shown that Ak has its eigenvalues
at λki , and keeps the same eigenvectors as A.
10 1 MATH FACTS
G = U ΣV ∗ ,
where U is m × m, Σ is m × n, and V is n × n. The star notation indicates a

complex-conjugate transpose. The matrix Σ is diagonal, with the form
σ1 0 0 0
 
 0 · 0 0 
Σ= ,
 
 0 0 σp 0 
0 0 0 0
where p = min(m, n). Each nonzero entry on the diagonal is a real, positive
singular value, ordered such that σ1 > σ2 > · · · σp . The notation is common
that σ1 = σ, the maximum singular value, and σp = σ, the minimum singular
value. The auxiliary matrices U and V are unitary, i.e., they satisfy X ∗ =
X −1 . Like eigenvalues, the singular values of G are related to projections. σi
represents the Euclidean size of the matrix G along the i’th singular vector:
σ = max||x||=1 ||Gx||
σ = min||x||=1 ||Gx||.
Other properties of the singular value include:
• σ(AB) ≤ σ(A)σ(B).
q
• σ(A) = λmax (A∗ A).
q
• σ(A) = λmin (A∗ A).
• σ(A) = 1/σ(A−1 ).
• σ(A) = 1/σ(A−1 ).
1.3 Laplace Transform 11
1.3 Laplace Transform

1.3.1 Definition
The Laplace transform converts time-domain signals into a frequency-domain
equivalent. The signal y(t) has transform Y (s) defined as follows:
Z ∞
Y (s) = L(y(t)) = y(τ )e−sτ dτ,
0
where s is an unspecified complex number; Y (s) is considered to be complex as

a result. Note that the Laplace transform is linear, and so it is is distributive:
L(x(t) + y(t)) = L(x(t)) + L(y(t)) will hold throughout. The following table
gives a list of some useful transform pairs and other properties, for reference.
The last two properties are of special importance: for control system design,
the differentiation of a signal is equivalent to multiplication of its Laplace
transform by s; integration of a signal is equivalent to division by s. The other
terms that arise will cancel if y(0) = 0, or if y(0) is finite, so they are usually
ignored.
1.3.2 Convergence
We note first that the value of s affects the convergence of the integral. For
instance, if y(t) = et , then the integral converges only for Re(s) > 1, since
the integrand is e1−s in this case. Convergence issues are not a problem for
evaluation of the Laplace transform, however, because of analytic continuation.
This result from complex analysis holds that if two complex functions are
equal on some arc (or line) in the complex plane, then they are equivalent
everywhere. This fact allows us to always pick a value of s for which the
integral above converges, and then by extension infer the existence of the
general transform.
1.3.3 Convolution Theorem

One of the main points of the Laplace transform is the ease of dealing with
dynamic systems. As with the Fourier transform, the convolution of two sig-
nals in the time domain corresponds with the multiplication of signals in the
frequency domain. Consider a system whose impulse response is g(t), being
12 1 MATH FACTS
y(t) ←→ Y (s)
(Impulse) δ(t) ←→ 1
1
(Unit Step) 1(t) ←→
s
1
(Unit Ramp) t ←→ 2
s
−αt 1
e ←→
s+α
ω
sin ωt ←→ 2
s + ω2
s
cos ωt ←→ 2
s + ω2
ω
e−αt sin ωt ←→
(s + α)2 + ω 2
s+α
e−αt cos ωt ←→
(s + α)2 + ω 2
1 1
(e−at − e−bt ) ←→
b−a (s + a)(s + b)
1 1 1

1+ (be−at − ae−bt ) ←→
ab a−b s(s + a)(s + b)
ωn −ζωn t
q ωn2
√ e sin ωn 1 − ζ 2 t ←→
1 − ζ2 s2 + 2ζωn s + ωn2
1 ωn2
q
−ζωn t
1− √ e sin ωn 1 − ζ 2t + φ ←→
1 − ζ2 s(s2 + 2ζωn s + ωn2 )
√ !
−1 1 − ζ2
φ = tan
ζ
(Pure Delay) y(t − τ )1(t − τ ) ←→ Y (s)e−sτ

dy(t)
(Time Derivative) ←→ sY (s) − y(0)
dt R 0+
Z t
Y (s) y(t)dt
(Time Integral) y(τ )dτ ←→ + 0−
0 s s
1.3 Laplace Transform 13
driven by an input signal x(t); the output is y(t) = g(t)∗x(t). The Convolution
Theorem is
Z t
y(t) = g(t − τ )x(τ )dτ ⇐⇒ Y (s) = G(s)X(s).
0
Here’s the proof given by Siebert:
Z ∞
Y (s) = y(t)e−st dt
0
Z ∞ Z t
= g(t − τ ) x(τ ) dτ e−st dt
0 0
Z ∞ Z ∞
= g(t − τ ) 1(t − τ ) x(τ ) dτ e−st dt
0 0
Z ∞ Z ∞
−st
= x(τ ) g(t − τ ) 1(t − τ ) e dt dτ
Z0∞ 0
= x(τ ) G(s)e−sτ dτ
0
= G(s)X(s)
When g(t) is the impulse response of a dynamic system, then y(t) represents
the output of this system when it is driven by the external signal x(t).
1.3.4 Solution of Differential Equations by Laplace Transform

The Convolution Theorem allows one to solve (linear time-invariant) differen-
tial equations in the following way:
1. Transform the system impulse response g(t) into G(s), and the input
signal x(t) into X(s), using the transform pairs.
2. Perform the multiplication in the Laplace domain to find Y (s).
3. Ignoring the effects of pure time delays, break Y (s) into partial fractions
with no powers of s greater than 2 in the denominator.
4. Generate the time-domain response from the simple transform pairs.
Apply time delay as necessary.
Specific examples of this procedure are given in a later section on transfer
functions.
14 2 KINEMATICS OF MOVING FRAMES
2 KINEMATICS OF MOVING FRAMES

2.1 Rotation of Reference Frames
We say that a vector expressed in the inertial frame has coordinates ~x, and
in a body-reference frame ~xb . For the moment, we assume that the origins of
these frames are coincident, but that the body frame has a different angular
orientation. The angular orientation has several well-known descriptions, in-
cluding the Euler angles and the Euler parameters (quaternions). The former
method involves successive rotations about the principle axes, and has a solid
link with the intuitive notions of roll, pitch, and yaw. Quaternions present a
more elegant and robust method, but with more abstraction. We will develop
the equations of motion using Euler angles.
Tape three pencils together to form a right-handed three-dimensional coordi-
nate system. Successively rotating the system about three of its own principle
axes, it is easy to see that any possible orientation can be achieved. For ex-
ample, consider the sequence of [yaw, pitch, roll]: starting from an orientation
identical to some inertial frame, rotate the movable system about its yaw axis,
then about the new pitch axis, then about the newer still roll axis. Needless
to say, there are many valid Euler angle rotation sets possible to reach a given
orientation; some of them might use the same axis twice.
z=z’ z’’
z’’’
y’’’ x’
y x
y’=y’’ x’’=x’’’
Figure 1: Successive application of three Euler angles transforms the original

coordinate frame into an arbitrary orientation.
A first question is: what is the coordinate of a point fixed in inertial space,
referenced to a rotated body frame? The transformation takes the form of
a 3×3 matrix, which we now derive through successive rotations of the three
2.1 Rotation of Reference Frames 15
Euler angles. Before the first rotation, the body-referenced coordinate matches
that of the inertial frame: ~xb0 = ~x. Now rotate the movable frame yaw axis
(z) through an angle φ. We have
 
cos φ sin φ 0
~xb1 =  − sin φ cos φ 0  ~xb0 = R(φ)~xb0 . (1)
 
0 0 1
Rotation about the z-axis does not change the z-coordinate of the point; the
other axes are modified according to basic trigonometry. Now apply the second
rotation, pitch about the new y-axis by the angle θ:
 
cos θ 0 − sin θ
2
~xb = 

0 1 0 
~xb1 = R(θ)~xb1 . (2)
sin θ 0 cos θ
Finally, rotate the body system an angle ψ about its newest x-axis:
 
1 0 0
3
~xb =  0 cos ψ sin ψ0 
~xb2 = R(ψ)~xb2 . (3)

0 − sin ψ cos ψ
This represents the location of the original point, in the fully-transformed
body-reference frame, i.e., ~xb3 . We will use the notation ~xb instead of ~xb3 from
here on. The three independent rotations can be cascaded through matrix
multiplication (order matters!):
~xb = R(ψ)R(θ)R(φ)~x (4)

 
cθcφ cθsφ −sθ
=  −cψsψ + sψsθcφ cψcφ + sψsθsφ sψcθ 
~x

sψsφ + cψsθcφ −sψcφ + cψsθsψ cψcθ

= R(φ, θ, ψ)~x.
All of the transformation matrices, including R(φ, θ, ψ), are orthonormal: their
inverse is equivalent to their transpose. Additionally, we should note that the
rotation matrix R is universal to all representations of orientation, including
quaternions. The roles of the trigonometric functions, as written, are specific
to Euler angles, and to the order in which we performed the rotations.
In the case that the movable (body) reference frame has a different origin than
the inertial frame, we have
~x = ~x0 + RT ~xb , (5)

where ~x0 is the location of the moving origin, expressed in inertial coordinates.
2.2 Differential Rotations

Now consider small rotations from one frame to another; using the small angle
assumption to ignore higher-order terms gives
 
1 δφ −δθ
R '  −δφ

1 δψ  (6)
δθ −δψ 1
 
0 δφ −δθ
=  −δφ 0 δψ  + I3×3 .

δθ −δψ 0
R comprises the identity plus a part equal to the (negative) cross-product

~
operator [−δ E×], ~ = [δψ, δθ, δφ], the vector of Euler angles ordered
where δ E
with the axes [x, y, z]. Small rotations are completely decoupled; the order of
the small rotations does not matter. Since R−1 = RT , we have also R−1 =
~
I3×3 + δ E×;
~xb = ~x − δ E~ × ~x (7)
~ × ~xb .
~x = ~xb + δ E (8)
We now fix the point of interest on the body, instead of in inertial space,
calling its location in the body frame ~r (radius). The differential rotations
occur over a time step δt, so that we can write the location of the point before
and after the rotation, with respect to the first frame as follows:
~x(t) = ~r (9)
T ~ × ~r.
~x(t + δt) = R ~r = ~r + δ E
Dividing by the differential time step gives

2.2 Differential Rotations 17
δ~x δE~
= × ~r (10)
δt δt
= ω × ~r,
~
where the rotation rate vector ω ' dE/dt because the Euler angles for this
infinitesimal rotation are small and decoupled. This same cross-product rela-
tionship can be derived in the second frame as well:
~ × ~r
~xb (t) = R~r = ~r − δ E (11)
~xb (t + δt) = ~r.
such that
δ~xb δE~
= × ~r (12)
δt δt
= ω × ~r,
On a rotating body whose origin point is fixed, the time rate of change of a
constant radius vector is the cross-product of the rotation rate vector ω~ and
the radius vector itself. The resultant derivative is in the moving body frame.
In the case that the radius vector changes with respect to the body frame, we
need an additional term:
d~xb ∂~r
= ω × ~r + . (13)
dt ∂t
Finally, allowing the origin to move as well gives
d~xb ∂~r d~xo

= ω × ~r + + . (14)
dt ∂t dt
This result is often written in terms of body-referenced velocity ~v :
∂~r
~v = ω × ~r + + ~vo , (15)
∂t
where ~vo is the body-referenced velocity of the origin. The total velocity of the
particle is equal to the velocity of the reference frame origin, plus a component
due to rotation of this frame. The velocity equation can be generalized to any
body-referenced vector f~:
df~ ∂f
= ~ × f~.
+ω (16)
dt ∂t
2.3 Rate of Change of Euler Angles

Only for the case of infinitesimal Euler angles is it true that the time rate
of change of the Euler angles equals the body-referenced rotation rate. For
example, with the sequence [yaw,pitch,roll], the Euler yaw angle (applied first)
is definitely not about the final body yaw axis; the pitch and roll rotations
moved the axis. An important part of any simulation is the evolution of the
Euler angles. Since the physics determine rotation rate ω ~ , we seek a mapping
~
~ → dE/dt.
ω
The idea is to consider small changes in each Euler angle, and determine the
effects on the rotation vector. The first Euler angle undergoes two additional
rotations, the second angle one rotation, and the final Euler angle no additional
rotations:
dψ
     
 0 
   0 
  
 dt


dθ
ω
~ = R(ψ)R(θ) 0 + R(ψ) dt
+ 0  (17)
 dφ    

dt
 
0  
0 
 dψ
 
1 0 − sin θ   dt


dθ
 0
=  cos ψ sin ψ cos θ 
 dt
.
dφ
0 − sin ψ cos ψ cos θ

 

dt
Taking the inverse gives
 
~ 1 sin ψ tan θ cos ψ tan θ
dE
=  0

cos ψ − sin ψ ω
~ (18)
dt
0 sin ψ/ cos θ cos ψ/ cos θ
~ ω.
= Γ(E)~
Singularities exist in Γ at θ = {π/2, 3π/2}, because of the division by cos θ, and

hence this otherwise useful equation for propagating the angular orientation of
a body fails when the vehicle rotates about the intermediate y-axis by ninety
2.4 Dead Reckoning 19
degrees. In applications where this is a real possibility, for example in orbiting

satellites and robotic arms, quaternions provide a seamless mapping. For most
ocean vessels, the singularity is acceptable, as long as it is not on the yaw axis!
2.4 Dead Reckoning

The measurement of heading and longitudinal speed gives rise to one of the
oldest methods of navigation: dead reckoning. Quite simply, if the estimated
longitudinal speed over ground is U , and the estimated heading is φ, ignoring
the lateral velocity leads to the evolution of Cartesian coordinates:
ẋ = U cos φ
ẏ = U sin φ.
Needless to say, currents and vehicle sideslip will cause this to be in error.
Nonetheless, some of the most remarkable feats of navigation in history have
depended on dead reckoning.
20 3 VESSEL INERTIAL DYNAMICS
3 VESSEL INERTIAL DYNAMICS

We consider the rigid body dynamics with a coordinate system affixed on the
body. A common frame for ships, submarines, and other marine vehicles has
the body-referenced x-axis forward, y-axis to port (left), and z-axis up. This
will be the sense of our body-referenced coordinate system here.
3.1 Momentum of a Particle

Since the body moves with respect to an inertial frame, dynamics expressed in
the body-referenced frame need extra attention. First, linear momentum for
a particle obeys the equality
d
F~ = (m~v ) (19)
dt
A rigid body consists of a large number of these small particles, which can be
indexed. The summations we use below can be generalized to integrals quite
easily. We have
~ i = d (mi~vi ) ,
F~i + R (20)
dt
where F~i is the external force acting on the particle and R ~ i is the net force
exerted by all the other surrounding particles (internal forces). Since the
collection of particles is not driven apart by the internal forces, we must have
equal and opposite internal forces such that
N
~ i = 0.
X
R (21)
i=1
Then summing up all the particle momentum equations gives

N N
d
F~i =
X X
(mi~vi ) . (22)
i=1 i=1 dt
Note that the particle velocities are not independent, because the particles are
rigidly attached.
Now consider a body reference frame, with origin 0, in which the particle i
resides at body-referenced radius vector ~r; the body translates and rotates,
and we now consider how the momentum equation depends on this motion.
3.2 Linear Momentum in a Moving Frame 21
z, w, φ
x, u, ψ
y, v, θ
Figure 2: Convention for the body-referenced coordinate system on a vessel:

x is forward, y is sway to the left, and z is heave upwards. Looking forward
from the vessel bridge, roll about the x axis is positive counterclockwise, pitch
about the y-axis is positive bow-down, and yaw about the z-axis is positive
turning left.
3.2 Linear Momentum in a Moving Frame

The expression for total velocity may be inserted into the summed linear mo-
mentum equation to give
N N
d
F~i =
X X
~ × ~ri ))
(mi (~vo + ω (23)
i=1 i=1 dt
N
" #
∂~vo d X
= m + ~×
ω mi~ri ,
∂t dt i=1
where m = N vi = ~vo +~ω ×~ri . Further defining the center of gravity

P
i=1 mi , and ~
vector ~rG such that
N
X
m~rG = mi~ri , (24)
i=1
we have
N
∂~vo d
F~i = m
X
+ m (~ω × ~rG ). (25)
i=1 ∂t dt
Using the expansion for total derivative again, the complete vector equation
in body coordinates is
!
∂~vo d~ω
F~ =
X
N =m ~ × ~vo +
+ω × ~rG + ω
~ × (~ω × ~rG ) . (26)
i=1 ∂t dt
Now we list some conventions that will be used from here on:
~vo = {u, v, w} (body-referenced velocity)

~rG = {xG , yG , zg } (body-referenced location of center of mass)
ω
~ = {p, q, r} (rotation vector, in body coordinates)
F~ = {X, Y, Z} (external force, body coordinates).
The last term in the previous equation simplifies using the vector triple product
identity
~ × (~ω × ~rG ) = (~ω · ~rG )~ω − (~ω · ω

ω ~ )~rG ,
and the resulting three linear momentum equations are
" #
∂u dq dr
X = m + qw − rv + zG − yG + (qyG + rzG )p − (q 2 + r2 )xG (27)
∂t dt dt
" #
∂v dr dp 2 2
Y = m + ru − pw + xG − zG + (rzG + pxG )q − (r + p )yG
∂t dt dt
" #
∂w dp dq 2 2
Z = m + pv − qu + yG − xG + (pxG + qyG )r − (p + q )zG .
∂t dt dt
Note that about half of the terms here are due to the mass center being in a
different location than the reference frame origin, i.e., ~rG 6= ~0.
3.3 Example: Mass on a String

Consider a mass on a string, being swung around around in a circle at speed
U , with radius r. The centrifugal force can be computed in at least three
different ways. The vector equation at the start is
!
∂~vo d~ω
F~ = m ~ × ~vo +
+ω × ~rG + ω
~ × (~ω × ~rG ) .
∂t dt
3.3 Example: Mass on a String 23
3.3.1 Moving Frame Affixed to Mass

Affixing a reference frame on the mass, with the local x oriented forward and
y inward towards the circle center, gives
~vo = {U, 0, 0}T

ω
~ = {0, 0, U/r}T
~rG = {0, 0, 0}T
∂~vo
= {0, 0, 0}T
∂t
∂~ω
= {0, 0, 0}T ,
∂t
such that
F~ = m~ω × ~vo = m{0, U 2 /r, 0}T .
The force of the string pulls in on the mass to create the circular motion.
3.3.2 Rotating Frame Attached to Pivot Point

Affixing the moving reference frame to the pivot point of the string, with the
same orientation as above but allowing it to rotate with the string, we have
~vo = {0, 0, 0}T

ω
~ = {0, 0, U/r}T
~rG = {0, r, 0}T
∂~vo
= {0, 0, 0}T
∂t
∂~ω
= {0, 0, 0}T ,
∂t
giving the same result:
F~ = m~ω × (~ω × ~rG ) = m{0, U 2 /r, 0}T .

3.3.3 Stationary Frame

A frame fixed in inertial space, and momentarily coincident with the frame
on the mass (3.3.1), can also be used for the calculation. In this case, as the
string travels through a small arc δψ, vector subtraction gives
δ~v = {0, U sin δψ, 0}T ' {0, U δψ, 0}T .

Since ψ̇ = U/r, it follows easily that in the fixed frame d~v /dt = {0, U 2 /r, 0}T ,
as before.
3.4 Angular Momentum

For angular momentum, the summed particle equation is
N N
X
~ i + ~ri × F~i ) =
X d
(M ~ri × (mi~vi ), (28)
i=1 i=1 dt
~ i is an external moment on the particle i. Similar to the case for linear
where M
momentum, summed internal moments cancel. We have
N N N
" # !
X
~ i + ~ri × F~i ) =
X ∂~vo X ∂~ω
(M mi~ri × ~ × ~vo +
+ω mi~ri × × ~ri +
i=1 i=1 ∂t i=1 ∂t
N
X
mi~ri × (~ω × (~ω × ~ri )).
i=1
The summation in the first term of the right-hand side is recognized simply as
m~rG , and the first term becomes
" #
∂~vo
m~rG × +ω~ × ~vo . (29)
∂t
The second term expands as (using the triple product)
N N
! ! !
X ∂~ω X ∂~ω ∂~ω
mi~ri × × ~ri = mi (~ri · ~ri ) − · ~ri ~ri (30)
i=1 ∂t i=1 ∂t ∂t
 PN 2 2

 Pi=1 mi ((yi + zi )ṗ − (yi q̇ + zi ṙ)xi )
 

=  N 2 2
i=1 mi ((xi + zi )q̇ − (xi ṗ + zi ṙ)yi ) .
 PN 2 2

i=1 mi ((xi + yi )ṙ − (xi ṗ + yi q̇)zi ) 
3.4 Angular Momentum 25
Employing the definitions of moments of inertia,
 
Ixx Ixy Ixz
 Iyx Iyy Iyz 
I =  (inertia matrix)

Izx Izy Izz

N
mi (yi2 + zi2 )
X
Ixx =
i=1
N
mi (x2i + zi2 )
X
Iyy =
i=1
N
mi (x2i + yi2 )
X
Izz =
i=1
N
X
Ixy = Iyx = − mi xi yi (cross-inertia)
i=1
N
X
Ixz = Izx = − mi xi zi
i=1
XN
Iyz = Izy = − mi yi zi ,
i=1
the second term of the angular momentum right-hand side collapses neatly
into I∂~ω /∂t. The third term can be worked out along the same lines, but
offers no similar condensation:
N
X N
X
mi~ri × ((~ω · ~ri )~ω − (~ω · ω
~ )~ri ) = mi~ri × ω
~ (~ω · ~ri ) (31)
i=1 i=1
 PN 
 P i=1 mi (yi r − zi q)(xi p + yi q + zi r)
 

N
= i=1 mi (zi p − xi r)(xi p + yi q + zi r)
 N 
mi (xi q − yi p)(xi p + yi q + zi r) 
 P
i=1
2 2
 
 Iyz (q − r ) + Ixz pq − Ixy pr 
 
= I (r2 − p2 ) + I rq − I pq
xz xy yz +
Ixy (p2 − q 2 ) + Iyz pr − Ixz qr 

 
 
 (Izz − Iyy )rq 
 
(Ixx − Izz )rp .
(Iyy − Ixx )qp 

 
Letting M~ = {K, M, N } be the total moment acting on the body, i.e., the left
side of Equation 28, the complete moment equations are
K = Ixx ṗ + Ixy q̇ + Ixz ṙ + (32)

(Izz − Iyy )rq + Iyz (q 2 − r2 ) + Ixz pq − Ixy pr +
m [yG (ẇ + pv − qu) − zG (v̇ + ru − pw)]
M = Iyx ṗ + Iyy q̇ + Iyz ṙ +

(Ixx − Izz )pr + Ixz (r2 − p2 ) + Ixy qr − Iyz qp +
m [zG (u̇ + qw − rv) − xG (ẇ + pv − qu)]
N = Izx ṗ + Izy q̇ + Izz ṙ +

(Iyy − Ixx )pq + Ixy (p2 − q 2 ) + Iyz pr − Ixz qr +
m [xG (v̇ + ru − pw) − yG (u̇ + qw − rv)] .
3.5 Example: Spinning Book

Consider a homogeneous rectangular block with Ixx < Iyy < Izz and all off-
diagonal moments of inertia are zero. The linearized angular momentum equa-
tions, with no external forces or moments, are
dp
Ixx + (Izz − Iyy )rq = 0
dt
dq
Iyy + (Ixx − Izz )pr = 0
dt
dr
Izz + (Iyy − Ixx )qp = 0.
dt
We consider in turn the stability of rotations about each of the main axes,
with constant angular rate Ω. The interesting result is that rotations about
the x and z axes are unstable, while rotation about the y axis is not.
3.5.1 x-axis
In the case of the x-axis, p = Ω + δp, q = δq, and r = δr, where the δ prefix
indicates a small value compared to Ω. The first equation above is uncoupled
3.5 Example: Spinning Book 27
from the others, and indicates no change in δp, since the small term δqδr can
be ignored. Differentiate the second equation to obtain
∂ 2 δq ∂δr
Iyy 2
+ (Ixx − Izz )Ω =0
∂t ∂t
Substitution of this result into the third equation yields
∂ 2 δq
Iyy Izz + (Ixx − Izz )(Ixx − Iyy )Ω2 δq = 0.
∂t2
√
A simpler expression is δ q̈ + αδq = 0, which has response δq(t) = δq(0)e −αt ,
when δ q̇(0) = 0. For spin about the x-axis, both coefficients of the differential
equation are positive, and hence α > 0. The imaginary
√ exponent indicates
that the solution is of the form δq(t) = δq(0)cos αt, that is, it oscillates but
does not grow. Since the perturbation δr is coupled, it too oscillates.
3.5.2 y-axis
Now suppose q = Ω + δq: differentiate the first equation and substitute into
the third equation to obtain
∂ 2 δp
Izz Ixx + (Iyy − Ixx )(Iyy − Izz )Ω2 δp = 0.
∂t2
Here the second coefficient has negative sign, and therefore α < 0. The ex-
ponent√ is real now, and the solution grows without bound, following δp(t) =
δp(0)e −αt .
3.5.3 z-axis
Finally, let r = Ω + δr: differentiate the first equation and substitute into the
second equation to obtain
∂ 2 δp
Iyy Ixx + (Ixx − Izz )(Iyy − Izz )Ω2 δp = 0.
∂t2
The coefficients are positive, so bounded oscillations occur.
3.6 Parallel Axis Theorem

Often, the mass center of an body is at a different location than a more con-
venient measurement point, the geometric center of a vessel for example. The
parallel axis theorem allows one to translate the mass moments of inertia refer-
enced to the mass center into another frame with parallel orientation, and vice
versa. Sometimes a translation of coordinates to the mass center will make the
cross-inertial terms Ixy , Iyz , Ixz small enough that they can be ignored; in this
case ~rG = ~0 also, so that the equations of motion are significantly reduced, as
in the spinning book example.
The formulas are:
Ixx = I¯xx + m(δy 2 + δz 2 ) (33)

Iyy = I¯yy + m(δx2 + δz 2 )
Izz = I¯zz + m(δx2 + δy 2 )
Iyz = I¯yz − mδyδz
Ixz = I¯xz − mδxδz
Ixy = I¯xy − mδxδy,
where I¯ represents an MMOI in the axes of the mass center, and δx, for ex-
ample, is the translation of the x-axis to the new frame. Note that translation
of MMOI using the parallel axis theorem must be either to or from a frame
resting exactly at the center of gravity.
3.7 Basis for Simulation

Except for external forces and moments F~ and M ~ , we now have the neces-
sary terms for writing a full nonlinear simulation of a rigid body, in body
coordinates. There are twelve states, comprising the following components:
• ~vo , the vector of body-referenced velocities.
• ω
~ , body rotation rate vector.
• ~x, location of the body origin, in inertial space.

~ Euler angle vector.
• E,
3.7 Basis for Simulation 29
The derivatives of body-referenced velocity and rotation rate come from Equa-
tions 27 and 32, with some coupling which generally requires a 6 × 6 matrix
inverse. The Cartesian position propagates according to
~x˙ = RT (E)~
~ vo , (34)
while the Euler angles follow:
~˙ = Γ(E)~
E ~ ω. (35)
30 4 HYDRODYNAMICS: INTRODUCTION
4 HYDRODYNAMICS: INTRODUCTION
The forces and moments on a vessel are complicated functions of many fac-
tors, including water density, viscosity, surface tension, pressure, vapor pres-
sure, and motions of the body. The most important factors for large ocean
vehicles are density and motion, and we can make simplifications to param-
eterize the most prominent relationships. This section pertains to the use of
hydrodynamic coefficients for predicting hydrodynamic response.
4.1 Taylor Series and Hydrodynamic Coefficients

Recall the Taylor expansion of a function:
∂f (xo ) 1 ∂ 2 f (xo )
f (x) = f (xo ) + (x − xo ) + (x − xo )2 + · · · . (36)
∂x 2! ∂x2
We introduce the notation
∂f (xo )
fx =
∂x
1 ∂ 2 f (xo )
fx x = ,
2! ∂x2
and so on, so that a two-variable Taylor expansion would have the form
f (x, y) = f (xo , yo ) + (37)

fx (x − xo ) + fy (y − yo ) +
fxx (x − xo )2 + fyy (y − yo )2 + fxy (x − xo )(y − yo ) +
fxxx (x − xo )3 + · · · . (38)
Note that all of the factorials are included in the coefficients. This notation
covers some instances where the formal Taylor series is meaningless, but the
notation is still clear. As one example, fluid drag is often written as
1
F = ρCd Au|u| = Fu|u| u|u|.
2
4.2 Surface Vessel Linear Model 31
4.2 Surface Vessel Linear Model

We now discuss some of the hydrodynamic parameters which govern a ship
maneuvering in the horizontal plane. The body x-axis is forward and the y-
axis is to port, so positive r has the boat turning left. We will consider motions
only in the horizontal plane, which means θ = ψ = p = q = w = 0. Since the
vessel is symmetric about the x − z plane, yG = 0; zG is inconsequential. We
then have at the outset
!
∂u
X = m − rv − xG r2 (39)
∂t
!
∂v ∂r
Y = m + ru + xG
∂t ∂t
!
∂r ∂v
N = Izz + mxG + ru .
∂t ∂t
Letting u = U + u, where U >> u, and eliminating higher-order terms, this

set is
∂u
X = m (40)
∂t !
∂v ∂r
Y = m + rU + xG
∂t ∂t
!
∂r ∂v
N = Izz + mxG + rU .
∂t ∂t
A number of coefficients can be discounted. First, in a homogeneous sea, with

no current, wave, or wind effects, {Xx , Xy , Xφ , Yx , Yy , Yφ , Nx , Ny , Nφ } are all
zero. We assume that no hydrodynamic forces depend on the position of the
vessel.2 Second, consider Xv : since this longitudinal force would have the same
sign regardless of the sign of v (because of side-to-side hull symmetry), it must
have zero slope with v at the origin. Thus Xv = 0. The same argument shows
that {Xr , Xv̇ , Xṙ , Yu , Yu̇ , Nu , Nu̇ } = 0. Finally, since fluid particle acceleration
relates linearly with pressure or force, we do not consider nonlinear acceleration
2
Note that the linearized heave/pitch dynamics of a submarine do depend on the pitch
angle; this topic will be discussed later.
terms, or higher time derivatives. It should be noted that some nonlinear terms
related to those we have eliminated above are not zero. For instance, Yuu = 0
because of hull symmetry, but in general Xvv = 0 only if the vessel is bow-stern
symmetric.
We have so far, considering only the linear hydrodynamic terms,
(m − Xu̇ )u̇ = Xu u + X 0 (41)

(m − Yv̇ )v̇ + (mxG − Yṙ )ṙ = Yv v + (Yr − mU )r + Y 0 (42)
(mxG − Nv̇ )v̇ + (Izz − Nṙ )ṙ = Nv v − (Nr − mxG U )r + N 0 . (43)
The right side here carries also the imposed forces from a thruster(s) and
rudder(s) {X 0 , Y 0 , N 0 }. Note that the surge equation is decoupled from the
sway and yaw, but that sway and yaw themselves are coupled, and there-
fore are of immediate interest. With the state vector ~s = {v, r} and exter-
nal force/moment vector F~ = {Y 0 , N 0 }, a state-space representation of the
sway/yaw system is
" # " #
m − Yv̇ mxG − Yṙ d~s Yv Yr − mU
= ~s + F~ , or (44)
mxG − Nv̇ Izz − Nṙ dt Nv Nr − mxG U
M ~s˙ = P~s + F~
~s˙ = M −1 P~s + M −1 F~
~s˙ = A~s + B F~ . (45)
The matrix M is a mass or inertia matrix, which is always invertible. The

last form of the equation is a standard one wherein A represents the internal
dynamics of the system, and B is a gain matrix for the control and disturbance
inputs.
4.3 Stability of the Sway/Yaw System

Consider the homogeneous system ~s˙ = A~s:
ṡ1 = A11 s1 + A12 s2

ṡ2 = A21 s1 + A22 s2 .
4.3 Stability of the Sway/Yaw System 33
We can rewrite the second equation as

!−1
d(·)
s2 = − A22 A21 s1 (46)
dt
and substitute into the first equation to give
s̈1 + (−A11 − A22 )ṡ1 + (A11 A22 − A12 A21 )s1 = 0. (47)
Note that these operations are allowed because the derivative operator is linear;
in the language of the Laplace transform, we would simply use s. A necessary
and sufficient condition for stability of this ODE system is that each coefficient
must be greater than zero:
−A11 − A22 > 0 (48)

A11 A22 − −A12 A21 > 0
The components of A for the sway/yaw problem are
(Izz − Nṙ )Yv + (Yṙ − mxG )Nv

A11 = (49)
(m − Yv̇ )(Izz − Nṙ ) − (mxG − Yṙ )(mxG − Nv̇ )
−(Izz − Nṙ )(mU − Yr ) − (Yṙ − mxG )(mxG U − Nr )
A12 =
(Nv̇ − mxG )Yv + (m − Yv̇ )Nv
A21 =
−(Nv̇ − mxG )(mU − Yr ) − (m − Yv̇ )(mxG U − Nr )
A22 = .
The denominators are identical, and can be simplified. First, let xG ' 0;
valid for many vessels with the origin is at the geometric center. If the origin
is at the center of mass, xG = 0. Next, if the vessel is reasonably balanced
with regard to forward and aft areas with respect to the origin, the terms
{Nv̇ , Yṙ , Nv , Yr } take very small values in comparison with the others. To wit,
the added mass term −Yv̇ is of the order of the vessel’s material mass m,
and similarly Nṙ ' −Izz . Both Yv̇ and Nṙ take large negative values. Linear
drag and rotational drag are significant also; these are the terms Yv and Nr ,
both large and negative. The denominator for A’s components reduces to
(m − Yv̇ )(Izz − Nṙ ), and
Yv
A11 = m−Yv̇
< 0
Nr
A22 = Izz −Nṙ
< 0.
Hence the first condition for stability is met: −A11 − A22 > 0. For the second
condition, since the denominators of the Aij are identical, we have only to look
at the numerators. For stability, we require
(Izz − Nṙ )Yv (m − Yv̇ )Nr (50)

− [Nv̇ Yv + (m − Yv̇ )Nv ] [−(Izz − Nṙ )(mU − Yr ) + Yṙ Nr ] > 0.
The first term is the product of two large negative and two large positive
numbers. The second part of the second term contains mU , which has a large
positive value, generally making stability critical on the (usually negative) Nv .
When only the largest terms are considered for a vessel, a simpler form is
common:
C = Yv Nr + Nv (mU − Yr ) > 0. (51)

C is called the vessels stability parameter. The terms of C compete, and
yaw/sway stability depends closely on the magnitude and sign of Nv . Adding
more surface area aft drives Nv more positive, increasing stability as ex-
pected. Stability can also be improved by moving the center of gravity forward.
Nonzero xG shows up as follows:
C = Yv (Nr − mxG U ) + Nv (mU − Yr ) > 0. (52)

Since Nr and Yv are both negative, positive xG increases the (positive) influence
of C’s first term.
4.4 Basic Rudder Action in the Sway/Yaw Model

Rudders are devices which develop large lift forces due to an angle of attack
with respect to the oncoming fluid. As in our discussion of lift on the body,
the form is as follows: L = 12 ρAU 2 Cl (α), where α is the angle of attack. The
4.4 Basic Rudder Action in the Sway/Yaw Model 35
δ x
r
Figure 3: Convention for positive rudder angle in the vessel reference system.
lift coefficient Cl is normally linear with α near α = 0, but the rudder stalls
when the angle of attack reaches a critical value, and thereafter develops much
less lift. We will assume that α is small enough that the linear relationship
applies:

∂Cl
Cl (α) = α. (53)
∂α α=0
Since the rudder develops force (and a small moment) far away from the body
origin, say a distance l aft, the moment equation is quite simple. We have

1 ∂Cl
Yα = ρA U2 (54)
2 ∂α α=0

1 ∂Cl
Nα = − ρA lU 2 . (55)
2 ∂α α=0
Note the difference between the rudder angle expressed in the body frame, δ,
and the total angle of attack α. Angle of attack is influenced by δ, as well
as v/U and lr. Thus, in tank testing with v = 0, δ = α and Nδ = Nα ,
etc., but in real conditions, other hydrodynamic derivatives are augmented to
capture the necessary effects, for example Nv and Nr . Generally speaking, the
hydrodynamic characteristics of the vessel depend strongly on the rudder, even
when δ = 0. In this case the rudder still opposes yaw and sway perturbations
and acts to stabilize the vessel.
A positive rudder deflection (defined to have the same sense as the yaw angle)
causes a negative yaw perturbation, and a very small positive sway perturba-
tion.
4.4.1 Adding Yaw Damping through Feedback
The stability coefficient C resulting from the addition of a control law δ = kr r,

where kr > 0 is a feedback gain, is
C = Yv (Nr − mxG U + kr Nδ ) + Nv (mU − Yr − kr Yδ ). (56)
Yδ is small positive, but Nδ is large and negative. Hence C becomes more

positive, since Yv is negative.
Control system limitations and the stalling of rudders make obvious the fact
that even a very large control gain kr cannot completely solve stability prob-
lems of a poorly-designed vessel with an inadequate rudder. On the other
hand, a vessel which is overly stable (C >> 0 with no rudder action) is unma-
neuverable. A properly-balanced vessel just achieves stability with zero rudder
action, so that a reasonable amount of control will provide good maneuvering
capabilities.
4.4.2 Heading Control in the Sway/Yaw Model
Considering just the yaw equation of motion, i.e., v = 0, with a rudder, we

have
(Izz − Nṙ )φ̈ + (mxG U − Nr )φ̇ = Nδ δ. (57)
Employing the control law δ = kφ φ, the system equation becomes a homoge-

neous, second-order ODE:
(Izz − Nṙ )φ̈ + (mxG U − Nr )φ̇ − Nδ kφ φ = 0. (58)
Since all coefficients are positive (recall Nδ < 0), the equation gives a stable θ
response, settling under second-order dynamics to θ(∞) = 0. The control law
δ = kφ (φ − φdesired ) + kr r is the basis for heading autopilots, which are used
to track φdesired . This use of an error signal to drive an actuator is in fact the
essence of feedback control. In this case, we require sensors to obtain r and φ,
a controller to calculate δ, and an actuator to implement the corrective action.
4.5 Response of the Vessel to Step Rudder Input 37
4.5 Response of the Vessel to Step Rudder Input

4.5.1 Phase 1: Accelerations Dominate
When the rudder first moves, acceleration terms dominate, since the velocities
are zero. The equation looks like this:
" #( ) ( )
m − Yv̇ mxG − Yṙ v̇ Yδ
= δ. (59)
mxG − Nv̇ Izz − Nṙ ṙ Nδ
Since Yṙ and mxG are comparatively small in the first row, we have
Yδ δ
v̇(0) = , (60)
m − Yv̇
and the vessel moves to the left, the positive v-direction. The initial yaw is in
the negative r-direction, since Nδ < 0:
Nδ δ
ṙ(0) = . (61)
Izz − Nṙ
The first phase is followed by a period (Phase 2), in which many terms are
competing and contributing to the transient response.
4.5.2 Phase 3: Steady State

When the transients have decayed, the vessel is in a steady turning condition,
and the accelerations are zero. The system equations reduce to
( ) ( )
v δ (mxG U − Nr )Yδ + (Yr − mU )Nδ
= . (62)
r C N v Y δ − Y v Nδ
Note that the denominator is the stability parameter. The steady turning rate
is thus approximated by
Yv N δ
r=− δ. (63)
C
With C > 0, the steady-state yaw rate is negative. If the vessel is unstable
(C < 0), it turns in the opposite direction than expected. This turning rate
equation can also be used to estimate turning radius R:
U UC
R= = . (64)
r −Yv Nδ δ
The radius goes up directly with C, indicating that too stable a ship has poor
turning performance. We see also that increasing the rudder area increases
Nδ , decreasing R as desired. Increasing the deflection δ to reduce R works
only to the point of stalling.
4.6 Summary of the Linear Maneuvering Model

We conclude our discussion of the yaw/sway model by noting that
1. The linearized sway/yaw dynamics of a surface vessel are strongly cou-
pled, and they are independent of the longitudinal dynamics.
2. The design parameter C should be slightly greater than zero for easy
turning, and “ hands-off” stability. The case C < 0 should only be
considered under active feedback control.
3. The analysis is valid only up to small angles of attack and turning rates.
Very tight maneuvering requires the nonlinear inertial components and
hydrodynamic terms. Among other effects, the nonlinear equations cou-
ple surge to the other motions, and the actual vessel loses forward speed
during maneuvering.
4.7 Stability in the Vertical Plane

Stability in the horizontal plane changes very little as a function of speed,
because drag and lift effects generally scale with U 2 . This fact is not true
in the vertical plane, for which the dimensional weight/buoyancy forces and
moments are invariant with speed. For example, consider the case of heave
and pitch, with xG = 0 and no actuation:
!
∂w
m − Uq = Zẇ ẇ + Zw w + Zq̇ q̇ + Zq q + (B − W ) (65)
∂t
dq
Iyy = Mẇ ẇ + Mw w + Mq̇ q̇ + Mq q − Blb sin θ. (66)
dt
The last term in each equation is a hydrostatic effect induced by opposing net
buoyancy B and weight W . lb denotes the vertical separation of the center
of gravity and the center of buoyancy, creating the so-called righting moment
which nearly all underwater vehicles possess. Because buoyancy effects do not
4.7 Stability in the Vertical Plane 39
change with speed, the dynamic properties and hence stability of the vehicle
may change with speed.
40 5 SIMILITUDE
5 SIMILITUDE
5.1 Use of Nondimensional Groups
For a consistent description of physical processes, we require that all terms
in an equation must have the same units. On the basis of physical laws,
some quantities are dependent on other, independent quantities. We form
nondimensional groups out of the dimensional ones in this section, and apply
the technique to maneuvering.
The Buckingham π-theorem provides a basis for all nondimensionalization.
Let a quantity Qn be given as a function of a set of n − 1 other quantities:
Qn = fQ (Q1 , Q2 , · · · , Qn−1 ). (67)

There are n variables here, but suppose also that there are only k independent
ones; k is equivalent to the number of physical unit types encountered. The
theorem asserts that there are n − k dimensionless groups πi that can be
formed, and the functional equivalence is reduced to
πn−k = fπ (π1 , π2 , · · · , πn−k−1 ). (68)

Example. Suppose we have a block of mass m resting on a frictionless hor-
izontal surface. At time zero, a steady force of magnitude F is applied. We
want to know X(T ), the distance that the block has moved as of time T . The
dimensional function is X(T ) = fQ (m, F, T ), so n = 4. The (MKS) units are
[X(·)] = m
[m] = kg
[F ] = kgm/s2
[T ] = s,
and therefore k = 3. There is just one nondimensional group in this re-

lationship; π1 assumes only a constant (but unknown) value. Simple term-
cancellation gives π1 = X(T )m/F T 2 , not far at all from the known result that
X(T ) = F T 2 /2m!
Example. Consider the flow rate Q of water from an open bucket of height
h, through a drain nozzle of diameter d. We have
5.1 Use of Nondimensional Groups 41
Q = fQ (h, d, ρ, µ, g),
where the water density is ρ, and its absolute viscosity µ; g is the acceleration
due to gravity. No other parameters affect the flow rate. We have n = 6, and
the (MKS) units of these quantities are:
[Q] = m3 /s
[h] = m
[d] = m
[ρ] = kg/m3
[µ] = kg/ms
[g] = m/s2
There are only three units that appear: [length, time, mass], and thus k = 3.
Hence, only three non-dimensional groups exist, and one is a unique func-
tion of the other two. To arrive at a set of valid groups, we must create
three nondimensional quantities, making sure that each of the original (di-
mensional) quantities is represented. Intuition and additional manipulations
come in handy, as we now show. √
Three plausible first groups are: π1 = ρQ/dµ, π2 = dρ gh/µ, and π3 = h/d.
Note that all six quantities appear at least once. Since h and d have the same
units, they could easily change places in the first two groups. However, π1 is
recognized as a Reynolds number pertaining to the orifice flow. π2 is more
awkward, but products and fractions of groups √ are themselves valid groups,
2
and we may construct π4 = π1 /π2 = Q/d gh √ to nondimensionalize Q with
a pressure velocity, and then π5 = π1 /π4 = ρd gh/µ to establish an orifice
Reynolds number independent of Q. We finally have the useful result
π4 = fπ (π5 , π2 ) −→
√ !
Q ρd gh h
√ = fπ , .
d2 gh µ d
The uncertainty about where to use h and d, and the questionable importance
of h/d as a group are remnants of the theorem. Intuition is that h/d is im-
42 5 SIMILITUDE
material, and the other two terms have a nice physical meaning, e.g., π5 is a
Reynolds number.
The power of the π-theorem is primarily in reducing the number of parameters
which must be considered independently to characterize a process. In the flow
example, the theorem reduced the number of independent parameters from
five to two, with no constraints about the actual physics taking place.
5.2 Common Groups in Marine Engineering

One frequently encounters the following groups in fluid mechanics and marine
engineering:
1. Froude number:
U
Fr = √ , (69)
gL
where U is the speed of the vessel, g is the acceleration due to gravity,

and L is the waterline length of the vessel. The Froude number appears
in problems involving pressure boundary conditions, such as in waves on
the ocean surface. Roughly speaking, it relates the vessel speed U to
waterqwave speeds of wavelength L; the phase speed of a surface wave is
V = λg/2π, where λ is the wavelength.
2. Cavitation number:
P∞ − P v
δ= 1 , (70)
2
ρU 2
where P∞ represents the ambient total pressure, Pv the vapor pressure

of the fluid, and U the propeller inlet velocity. A low cavitation number
means that the Bernoulli pressure loss across the lifting surface will cause
the fluid to vaporize, causing bubbles, degradation of performance, and
possible deterioration of the material.
3. Reynolds number:
Ul
Re = , (71)
µ/ρ
5.2 Common Groups in Marine Engineering 43
where U is velocity, µ is absolute viscosity, and ρ is density. Since Re

appears in many applications, l represents one of many length scales.
Reynolds number is a ratio of fluid inertial pressures to viscous pres-
sures: When Re is high, viscous effects are negligible. Re can be used to
characterize pipe flow, bluff body wakes, and flow across a plate, among
others.
4. Weber number:
ρU 2 l
W = , (72)
σ
where σ is the surface tension of a fluid. Given that [σ] = N/m (MKS),
ρU 2 normalizes pressure, and l normalizes length. The Weber number
indicates the importance of surface tension.
To appreciate the origins of these terms from a fluid particle’s point of view,
consider a box having side lengths [dx, dy, dz]. Various forces on the box scale
as
∂v ∂v
(inertia) Fi = ρ dxdydz + ρv dxdydz ' ρU 2 l2
∂t ∂x
3
(gravity) Fg = ρgdxdydz ' ρgl
(pressure) Fp = P dxdy ' P l2
∂v
(shear) Fs = µ dxdy ' µU l
∂z
(surface tension) Fσ = σdx ' σl.
Thus the groups listed above can be written as
Fi U2
Fr = '
Fg gl
Fp P
δ= '
Fi ρU 2
Fi ρU l
Re = '
Fs µ
Fi ρU 2 l
W = '
Fσ σ
44 5 SIMILITUDE
When testing models, it is imperative to maintain as many of the nondimen-

sional groups as possible of the full-scale system. This holds for the geometry
of the body and the kinematics of the flow, the surface roughness, and the
all of the relevant groups governing fluid dynamics. Consider the example of
nozzle flow from a bucket. Suppose that we conduct a model test in which
Re is abnormally large, i.e., the viscous
√ effects are negligible. Under inviscid
conditions, the flow rate is Q = πd2 2gh/4. This rate cannot be achieved for
lower-Re conditions because of fluid drag in the orifice, however.
In a vessel, we write the functional relationship for drag as a starting point:
D
Cr = 1 = fQ (ρ, µ, g, σ, U, l)
2
ρAU 2
= fπ (Re, F r, W ).
First, since l is large, W is very large, and hence
√ surface tension plays no role.
Next, we look at Re = U l/ν and F r = U/ gl, both of which are important
for surface vessels. Suppose that lship = λlmodel , so that usually λ >> 1;
additionally, we set gmodel = gship , i.e., the model and the true vessel operate
in the same gravity field. √
Froude number similitude requires Umodel = Uship / λ. Then Reynolds number
scaling implies directly νmodel = νship /λ3/2 . Unfortunately, few fluids with this
property are workable in a large testing tank. As a result, accurate scaling of
Re for large vessels to model scale is quite difficult.
For surface vessels, and submarines near the surface, it is a routine procedure
to employ turbulence stimulators to achieve flow that would normally occur
with ship-scale Re. Above a critical value Re ' 500, 000, Cf is not sensitive
to Re. With this achieved, one then tries to match F r closely.
5.3 Similitude in Maneuvering

The linear equations of motion for the horizontal yaw/sway problem are:
(m − Yv̇ )v̇ − Yv v + (mU − Yr )r + (mxG − Yṙ )ṙ = Y

(Izz − Nṙ )ṙ + (mxG U − N r)r + (mxG − Nv̇ )v̇ − Nv v = N.
These equations can be nondimensionalized in a standard way, by using the

quantities [U, L, ρ]: these three values provide the necessary units of length,
5.3 Similitude in Maneuvering 45
time, and mass, and furthermore are readily accessible to the user. First, we
create nondimensional states, denoted with a prime symbol:
L
v̇ 0 = v̇ (73)
U2
1
v0 = v
U
L2
ṙ0 = ṙ
U2
L
r0 = r.
U
We follow a similar procedure for the constant terms as follows, including a
factor of 1/2 with ρ, for consistency with our previous expressions:
m
m0 = 1 (74)
2
ρL3
0 Izz
Izz = 1
2
ρL5
xG
x0G =
L
0 U
U = =1
U
Yv̇
Yv̇0 = 1
2
ρL3
Yv
Yv0 = 1
2
ρU L2
Yṙ
Yṙ0 = 1
2
ρL4
Yr
Yr0 = 1
2
ρU L3
Y
Y0 = 1
2
ρU 2 L2
Nv̇
Nv̇0 = 1
2
ρL4
Nv
Nv0 = 1
2
ρU L3
46 5 SIMILITUDE
Nṙ
Nṙ0 = 1
2
ρL5
Nr
Nr0 = 1
2
ρU L4
N
N0 = 1 .
2
ρU 2 L3
Note that every force has been normalized with 12 ρU 2 L2 , and every moment
with 21 ρU 2 L3 ; time has been also nondimensionalized with L/U . Thus we
arrive at a completely equivalent set of nondimensional system equations,
(m0 − Yv̇0 )v̇ 0 − Yv0 v 0 + (m0 U 0 − Yr0 )r0 + (m0 x0G − Yṙ0 )ṙ0 = Y 0 (75)
0
(Izz − Nṙ0 )ṙ0 + (m0 x0G U 0 − N 0 r0 )r0 + (m0 x0G − Nv̇0 )v̇ 0 − Nv0 v 0 = N 0 .
Since fluid forces and moments generally scale with U 2 , the nondimensionalized
description holds for a range of velocities.
5.4 Roll Equation Similitude

Certain nondimensional coefficients may arise which depend explicitly on U ,
and therefore require special attention. Let us carry out a similar normaliza-
tion of the simplified roll equation
(Ixx − Kṗ )ṗ − Kp p − Kψ ψ = K. (76)

For a surface vessel, the roll moment Kψ is based on metacentric stability, and
has the form Kψ = −ρg∇(GM ), where ∇ is the displaced fluid volume of the
vessel, and GM is the metacentric height. The nondimensional terms are
0 Ixx
Ixx = 1 (77)
2
ρL5
Kṗ
Kṗ0 = 1
2
ρL5
Kp
Kp0 = 1
2
ρU L4
Kψ
Kψ0 = 1
2
ρU 2 L3
5.4 Roll Equation Similitude 47
0 L2
ṗ = ṗ
U2
L
p0 = p,
U
leading to the equivalent system
2gL

0
(Ixx − Kṗ0 )ṗ0
− Kp0 p0
− ∇0 (GM 0 )φ = K 0 . (78)
U2
Note that the roll angle φ was not nondimensionalized. The Froude number
has a very strong influence on roll stability, since it appears explicitly in the
nondimensional righting moment term, and also has a strong influence on Kp0 .
In the case of a submarine, the righting moment has the form Kψ = −Bh,
where B is the buoyant force, and h is the righting arm. The nondimensional
coefficient becomes
Bh
Kψ0 = − 1 .
2
ρU 2 L3
Kψ0 again depends strongly on U , since B and h are fixed; this Kψ0 needs to be
maintained in model tests.
48 6 CAPTIVE MEASUREMENTS
6 CAPTIVE MEASUREMENTS
Before making the decision to measure hydrodynamic derivatives, a prelimi-
nary search of the literature may turn up useful estimates. For example, test
results for many hull-forms have already been published, and the basic lift-
ing surface models are not difficult. The available computational approaches
should be considered as well; these are very good for predicting added mass
in particular. Finally, modern sensors and computer control systems make
possible the estimation of certain coefficients based on open-water tests of a
model or a full-scale design.
In model tests, the Froude number F r = √UgL , which scales the influence
of surface waves, must be maintained between model and full-scale surface
vessels. Reynolds number Re = UνL , which scales the effect of viscosity, need
not be matched as long as the scale model attains turbulent flow (supercritical
Re). One can use turbulence stimulators near the bow if necessary. Since the
control surface(s) and propeller(s) affect the coefficients, they should both be
implemented in model testing.
6.1 Towtank
In a towtank, tow the vehicle at different angles of attack, measuring sway
force and yaw moment. The slope of the curve at zero angle determines Yv
and Nv respectively; higher-order terms can be generated if the points deviate
from a straight line. Rudder derivatives can be computed also by towing with
various rudder angles.
6.2 Rotating Arm Device

On a rotating arm device, the vessel is fixed on an arm of length R, rotating at
constant rate r: the vessel forward speed is U = rR. The idea is to measure the
crossbody force and yaw moment as a function of r, giving the coefficients Yr
and Nr . Note that the lateral force also contains the component (m − Yv̇ )r2 R.
The coefficients Yv and Nv can also be obtained by running with a fixed angle
of attack. Finally, the measurement is made over one rotation only, so that
the vessel does not re-enter its own wake.
6.3 Planar-Motion Mechanism 49
6.3 Planar-Motion Mechanism

With a planar motion mechanism, the vessel is towed at constant forward
speed U , but is held by two posts, one forward and one aft, which can each
impose independent sway motions, therefore producing variable yaw. The
model moves in pure sway if ya (t) = yb (t), in a pure yaw motion about the
mid-length point if ya (t) = −yb (t), or in a combination sway and yaw motion.
The connection points are a distance l forward and aft from the vessel origin.
Usually a sinusoidal motion is imposed:
ya (t) = a cos ωt (79)

yb (t) = b cos(ωt + ψ),
and the transverse forces on the posts are measured and approximated as
Ya (t) = Fa cos(ωt + θa ) (80)

Yb (t) = Fb cos(ωt + θb ).
If linearity holds, then
(m − Yv̇ )v̇ − Yv v + (mU − Yr )r + (mxG − Yṙ )ṙ = Ya + Yb (81)

(Izz − Nṙ )ṙ + (mxG U − Nr )r + (mxG − Nv̇ )v̇ − Nv v = (Yb − Ya )l.
We have v = (ẏa + ẏa )/2 and r = (ẏb − ẏa )/2l. When a = b, these become
aω
v = − (sin ωt(1 + cos ψ) + cos ωt sin ψ) (82)
2
aω 2
v̇ = − (cos ωt(1 + cos ψ) − sinωt sin ψ)
2
aω
r = − (sin ωt(cos ψ − 1) + cos ωt sin ψ)
2l
aω 2
ṙ = − (cos ωt(cos ψ − 1) − sin ωt sin ψ) .
2l
Equating the sine terms and then the cosine terms, we obtain four independent
equations:
50 6 CAPTIVE MEASUREMENTS
aω 2
!
(m − Yv̇ ) − (1 + cos ψ) − (83)
2
aω

Yv − sin ψ +
2
aω

(mU − Yr ) − sin ψ +
2l
aω 2
!
(mxG − Yṙ ) − (cos ψ − 1) = Fa cos θa + Fb cos θb
2l
aω 2
!
(m − Yv̇ ) − (− sin ψ) − (84)
2
aω

Yv − (1 + cos ψ) +
2
aω

(mU − Yr ) − (cos ψ − 1) +
2l
aω 2
!
(mxG − Yṙ ) − (− sin ψ) = −Fa sin θa − Fb sin θb
2l
aω 2
!
(Izz − Nṙ ) − (cos ψ − 1) + (85)
2l
aω

(mxG U − Nr ) − sin ψ +
2l
aω 2
!
(mxG − Nv̇ ) − (1 + cos ψ) −
2
aω

Nv − sin ψ = l(Fb cos θb − Fa cos θa )
2
aω 2
!
(Izz − Nṙ ) − (− sin ψ) + (86)
2l
aω

(mxG U − Nr ) − (cos ψ − 1) +
2l !
aω 2
(mxG − Nv̇ ) − (− sin ψ) −
2
aω

Nv − (1 + cos ψ) = l(−Fb sin θb + Fa sin θa )
2
6.3 Planar-Motion Mechanism 51
In this set of four equations, we know from the imposed motion the values
[U, ψ, a, ω]. From the experiment, we obtain [Fa , Fb , θa , θb ], and from the rigid-
body model we have [m, Izz , xG ]. It turns out that the two cases of ψ = 0
(pure sway motion) and ψ = 180o (pure yaw motion) yield a total of eight
independent equations, exactly what is required to find the eight coefficients
[Yv̇ , Yv , Yṙ , Yr , Nv̇ , Nv , Nṙ , Nr ]. Remarkably, we can write the eight solutions
directly: For ψ = 0,
aω 2
!
(m − Yv̇ ) − (2) = Fa cos θa + Fb cos θb (87)
2
aω

−Yv − (2) = −Fa sin θa − Fb sin θb
2!
aω 2
(mxG − Nv̇ ) − (2) = l(Fb cos θb − Fa cos θa )
2
aω

−Nv − (2) = l(−Fb sin θb + Fa sin θa ),
2
to be solved respectively for [Yv̇ , Yv , Nv̇ , Nv ]. For ψ = 180o , we have
aω 2
!
(mxG − Yṙ ) − (−2) = Fa cos θa + Fb cos θb (88)
2l
aω

(mU − Yr ) − (−2) = −Fa sin θa − Fb sin θb
2l !
aω 2
(Izz − Nṙ ) − (−2) = l(Fb cos θb − Fa cos θa )
2l
aω

(mxG U − Nr ) − (−2) = l(−Fb sin θb + Fa sin θa ),
2l
to be solved for [Yṙ , Yr , Nṙ , Nr ]. Thus, the eight linear coefficients for a surface
vessel maneuvering, for a given speed, can be deduced from two tests with
a planar motion mechanism. We note that the nonlinear terms will play a
significant role if the motions are too large, and that some curve fitting will be
needed in any event. The PMM can be driven with more complex trajectories
which will target specific nonlinear terms.
52 7 STANDARD MANEUVERING TESTS
7 STANDARD MANEUVERING TESTS

This section describes some of the typical maneuvering tests which are per-
formed on full-scale vessels, to assess stability and performance.
7.1 Dieudonné Spiral

1. Achieve steady speed and direction for one minute. No changes in speed
setting are made after this point.
2. Turn rudder quickly by 15◦ , and keep it there until steady yaw rate is
maintained for one minute.
3. Reduce rudder angle by 5◦ , and keep it there until steady yaw rate is
maintained for one minute.
4. Repeat in decrements of -5◦ , to -15◦ .
5. Proceed back up to 15◦ .
The Dieudonné maneuver has the vessel path following a growing spiral, and
then a contracting spiral in the opposite direction. The test reveals if the vessel
has a memory effect, manifested as a hysteresis in yaw rate r. For example,
suppose that the first 15◦ rudder deflection causes the vessel to turn right, but
that the yaw rate at zero rudder, on the way negative, is still to the right. The
vessel has gotten “stuck” here, and will require a negative rudder action to
pull out of the turn. But if the corrective action causes the vessel to turn left
at all, the same memory effect may occur. It is easy to see that the rudder in
this case has to be used excessively driving the vessel back and forth. We say
that the vessel is unstable, and clearly a poor design.
7.2 Zig-Zag Maneuver

1. With zero rudder, achieve steady speed for one minute.
2. Deflect the rudder to 20◦ , and hold until the vessel turns 20◦ .
3. Deflect the rudder to -20◦ , and hold until the vessel turns to -20◦ with
respect to the starting heading.
4. Repeat.
7.3 Circle Maneuver 53
This maneuver establishes several important characteristics of the yaw re-

sponse. These are: the response time (time to reach a given heading), the yaw
overshoot (amount the vessel exceeds ±20◦ when the rudder has turned the
other way), and the total period for the 20◦ oscillations. Of course, similar
tests can be made with different rudder angles and different threshold vessel
headings.
7.3 Circle Maneuver

From a steady speed, zero yaw rate condition, the rudder is moved to a new
setting. The vessel responds by turning in a circle. After steady state is
reached again, parameters of interest are the turning diameter, the drift angle
β, the speed loss, and the angle of heel ψ.
7.3.1 Drift Angle

The drift angle is the equivalent to angle of attack for lifting surfaces, and
describes how the vessel “skids” during a turn. If the turning circle has radius
R (measured from the vessel origin), then the speed tangential to the circle
is U = rR. The vessel-reference velocity components are thus u = U cos β
and v = −U sin β. A line along the vessel centerline reaches closest to the
true center of the turning circle at a point termed the turning center. At
this location, which may or may not exist on the physical vessel, there is no
apparent lateral velocity, and it appears to an observer there that the vessel
turns about this point.
7.3.2 Speed Loss

Speed loss occurs primarily because of drag induced by the drift angle. A
vessel which drifts very little may have very little speed loss.
7.3.3 Heel Angle

Heel during turning occurs as a result of the intrinsic coupling of sway, yaw,
and roll caused by the center of gravity. In a surface vessel, the fluid forces act
below the waterline, but the center of gravity is near the waterline or above.
When the rudder is first deflected, inertial terms dominate (Phase 1) and the
sway equation is
54 7 STANDARD MANEUVERING TESTS
(m − Yv̇ )v̇ − (Yṙ − mxG )ṙ = Yδ δ. (89)

The coefficients for ṙ are quite small, and thus the vessel first rolls to starboard
(positive) for a positive rudder action.
When steady turning conditions are reached (Phase 3), hydrodynamic forces
equalize the centrifugal force mU r and the rudder force Yδ δ. The sway equation
is
−Yv v + (mU − Yr )r = Yδ δ, (90)

with Yr small, v < 0 when r > 0 for most vessels, and |Yv v| > |Yδ δ|. Because
the centrifugal force acts above the waterline, the vessel ultimately rolls to
port (negative) for a positive rudder action.
The transition between the inertially-dominated and steady-turning regimes
includes an overshoot: in the above formulas, the vessel overshoots the final
port roll angle as the vessel slows. From the sway equation, we see that if
the rudder is straightened out at this point, the roll will momentarily be even
worse!
In summary, the vessel rolls into the turn initially, but then out of the turn in
the steady state.
7.3.4 Heeling in Submarines with Sails

Submarines typically roll into a turn during all phases. Unlike surface vessels,
which have the rigid mass center above the center of fluid forcing, submarines
have the mass center below the rudder action point, and additionally feel the
effects of a large sail above both. The inertial equation
(m − Yv̇ )v̇ − (Yṙ − mxG )ṙ = Yδ δ (91)

is dominated by mv̇ (acting low), −Yv̇ v̇ (acting high), and Yδ δ (intermediate).
Because |Yv̇ | > m, the vessel rolls under the sail, the keel out of the turn. In
the steady state,
−Yv v + (mU − Yr )r = Yδ δ. (92)

The drift angle β keeps the Yv -force, acting high, toward the center of the turn,
and again centrifugal force mU r causes the bottom of the submarine to move
out of the turn. Hence, the roll angle of a submarine with a sail is always into
7.3 Circle Maneuver 55
the turn, both initially and in the steady state. The heel angle declines as the
speed of the submarine drops.
56 8 STREAMLINED BODIES
8 STREAMLINED BODIES
8.1 Nominal Drag Force
A symmetric streamlined body at zero angle of attack experiences only a drag
force, which has the form
1
FA = − ρCA Ao U 2 . (93)
2
The drag coefficient CA has both pressure and skin friction components, and
hence area Ao is usually that of the wetted surface. Note that the A-subscript
will be used to denote zero angle of attack conditions; also, the sign of FA is
negative, because it opposes the vehicle’s x-axis.
8.2 Munk Moment

Any shape other than a sphere generates a moment when inclined in an invis-
cid flow. d’Alembert’s paradox predicts zero net force, but not necessarily a
zero moment. This Munk moment arises for a simple reason, the asymmetric
location of the stagnation points, where pressure is highest on the front of
the body (decelerating flow) and lowest on the back (accelerating flow). The
Munk moment is always destabilizing, in the sense that it acts to turn the
vehicle perpendicular to the flow.
Consider a symmetric body with added mass components Axx along the vehicle
(slender) x-axis (forward), and Azz along the vehicle’s z-axis z (up). We will
limit the present discussion to the vertical plane, but similar arguments can
be used to describe the horizontal plane. Let α represent the angle of attack,
taken to be positive with the nose up – this equates to a negative pitch angle
φ in vehicle coordinates, if it is moving horizontally. The Munk moment is:
1
Mm = − (Azz − Axx )U 2 sin 2α (94)
2
' −(Azz − Axx )U 2 α.
Azz > Axx for a slender body, and the negative sign indicates a negative
pitch with respect to the vehicle’s pitch axis. The added mass terms Azz and
Axx can be estimated from analytical expressions (available only for regular
shapes such as ellipsoids), from numerical calculation, or from slender body
approximation (to follow).
8.3 Separation Moment 57
8.3 Separation Moment

In a viscous fluid, flow over a streamlined body is similar to that of potential
flow, with the exceptions of the boundary layer, and a small region near the
trailing end. In this latter area, a helical vortex may form and convect down-
stream. Since vortices correlate with low pressure, the effect of such a vortex
is stabilizing, but it also induces drag. The formation of the vortex depends on
the angle of attack, and it may cover a larger area (increasing the stabilizing
moment and drag) for a larger angle of attack. For a small angle of attack, the
transverse force Fn can be written in the same form as for control surfaces:
1
Fn = ρCn Ao U 2 (95)
2
1 ∂Cn
' ρ αAo U 2 .
2 ∂α
With a positive angle of attack, this force is in the positive z-direction. The
zero-α drag force FA is modified by the vortex shedding:
1
Fa = − ρCa Ao U 2 , where (96)
2
Ca = CA cos2 φ.
The last relation is based on writing CA (U cos φ)2 as (CA cos2 φ)U 2 , i.e., a
decomposition using apparent velocity.
8.4 Net Effects: Aerodynamic Center

The Munk moment and the moment induced by separation are competing,
and their magnitudes determine the stability of a hullform. First we simplify:
Fa = −γa
Fn = γn α
Mm = −γm α.
Each constant γ is taken as positive, and the signs reflect orientation in the
vehicle reference frame, with a nose-up angle of attack. The Munk moment is a
pure couple which does not depend on a reference point. We pick a temporary
origin O for Fn however, and write the total pitch moment about O as:
M = Mm + Fn ln (97)
= (−γm + γn ln )α.
where ln denotes the (positive) distance between O and the application point
of Fn . The net moment about O is zero if we select
γm
ln = , (98)
γn
and the location of O is then called the aerodynamic center or AC.
The point AC has an intuitive explanation: it is the location on the hull where
Fn would act to create the total moment. Hence, if the vehicle’s origin lies in
front of AC, the net moment is stabilizing. If the origin lies behind AC, the
moment is destabilizing. For self-propelled vehicles, the mass center must be
forward of AC for stability. Similarly, for towed vehicles, the towpoint must
be located forward of AC. In many cases with very streamlined bodies, the
aerodynamic center is significantly ahead of the nose, and in this case, a rigid
sting would have to extend at least to AC in order for stable towing. As a
final note, since the Munk moment persists even in inviscid flow, AC moves
infinitely far forward as viscosity effects diminish.
8.5 Role of Fins in Moving the Aerodynamic Center

Control surfaces or fixed fins are often attached to the stern of a slender vehicle
to enhance directional stability. Fixed surfaces induce lift and drag on the
body:
1
L = ρAf U 2 Cl (α) ' γL α (99)
2
1
D = − ρAf U 2 Cd ' −γD (constant)
2
These forces act somewhere on the fin, and are signed again to match the
vehicle frame, with γ > 0 and α > 0. The summed forces on the body are
thus:
8.5 Role of Fins in Moving the Aerodynamic Center 59
X = Fa − |D| cos α + |L| sin α (100)

' −γa − γD + γL α2
Z = Fn + |L| cos α + |D| sin α
' γn α + γL α + γD α.
All of the forces are pushing the vehicle up. If we say that the fins act a
distance lf behind the temporary origin O, and that the moment carried by
the fins themselves is very small (compared to the moment induced by Llf )
the total moment is as follows:
M = (−γm + γn ln )α + (γL + γD )lf α. (101)

The moment about O vanishes if
γm = γn ln + lf (γL + γD ). (102)
The Munk moment γm opposes the aggregate effects of vorticity lift γn and
the fins’ lift and drag γL + γD . With very large fins, this latter term is large, so
that lf might be very small; this is the case of AC moving aft toward the fins.
A vehicle with excessively large fins will be difficult to turn and maneuver.
Equation 102 contains two length measurements, referenced to an arbitrary
body point O. To solve it explicitly, let lf n denote the (positive) distance that
the fins are located behind Fn ; this is likely a small number, since both effects
usually act near the stern. We solve for lf :
γm + γn lf n
lf = . (103)
γn + γL + γD
This is the distance that AC is located forward of the fins, and thus AC can
be referenced to any other fixed point easily. Without fins, it can be recalled
that
γm
ln = .
γn
Hence, the fins act directly in the denominator to shorten lf . Note that if
the fins are located forward of the vortex shedding force Fn , i.e., lf n < 0, lf is
reduced, but since AC is referenced to the fins, there is no net gain in stability.
8.6 Aggregate Effects of Body and Fins

Since all of the terms discussed so far have the same dependence on α, it is
possible to group them into a condensed form. Setting F̂a and F̂n to account
for the fuselage and fins, we have
1
X = F̂a ' − ρĈa Âo U 2 (104)
2
1 0
Z = F̂n ' ρĈn Âo U 2 α
2
1
M = −F̂n xAC ' − Ĉn0 Âo U 2 xAC α.
2
8.7 Coefficients Zw , Mw , Zq , and Mq for a Slender Body

The angle of attack α is related to the cross-body velocity w as follows:
w

−1
α = − tan (105)
u
w
' − for U >> w.
U
We can then write several linear hydrodynamic coefficients easily:
1
Zw = − ρĈn0 Âo U (106)
2
1 0
Mw = ρĈ Âo U xAC .
2 n
The rotation of the vessel involves complex flow, which depends on both w and
q, as well as their derivatives. To start, we consider the contribution of the fins
only – slender body theory, introduced shortly, provides good results for the
hull. The fin center of pressure is located a distance lf aft of the body origin,
and we assume that the vehicle is moving horizontally, with an instantaneous
pitch angle of θ. The angle of attack seen by the fin is a combination of a part
due to θ and a part linear with q:
lf q
αf ' −θ + (107)
U
8.7 Coefficients Zw , Mw , Zq , and Mq for a Slender Body 61
and so lateral force and moment derivatives (for the fin alone) emerge as
1
Zq = − ρCl0 Af U lf (108)
2
1
Mq = − ρCl0 Af U lf2 .
2
62 9 SLENDER-BODY THEORY
9 SLENDER-BODY THEORY
9.1 Introduction
Consider a slender body with d << L, that is mostly straight. The body
could be asymmetric in cross-section, or even flexible, but we require that
the lateral variations are small and smooth along the length. The idea of
the slender-body theory, under these assumptions, is to think of the body as
a longitudinal stack of thin sections, each having an easily-computed added
mass. The effects are integrated along the length to approximate lift force and
moment. Slender-body theory is accurate for small ratios d/L, except near
the ends of the body.
As one example, if the diameter of a body of revolution is d(s), then we can
compute δma (x), where the nominal added mass value for a cylinder is
π
δma = ρ d2 δx. (109)
4
The added mass is equal to the mass of the water displaced by the cylinder.
The equation above turns out to be a good approximation for a number of
two-dimensional shapes, including flat plates and ellipses, if d is taken as the
width dimension presented to the flow. Many formulas for added mass of two-
dimensional sections, as well as for simple three-dimensional bodies, can be
found in the books by Newman and Blevins.
9.2 Kinematics Following the Fluid

The added mass forces and moments derive from accelerations that a fluid
particles experience when they encounter the body. We use the notion of a
fluid derivative for this purpose: the operator d/dt indicates a derivative taken
in the frame of the passing particle, not the vehicle. Hence, this usage has an
indirect connection with the derivative described in our previous discussion of
rigid-body dynamics.
For the purposes of explaining the theory, we will consider the two-dimensional
heave/surge problem only. The local geometry is described by the location of
the centerline; it has vertical location (in body coordinates) of zb (x, t), and
local angle α(x, t). The time-dependence indicates that the configuration is
free to change with time, i.e., the body is flexible. Note that the curvilinear
coordinate s is nearly equal to the body-reference (linear) coordinate x.
The velocity of a fluid particle normal to the body at x is wn (t, x):
9.3 Derivative Following the Fluid 63
∂zb
wn = cos α − U sin α. (110)
∂t
The first component is the time derivative in the body frame, and the second
due to the deflection of the particle by the inclined body. If the body reference
frame is rotated to the flow, that is, if w 6= 0, then ∂zb /∂t will contain w. For
small angles, sin α ' tan α = ∂zb /∂x, and we can write
∂zb ∂zb
wn ' −U .
∂t ∂x
The fluid derivative operator in action is as follows:
!
dzb ∂ ∂
wn = = −U zb .
dt ∂t ∂x
9.3 Derivative Following the Fluid

A more formal derivation for the fluid derivative operator is quite simple. Let
µ(x, t) represent some property of a fluid particle.
d 1
[µ(x, t)] = lim [µ(x + δx, t + δt) − µ(x, t)]
dt δt→0 δt
" #
∂µ ∂µ
= −U .
∂t ∂x
The second equality can be verified using a Taylor series expansion of µ(x +
δx, t + δt):
∂µ ∂µ
µ(x + δx, t + δt) = µ(x, t) + δt + δx + h.o.t.,
∂t ∂x
and noting that δx = −U δt. The fluid is convected downstream with respect
to the body.
9.4 Differential Force on the Body

If the local transverse velocity is wn (x, t), then the differential inertial force on
the body here is the derivative (following the fluid) of the momentum:
d
δF = − [ma (x, t)wn (x, t)] δx. (111)
dt
Note that we could here let the added mass vary with time also – this is the
case of a changing cross-section! The lateral velocity of the point zb (x) in the
body-reference frame is
∂zb
= w − xq, (112)
∂t
such that
wn = w − xq − U α. (113)
Taking the derivative, we have
!
δF ∂ ∂
=− −U [ma (x, t)(w(t) − xq(t) − U α(x, t))] .
δx ∂t ∂x
We now restrict ourselves to a rigid body, so that neither ma nor α may change
with time.
δF ∂
= ma (x)(−ẇ + xq̇) + U [ma (x)(w − xq − U α)] . (114)
δx ∂x
9.5 Total Force on a Vessel

The net lift force on the body, computed with strip theory is
Z xN
Z = δF dx (115)
xT
where xT represents the coordinate of the tail, and xN is the coordinate of the
nose. Expanding, we have
9.6 Total Moment on a Vessel 65
Z xN ∂ Z xN
Z = ma (x) [−ẇ + xq̇] dx + U [ma (x)(w − xq − U α)] dx
xT xT ∂x
= −m33 ẇ − m35 q̇ + U ma (x)(w − xq − U α)|x=x
x=xT .
N
We made use here of the added mass definitions
Z xN
m33 = ma (x)dx
xT
Z xN
m35 = − xma (x)dx.
xT
Additionally, for vessels with pointed noses and flat tails, the added mass ma
at the nose is zero, so that a simpler form occurs:
Z = −m33 ẇ − m35 q̇ − U ma (xT )(w − xT q − U α(xT )). (116)

In terms of the linear hydrodynamic derivatives, the strip theory thus provides
Zẇ = −m33
Zq̇ = −m35
Zw = −U ma (xT )
Zq = U xT ma (xT )
Zα(xT ) = U 2 ma (xT ).
It is interesting to note that both Zw and Zα(xT ) depend on a nonzero base
area. In general, however, potential flow estimates do not create lift (or drag)
forces for a smooth body, so this should come as no surprise. The two terms are
clearly related, since their difference depends only on how the body coordinate
system is oriented to the flow. Another noteworthy fact is that the lift force
depends only on α at the tail; α could take any value(s) along the body, with
no effect on Z.
9.6 Total Moment on a Vessel

A similar procedure can be applied to the moment predictions from slender
body theory (again for small α):
Z xN
M = − xδF dx
xT
!
Z xN∂ ∂
= x −U [ma (x)(w − xq − U α)] dx
xT ∂t ∂x
Z xN Z xN
∂
= xma (x)(ẇ − xq̇)dx − U x [ma (x)(w − xq + U α)] dx.
xT xT ∂x
Then we make the further definition
Z xN
m55 = x2 ma (x)dx,
xT
(note that m35 = m53 ) and use integration by parts to obtain
M = −m35 ẇ − m55 q̇ − U xma (x)(w − xq − U α)|x=x

x=xT +
N
Z xN
U ma (x)(w − xq − U α)dx.
xT
The integral above contains the product ma (x)α(x), which must be calculated
if α changes along the length. For simplicity, we now assume that α is in fact
constant on the length, leading to
M = −m35 ẇ − m55 q̇ + U xT ma (xT )(w − xT q − U α) +

U m33 w + U m35 q − U 2 m33 α.
Finally, the linear hydrodynamic moment derivatives are
Mẇ = −m35
Mq̇ = −m55
Mw = U xT ma (xT ) + U m33
Mq = −U x2T ma (xT ) + U m35
Mα = −U 2 xT ma (xT ) − U 2 m33 .
The derivative Mw is closely-related to the Munk moment, whose linearization
would provide Mw = (m33 − m11 )U . The Munk moment (an exact result) may
therefore be used to make a correction to the second term in the slender-body
approximation above of Mw . As with the lift force, Mw and Mα are closely
related, depending only on the orientation of the body frame to the flow.
9.7 Relation to Wing Lift 67
9.7 Relation to Wing Lift

There is an important connection between the slender body theory terms in-
volving added mass at the tail (ma (xT )), and low aspect-ratio wing theory.
The lift force from the latter is of the form L = − 12 ρU ACl0 w, where A = cs,
the product of chord (long) and span (short).
The lift coefficient slope is approximated by (Hoerner)
1
Cl0 ≈ π(AR), (117)
2
where (AR) is the aspect ratio. Inserting this approximation into the lift
formula, we obtain
π
L = − ρs2 U w. (118)
4
Now we look at a slender body approximation of the same force: The added
mass at the tail is ma (xT ) = ρs2 π/4, and using the slender-body estimate for
Zw , we calculate for lift:
Z = −ma (xT )U w
π
= − ρs2 U w.
4
Slender-body theory is thus able to recover exactly the lift of a low-aspect
ratio wing. Where does the slender-body predict the force will act? Recalling
that Mw = U m33 + U xT ma (xT ), and since m33 = 0 for a front-back symmetric
wing, the estimated lift force acts at the trailing edge. This location will tend
to stabilize the wing, in the sense that it acts to orient the wing parallel to
the incoming flow.
9.8 Convention: Hydrodynamic Mass Matrix A

Hydrodynamic derivatives that depend on accelerations are often written as
components of a mass matrix A. By listing the body-referenced velocities
in the order ~s = [u, v, w, p, q, r], we write (M + A)~s˙ = F~ , where M is the
mass matrix of the material vessel and F is a generalized force. Therefore
A33 = −Zẇ , A5,3 = −Mẇ , and so on.
68 10 PRACTICAL LIFT CALCULATIONS
10 PRACTICAL LIFT CALCULATIONS

10.1 Characteristics of Lift-Producing Mechanisms
At a small angle of attack, a slender body experiences transverse force due to:
helical body vortices, the blunt trailing end, and fins. The helical body vortices
are stable and symmetric in this condition, and are convected continuously into
the wake. The low pressure associated with the vortices provides the suction
force, usually toward the stern of the vehicle. The blunt trailing end induces
lift as a product of added mass effects, and can be accurately modeled with
slender body theory. A blunt trailing edge also induces some drag, which itself
is stabilizing. The fins can often be properly modeled using experimental data
parametrized with aspect ratio and several other geometric quantities.
As the angle of attack becomes larger, the approximations in the fin and
slender-body analysis will break down. The helical vortices can become bigger
while remaining stable, but eventually will split randomly. Some of it convects
downstream, and the rest peels away from the body; this shedding is nonsym-
metric, and greatly increases drag by widening the wake. In the limit of a 90◦
angle of attack, vorticity sheds as if from a bluff-body, and there is little axial
convection.
10.2 Jorgensen’s Formulas

There are some heuristic and theoretical formulas for predicting transverse
force and moment on a body at various angles of attack, and we now present
one of them due to Jorgensen. The formulas provide a good systematic proce-
dure for design, and are best suited to vehicles with a blunt trailing edge. We
call the area of the stern the base area.
Let the body have length L, and reference area Ar . This area could be the
frontal projected area, the planform area, or the wetted area. The body travels
at speed U , and angle of attack α. The normal force, axial force, and moment
coefficients are defined as follows:
FN
CN = 1 (119)
2
ρU 2 Ar
FA
CA = 1
2
ρU 2 Ar
10.2 Jorgensen’s Formulas 69
M xm
CM = 1 .
2
ρU 2 Ar L
The moment Mxm is taken about a point xm , measured back from the nose; this
location is arbitrary, and appears explicitly in the formula for Cm . Jorgensen
gives the coefficients as follows:
Ab α Ap
CN = sin 2α cos + Cd sin2 α (120)
Ar 2 Ar n
CA = CAo cos2 α (121)
∇ − Ab (L − xm ) α Ap x m − x c

CM = − sin 2α cos − Cdn sin2 α.(122)
Ar L 2 Ar L
We have listed only the formulas for the special case of circular cross-section,
although the complete equations do account for more complex shapes. Further,
we have assumed that L >> D, which is also not a constraint in the complete
equations. The parameters used here are
• Ab : stern base area. Ab = 0 for a body that tapers to a point at the

stern.
• Cdn : crossflow drag coefficient; equivalent to that of an infinite circular

cylinder. If “normal” Reynolds number Ren = U sin αD/ν,
– Cd ≈ 1.2, Ren < 3 × 105

– Cd ≈ 0.3, 3 × 105 < Ren < 7 × 105
– Cd ≈ 0.6, 7 × 105 < Ren .
• Ap : planform area.
• CAo : axial drag at zero angle of attack, both frictional and form. CAo '
0.002 − 0.006 for slender streamlined bodies, based on wetted surface
area. It depends on Re = U L/ν.
• ∇: body volume.
• xc : distance from the nose backwards to the center of the planform area.
In the formula for normal force, we see that if Ab = 0, only drag forces act
to create lift, through a sin2 α-dependence. Similarly, the axial force is simply
the zero-α result, modified by cos2 α. In both cases, scaling of U 2 into the
body principle directions is all that is required.
There are several terms that match exactly the slender-body theory approxi-
mations for small α. These are the first term in the normal force (CN ), and
the entire first term in the moment (CM ), whether or not Ab = 0. Finally, we
note that the second term in CM disappears if xm = xc , i.e., if the moment is
referenced to the center of the planform area. The idea here is that the fore
and aft components of crossflow drag cancel out.
The aerodynamic center (again referenced toward the stern, from the nose)
can be found after the coefficients are computed:
CM
xAC = xm + L. (123)
CN
As written, the moment coefficient is negative if the moment destabilizes the
body, while CN is always positive. Thus, the moment seeks to move the AC
forward on the body, but the effect is moderated by the lift force.
10.3 Hoerner’s Data: Notation

An excellent reference for experimental data is the two-volume set by S. Ho-
erner. It contains a large amount of aerodynamic data from many different
types of vehicles, wings, and other common engineering shapes. A few nota-
tions are used throughout the books, and are described here.
First, dynamic pressure is given as q = 12 ρU 2 , such that two typical body lift
coefficients are:
Y
CY =
DLq
Y
CYd = 2 .
D q
The first version uses the rectangular planform area as a reference, while the
second uses the square frontal area. Hence, CY = CYd D/L. Two moment
coefficients are:
10.4 Slender-Body Theory vs. Experiment 71
N
CN =
LD2 q
N1
CN1 = ,
LD2 q
where N is the moment taken about the body mid-point, and N1 is taken
about the nose. Note that the reference area for moment is the square frontal
projection, and the reference length is body length L. The following relation
holds for these definitions
1
CN1 = CN + CYd .
2
The lift and moment coefficients are strongly dependent on angle of attack;
Hoerner uses the notation
∂CN
Cnb =
∂b
∂CN1
Cn·b =
∂b
∂Cy
Cyb = ,
∂b
and so on, where b is the angle of attack, usually in degrees. It follows from
above that Cn·b = Cnb + Cydb /2.
10.4 Slender-Body Theory vs. Experiment

In an experiment, the net moment is measured, comprising both the desta-
bilizing part due to the potential flow, and the stabilizing part due to vortex
shedding or a blunt tail. Comparison of the measurements and the theory
allows us to place the action point of the suction force. This section gives the
formula for this location in Hoerner’s notation, and gives two further examples
of how well the slender-body theory matches experiments.
For L/D > 6, the slender-body (pure added mass) estimates give C̃nb '
−0.015/deg, acting to destabilize the vehicle. The value compares well with
−0.027/deg for a long cylinder and −0.018/deg for a long ellipsoid; it also
reduces to −0.009 for L/D = 4. Note that the negative sign here is consistent
with Hoerner’s convention that destabilizing moments have negative sign.
The experimental lift force is typically given by CY b ' 0.003/deg; this acts at
a point on the latter half of the vehicle, stabilizing the angle. Because this
coefficient scales roughly with wetted area, proportional to LD, it changes
little with L/D. It can be compared with a low-aspect ratio wing, which
achieves an equivalent lift of π(AR)/2 = 0.0027 for (AR) = 10 ' D/L.
The point at which the viscous forces act can then be estimated as the following
distance aft of the nose:
x Cn·b − C̃nb
= (124)
L Cydb
The calculation uses experimental values of Cydb and Cn·b , the moment slope
referenced to the nose. In the table following are values from Hoerner (p. 13-2,
Figure 2) for a symmetric and a blunt-ended body.
symmetric blunt
L/D 6.7 6.7
C̃nb -0.012 -0.012
Cyb 0.0031 0.0037
Cydb 0.021 0.025
Cn·b 0.0012 (stable) 0.0031 (stable)
Cnb -0.0093 (unstable) -0.0094 (unstable)
x/L 0.63 0.60
In comparing the two body shapes, we see that the moment at the nose is
much more stable (positive) for the body with a blunt trailing edge. At the
body midpoint, however, both vehicles are equally unstable. The blunt-tailed
geometry has a much larger lift force, but it acts too close to the midpoint to
add any stability there.
The lift force dependence on the blunt tail is not difficult to see, using slender-
body theory. Consider a body, with trailing edge radius r. The slender-body
lift force associated with this end is simply the product of speed U and local
added mass ma (xT ) (in our previous notation). It comes out to be
1
Z = ρU 2 (πr2 )(2α), (125)
2
such that the first term in parentheses is an effective area, and the second is
a lift coefficient. With respect to the area πr2 , the lift curve slope is therefore
10.5 Slender-Body Approximation for Fin Lift 73
2/rad. Expressed in terms of the Hoerner reference area D2 , the equivalent

lift coefficient is Cydb = 0.0044/deg, where we made the assumption here that
2r/D = 0.4 for the data in the table. This lift difference, due solely to the
blunt end condition, is consistent with measurements.
10.5 Slender-Body Approximation for Fin Lift

Let us now consider two fins of span s each, acting at the tail end of the
vehicle. This is the case if the vehicle body tapers to a point where the fins
have their trailing edge. The slender-body approximation of lift as a result of
blunt-end conditions is
Z = πs2 ρU 2 α. (126)
Letting the aspect ratio be (AR) = (2s)2 /Af , where Af is the total area of the
fin pair, substitution will give a lift curve slope of
π
Cl0 = (AR).
2
This is known as Jones’ formula, and is quite accurate for (AR) ' 1. It is
inadequate for higher-aspect ratio wings however, overestimating the lift by
about 30% when (AR) = 2, and worsening further as (AR) grows.
74 11 FINS AND LIFTING SURFACES
11 FINS AND LIFTING SURFACES

Vessels traveling at significant speed typically use rudders, elevators, and other
streamlined control surfaces to maneuver. Their utility arises mainly from the
high lift forces they can develop, with little drag penalty. Lift is always defined
to act in a direction perpendicular to the flow, and drag in the same direction
as the flow.
11.1 Origin of Lift

A lifting surface is nominally an extrusion of a streamlined cross-section: the
cross-section has a rounded leading edge, sharp trailing edge, and a smooth
surface. The theory of lifting surfaces centers on the Kutta condition, which
requires that fluid particle streamlines do not wrap around the trailing edge
of the surface, but instead rejoin with streamlines from the other side of the
wing at the trailing edge. This fact is true for a non-stalled surface at any
angle of attack.
Since the separation point on the front of the section rotates with the angle
of attack, it is clear that the fluid must travel faster over one side of the
surface than the other. The reduced Bernoulli pressure this induces can then
be thought of as the lift-producing mechanism. More formally, lift arises from
circulation Γ:
I
Γ= V~ · d~s. (127)
and then L = −ρU Γ. Circulation is the integral of velocity around the cross-
section, and a lifting surface requires circulation in order to meet the Kutta
condition.
11.2 Three-Dimensional Effects: Finite Length

Since all practical lifting surfaces have finite length, the flow near the ends
may be three-dimensional. Prandtl’s inviscid theory provides some insight.
Since bound circulation cannot end abruptly at the wing end, it continues on
in the fluid, leading to so-called wing-tip vortices. This continuation causes
induced velocities at the tips, and some induced drag. Another description for
the wing-tip vortices is that the pressure difference across the surface simply
causes flow around the end.
11.3 Ring Fins 75
A critical parameter which governs the extent of three-dimensional effects is

the aspect ratio:
span span2
AR = = . (128)
chord area
The second representation is useful for non-rectangular control surfaces. The
effective span is taken to be the length between the free ends of a symmetric
wing. If the wing is attached to a wall, the effective span is twice the physical
value, by reflection, and in this case the effective aspect ratio is therefore twice
the physical value.
The aspect ratio is a strong determinant of wing performance: for a given
angle of attack, a larger aspect ratio achieves a higher lift value, but also stalls
earlier.
Lift is written as
1
L = ρU 2 ACl , (129)
2
where A is the single-side area of the surface. For angles of attack α below
stall, the lift coefficient Cl is nearly linear with α: Cl = Cl0 α, where Cl0 is called
the lift coefficient slope, and has one empirical description
1
Cl0 = 1 1 1 , (130)
2π ᾱ
+ π(AR)
+ 2π(AR)2
where α is in radians, ᾱ ' 0.90, and AR is the effective aspect ratio. When
AR → ∞, the theoretical and maximum value for Cl0 is 2π.
The lift generated on a surface is the result of a distributed pressure field; this
fact creates both a net force and a net moment. A single equivalent force acts
at a so-called center of action xA , which depends mainly on the aspect ratio.
For high AR, xA ' c/4, measured back from the front of the wing. For low
AR, xA ' c/2.
11.3 Ring Fins

Ring fins are useful when space allows, since they are omnidirectional, and
structurally more robust than cantilevered plane surfaces. The effective aspect
ratio for a ring of diameter d is given as
4d
AR = . (131)
πc
76 11 FINS AND LIFTING SURFACES
The effective area of the ring is taken as

π
Ae = dc, (132)
2
and we thus have L = 21 ρU 2 Ae Cl0 α, where one formula for Cl0 is
1
Cl0 = 1 . (133)
0.63 + π(AR)
77
12 PROPELLERS AND PROPULSION

12.1 Introduction
We discuss in this section the nature of steady and unsteady propulsion. In
many marine vessels and vehicles, an engine (diesel or gas turbine, say) or an
electric motor drives the propeller through a linkage of shafts, reducers, and
bearings, and the effects of each part are important in the response of the net
system. Large, commercial surface vessels spend the vast majority of their time
operating in open-water and at constant speed. In this case, steady propulsion
conditions are generally optimized for fuel efficiency. An approximation of the
transient behavior of a system can be made using the quasi-static assumption.
In the second section, we list several low-order models of thrusters, which have
recently been used to model and simulate truly unsteady conditions.
12.2 Steady Propulsion of Vessels

The notation we will use is as given in Table 1, and there are two different flow
conditions to consider. Self-propelled conditions refer to the propeller being
installed and its propelling the vessel; there are no additional forces or moments
on the vessel, such as would be caused by a towing bar or hawser. Furthermore,
the flow around the hull interacts with the flow through the propeller. We use
an sp subscript to indicate specifically self-propulsion conditions. Conversely,
when the propeller is run in open water, i.e., not behind a hull, we use an o
subscript; when the hull is towed with no propeller we use a t subscript. When
subscripts are not used, generalization to either condition is implied. Finally,
because of similitude (using diameter D in place of L when the propeller is
involved), we do not distinguish between the magnitude of forces in model and
full-scale vessels.
12.2.1 Basic Characteristics

In the steady state, force balance in self-propulsion requires that
Rsp = Tsp . (134)

The gear ratio λ is usually large, indicating that the propeller turns much
more slowly than the driving engine or motor. The following relations define
the gearbox:
78 12 PROPELLERS AND PROPULSION
Rsp N hull resistance under self-propulsion

Rt N towed hull resistance (no propeller attached)
T N thrust of the propeller
ne Hz rotational speed of the engine
nm Hz maximum value of ne
np Hz rotational speed of the propeller
λ gear ratio
Qe Nm engine torque
Qp Nm propeller torque
ηg gearbox efficiency
Pe W engine power
Pp W propeller shaft power
D m propeller diameter
U m/s vessel speed
Up m/s water speed seen at the propeller
Qm Nm maximum engine torque
f kg/s fuel rate (or energy rate in electric motor)
fm kg/s maximum value of f
Table 1: Nomenclature
ne = λnp (135)
Qp = ηg λQe ,
and power follows as Pp = ηg Pe , for any flow condition. We call J = Up /np D

the advance ratio of the prop when it is exposed to a water speed Up ; note
that in the wake of the vessel, Up may not be the same as the speed of the
vessel U . A propeller operating in open water can be characterized by two
nondimensional parameters which are both functions of J:
To
KT = (thrust coefficient) (136)
ρn2p D4
Qp
KQ = 2 o 5 (torque coefficient). (137)
ρnp D
12.2 Steady Propulsion of Vessels 79
1.0 0.10
η0
0.8 0.08
K
T 0.6 0.06 K Q
η
0 0.4 K 0.04
Q
0.2 0.02
KT
0.0 0.00
0.0 0.2 0.4 0.6 0.8 1.0
J
Figure 4: Typical thrust and torque coefficients.
The open propeller efficiency can be written then as
To U J(U )KT
ηo = = . (138)
2πnp Qpo 2πKQ
This efficiency divides the useful thrust power by the shaft power. Thrust and
torque coefficients are typically nearly linear over a range of J, and therefore
fit the approximate form:
KT (J) = β1 − β2 J (139)
KQ (J) = γ1 − γ2 J.
As written, the four coefficients [β1 , β2 , γ1 , γ2 ] are usually positive, as shown in

the figure.
We next introduce three factors useful for scaling and parameterizing our
mathematical models:
• Up = U (1 − w); w is referred to as the wake fraction. A typical wake

fraction of 0.1, for example, indicates that the incoming velocity seen by
the propeller is only 90% of the vessel’s speed. The propeller is operating
in a wake.
In practical terms, the wake fraction comes about this way: Suppose the
open water thrust of a propeller is known at a given U and np . Behind
a vessel moving at speed U , and with the propeller spinning at the same
np , the prop creates some extra thrust. w scales U at the prop and thus
J; w is then chosen so that the open water thrust coefficient matches
what is observed. The wake fraction can also be estimated by making
direct velocity measurements behind the hull, with no propeller.
• Rt = Rsp (1 − t). Often, a propeller will increase the resistance of the
vessel by creating low-pressure on its intake side (near the hull), which
makes Rsp > Rt . In this case, t is a small positive number, with 0.2 as a
typical value. t is called the thrust deduction even though it is used to
model resistance of the hull; it is obviously specific to both the hull and
the propeller(s), and how they interact.
The thrust deduction is particularly useful, and can be estimated from
published values, if only the towed resistance of a hull is known.
• Qpo = ηR Qpsp . The rotative efficiency ηR , which may be greater than
one, translates self-propelled torque to open water torque, for the same
incident velocity Up , thrust T , and rotation rate np . ηR is meant to
account for spatial variations in the wake of the vessel which are not
captured by the wake fraction, as well as the turbulence induced by the
hull. Note that in comparison with the wake fraction, rotative efficiency
equalizes torque instead of thrust.
A common measure of efficiency, the quasi-propulsive efficiency, is based on

the towed resistance, and the self-propelled torque.
Rt U
ηQP = (140)
2πnp Qpsp
To (1 − t)Up ηR
=
2πnp (1 − w)Qpo
(1 − t)
= ηo ηR .
(1 − w)
To and Qpo are values for the inflow speed Up , and thus that ηo is the open
propeller efficiency at this speed. It follows that To (Up ) = Tsp , which was used
to complete the above equation. The quasi-propulsive efficiency can be greater
than one, since it relies on the towed resistance and in general Rt > Rsp . The
ratio (1 − t)/(1 − w) is often called the hull efficiency, and we see that a
12.2 Steady Propulsion of Vessels 81
small thrust deduction t and a large wake fraction w are beneficial effects, but
which are in competition. A high rotative efficiency and open water propeller
efficiency (at Up ) obviously contribute to an efficient overall system.
12.2.2 Solution for Steady Conditions

The linear form of KT and KQ (Equation 140) allows a closed-form solution
for the steady-operating conditions. Suppose that the towed resistance is of
the form
1
Rt = ρCr Aw U 2 , (141)
2
where Cr is the resistance coefficient (which will generally depend on Re and
F r), and Aw is the wetted area. Equating the self-propelled thrust and resis-
tance then gives
Tsp = Rsp
To = Rt /(1 − t)
1
KT (J(Up ))ρn2p D4 (1 − t) = ρCr Aw U 2
2
1 Up2
(β1 − β2 J(Up ))ρn2p D4 (1 − t) = ρCr Aw
2 (1 − w)2
Cr Aw
β1 − β2 J(Up ) = J(Up )2
2D (1 − t)(1 − w)2
2
| {z }
δ
q
−β2 + β22 + 4β1 δ
J(Up ) = . (142)
2δ
The last equation predicts the steady-state advance ratio of the vessel, depend-
ing only on the propeller open characteristics, and on the hull. The vessel speed
can be computed by recalling that J(U ) = U/np D and Up = U (1 − w), but it
is clear that we need now to find np . This requires a torque equation, which
necessitates a model of the drive engine or motor.
12.2.3 Engine/Motor Models

The torque-speed maps of many engines and motors fit the form
Q e /Q
m
f3
f2
f1
n e /n m
Figure 5: Typical gas turbine engine torque-speed characteristic for increasing

fuel rates f1 , f2 , f3 .
Qe = Qm F (f /fm , ne /nm ), (143)

where F () is the characteristic function. For example, gas turbines roughly fit
the curves shown in the figure (Rubis). More specifically, if F () has the form
! !
f ne f
F (f /ff m, ne /nm ) = − a +b + c +d (144)
fm nm fm
np
= −α1 + α2 .
nm /λ
then a closed-form solution for ne (and thus np ) can be found. The manipula-
tions begin by equating the engine and propeller torque:
Qpo (J(Up )) = ηR Qps p (J(Up ))

2 5
ρnp D KQ (J(Up )) = ηR ηg λQe
ρn2p D5 (γ1 − γ2 J(Up )) = ηR ηg λQm F (f /fm , ne /nm )
n2p
!
ηR ηg λQm np
= −α1 + α2
(nm /λ)2 5
ρD (γ1 − γ2 J(Up ))(nm /λ)2 (nm /λ)
| {z }

q
np −α1 + 2 α12 + 4α2
= . (145)
(nm /λ) 2
12.3 Unsteady Propulsion Models 83
Note that the fuel rate enters through both α1 and α2 .

The dynamic response of the coupled propulsion and ship systems, under the
assumption of quasi-static propeller conditions, is given by
(m + ma )u̇ = Tsp − Rsp (146)

2πIp ṅp = ηg λQe − Qps p .
Making the necessary substitutions creates a nonlinear model with f as the

input; this is left as a problem for the reader.
12.3 Unsteady Propulsion Models

When accurate positioning of the vehicle is critical, the quasi-static assumption
used above does not suffice. Instead, the transient behavior of the propulsion
system needs to be considered. The problem of unsteady propulsion is still in
development, although there have been some very successful models in recent
years. It should be pointed out that the models described below all pertain to
open-water conditions and electric motors, since the positioning problem has
been central to bluff vehicles with multiple electric thrusters.
We use the subscript m to denote a quantity in the motor, and p for the
propeller.
12.3.1 One-State Model: Yoerger et al.

The torque equation at the propeller and the thrust relation are
Ip ω̇p = λQm − Kω ωp |ωp | (147)

T = Ct ωp |ωp |. (148)
where Ip is the total (material plus fluid) inertia reflected to the prop;the
propeller spins at ωp radians per second. The differential equation in ωp pits
the torque delivered by the motor against a quadratic-drag type loss which
depends on rotation speed. The thrust is then given as a static map directly
from the rotation speed.
This model requires the identification of three parameters: Ip , Kω , and Ct .
It is a first-order, nonlinear, low-pass filter from Qm to T , whose bandwidth
depends directly on Qm .
12.3.2 Two-State Model: Healey et al.

The two-state model includes the velocity of a mass of water moving in the
vicinity of the blades. It can accommodate a tunnel around the propeller,
which is very common in thrusters for positioning. The torque equation, sim-
ilarly to the above, is referenced to the motor and given as
Im ω̇m = −Kω ωm + Kv V − Qp /λ. (149)

Here, Kω represents losses in the motor due to spinning (friction and resistive),
and Kv is the gain on the input voltage (so that the current amplifier is included
in Kv ). The second dynamic equation is for the fluid velocity at the propeller:
ρALγ U̇p = −ρA∆β(Up − U )|Up − U | + T. (150)

Here A is the disc area of the tunnel, or the propeller disk diameter if no tunnel
exists. L denotes the length of the tunnel, and γ is the effective added mass
ratio. Together, ρALγ is the added mass that is accelerated by the blades; this
mass is always nonzero, even if there is no tunnel. The parameter ∆β is called
the differential momentum flux coefficient across the propeller; it may be on
the order of 0.2 for propellers with tunnels, and up to 2.0 for open propellers.
The thrust and torque of the propeller are approximated using wing theory,
which invokes lift and drag coefficients, as well as an effective angle of attack
and the propeller pitch. However, these formulae are static maps, and therefore
introduce no new dynamics. As with the one-state model of Yoerger et al., this
version requires the identification of the various coefficients from experiments.
This model has the advantage that it creates a thrust overshoot for a step
input, which is in fact observed in experiments.
85
13 TOWING OF VEHICLES
Vehicles which are towed have some similarities to the vehicles that have been
discussed so far. For example, towed vehicles are often streamlined, and usu-
ally need good directional stability. Some towed vehicles might have active
lifting surfaces or thrusters for attitude control. On the other hand, if they
are to be supported by a cable, towed vehicles may be quite heavy in water,
and do not have to be self-propelled. The cable itself is an important factor in
the behavior of the complete towed system, and in this section, we concentrate
on cable mechanics more than vehicle characteristics, which can generally be
handled with the same tools as other vehicles, i.e., slender-body theory, wing
theory, linearization, etc.. Some basic guidelines for vehicle design are given
at the end of this section.
Modern cables can easily exceed 5000m in length, even a heavy steel cable with
2cm diameter. The cables are generally circular in cross section, and may
carry power conductors and multiple communication channels (fiber optic).
The extreme L/D ratio for these cables obviates any bending stiffness effects.
Cable systems come in a variety of configurations, and one main division may
be made simply of the density of the cable. Light-tether systems are character-
ized by neutrally-buoyant (or nearly so) cables, with either a minimal vehicle
at the end, as in a towed array, or a vehicle capable of maneuvering itself, such
as a remotely-operated vehicle. The towed array is a relatively high-velocity
system that nominally streams out horizontally behind the vessel. An ROV,
on the other hand, operates at low speed, and must have large propulsors to
control the tether if there are currents. Heavy systems, in contrast, employ a
heavy cable and possibly a heavy weight; the rationale is that gravity will tend
to keep the cable vertical and make the deployment robust against currents
and towing speed. The heavy systems will generally transmit surface motions
and tensions to the towed vehicle much more easily than light-tether systems.
We will not discuss light systems specifically here, but rather look at heavy
systems. Most of the analysis can be adapted to either case, however.
13.1 Statics
13.1.1 Force Balance
For the purposes of deriving the static configuration of a cable in a flow, we
assume for the moment that that it is inextensible. Tension and hydrostatic
86 13 TOWING OF VEHICLES
y s+ T+dT
ds
x ds Rt φ U
T ds Rn
ds Wn
pressure will elongate a cable, but the effect is usually a small percentage of
the total length.
We employ the curvilinear axial coordinate s, which we take to be zero at the
bottom end of the cable; upwards along the cable is the positive direction.
The free-body diagram shown has the following components:
• Wn : net in-water weight of the cable per unit length.

• Rn (s): external normal force, per unit length.
• Rt (s): external tangential force, per unit length.
• T (s): local tension.
• φ(s): local inclination angle.
Force balance in the tangential and normal coordinates gives two coupled
equations for T and φ:
dT
amp; = amp; Wn sin φ − Rt (151)
ds
dφ
T amp; = amp; Wn cos φ + Rn . (152)
ds
The external forces are primarily fluid drag; the tangential drag is controlled
by a frictional drag coefficient Ct , and the normal drag scales with a crossflow
drag coefficient Cn . In both cases, the fluid velocity vector, U horizontal
toward the left, is to be projected onto the relevant axes, leading to
13.1 Statics 87
1
Rt = − ρCt dU 2 cos2 φ (153)
2
1
Rn = − ρCn dU 2 sin2 φ. (154)
2
Note that we simplified the drag laws from a usual form v|v| to v 2 , since as
drawn, 0 ≤ φ ≤ π/2.
The equations for T and φ can be integrated along the cable coordinate s to
find the cable’s static configuration. Two boundary conditions are needed, and
the common case is that a force balance on the vehicle, dominated by drag,
weight, and the cable tension, provides both T (0) and φ(0). For example, a
very heavy but low-drag vehicle will impose φ(0) ' π/2, with T (0) equal to
the in-water weight of the vehicle.
With regard to Cartesian coordinates x, y, the cable configuration follows
dx
amp; = amp; cos φ (155)
ds
dy
amp; = amp; sin φ. (156)
ds
The simultaneous integration of all four equations (T, φ, x, y) defines the cable
configuration, and current dependency may be included, say U is a function
of y.
13.1.2 Critical Angle

For very deep systems, the total weight of cable will generally exceed that the
vehicle. This gives rise to a configuration in which the cable is straight for
a majority of its length, but turns as necessary at the vehicle end, to meet
the bottom boundary condition. In the straight part of the cable, normal
weight and drag components are equalized. The uniform angle is called the
critical angle φc , and can be approximated easily. Let the relative importance
of weight be given as
Wn
δ= ,
ρCn dU 2
85
13 TOWING OF VEHICLES
Vehicles which are towed have some similarities to the vehicles that have been
discussed so far. For example, towed vehicles are often streamlined, and usu-
ally need good directional stability. Some towed vehicles might have active
lifting surfaces or thrusters for attitude control. On the other hand, if they
are to be supported by a cable, towed vehicles may be quite heavy in water,
and do not have to be self-propelled. The cable itself is an important factor in
the behavior of the complete towed system, and in this section, we concentrate
on cable mechanics more than vehicle characteristics, which can generally be
handled with the same tools as other vehicles, i.e., slender-body theory, wing
theory, linearization, etc.. Some basic guidelines for vehicle design are given
at the end of this section.
Modern cables can easily exceed 5000m in length, even a heavy steel cable with
2cm diameter. The cables are generally circular in cross section, and may
carry power conductors and multiple communication channels (fiber optic).
The extreme L/D ratio for these cables obviates any bending stiffness effects.
Cable systems come in a variety of configurations, and one main division may
be made simply of the density of the cable. Light-tether systems are character-
ized by neutrally-buoyant (or nearly so) cables, with either a minimal vehicle
at the end, as in a towed array, or a vehicle capable of maneuvering itself, such
as a remotely-operated vehicle. The towed array is a relatively high-velocity
system that nominally streams out horizontally behind the vessel. An ROV,
on the other hand, operates at low speed, and must have large propulsors to
control the tether if there are currents. Heavy systems, in contrast, employ a
heavy cable and possibly a heavy weight; the rationale is that gravity will tend
to keep the cable vertical and make the deployment robust against currents
and towing speed. The heavy systems will generally transmit surface motions
and tensions to the towed vehicle much more easily than light-tether systems.
We will not discuss light systems specifically here, but rather look at heavy
systems. Most of the analysis can be adapted to either case, however.
13.1 Statics
13.1.1 Force Balance
For the purposes of deriving the static configuration of a cable in a flow, we
assume for the moment that that it is inextensible. Tension and hydrostatic
so that the condition dφ/ds = 0 requires from the force balance
1 2
δ cos φc − sin φc = 0.
2
We are considering the case of 0 < φc < π/2. Substituting sin2 φc = 1−cos2 φc ,
we solve a quadratic equation and keep only the positive solution:
√
cos φc = δ 2 + 1 − 1. (157)
In the case of a √
very heavy cable, δ is large, and the linear approximation of
the square root 1 + ≈ 1 + /2 gives
1
cos φc amp; ' amp; −→
2δ
π 1
φc amp; ' amp; − . (158)
2 2δ
For a very light cable, δ is small; the same approximation gives
13.2 Linearized Dynamics 89
√
cos φc ' 1 − δ −→ φc ' 2δ.
The table below gives some results of the exact solution, and the approxima-
tions.
δ amp; exact amp; δ >> 1 amp; δ << 1

0.1 amp; 0.44 amp; - amp; 0.45
0.2 amp; 0.61 amp; - amp; 0.63
0.5 amp; 0.91 amp; 0.57 amp; 1.00
1.0 amp; 1.14 amp; 1.07 amp; 1.41
2.0 amp; 1.33 amp; 1.32 amp; -
5.0 amp; 1.47 amp; 1.47 amp; -
13.2 Linearized Dynamics

13.2.1 Derivation
The most direct procedure for deriving useful linear dynamic equations for a
planar cable problem is to consider the total tension and angle as made up of
static parts summed with dynamic parts:
T (s, t) = T̄ (s) + T̃ (s, t)

φ(s, t) = φ̄(s) + φ̃(s, t).
We also write the axial deflection with respect to the static configuration as
p(s, t), and the lateral deflection q(s, t). It follows that φ̃ = ∂q/∂s. Now
augment the two static configuration equations with inertial components:
!2
∂2p ∂ T̄ ∂ T̃ 1 ∂p
m 2 amp; = amp; + − Wn sin(φ̄ + φ̃) − ρCt d U cos φ +
∂t ∂s ∂s 2 ∂t
∂2q
!
∂ φ̄ ∂ φ̃
(m + ma ) 2 amp; = amp; (T̄ + T̃ ) + − Wn cos(φ̄ + φ̃) +
∂t ∂s ∂s
!2
1 ∂q
amp; amp; ρCn d U sin φ − .
2 ∂t
Here the material mass of the cable per unit length is m, and its transverse
added mass is ma . Note that avoiding the drag law form v|v| again, we have
implicitly assumed that U cos φ > |∂p/∂t| and U sin φ > |∂q/∂t|. If it is not
the case, say U = 0, then equivalent linearization can be used for the quadratic
drag.
Now we perform the trigonometry substitutions in the weight terms, let φ ' φ̄
for the calculation of drag, and substitute the constitutive (Hooke’s) law
∂ T̃ ∂2p
= EA 2 .
∂s ∂s
The static solution cancels out of both governing equations, and keeping only
linear terms we obtain
∂2p ∂2p ∂q ∂p
m 2
amp; = amp; EA 2
− Wn cos φ̄ − ρCt dU cos φ̄
∂t ∂s ∂s ∂t
2 2
∂ q ∂ q ∂p ∂ φ̄ ∂q ∂q
(m + ma ) 2 amp; = amp; T̄ 2 + EA + Wn sin φ̄ − ρCn dU sin φ̄ .
∂t ∂s ∂s ∂s ∂s ∂t
The axial dynamics (p) couples with the lateral equation through the weight
term −Wn cos φ̄φ̃. The lateral dynamics (q) couples with the axial through the
term T̃ ∂ φ̄/∂s. An additional weight term Wn sin φ̄∂q/∂s also appears. The
uncoupled dynamics are both in the form of damped wave equations
∂2p ∂p ∂2p
m + b t amp; = amp; EA
∂t2 ∂t ∂s2
2 2
∂ q ∂q ∂ q ∂q
(m + ma ) 2 + bn amp; = amp; T̄ 2 + Wn sin φ̄ ,
∂t ∂t ∂s ∂s
where we made the substitution bt = ρCt dU cos φ̄ and bn = ρCn dU sin φ̄. To a
linear approximation, the out-of-plane vibrations of a cable are also governed
by the second equation above.
Because of light damping in the tangential direction, heavy cables easily trans-
mit motions and tensions along their length, and can develop longitudinal
resonant conditions (next section). In contrast, the lateral cable motions are
heavily damped, such that disturbances only travel a few tens or hundreds of
meters before they dissipate. The nature of the lateral response, in and out
of the towing plane, is a very slow, damped nonlinear filter. High-frequency

vessel motions in the horizontal plane are completely missed by the vehicle,
while low-frequency motions occur sluggishly, and only after a significant delay
time.
amp; axial amp; lateral

q r
EA T̄ (s)
wave speed amp; m
amp; m+ma
amp; FAST amp; SLOW
r !
q
nπ EA nπ T̄ (L/2)
natural frequency amp; L m
amp; O L m+ma
mode shape amp; sine/cosine amp; Bessel function

damping amp; Ct ' O(0.01) amp; Cn ' O(1)
disturbances amp; amp;
travel down amp; YES amp; NO
cable? amp; amp;
13.2.2 Damped Axial Motion

Mode Shape. The axial direction is of particular interest, since it is lightly
damped and forced by the heaving of vessels in seas. Consider a long cable
governed by the damped wave equation
mp̈ + bt ṗ = EAp00 (159)

We use over-dots to indicate time derivatives, and primes to indicate spatial
derivatives. At the surface, we impose the motion
p(L, t) = P cos ωt, (160)

while the towed vehicle, at the lower end, is an undamped mass responding to
the local tension variations:
∂p(0, t)
EA = M p̈(0, t). (161)
∂s
These top and bottom behaviors comprise the boundary conditions for the
wave equation. We let p(s, t) = p̃(s) cos ωt, so that
mω 2 − iωbt
!
00
p̃ + p̃ = 0. (162)
EA
This admits the solution p̃(s) = c1 cos ks + c2 sin ks, where
s
mω 2 − iωbt
k= . (163)
EA
6 0. The top and bottom boundary conditions
Note that k is complex when bt =
give, respectively,
P amp; = amp; c1 cos kL + c2 sin kL

0 amp; = amp; c1 + δc2 ,
where δ = EAk/ω 2 M . These can be combined to give the solution

δ cos ks − sin ks
p̃ = P . (164)
δ cos kL − sin kL
In the case that M → 0, the scalar δ → ∞, simplifying the result to p̃ =
P cos ks/ cos kL.
Dynamic Tension. It is possible to compute the dynamic tension via T̃ =

EAp̃0 . We obtain
δ sin ks + cos ks
T̃ = −EAP k . (165)
δ cos kL − sin kL
There are two dangerous situations:
• The maximum tension is T̄ + |T̃ | and must be less than the working load
of the cable. This is normally problematic at the top of the cable, where
the static tension is highest.
• If |T̃ | > T̄ , the cable will unload completely and then reload with ex-
tremely high impulsive forces. This is known as snap loading; it occurs
primarily at the vehicle, where T̄ is low.
Natural Frequency. The natural frequency can be found by letting bt = 0,

and investigating the singularity of p̃, for which δ cos kL = sin kL. In general,
q
kL << 1, but we find that a first-order approximation yields ω = EA/LM ,
which is only a correct answer if M >> mL, i.e., the system is dominated by
the vehicle mass. Some higher order terms need to be kept. We start with
better approximations for sin() and cos():
(kL)2 (kL)2
! !
δ 1− = kL 1 − .
2 6
Employing the definition for δ, and recalling that ω 2 = k 2 EA/m, we arrive at
(kL)2 (kL)2
! !
mL
1− = (kL)2 1− .
M 2 6
If we match up to second order in kL, then
v
EA/L
u
u
ω=t .
M + mL/2
This has the familiar form of the square root of a stiffness divided by a mass:
the stiffness of the cable is EA/L, and the mass that is oscillating is M +mL/2.
In very deep water, the effects of mL/2 dominate; if ρc = m/A is the density
of the cable, we have the approximation
s
1 2E
ω' .
L ρc
A few examples are given below for a steel cable with E = 200 × 109 P a, and
ρc = 7000kg/m3 . The natural frequencies near wave excitation at the surface
vessel must be taken into account in any design or deployment. Even if a
cable can withstand the effects of resonance, it may be undesirable to expose
the vehicle to these motions. Some solutions in use today are: stable vessels
(e.g., SWATH), heave compensation through an active crane, a clump weight
below which a light cable is employed, and an S-shaped length of cable at the
bottom formed with flotation balls.
L = 500m amp; ωn = 15.0rad/s

1000m amp; 7.6rad/s
2000m amp; 3.7rad/s
5000m amp; 1.5rad/s
13.3 Cable Strumming

Cable strumming causes a host of problems, including obvious fatigue when
the amplitudes and frequencies are high. The most noteworthy issue with
towing is that the vibrations may cause the normal drag coefficient Cn to
increase dramatically – from about 1.2 for a non-oscillating cable to as high
as 3.5. This drag penalty decreases the critical angle of towing, so that larger
lengths of cable are needed to reach a given depth, and the towed system lags
further and further behind the surface vessel. The static tension will increase
accordingly as well.
Strumming of cables is caused by the proximity of a preferred vortex forma-
tion frequency ωS to the natural frequency of the structure ωn . This latter
frequency can be obtained as a zero of the lightly-damped Bessel function
solution of the lateral dynamics equation above. The preferred frequency of
vortex formation is given by the empirical relation ωS = 2πSU/d, where S
is the Strouhal number, about 0.16-0.20 for a large range of Re. Strumming
of amplitude d/2 or greater can occur for 0.6 < ωS /ωn < 2.0. The book by
Blevins is a good general reference.
13.4 Vehicle Design

The physical layout of a towed vehicle is amenable to the analysis tools of
self-propelled vehicles, with the main exceptions that the towpoint presents
a large mean force as well as some disturbances, and that the vehicle can be
quite heavy in water. Here are basic guidelines to be considered:
1. The towpoint must be located above the vehicle center of in-water weight,
for basic roll and pitch stability.
2. The towpoint should be forward of the aerodynamic center, for towing

stability reasons.
3. The combined center of mass (material and added mass) should be longi-
tudinally between the towpoint and the aerodynamic center, and nearer
13.4 Vehicle Design 95
the towpoint. This will ensure that high-frequency disturbances do not

induce excessive pitching.
4. The towpoint should be longitudinally forward of the center of in-air

weight, so that the vehicle enters the water fins first, and self-stabilizes
with U > 0.
5. The center of buoyancy should be behind the in-water center of weight,

so that the vehicle pitches downward at small U , and hence the net lift
force is downward, away from the surface.
Meeting all of these criteria simultaneously is no small feat, and the perfor-
mance of the device is very sensitive to small perturbations in the geometry.
For this reason, full-scale experiments are commonly used in the design pro-
cess.
96 14 TRANSFER FUNCTIONS & STABILITY
14 TRANSFER FUNCTIONS & STABILITY

The reader is referred to Laplace Transforms in the section MATH FACTS for
preliminary material on the Laplace transform. Partial fractions are presented
here, in the context of control systems, as the fundamental link between pole
locations and stability.
14.1 Partial Fractions

Solving linear time-invariant systems by the Laplace Transform method will
generally create a signal containing the (factored) form
K(s + z1 )(s + z2 ) · · · (s + zm )
Y (s) = . (166)
(s + p1 )(s + p2 ) · · · (s + pn )
Although for the moment we are discussing the signal Y (s), later we will see
that dynamic systems are described in the same format: in that case we call
the impulse response G(s) a transfer function. A system transfer function is
identical to its impulse response, since L(δ(t)) = 1.
The constants −zi are called the zeros of the transfer function or signal, and
−pi are the poles. Viewed in the complex plane, it is clear that the magnitude
of Y (s) will go to zero at the zeros, and to infinity at the poles.
Partial fraction expansions alter the form of Y (s) so that the simple transform
pairs can be used to find the time-domain output signals. We must have
m < n; if this is not the case, then we have to divide the numerator by the
denominator as necessary to find a simple form.
14.2 Partial Fractions: Unique Poles

Under the condition m < n, it is a fact that Y (s) is equivalent to
a1 a2 an
Y (s) = + + ··· , (167)
s + p1 s + p2 s + pn
in the special case that all of the poles are unique and real. The coefficient
ai is termed the residual associated with the i’th pole, and once all these are
found it is a simple matter to go back to the transform table and look up the
time-domain responses.
14.3 Example: Partial Fractions with Unique Real Poles 97
How to find ai ? A simple rule applies: multiply the right-hand sides of the
two equations above by (s + pi ), evaluate them at s = −pi , and solve for ai ,
the only one left.
14.3 Example: Partial Fractions with Unique Real Poles
s(s + 6)
G(s) = e−2s .
(s + 4)(s − 1)
Since we have a pure delay and m = n, we can initially work with G(s)/se−2s .
We have
s+6 a1 a2
= + , giving
(s + 4)(s − 1) s+4 s−1
h i
(s+6)(s+4) 2
a1 = (s+4)(s−1) s=−4
= −
5
h
(s+6)(s−1)
i 7
a2 = (s+4)(s−1) s=1
=
5
Thus
2 7
L−1 (G(s)/se−2s ) = − e−4t + et −→
5 5
8 7
g(t) = δ(t − 2) + e−4(t−2) + et−2 .
5 5
The impulse response is needed to account for the step change at t = 2. Note
that in this example, we were able to apply the derivative operator s after
expanding the partial fractions. For cases where a second derivative must be
taken, i.e., m ≥ n + 1, special care should be used when accounting for the
signal slope discontinuity at t = 0. The more traditional method, exemplified
by Ogata, may prove easier to work through.
The case of repeated real roots may be handled elegantly, but this condition
rarely occurs in applications.
98 14 TRANSFER FUNCTIONS & STABILITY
14.4 Partial Fractions: Complex-Conjugate Poles

A complex-conjugate pair of poles should be kept together, with the following
procedure: employ the form
b1 s + b2 a3
Y (s) = + + ···, (168)
(s + p1 )(s + p2 ) s + p3
where p1 = p∗2 (complex conjugate). As before, multiply through by (s +
p1 )(s + p2 ), and then evaluate at s = −p1 .
14.5 Example: Partial Fractions with Complex Poles
s+1 b1 s + b2 a3
G(s) = = + :
s(s + j)(s − j) (s + j)(s − j) s
s+1

= [b1 s + b2 ]s=−j −→
s s=−j
1 + j = −b1 j + b2 −→
b1 = −1
b2 = 1; also
" #
s+1
= a3 = 1.
(s + j)(s − j) s=0
Working out the inverse transforms from the table of pairs, we have simply
(noting that ζ = 0)
g(t) = − cos t + sin t + 1(t).
14.6 Stability in Linear Systems

In linear systems, exponential stability occurs when all the real exponents of e
are strictly negative. The signals decay within an exponential envelope. If one
exponent is 0, the response never decays or grows in amplitude; this is called
marginal stability. If at least one real exponent is positive, then one element
of the response grows without bound, and the system is unstable.
14.7 Stability ⇐⇒ Poles in LHP 99
14.7 Stability ⇐⇒ Poles in LHP

In the context of partial fraction expansions, the relationship between stability
and pole locations is especially clear. The unit step function 1(t) has a pole at
zero, the exponential e−at has a pole at −a, and so on. All of the other pairs
exhibit the same property: A system is stable if and only if all of the poles
occur in the left half of the complex plane. Marginally stable parts correlate
with a zero real part, and unstable parts to a positive real part.
14.8 General Stability

There are two definitions, which apply to systems with input ~u(t) and output
~y (t).
1. Exponential. If ~u(t) = ~0 and ~y (0) = ~yo , then |yi (t)| < αe−γt , for finite
α and γ > 0. The output asymptotically approaches zero, within a
decaying exponential envelope.
2. Bounded-Input Bounded-Output (BIBO). If ~y (0) = ~0, and |fi (t)| <

γ, γ > 0 and finite, then |yi (t)| < α, α > 0 and finite.
In linear time-invariant systems, the two definitions are identical. Exponential

stability is easy to check for linear systems, but for nonlinear systems, BIBO
stability is usually easier to achieve.
100 15 CONTROL FUNDAMENTALS
15 CONTROL FUNDAMENTALS
15.1 Introduction
15.1.1 Plants, Inputs, and Outputs
Controller design is about creating dynamic systems that behave in useful

ways. Many target systems are physical; we employ controllers to steer ships,
fly jets, position electric motors and hydraulic actuators, and distill alcohol.
Controllers are also applied in macro-economics and many other important,
non-physical systems. It is the fundamental concept of controller design that
a set of input variables acts through a given “plant” to create an output.
Feedback control then uses sensed plant outputs to apply corrective inputs:
Plant Inputs Outputs Sensors

Jet aircraft elevator, rudder, etc. altitude, hdg altimeter, GPS
Marine vessel rudder angle heading gyrocompass
Hydraulic robot valve position tip position joint angle
U.S. economy fed interest rate, etc. prosperity inflation, M1
Nuclear reactor cooling, neutron flux power level temp., pressure
15.1.2 The Need for Modeling
Effective control system design usually benefits from an accurate model of

the plant, although it must be noted that many industrial controllers can be
tuned up satisfactorily with no knowledge of the plant. Ziegler and Nichols, for
example, developed a general recipe which we detail later. In any event, plant
models simply do not match real-world systems exactly; we can only hope to
capture the basic components in the form of differential or integro-differential
equations.
Beyond prediction of plant behavior based on physics, the process of system
identification generates a plant model from data. The process is often prob-
lematic, however, since the measured response could be corrupted by sensor
noise or physical disturbances in the system which cause it to behave in unpre-
dictable ways. At some frequency high enough, most systems exhibit effects
that are difficult to model or reproduce, and this is a limit to controller per-
formance.
15.2 Representing Linear Systems 101
15.1.3 Nonlinear Control

The bulk of this subject is taught using the tools of linear systems analysis.
The main reason for this restriction is that nonlinear systems are difficult
to model, difficult to design controllers for, and difficult overall! Within the
paradigm of linear systems, there are many sets of powerful tools available.
The reader interested in nonlinear control is referred to the book by Slotine
and Li.
15.2 Representing Linear Systems

Except for the most heuristic methods of tuning up simple systems, control
system design depends on a model of the plant. The transfer function de-
scription of linear systems has already been described in the discussion of the
Laplace transform. The state-space form is an entirely equivalent time-domain
representation that makes a clean extension to systems with multiple inputs
and multiple outputs, and opens the way to standard tools from linear algebra.
15.2.1 Standard State-Space Form

We write a linear system in a state-space form as follows
ẋ = Ax + Bu + Gw (169)
y = Cx + Du + v
where
• x is a state vector, with as many elements as there are orders in the

governing differential equations.
• A is a matrix mapping x to its derivative; A captures the natural dy-
namics of the system without external inputs.
• B is an input gain matrix for the control input u.
• G is a gain matrix for unknown disturbance w; w drives the state just
like the control u.
• y is the observation vector, comprised mainly of a linear combination of
states Cx (where C is a matrix.
• Du is a direct map from input to output (usually zero for physical sys-
tems).
• v is an unknown sensor noise which corrupts the measurement.
w G
v
+ +
+ x’ x +
u B 1/s C y
+ +
15.2.2 Converting a State-Space Model into a Transfer Function

There are a number of canonical state-space forms available, which can create
the same transfer function. In the case of no disturbances or noise, the transfer
function (or transfer matrix) can be written as
y(s)
G(s) = = C(sI − A)−1 B + D, (170)
u(s)
where I is the identity matrix with the same size as A. A similar equation
holds for y(s)/w(s), and clearly y(s)/v(s) = I.
15.2.3 Converting a Transfer Function into a State-Space Model

It may be possible to write the corresponding differential equation along one
row of the state vector, and then cascade derivatives. For example, consider
the following system:
my 00 (t) + by 0 (t) + ky(t) = u0 (t) + u(t) (mass-spring-dashpot)

s+1
G(s) =
ms2 + bs + k
15.3 PID Controllers 103
Setting ~x = [y 0 , y]T , we obtain the system

" # " #
d~x −b/m −k/m 1/m
= ~x + u
dt 1 0 0
y = [1 1] ~x
Note specifically that dx2 /dt = x1 , leading to an entry of 1 in the off-diagonal
of the second row in A. Entries in the C-matrix are easy to write in this case
because of linearity; the system response to u0 is the same as the derivative of
the system response to u.
15.3 PID Controllers

The most common type of industrial controller is the proportional-integral-
derivative (PID) design. If u is the output from the controller, and e is the
error signal it receives, this control law has the form
Z t
u(t) = kp e(t) + ki e(τ )dτ + kd e0 (t),
0
U (s) ki
C(s) = = kp + + k d s (171)
E(s) s
1

= kp 1 + + τd s ,
τi s
where the last line is written using the conventions of one overall gain kp , plus
a time characteristic to the integral part (τi ) and and time characteristic to
the derivative part (τd ).
In words, the proportional part of this control law will create a control action
that scales linearly with the error – we often think of this as a spring-like action.
The integrator is accumulating the error signal over time, and so the control
action from this part will continue to grow as long as an error exists. Finally,
the derivative action scales with the derivative of the error. The controller will
retard motion toward zero error, which helps to reduce overshoot.
The common variations are: P , P D, P I, P ID.
15.4 Example: PID Control

Consider the case of a mass (m) sliding on a frictionless table. It has a per-
fect thruster that generates force u(t), but is also subject to an unknown
disturbance d(t). If the linear position of the mass is y(t), and it is perfectly
measured, we have the plant
my 00 (t) = u(t) + d(t).
Suppose that the desired condition is simply y(t) = 0, with initial conditions
y(0) = yo and y 0 (0) = 0.
15.4.1 Proportional Only

A proportional controller alone invokes the control law u(t) = −kp y(t), so that
the closed-loop dynamics follow
my 00 (t) = −kp y(t) + d(t).

q
kp
In the absence of d(t), we see that y(t) = yo cos m
t, a marginally stable
response that is undesirable.
15.4.2 Proportional-Derivative Only

Let u(t) = −kp y(t) − kd y 0 (t), and it follows that
my 00 (t) = −kp y(t) − kd y 0 (t) + d(t).
The system now resembles a second-order mass-spring-dashpot system where

kp plays the part of the spring, and kd the part of the dashpot. With an
excessively large value for kd , the system would be overdamped and very slow
to respond to any command. In most applications, a small amount of overshoot
is employed because the response q time is shorter. The kd valueqfor critical
damping in this example is 2 mkp , and so the rule is kd < 2 mkp . The
result, easily found using the Laplace transform, is
" #
−kd
t kd
y(t) = yo e 2m cos ωd t + sin ωd t ,
2mωd
15.5 Heuristic Tuning 105
q
where ωd = 4mkp − kd2 /2m. This response is exponentially stable as desired.
Note that if the mass had a very large amount of natural damping, a negative
kd could be used to cancel some of its effect and speed up the system response.
Now consider what happens if d(t) has a constant bias do : it balances exactly
the proportional control part, eventually settling out at y(t = ∞) = do /kp . To
achieve good rejection of do with a P D controller, we would need to set kp very
large. However, very large values of kp will also drive the resonant frequency
ωd up, which is unacceptable.
15.4.3 Proportional-Integral-Derivative
Rt
Now let u(t) = −kp y(t) − ki 0 y(τ )dτ − kd y 0 (t): we have
Z t
00
my (t) = −kp y(t) − ki y(τ )dτ − kd y 0 (t) + d(t).
0
The control system has now created a third-order closed-loop response. If

d(t) = do , a time derivative leads to
my 000 (t) + kp y 0 (t) + ki y(t) + kd y 00 (t) = 0,
so that y(t = ∞) = 0, as desired, provided the roots are stable.
15.5 Heuristic Tuning

For many practical systems, tuning of a PID controller may proceed without
any system model. This is especially pertinent for plants which are open-loop
stable, and can be safely tested with varying controllers. One useful approach
is due to Ziegler and Nichols (1942), which transforms the basic characteristics
of a step response (e.g., the input is 1(t)) into a reasonable PID design. The
idea is to approximate the response curve by a first-order lag (gain k and time
constant τ ) and a pure delay T :
ke−T s
G(s) ' (172)
τs + 1
The following rules apply only if the plant contains no dominating, lightly-
damped complex poles, and has no poles at the origin:
P kp = 1.0τ /T
PI kp = 0.9τ /T ki = 0.27τ /T 2
PID kp = 1.2τ /T ki = 0.60τ /T 2 kd = 0.60τ
Note that if no pure time delay exists (T = 0), this recipe suggests the pro-
portional gain can become arbitrarily high! Any characteristic other than a
true first-order lag would therefore be expected to cause a measurable delay.
15.6 Block Diagrams of Systems

15.6.1 Fundamental Feedback Loop
The topology of a feedback system can be represented graphically by consid-
ering each dynamical system element to reside within a box, having an input
line and an output line. For example, the plant used above (a simple mass)
has transfer function P (s) = 1/ms2 , which relates the input, force u(s), into
the output, position y(s). In turn, the PD-controller has transfer function
C(s) = kp + kd s; its input is the error signal E(s) = −y(s), and its output is
force u(s). The feedback loop in block diagram form is shown below.
u(s) y(s)
C(s) P(s)
15.6.2 Block Diagrams: General Case

The simple feedback system above is augmented in practice by three external
inputs. The first is a process disturbance, which can be taken to act at the
input of the physical plant, or at the output. In the former case, it is additive
with the control action, and so has some physical meaning. In the second case,
the disturbance has the same units as the plant output.
Another external input is the reference command or setpoint, used to create a
more general error signal e(s) = r(s) − y(s). Note that the feedback loop, in
trying to force e(s) to zero, will necessarily make y(s) approximate r(s).
The final input is sensor noise, which usually corrupts the feedback signal y(s),
causing some error in the evaluation of e(s), and so on. Sensors with very poor
15.6 Block Diagrams of Systems 107
noise properties can ruin the performance of a control system, no matter how
perfectly understood are the other components.
du dy
+ e u + + y
r C P
- + +
+
n
+
15.6.3 Primary Transfer Functions

Some algebra shows that
e 1
= =S
r 1 + PC
y PC
= =T
r 1 + PC
u C
= = U.
r 1 + CP
e/r = S relates the reference input and noise to the error, and is known as the
sensitivity function. We would generally like S to be small at low frequencies,
so that the tracking error there is small. y/r = T is called the complementary
sensitivity function. Note that S + T = 1, implying that these two functions
must always trade off; they cannot both be small or large at the same time.
Other systems we encounter again later are the (forward) loop transfer function
P C, the loop transfer function broken between C and P : CP , and
e −P
=
du 1 + PC
y P
=
du 1 + PC
u −CP
=
du 1 + CP
e −1
= = −S
dy 1 + PC
y 1
= =S
dy 1 + PC
u −C
= = −U
dy 1 + CP
e −1
= = −S
n 1 + PC
y −P C
= = −T
n 1 + PC
u −C
= = −U.
n 1 + CP
If the disturbance is taken at the plant output, then the three functions S, T ,
and U (control action) completely describe the system. This will in fact be
the procedure when we address loopshaping.
109
16 MODAL ANALYSIS
16.1 Introduction
The evolution of states in a linear system occurs through independent modes,
which can be driven by external inputs, and observed through plant output.
This section provides the basis for modal analysis of systems. Throughout, we
use the state-space description of a system with D = 0:
~x˙ = A~x + B~u

~y = C~x.
16.2 Matrix Exponential

16.2.1 Definition
In the instance of an unforced response to initial conditions, consider the
system
~x˙ = A~x, ~x(t = 0) = χ

~.
In the scalar case, the response is x(t) = χeat , giving a decaying exponential
if a < 0. The same notation holds for the case of a vector ~x, and matrix A:
~x(t) = eAt χ
~ , where
(At)2
eAt = I + At + + ···
2!
eAt is usually called the matrix exponential.
16.2.2 Modal Canonical Form

Introductory material on the eigenvalue problem and modal decomposition
can be found in the MATH FACTS section. This modal decomposition of A
leads to a very useful state-space representation. Namely, since A = V ΛV −1 ,
a transformation of state variables can be made, ~x = V ~z, leading to
110 16 MODAL ANALYSIS
~z˙ = Λ~z + V −1 B~u (173)

~y = CV ~z.
This is called the modal canonical form, since the states are simply the modal
amplitudes. These states are uncoupled in Λ, but may be coupled through the
input (V −1 B) and output (CV ) mappings. The modal form is numerically
robust for computations.
16.2.3 Modal Decomposition of Response

Now we are ready to look at the matrix exponential eAt in terms of its con-
stituent modes. Employing the above form for A, we find that
(At)2
eAt = I + At + + ···
2!
(Λt)2
!
= V I + Λt + + ··· WT
2!
= V eΛt W T
n
eλi t~vi w
~ iT .
X
=
i=1
In terms of the response to an initial condition χ

~ , we have
n
eλi t~vi (w
~ iT χ
X
~x(t) = ~ ).
i=1
The product w ~ iT χ
~ is a scalar, the projection of the initial conditions onto the
i’th mode. If χ~ is perpendicular to w ~ i , then the product is zero and the i’th
mode does not respond. Otherwise, the i’th mode does participate in the
response. The projection of the i’th mode onto the states ~x is through the
right eigenvector ~vi .
For stability of the system, the eigenvalues of A, that is, λi , must have neg-
ative real parts; they are in fact the poles of the equivalent transfer function
description.
16.3 Forced Response and Controllability 111
16.3 Forced Response and Controllability

Now consider the system with an external input ~u:
~x˙ = A~x + B~u, ~x(t = 0) = χ

~.
Taking the Laplace transform of the system, taking into account the initial
condition for the derivative, we have
s~x(s) − χ~ = A~x(s) + B~u(s) −→

~x(s) = (sI − A)−1 χ
~ + (sI − A)−1 B~u(s).
Thus (sI − A)−1 can be recognized as the Laplace transform of the matrix
exponential eAt . In the time domain, the second term then has the form of a
convolution of the matrix exponential and the net input B~u:
Z t
~x(t) = eA(t−τ ) B~u(τ )dτ
0
n Z t
eλi (t−τ )~vi w
~ iT B~u(τ )dτ.
X
=
i=1 0
Suppose now that there are m inputs, such that B = [~b1 , ~b2 , · · · , ~bm ]. Then
some rearrangement will give
n m Z t
~ iT~bk ) eλi (t−τ )~uk (τ )dτ.
X X
~x(t) = ~vi (w
i=1 k=1 0
The product w ~ iT~bk , a scalar, represents the projection of the k’th control chan-
nel onto the i’th mode. We say that the i’th mode is controllable from the
k’th input if the product is nonzero. If a given mode i has w ~ iT~bk = 0 for all
input channels k, then the mode is uncontrollable.
In normal applications, controllability for the entire system is checked using
the following test: Construct the so-called controllability matrix:
Mc = [B, AB, · · · , An−1 B]. (174)

This matrix has size n × (nm), where m is the number of input channels. If
Mc has rank n, then the system is controllable, i.e., all modes are controllable.
112 16 MODAL ANALYSIS
16.4 Plant Output and Observability

We now turn to a related question: can the complete state vector of the system
be observed given only the output measurements ~y , and the known control ~u?
The response due to the external input is easy to compute deterministically,
through the convolution integral. Consider the part due to initial conditions
χ
~ . We found above
n
eλi t~vi w
~ iT χ
X
~x(t) = ~.
i=1
The observation is ~y = C~x (r channels of output), and writing
~c1T
 
C =  · .
 
~crT
the k’th channel of the output is
n
(~ckT ~vi )eλi t (w
~ iT χ
X
yk (t) = ~ ).
i=1
The i’th mode is observable in the k’th output if the product ~ckT ~vi 6= 0. We
say that a system is observable if every mode can be seen in at least one output
channel. The usual test for system observability requires computation of the
observability matrix:
h i
Mo = C T , AT C T , · · · , (AT )n−1 C T . (175)
This matrix has size n × (rn); the system is observable if Mo has rank n.
113
17 CONTROL SYSTEMS – LOOPSHAPING

17.1 Introduction
This section formalizes the notion of loopshaping for linear control system
design. The loopshaping approach is inherently two-fold. First, we shape the
open-loop transfer function (or matrix) P (s)C(s), to meet performance and
robustness specifications. Once this is done, then the compensator must be
computed, from from knowing the nominal product P (s)C(s), and the nominal
plant P (s).
Most of the analysis here is given for single-input, single-output systems, but
the link to multivariable control is not too difficult. In particular, absolute
values of transfer functions are replaced with the maximum singular values of
transfer matrices. Design based on singular values is the idea of L2 -control, or
LQG/LTR, to be presented in the next lectures.
17.2 Roots of Stability – Nyquist Criterion

We consider the SISO feedback system with reference trajectory r(s) and plant
output y(s), as given previously. The tracking error signal is defined as e(s) =
r(s) − y(s), thus forming the negative feedback loop. The sensitivity function
is written as
e(s) 1
S(s) = = ,
r(s) 1 + P (s)C(s)
where P (s) represents the plant transfer function, and C(s) the compensator.
The closed-loop characteristic equation, whose roots are the poles of the closed-
loop system, is 1 + P (s)C(s) = 0, equivalent to P (s)C(s) + P (s)C(s) =
0, where the underline and overline denote the denominator and numerator,
respectively. The Nyquist criterion allows us to assess the stability properties
of a system based on P (s)C(s) only. This method for design involves plotting
the complex loci of P (s)C(s) for the range ω = [−∞, ∞]. There is no explicit
calculation of the closed-loop poles, and in this sense the design approach is
quite different from the root-locus method (see Ogata).
114 17 CONTROL SYSTEMS – LOOPSHAPING
17.2.1 Mapping Theorem

We impose a reasonable assumption from the outset: The number of poles in
P (s)C(s) exceeds the number of zeros. It is a reasonable constraint because
otherwise the loop transfer function could pass signals with infinitely high
frequency. In the case of a PID controller (two zeros) and a second-order zero-
less plant, this constraint can be easily met by adding a high-frequency rolloff
to the compensator, the equivalent of low-pass filtering the error signal.
Let F (s) = 1 + P (s)C(s). The heart of the Nyquist analysis is the mapping
theorem, which answers the following question: How do paths in the s-plane
map into paths in the F -plane? We limit ourselves to closed, clockwise(CW)
paths in the s-plane, and the remarkable result of the mapping theorem is
Every zero of F (s) enclosed in the s-plane generates exactly one CW encir-
clement of the origin in the F (s)-plane. Conversely, every pole of F (s) enclosed
in the s-plane generates exactly one CCW encirclement of the origin in the
F (s)-plane. Since CW and CCW encirclements of the origin may cancel, the
relation is often written Z − P = CW .
The trick now is to make the trajectory in the s-plane enclose all unstable poles,
i.e., the path encloses the entire right-half plane, moving up the imaginary
axis, and then proceeding to the right at an arbitrarily large radius, back to
the negative imaginary axis.
Since the zeros of F (s) are in fact the poles of the closed-loop transfer function,
e.g., S(s), stability requires that there are no zeros of F (s) in the right-half
s-plane. This leads to a slightly shorter form of the above relation:
P = CCW. (176)
In words, stability requires that the number of unstable poles in F (s) is equal
to the number of CCW encirclements of the origin, as s sweeps around the
entire right-half s-plane.
17.2.2 Nyquist Criterion

The Nyquist criterion now follows from one translation. Namely, encirclements
of the origin by F (s) are equivalent to encirclements of the point (−1 + 0j)
by F (s) − 1, or P (s)C(s). Then the stability criterion can be cast in terms of
the unstable poles of P (s)C(s), instead of those of F (s):
P = CCW ←→ closed-loop stability (177)

17.2 Roots of Stability – Nyquist Criterion 115
This is in fact the complete Nyquist criterion for stability. It is a necessary and
sufficient condition that the number of unstable poles in the loop transfer func-
tion P (s)C(s) must be matched by an equal number of CCW encirclements
of the critical point (−1 + 0j).
There are several details to keep in mind when making Nyquist plots:
• If neither the plant nor the controller have unstable modes, then the loci
of P (s)C(s) must not encircle the critical point at all.
• Because the path taken in the s-plane includes negative frequencies (i.e.,
the negative imaginary axis), the loci of P (s)C(s) occur as complex
conjugates – the plot is symmetric about the real axis.
• The requirement that the number of poles in P (s)C(s) exceeds the num-
ber of zeros means that at high frequencies, P (s)C(s) always decays such
that the loci go to the origin.
• For the multivariable (MIMO) case, the procedure of looking at individ-
ual Nyquist plots for each element of a transfer matrix is unreliable and
outdated. Referring to the multivariable definition of S(s), we should
count the encirclements for the function [det(I + P (s)C(s)) − 1] instead
of P (s)C(s). The use of gain and phase margin in design is similar to
the SISO case.
17.2.3 Robustness on the Nyquist Plot

The question of robustness in the presence of modelling errors is central to
control system design. There are two natural measures of robustness for the
Nyquist plot, each having a very clear graphical representation. The loci need
to stay away from the critical point; how close the loci come to it can be
expressed in terms of magnitude and angle.
• When the angle of P (s)C(s) is −180◦ , the magnitude |P (s)C(s)| should

not be near one.
• When the magnitude |P (s)C(s)| = 1, its angle should not be −180◦ .
These notions lead to definition of the gain margin kg and phase margin γ for
a design. As the figure shows, the definition of kg is different for stable and
unstable P (s)C(s). Rules of thumb are as follows. For a stable plant, kg ≥ 2
and γ ≥ 30◦ ; for an unstable plant, kg ≤ 0.5 and γ ≥ 30◦ .
1/kg Im(s) P(s)C(s) Im(s)
Re(s) Re(s)
γ
1/kg
P(s)C(s)
Stable P(s)C(s) Unstable P(s)C(s)
17.3 Design for Nominal Performance

Performance requirements of a feedback controller, using the nominal plant
model, can be cast in terms of the Nyquist plot. We restrict the discussion to
the scalar case in the following sections.
Since the sensitivity function maps reference input r(s) to tracking error e(s),
we know that |S(s)| should be small at low frequencies. For example, if one-
percent tracking is to be maintained for all frequencies below ω = λ, then
|S(s)| < 0.01, ∀ω < λ. This can be formalized by writing
|W1 (s)S(s)| < 1, (178)

where W1 (s) is a stable weighting function of frequency. To force S(s) to be
small at low ω, W1 (s) should be large in the same range. The requirement
|W1 (s)S(s)| < 1 is equivalent to |W1 (s)| < |1 + P (s)C(s)|, and this latter
condition can be interpreted as: The loci of P (s)C(s) must stay outside the
disk of radius W1 (s), which is to be centered on the critical point (−1 + 0j).
The disk is to be quite large, possibly infinitely large, at the lower frequencies.
17.4 Design for Robustness

It is a ubiquitous observation that models of plants degrade with increasing
frequency. For example, the DC gain and slow, lightly-damped modes or zeros
are easy to observe, but higher-frequency components in the response may
17.4 Design for Robustness 117
be hard to capture or even excite repeatably. Higher-frequency behavior may

have more nonlinear properties as well.
The effects of modeling uncertainty can be considered to enter the nominal
feedback system as a disturbance at the plant output, dy . One of the most
useful descriptions of model uncertainty is the multiplicative uncertainty:
P̃ (s) = (1 + ∆(s)W2 (s))P (s). (179)

Here, P (s) represents the nominal plant model used in the design of the control
loop, and P̃ (s) is the actual, perturbed plant. The perturbation is of the
multiplicative type, ∆(s)W2 (s)P (s), where ∆(s) is an unknown but stable
function of frequency for which |∆(s)| ≤ 1. The weighting function W2 (s)
scales ∆(s) with frequency; W2 (s) should be growing with increasing frequency,
since the uncertainty grows. However, W2 (s) should not grow any faster than
necessary, since it will turn out to be at the cost of nominal performance.
In the scalar case, the weight can be estimated as follows: since P̃ /P − 1 =
∆W2 , it will suffice to let |P̃ /P − 1| < |W2 |.
Example: Let P̃ = k/(s − 1), where k is in the range 2–5. We need to create
a nominal model P = k0 /(s − 1), with the smallest possible value of W2 , which
will not vary with frequency in this case. Two equations can be written using
the above estimate, for the two extreme values of k, yielding k0 = 7/2, and
W2 = 3/7.
For constructing the Nyquist plot, we observe that
P̃ (s)C(s) = (1 + ∆(s)W2 (s))P (s)C(s). The path of the perturbed plant could
be anywhere on a disk of radius |W2 (s)P (s)C(s)|, centered on the nominal loci
P (s)C(s). The robustness condition is that this disk should not intersect the
critical point. This can be written as
|1 + P C| > |W2 P C| ←→
|W2 P C|
1 > ←→
|1 + P C|
1 > |W2 T |, (180)
where T is the complementary sensitivity function. The last inequality is thus

a condition for robust stability in the presence of multiplicative uncertainty
parametrized with W2 .
17.5 Robust Performance

The condition for good performance with plant uncertainty is a combination
of the above two conditions. Graphically, the disk at the critical point, with
radius |W1 |, should not intersect the disk of radius |W2 P C|, centered on the
nominal locus P C. This is met if
|W1 S| + |W2 T | < 1. (181)

The robust performance requirement is related to the magnitude |P C| at dif-
ferent frequencies, as follows:
1. At low frequency, |W1 S| ' |W1 /P C|, since |P C| is large. This leads
directly to the performance condition |P C| > |W1 | in this range.
2. At high frequency, W2 T | ' |W2 P C|, since |P C| is small. We must

therefore have |P C| < 1/|W2 |, for robustness.
17.6 Implications of Bode’s Integral

The loop transfer function P C cannot roll off too rapidly in the crossover
region. The simple reason is that a steep slope induces a large phase loss,
which in turn degrades the phase margin. To see this requires a short foray
into Bode’s integral. For a transfer function H(s), the crucial relation is
1Z∞ d
angle(H(jω0 )) = (ln|H(jω)) · ln(coth(|ν|/2))dν, (182)
π −∞ dν
where ν = ln(ω/ω0 ). The integral is hence taken over the log of a frequency
normalized with ω0 . It is not hard to see how the integral controls the angle:
the function ln(coth(|ν|/2)) is nonzero only near ν = 0, implying that the
angle depends only on the local slope d(ln|H|)/dν. Thus, if the slope is large,
the angle is large.
Example: Suppose H(s) = ω0n /sn , i.e., it is a simple function with n poles at
the origin, and no zeros; ω0 is a fixed constant. It follows that |H| = ω0n /ω n ,
and ln|H| = −nln(ω/ω0 ), so that d(ln|H|)/dν = −n. Then we have just
nZ ∞ nπ
angle(H) = − ln(coth(|ν|/2))dν = − . (183)
π −∞ 2
17.7 The Recipe for Loopshaping 119
This integral is trivial to look up or compute. Each pole at the origin clearly
induces 90◦ of phase loss. In the general case, each pole not at the origin
induces 90◦ of phase loss for frequencies above the pole. Each zero at the
origin adds 90◦ phase lead, while zeros not at the origin at 90◦ of phase lead
for frequencies above the zero. In the immediate neighborhood of these poles
and zeros, the phase may vary significantly with frequency.
The Nyquist loci are clearly susceptible to these variations is phase, and
the phase margin can be easily lost if the slope of P C at crossover (where
the magnitude is unity) is too steep. The slope can safely be first-order
(−20dB/decade, equivalent to a single pole), and may be second-order
(−40dB/decade) if an adequate phase angle can be maintained near crossover.
17.7 The Recipe for Loopshaping

In the above analysis, we have extensively described what the open loop trans-
fer function P C should look like, to meet robustness and performance speci-
fications. We have said very little about how to get the compensator C, the
critical component. For clarity, let the designed loop transfer function be re-
named, L = P C. We will use concepts from optimal linear control for the
MIMO case, but in the scalar case, it suffices to just pick
C = L/P. (184)
This extraordinarily simple step involves a plant inversion.
The overall idea is to first shape L as a stable transfer function meeting the
requirements of stability and robustness, and then divide through by the plant.
• When the plant is stable and has stable zeros (minimum-phase), the
division can be made directly.
• One caveat for the well-behaved plant is that lightly-damped poles or

zeros should not be cancelled verbatim by the compensator, because
the closed-loop response will be sensitive to any slight change in the
resonant frequency. The usual procedure is to widen the notch or pole
in the compensator, through a higher damping ratio.
• Non-minimum phase or unstable behavior in the plant can usually be

handled by performing the loopshaping for the closest stable model, and
then explicitly considering the effects of adding the unstable parts. In
the case of unstable zeros, we find that they impose an unavoidable

frequency limit for the crossover. In general, the troublesome zeros must
be faster than the closed-loop frequency response.
In the case of unstable poles, the converse is true: The feedback system
must be faster than the corresponding frequency of the unstable mode.
When a control system involves multiple inputs and outputs, the ideas from
scalar loopshaping can be adapted using the singular value. We list below
some basic properties of the singular value decomposition, which is analogous
to an eigenvector, or modal, analysis. Useful properties and relations for the
singular value are found in the section MATH FACTS.
The condition for MIMO robust performance can be written in many ways,
including a direct extension of our scalar condition
σ(W1 S) + σ(W2 T ) < 1. (185)

The open-loop transfer matrix L should be shaped accordingly. In the follow-
ing sections, we use the properties of optimal state estimation and control to
perform the plant inversion for MIMO systems.
121
18 LINEAR QUADRATIC REGULATOR

18.1 Introduction
The simple form of loopshaping in scalar systems does not extend directly
to multivariable (MIMO) plants, which are characterized by transfer matrices
instead of transfer functions.
The notion of optimality is closely tied to MIMO control system design. Op-
timal controllers, i.e., controllers that are the best possible, according to some
figure of merit, turn out to generate only stabilizing controllers for MIMO
plants. In this sense, optimal control solutions provide an automated design
procedure – we have only to decide what figure of merit to use. The lin-
ear quadratic regulator (LQR) is a well-known design technique that provides
practical feedback gains.
18.2 Full-State Feedback

For the derivation of the linear quadratic regulator, we assume the plant to be
written in state-space form ẋ = Ax + Bu, and that all of the n states x are
available for the controller. The feedback gain is a matrix K, implemented as
u = −K(x − xdesired ). The system dynamics are then written as:
ẋ = (A − BK)x + Kxdesired . (186)

xdesired represents the vector of desired states, and serves as the external input
to the closed-loop system. The “A-matrix” of the closed loop system is (A −
BK), and the “B-matrix” of the closed-loop system is K. The closed-loop
system has exactly as many outputs as inputs: n. The column dimension
of B equals the number of channels available in u, and must match the row
dimension of K. Pole-placement is the process of placing the poles of (A−BK)
in stable, suitably-damped locations in the complex plane.
18.3 Dynamic Programming

There are at least two conventional derivations for the LQR; we present here
one based on dynamic programming, due to R. Bellman. The key observation
is best given through a loose example:
Suppose that we are driving from Point A to Point C, and we ask what is
the shortest path in miles. If A and C represent Los Angeles and Boston,
122 18 LINEAR QUADRATIC REGULATOR
B1 C1
A B2 C2 D
B3 C3 n=3
s=2
for example, there are many paths to choose from! Assume that one way or
another we have found the best path, and that a Point B lies along this path,
say Las Vegas. Let X be an arbitrary point east of Las Vegas. If we were to
now solve the optimization problem for getting from only Las Vegas to Boston,
this same arbitrary point X would be along the new optimal path as well.
The point is a subtle one: the optimization problem from Las Vegas to Boston
is easier than that from Los Angeles to Boston, and the idea is to use this
property backwards through time to evolve the optimal path, beginning in
Boston.
Example: Nodal Travel. We now add some structure to the above experi-
ment. Consider now traveling from point A (Los Angeles) to Point D (Boston).
Suppose there are only three places to cross the Rocky Mountains, B1 , B2 , B3 ,
and three places to cross the Mississippi River, C1 , C2 , C3 .3 By way of no-
tation, we say that the path from A to B1 is AB1 . Suppose that all of the
paths (and distances) from A to the B-nodes are known, as are those from the
B-nodes to the C-nodes, and the C-nodes to the terminal point D. There are
nine unique paths from A to D.
A brute-force approach sums up the total distance for all the possible paths,
and picks the shortest one. In terms of computations, we could summarize that
this method requires nine additions of three numbers, equivalent to eighteen
additions of two numbers. The comparison of numbers is relatively cheap.
The dynamic programming approach has two steps. First, from each B-node,
pick the best path to D. There are three possible paths from B1 to D, for
example, and nine paths total from the B-level to D. Store the best paths
as B1 D|opt , B2 D|opt , B3 D|opt . This operation involves nine additions of two
3
Apologies to readers not familiar with American geography.
18.4 Dynamic Programming and Full-State Feedback 123
numbers.
Second, compute the distance for each of the possible paths from A to D,
constrained to the optimal paths from the B-nodes onward: AB1 + B1 D|opt ,
AB2 + B2 D|opt , or AB3 + B3 D|opt . The combined path with the shortest
distance is the total solution; this second step involves three sums of two
numbers, and total optimization is done in twelve additions of two numbers.
Needless to say, this example gives only a mild advantage to the dynamic
programming approach over brute force. The gap widens vastly, however, as
one increases the dimensions of the solution space. In general, if there are s
layers of nodes (e.g., rivers or mountain ranges), and each has width n (e.g., n
river crossing points), the brute force approach will take (sns ) additions, while
the dynamic programming procedure involves only (n2 (s − 1) + n) additions.
In the case of n = 5, s = 5, brute force requires 6250 additions; dynamic
programming needs only 105.
18.4 Dynamic Programming and Full-State Feedback

We consider here the regulation problem, that is, of keeping xdesired = 0.
The closed-loop system thus is intended to reject disturbances and recover
from initial conditions, but not necessarily follow y-trajectories. There are
several necessary definitions. First we define an instantaneous penalty function
l(x(t), u(t)), which is to be greater than zero for all nonzero x and u. The cost
associated with this penalty, along an optimal trajectory, is
Z ∞
J= l(x(t), u(t))dt, (187)
0
i.e., the integral over time of the instantaneous penalty. Finally, the optimal
return is the cost of the optimal trajectory remaining after time t:
Z ∞
V (x(t), u(t)) = l(x(τ ), u(τ ))dτ. (188)
t
.
We have directly from the dynamic programming principle
V (x(t), u(t)) = min {l(x(t), u(t))δt + V (x(t + δt), u(t + δt))} . (189)
u
The minimization of V (x(t), u(t)) is made by considering all the possible con-
trol inputs u in the time interval (t, t + δt). As suggested by dynamic pro-
gramming, the return at time t is constructed from the return at t + δt, and
the differential component due to l(x, u). If V is smooth and has no explicit
dependence on t, as written, then
∂V dx
V (x(t + δt), u(t + δt)) = V (x(t), u(t)) + δt + h.o.t. −→ (190)
∂x dt
∂V
= V (x(t), u(t)) + (Ax(t) + Bu(t))δt.
∂x
Now control input u in the interval (t, t + δt) cannot affect V (x(t), u(t)), so
inserting the above and making a cancellation gives
( )
∂V
0 = min l(x(t), u(t)) + (Ax(t) + Bu(t)) . (191)
u ∂x
We next make the assumption that V (x, u) has the following form:
1
V (x, u) = xT P x, (192)
2
where P is a symmetric matrix, and positive definite.45 It follows that
∂V
= xT P −→ (193)
∂x n o
0 = min l(x, u) + xT P (Ax + Bu) .
u
We finally specify the instantaneous penalty function. The LQR employs the
special quadratic form
1 1
l(x, u) = xT Qx + uT Ru, (194)
2 2
where Q and R are both symmetric and positive definite. The matrices Q
and R are to be set by the user, and represent the main “tuning knobs” for
the LQR. Substitution of this form into the above equation, and setting the
derivative with respect to u to zero gives
4
Positive definiteness means that xT P x > 0 for all nonzero x, and xT P x = 0 if x = 0.
5
This suggested form for the optimal return can be confirmed after the optimal controller
is derived.
18.5 Properties and Use of the LQR 125
0 = uT R + x T P B (195)
uT = −xT P BR−1
u = −R−1 B T P x.
The gain matrix for the feedback control is thus K = R−1 B T P . Inserting
this solution back into equation 194, and eliminating u in favor of x, we have
1 1
0 = xT Qx − xT P BR−1 B T P + xT P Ax.
2 2
All the matrices here are symmetric except for P A; since xT P Ax = xT AT P x,

we can make its effect symmetric by letting
1 1
xT P Ax = xT P Ax + xT AT P x,
2 2
leading to the final matrix-only result
0 = Q + P A + AT P − P BR−1 B T P. (196)
This equation is the matrix algebraic Riccati equation (MARE), whose solution
P is needed to compute the optimal feedback gain K. The MARE is easily
solved by standard numerical tools in linear algebra.
18.5 Properties and Use of the LQR

Static Gain. The LQR generates a static gain matrix K, which is not a
dynamical system. Hence, the order of the closed-loop system is the same as
that of the plant.
Robustness. The LQR achieves infinite gain margin: kg = ∞, implying that

the loci of (P C) (scalar case) or (det(I + P C) − 1) (MIMO case) approach
the origin along the imaginary axis. The LQR also guarantees phase margin
γ ≥ 60 degrees. This is in good agreement with the practical guidelines for
control system design.
Output Variables. In many cases, it is not the states x which are to be

minimized, but the output variables y. In this case, we set the weighting
matrix Q = C T Q0 C, since y = Cx, and the auxiliary matrix Q0 weights the
plant output.
Behavior of Closed-Loop Poles: Expensive Control. When R >>

C T Q0 C, the cost function is dominated by the control effort u, and so the
controller minimizes the control action itself. In the case of a completely stable
plant, the gain will indeed go to zero, so that the closed-loop poles approach
the open-loop plant poles in a manner consistent with the scalar root locus.
The optimal control must always stabilize the closed-loop system, however,
so there should be some account made for unstable plant poles. The expen-
sive control solution puts stable closed-loop poles at the mirror images of the
unstable plant poles.
Behavior of Closed-Loop Poles: Cheap Control. When R << C T Q0 C,

the cost function is dominated by the output errors y, and there is no penalty
for using large u. There are two groups of closed-loop poles. First, poles
are placed at stable plant zeros, and at the mirror images of the unstable
plant zeros. This part is akin to the high-gain limiting case of the root locus.
The remaining poles assume a Butterworth pattern, whose radius increases to
infinity as R becomes smaller and smaller.
The Butterworth pattern refers to an arc in the stable left-half plane, as shown
in the figure. The angular separation of n closed-loop poles on the arc is
constant, and equal to 180◦ /n. An angle 90◦ /n separates the most lightly-
damped poles from the imaginary axis.
18.5 Properties and Use of the LQR 127
X
φ /2
n=4
X φ = 45 degrees
φ
φ
φ
X
φ /2
X
128 19 KALMAN FILTER
19 KALMAN FILTER
19.1 Introduction
In the previous section, we derived the linear quadratic regulator as an optimal
solution for the full-state feedback control problem. The inherent assumption
was that each state was known perfectly. In real applications, the measure-
ments are subject to disturbances, and may not allow reconstruction of all the
states. This state estimation is the task of a model-based estimator having
the form:
x̂˙ = Ax̂ + Bu + H(y − C x̂) (197)

The vector x̂ represents the state estimate, whose evolution is governed by
the nominal A and B matrices of the plant, and a correction term with the
estimator gain matrix H. H operates on the estimation error mapped to
the plant output y, since C x̂ = ŷ. Given statistical properties of real plant
disturbances and sensor noise, the Kalman Filter designs an optimal H.
y
C
Bu
y - +
+ x x
H 1/s
+
+
19.2 Problem Statement

We consider the state-space plant model given by:
ẋ = Ax + Bu + W1 (198)
y = Cx + W2 .
There are n states, m inputs, and l outputs, so that A has dimension n × n, B
is n × m, and C is l × n. The plant subject to two random input signals, W1
19.3 Step 1: An Equation for Σ̇ 129
and W2 . W1 represents disturbances to the plant, since it drives ẋ directly;

W2 denotes sensor noise, which corrupts the measurement y.
An important assumption of the Kalman Filter is that W1 and W2 are each
vectors of unbiased, independent white noise, and that all the n + l channels
are uncorrelated. Hence, if E(·) denotes the expected value,
E(W1 (t1 )W1 (t2 )T ) = V1 δ(t1 − t2 ) (199)

E(W2 (t1 )W2 (t2 )T ) = V2 δ(t1 − t2 ) (200)
E(W1 (t)W2 (t)T ) = 0n×l . (201)
Here δ(t) represents the impulse (or delta) function. V1 is an n × n diagonal

matrix of intensities, and V2 is an l × l diagonal matrix of intensities.
The estimation error may be defined as e = x − x̂. It can then be verified that
ė = [Ax + Bu + W1 ] − [Ax̂ + Bu + H(y − C x̂)] (202)

= (A − HC)e + (W1 − HW2 ).
The eigenvalues of the matrix A − HC thus determine the stability properties

of the estimation error dynamics. The second term above, W1 + HW2 is
considered an external input.
The Kalman filter design provides H that minimizes the scalar cost function
J = E(eT W e), (203)
where W is an unspecified symmetric, positive definite weighting matrix. A

related matrix, the symmetric error covariance, is defined as
Σ = E(eeT ). (204)
There are two main steps for deriving the optimal gain H.
19.3 Step 1: An Equation for Σ̇

The evolution of Σ follows from some algebra and the convolution form of e(t).
We begin with
Σ̇ = E(ėeT + eėT ) (205)

T T T T T T
= E[(A − HC)ee + (W1 − HW2 )e + ee (A − C H ) +
e(W1T − W2T H T )].
The last term above can be expanded, using the property that
Z t
e(t) = e(A−HC)t e(0) + e(A−HC)(t−τ ) (W1 (τ ) − HW2 (τ )) dτ.
0
We have
E(e(W1T − W2T H T )) = e(A−HC)t E(e(0)(W1T − W2T H T )) +

Z t
e(A−HC)(t−τ ) E((W1 (τ ) − HW2 (τ ))
0
(W1T (t) − W2T (t)H T )) dτ
Z t
= e(A−HC)(t−τ ) E((W1 (τ ) − HW2 (τ ))
0
(W1T (t) − W2T (t)H T )) dτ
Z t
= e(A−HC)(t−τ ) (V1 δ(t − τ ) + HV2 H T δ(t − τ )) dτ
0
1 1
= V1 + HV2 H T .
2 2
To get from the first right-hand side to the second, we note that the initial
condition e(0) is uncorrelated with W1T − W2T H T . The fact that W1 and HW2
are uncorrelated leads to the third line, and the final result follows from
Z t 1
δ(t − τ )dτ = ,
0 2
i.e., the written integral includes only half of the impulse.
The final expression for E(e(W1T − W2T H T )) is symmetric, and therefore ap-
pears in Equation 205 twice, leading to
Σ̇ = (A − HC)Σ + Σ(AT − C T H T ) + V1 + HV2 H T . (206)

This equation governs propagation of the error covariance. It is independent
of the initial condition e(0), and depends on the (as yet) unknown estimator
gain matrix H.
19.4 Step 2: H as a Function of Σ 131
19.4 Step 2: H as a Function of Σ

We now make the connection between Σ = E(eeT ) (a matrix) and J =
E(eT W e) (a scalar). The trace of a matrix is the sum of its diagonal ele-
ments, and it can be verified that
J = E(eT W e) = trace(ΣW ). (207)

We now introduce an auxiliary cost function defined as
J 0 = trace(ΣW + ΛF ), (208)
where F is an n × n matrix of zeros, and Λ is an n × n matrix of unknown
Lagrange multipliers. Note that since F is zero, J 0 = J, so minimizing J 0 solves
the same problem. Lagrange multipliers provide an ingenious mechanism for
drawing constraints into the optimization; the constraint we invoke is the
evolution of Σ, Equation 206:

J 0 = trace ΣW + Λ(−Σ̇ + AΣ − HCΣ + ΣAT − ΣC T H T + V1 + HV2 H T )
(209)
If J 0 is an optimal cost, it follows that ∂J 0 /∂H = 0, i.e., the correct choice of
H achieves an extremal value. We need the following lemmas, which give the
derivative of a trace with respect to a constituent matrix:
∂
trace(−ΛHCΣ) = −ΛT ΣC T
∂H
∂
trace(−ΛΣC T H T ) = −ΛΣC T
∂H
∂
trace(ΛHV2 H T ) = ΛT HV2 + ΛHV2 .
∂H
Proofs of the first two are given at the end of this section; the last lemma uses
the chain rule, and the previous two lemmas. Next, we enforce Λ = ΛT , since
the values are arbitrary. Then the condition ∂J 0 /∂H = 0 leads to
0 = 2Λ(−ΣC T + HV2 ), satisfied if

H = ΣC T V2−1 . (210)
Duality of Linear Quadratic Regulator and Kalman Filter

Linear Quadratic Regulator Kalman Filter
ẋ = Ax + Bu ẋ = Ax + Bu + W1
y = Cx + W2
u = −Kx x̂˙ = Ax̂ + Bu + H(y − C x̂)
R∞ T T
2J = 0 (x Qx + u Ru)dt J = E(eT W e)
Q ≥ 0, R > 0 V1 ≥ 0, V2 > 0
K = R−1 B T P H = ΣC T V2−1
P A + AT P + Q − P BR−1 B T P = 0 ΣAT + AΣ + V1 − ΣC T V2−1 CΣ = 0
Hence the estimator gain matrix H can be written as a function of Σ. Inserting

this back into Equation 206, we obtain
Σ̇ = AΣ + ΣAT + V1 − ΣC T V2−1 CΣ. (211)
Equations 210 and 211 represent the practical solution to the Kalman filtering
problem, which minimizes the squared-norm of the estimation error. The
evolution of Σ is always stable, and depends only on the constant matrices
[A, C, V1 , V2 ]. Notice also that the result is independent of the weighting matrix
W , which might as well be the identity.
19.5 Properties of the Solution

The solution involves a matrix Riccati equation, like the LQR, suggesting a
duality with the LQR problem. This is in fact the case, and the same analysis
and numerical tools can be applied to both methodologies.
The steady-state solution for Σ is valid for time-invariant systems, leading to
a more common MARE form of Equation 211:
0 = AΣ + ΣAT + V1 − ΣC T V2−1 CΣ. (212)
The Kalman Filter is guaranteed to create a stable nominal dynamics A−HC,

as long as the plant is fully state-observable. This is dual to the stability
guarantee of the LQR loop, when the plant is state-controllable. Furthermore,
like the LQR, the KF loop achieves 60◦ phase margin, and infinite gain margin,
for all the channels together or independently.
19.6 Combination of LQR and KF 133
The qualitative dependence of the estimator gain H = ΣC T V2−1 on the other

parameters can be easily seen. Recall that V1 is the intensity matrix of the
plant disturbance, V2 is the intensity of the sensor noise, and Σ is the error
covariance matrix.
• A large uncertainty Σ creates large H, placing emphasis on the corrective

action of the filter.
• A small disturbance V1 , and large sensor noise V2 creates a small H,

weighting the model dynamics Ax̂ + Bu more.
• A large disturbance V1 , and small sensor noise V2 creates a large H, so

that the filter’s correction is dominant.
The limiting closed-loop poles of the Kalman filter are similar, and dual to
those of the LQR:
• V2 << V1 : good sensors, large disturbance, H >> 1, dual to cheap-

control problem. Some closed-loop poles go to the stable plant zeros, or
the mirror image of unstable plant zeros. The remaining poles follow a
Butterworth pattern whose radius increases with increasing V1 /V2 .
• V2 >> V1 : poor sensors, small disturbance, H small, dual to expensive-

control problem. Closed-loop poles go to the stable plant poles, and the
mirror images of the unstable plant poles.
19.6 Combination of LQR and KF

An optimal output feedback controller is created through the use of a Kalman
filter coupled with an LQR full-state feedback gain. This combination is usu-
ally known as the Linear Quadratic Gaussian design, or LQG. For the plant
given as
ẋ = Ax + Bu + W1
y = Cx + W2 ,
we put the Kalman Filter and controller gain G together as follows:

x̂˙ = Ax̂ + Bu + H(y − C x̂) (213)

u = −K x̂. (214)
C(s)
B
+ +
y H φ -K u
- +
φ = (sI-A) -1
There are two central points to this construction:

1. Separation Principle: The eigenvalues of the nominal closed-loop sys-
tem are made of up the eigenvalues of (A − HC) and the eigenvalues of
(A − BK), separately. See proof below.
2. Output Tracking: This compensator is a stand-alone system that,
as written, tries to drive its input y to zero. It can be hooked up to
receive tracking error e(s) = r(s) − y(s) as an input instead, so that it
is not limited to the regulation problem alone. In this case, x̂ no longer
represents an estimated state, but rather an estimated state tracking
error. We use the output error as a control input in the next section, on
loopshaping via loop transfer recovery.
19.7 Proofs of the Intermediate Results

19.7.1 Proof that E(eT W e) = trace(ΣW )
  
n n
E(eT W e) = E  ei  Wij ej 
X X
i=1 j=1
n
XXn
= Σij Wji ,
i=1 j=1
19.7 Proofs of the Intermediate Results 135
the transpose of W being valid since it is symmetric. Now consider the diagonal
elements of the product ΣW :
 
Σ11 W11 + Σ12 W21 + · · · · ·
ΣW = 

· Σ21 W12 + Σ22 W22 + · · · ·  →
· · ···
n X
X n
trace(ΣW ) = Σij Wji . 2
i=1 j=1
∂
19.7.2 Proof that ∂H
trace(−ΛHCΣ) = −ΛT ΣC T
 
n l
Hjk Bkl  , the il0 th element
X X
trace(AHB) = trace  Aij
j=1 k=1
n X
X n l
X
= Aij Hjk Bki ,
i=1 j=1 k=1
where the second form is a sum over i of the ii’th elements. Now
n
∂ X
trace(AHB) = Aijo Bko i
∂Hjo ko i=1
= (BA)ko jo
= (BA)Tjo ko −→
∂
trace(AHB) = (BA)T
∂H
= AT B T . 2
∂
19.7.3 Proof that ∂H
trace(−ΛΣC T H T ) = −ΛΣC T
 
n
trace(AH T ) = trace  Aij HjlT  , the il0 th element
X
j=1
 
n
X
= trace  Aij Hlj 
j=1
n X
X n
= Aij Hij ,
i=1 j=1
where the last form is a sum over the ii’th elements. It follows that
∂
trace(AH T ) = Aio jo −→
∂Hio jo
∂
trace(AH T ) = A. 2
∂H
19.7.4 Proof of the Separation Principle

Without external inputs W1 and W2 , the closed-loop system evolves according
to
( ) " #( )
d x A −BK x
= .
dt x̂ HC A − BK − HC x̂
Using the definition of estimation error e = x − x̂, we can write the above in
another form:
( ) " #( )
d x A − BK BK x
= .
dt e 0 A − HC e
If A0 represents this compound A-matrix, then its eigenvalues are the roots of
det(sI − A0 ) = 0. However, the determinant of an upper triangular block
matrix is the product of the determinants of each block on the diagonal:
det(sI −A0 ) = det(sI −(A−BK))det(sI −(A−HC)), and hence the separation
of eigenvalues follows.
137
20 LOOP TRANSFER RECOVERY

20.1 Introduction
The Linear Quadratic Regulator(LQR) and Kalman Filter (KF) provide prac-
tical solutions to the full-state feedback and state estimation problems, respec-
tively. If the sensor noise and disturbance properties of the plant are indeed
well-known, then an LQG design approach, that is, combining the LQR and
KF into an output feedback compensator, may yield good results. The LQR
tuning matrices Q and R would be picked heuristically to give a reasonable
closed-loop response.
There are two reasons to avoid this kind of direct LQG design procedure,
however. First, although the LQR and KF each possess good robustness
properties, there do exist plants for which there is no robustness guarantee
for an LQG compensator. Even if one could steer clear of such pathological
cases, a second problem is that this design technique has no clear equivalent
in frequency space. It cannot be directly mapped to the intuitive ideas of
loopshaping and the Nyquist plot, which are at the root of feedback control.
We now reconsider just the feedback loop of the Kalman filter. The KF has
open-loop transfer function L(s) = Cφ(s)H, where φ(s) = (sI − A)−1 . This
follows from the estimator evolution equation
x̂˙ = Ax̂ + Bu + H(y − C x̂)
and the figure. Note that we have not included the factor Bu as part of the
figure, since it does not affect the error dynamics of the filter.
As noted previously, the KF loop has good robustness properties, specifically
to perturbations at the output ŷ, and further is amenable to output tracking.
In short, the KF loop is an ideal candidate for a loopshaping design. Supposing
that we have an estimator gain H which creates an attractive loop function
L(s), we would like to find the compensator C(s) that establishes
P (s)C(s) ≈ Cφ(s)H, or (215)

Cφ(s)BC(s) ≈ Cφ(s)H.
It will turn out that the LQR can be set up so that the an LQG-type com-
pensator achieves exactly this result. The procedure is termed Loop Transfer
138 20 LOOP TRANSFER RECOVERY
y ^x
H φ (s) C
Recovery (LTR), and has two main parts. First, one carries out a KF design for
H, so that the Kalman filter loop itself has good performance and robustness
properties. In this regard, the KF loop has sensitivity function S(s) = (I +
Cφ(s)H)−1 and complementary sensitivity T (s) = (I + Cφ(s)H)−1 Cφ(s)H.
The condition σ(W1 (s)S(s)) + σ(W2 (s)T (s)) < 1 is sufficient for robust per-
formance with multiplicative plant uncertainty at the output. Secondly, we
pick suitable parameters of the LQR design, so that the LQG compensator
satisfies the approximation of Equation 215.
LTR is useful as a SISO control technique, but has a much larger role in
multivariable control.
20.2 A Special Property of the LQR Solution

Letting Q = C T C and R = ρI, where I is the identity matrix, we will show
(roughly) that
√
lim ( ρK) = W C,
ρ→0
where K is the LQR gain matrix, and W is an orthonormal matrix, for which
W T W = I.
First recall the gain and Riccati equations for the LQR:
K = R−1 B T P
0 = Q + P A + AT P − P BR−1 B T P.
Now Q = C T C = C T W T W C = (W C)T W C. The Riccati equation becomes
0 = ρ(W C)T W C + ρP A + ρAT P − P BB T P = 0.
In the limit as ρ → 0, it must be the case that P → 0 also, and so in this limit
20.3 The Loop Transfer Recovery Result 139
ρ(W C)T W C ≈ P BB T P
= (B T P )T B T P
= (R−1 B T P )T RR(R−1 B T P )
= ρ2 K T K −→
√
WC ≈ ρK. 2
Note that another orthonormal matrix W 0 could be used in separating K T

from K in the last line. This matrix may be absorbed into W through a
matrix inverse, however, and so does not need to be written. The result of the
last line establishes that the plant must be square: the number of inputs (i.e.,
rows of K) is equal to the number of outputs (i.e., rows of C).
Finally, we note that the above property is true only for LQR designs with
minimum-phase plants, i.e., those with only stable zeros (Kwakernaak and
Sivan).
20.3 The Loop Transfer Recovery Result

√
The theorem is stated as: If limρ→0 ( ρK) = W C (the above result), with W
an orthonormal matrix, then the limiting LQG controller C(s) satisfies
lim P (s)C(s) = Cφ(s)H.

ρ→0
The LTR method is limited by two conditions:
• The plant has an equal number of inputs and outputs.
• The design plant has no unstable zeros. The LTR method can be in fact
be applied in the presence of unstable plant zeros, but the recovery is
not to the Kalman filter loop transfer function. Instead, the recovered
function will exhibit reasonable limitations inherent to unstable zeros.
See Athans for more details and references on this topic.
The proof of the LTR result depends on some easy lemmas, given at the end
of this section. First, we develop C(s), with the definitions φ(s) = (sI − A)−1
and X(s) = (φ−1 (s) + HC)−1 = (sI − A + HC)−1 .
C(s) = K(sI − A + BK + HC)−1 H

= K(X −1 (s) + BK)−1 H, then use Lemma 2 →
= K(X(s) − X(s)B(I + KX(s)B)−1 KX(s))H
= KX(s)H − KX(s)B(I + KX(s)B)−1 KX(s)H
= (I − KX(s)B(I + KX(s)B)−1 )KX(s)H, then use Lemma 3 →
= (I + KX(s)B)−1 KX(s)H
√ √ √
= ( ρI + ρKX(s)B)−1 ρKX(s)H.
Next we invoke the result from the LQR design, with ρ → 0, to eliminate
√
ρK:
lim C(s) = (W CX(s)B)−1 W CX(s)H

ρ→0
= (CX(s)B)−1 CX(s)H.
In the last expression, we used the assumption that W is square and invertible,
both properties of orthonormal matrices. Now we look at the product CX(s):
CX(s) = C(SI − A + HC)−1

= C(φ−1 (s) + HC)−1 , then use Lemma 2 →
= C(φ(s) − φ(s)H(I + Cφ(s)H)−1 Cφ(s))
= (I − Cφ(s)H(I + Cφ(s)H)−1 )Cφ(s), then use Lemma 3 →
= (I + Cφ(s)H)−1 Cφ(s).
This result, reintroduced into the limiting compensator, gives
lim C(s) = ((I + Cφ(s)H)−1 Cφ(s)B)−1 (I + Cφ(s)H)−1 Cφ(s)H

ρ→0
= (Cφ(s)B)−1 Cφ(s)H
= P −1 (s)Cφ(s)H.
Finally it follows that limρ→0 P (s)C(s) = Cφ(s)H, as desired.

20.4 Usage of the Loop Transfer Recovery 141
20.4 Usage of the Loop Transfer Recovery

The idea of LTR is to “recover” a Kalman filter loop transfer function L(s) =
Cφ(s)H, by using the limiting cheap-control LQR design, with Q = C T C and
R = ρI. The LQR design step is thus trivial.
Some specific techniques are useful.
• Scale the plant outputs (and references), so that one unit of error in one
channel is as undesirable as one unit of error in another channel. For
example, in depth and pitch control of a large submarine, one meter of
depth error cannot be compared directly with one radian of pitch error.
• Scale the plant inputs in the same way. One Newton of propeller thrust
cannot be compared with one radian of rudder angle.
• Design for crossover frequency. The bandwidth of the controller is

roughly equal to the frequency at which the (recovered) loop transfer
function crosses over 0dB. Often, the bandwidth of is a more intuitive
design parameter than is, for example, the high-frequency multiplicative
weighting W2 . Quantitative uncertainty models are usually at the cost
of a lengthy identification effort.
• Integrators should be part of the KF loop transfer function, if no steady-

state error is to be allowed. Since the Kalman filter loop has only as many
poles as the plant, the plant input channels must be augmented with the
necessary additional poles (at the origin). Then, once the KF design is
completed, and the compensator C(s) is constructed, the integrators are
moved from the plant over to the input side of the compensator. The
tracking errors will accrue as desired.
20.5 Three Lemmas

Lemma 1: Matrix Inversion
(A + BCD)−1 = A−1 − A−1 B(C −1 + DA−1 B)−1 DA−1 .
Proof:
(A + BCD)(A + BCD)−1 = I
A(A + BCD)−1 = I − BCD(A + BCD)−1
(A + BCD)−1 = A−1 − A−1 BCD(A + BCD)−1
= A−1 − A−1 BCD(I + A−1 BCD)−1 A−1
= A−1 − A−1 B(D−1 C −1 + A−1 B)−1 A−1
= A−1 − A−1 B(C −1 + DA−1 B)−1 DA−1 . 2
Lemma 2: Short Form of Lemma 1
(X −1 + BD)−1 = X − XB(I + DXB)−1 DX
Proof: substitute A = X −1 and C = I into Lemma 1.

Lemma 3
I − A(I + A)−1 = (I + A)−1
Proof:
I − A(I + A)−1 = (I + A)(I + A)−1 − A(I + A)−1

= (I + A − A)(I + A)−1
= (I + A)−1 . 2
143
21 SYSTEM IDENTIFICATION
21.1 Introduction
Model-based controller design techniques, such as the LQR and LTR, require
a plant model. The process of generating workable models from observed
data is the goal of system identification. Good controllers usually have some
reasonable robustness guarantees, which motivates identification with simple
methods. We discuss in this section four fundamental but useful techniques
for approximate system identification of single-input, single-output plants. It
should be noted that the area of system identification is a very rich one, and
that the methods are only a small subset of what is available.
Except for the last approach, time-domain simulation, the methods are limited
to linear models. If different inputs give different linear model coefficients, then
it is likely that nonlinear terms are playing a role. The user then has the choice
of ignoring the nonlinearity, for example if the operating point is controlled
very closely, or developing a controller which takes specific account of the it.
In any event, simulations with the nonlinear plant should always be performed
to assess the robustness of the control strategy.
21.2 Visual Output from a Simple Input

For low-order plants which can tolerate impulse or step input, a great amount
can be learned through step and impulse responses. The basic idea is to
express what is observed as a time signal whose Laplace-domain equivalent
can be recognized. As an example, consider the plant transfer function P (s) =
k/(τ s + 1), a first-order lag with gain k, and time constant τ . We have y(s) =
k/s(τ s + 1) for the step input u(s) = 1/s, and therefore y(t) = k(1 − e−t/τ ).
The gain k is evident as the maximum value taken by the measured output.
The time constant τ is equal to the time required for y(t) to reach the value
k(1 − e−1 ) = 0.632k. Similar estimates can be made for second-order systems,
especially with the help of step functions parametrized on damping ratio ζ.
Systems with order three or higher will usually be more difficult to assess with
this visual technique.
Example: Two raw step responses in heading were recorded for two dif-
ferent vessels. The first vessel was strongly unstable and had to be powered
down abruptly after one second; however, the smoothed data can still be fit-
ted to the curve y(t) = 0.58(et − 1). Give estimates of the two plant transfer
144 21 SYSTEM IDENTIFICATION
1.5 1
0.8
1
0.6
0.5
0.4
0
0.2
−0.5 0
0 0.2 0.4 0.6 0.8 1 0 1 2 3 4 5
seconds seconds
functions P (s), assuming that the step input was of magnitude one.
The time trace of the first system looks like the Laplace transform pair:
1 1
y(t) = (e−at − e−bt ) ↔ y(s) =
b−a (s + a)(s + b)
for the values b = 0 and a = −1. Thus for the experimental data,
0.58
y(s) = .
s(s − 1)
Since y(s) = P̂ (s)u(s), where u(s) = 1/s, we have P̂ (s) = 0.58/(s − 1). The
second trace looks like a second-order response of the form
y(s) kωn2
= 2 ,
u(s) s + 2ζωn s + ωn2
where gain k, undamped frequency ωn , and ζ are to be determined. First,

we note that the steady value of y(t) is about 0.5, so let k = 0.5. Next,
with respect to the steady value, the first overshoot is about 0.3, and the
second overshoot (same side) is about 0.1. The ratio is often written as the
logarithmic decrement δ = ln(0.3/0.1) ' 1.10, so that the damping ratio is
simply ζ = δ/2π ' 0.175. Finally, the damped natural period is about 1.7s,
leading to the damped natural frequency ωd = 2π/1.7 ' 3.70rad/s. The
undamped natural frequency ωn is related to ωd as follows:
21.3 Transfer Function Estimation – Sinusoidal Input 145
ωd
ωn = √ ' 3.75rad/s.
1 − ζ2
Inserting these into the template above, we have
7.0
P̂ (s) = .
s2 + 1.28s + 14.1
21.3 Transfer Function Estimation – Sinusoidal Input

The main idea of transfer function estimation is that P (s) can be estimated by
simply dividing the measured output by the measured input, in the frequency
domain:
y(s)
P̂ (s) = . (216)
u(s)
In applications, the input signal u(s) is known quite accurately because it is
generated by a computer. Two sources of error can corrupt the output signal
y(s), however: real disturbances and sensor noise. In some cases, disturbances
may be band-limited (e.g., water waves), and if these occur in a frequency
range far away from the dominant dynamics, the estimated transfer function
approach will succeed. A similar argument holds for sensor noise, which in
many devices is negligible for low frequencies. A sensor which has noise in the
frequency range of the system dynamics is problematic for obvious reasons.
This section discusses the use of periodic inputs to create a Bode plot, while
the next section generalizes to broadband input. Bode plots are figures of
transfer function magnitude and phase as functions of frequency, which can
be either parameterized in terms of poles, zeros, and a gain, or used directly
in a loopshaping or Nyquist plot approach.
Certain plants which cannot admit a step or impulsive input will tolerate a
sinusoidal input. The idea then is to drive the plant with a narrow-band, i.e.,
periodic, input signal u(s). For each such test at a specific frequency, compute
the magnitude and phase relating the input to the output. Conduct as many
tests as are necessary to build a Bode plot of the plant estimate P̂ (s).
Example: Sinusoidal rudder angles trajectories, of 10deg amplitude (|u(t)| =
10deg), were implemented on a vessel operating near its cruising speed. The
data are plotted, along with the magnitude and phase of the transfer function
0.093
P̂ (s) = ,
s2 + 0.83s + 0.21
which holds reasonably well at low frequencies. The phase angle of y(s) is taken
with respect to the input signal u(s). Note that the experimental magnitude
and phase in this example deteriorate at the higher frequencies. This is a
property of almost all physical systems, and an indicator that the plant model
cannot be trusted above a certain frequency range.
−10
|y(s)/u(s)|, dB
−20
−30
−40
0
angle(y(s)), deg
−50
−100
−150
−200
−3 −2 −1 0 1 2
10 10 10 10 10 10
rad/s
21.4 Transfer Function Estimation – Broadband Input

Many times one needs to deal with experimental data that is broadband, either
as the result of a closed-loop run, or of significant disturbances. In this case,
frequency-domain analysis can still be used, but elements of spectral analysis
are necessary.
21.4 Transfer Function Estimation – Broadband Input 147
21.4.1 Fourier Transform of Sampled Data

For the purpose of analyzing transfer functions, the discrete Fourier transform
(DFT) is a standard, optimized tool. It operates on a data vector x(n) of
length N :
N −1
x(n)e−j2πnm/N ,
X
X(m) = (217)
n=0
for m = [1, N ]. We review several points for dealing with the DFT calculation:
• The m’th DFT point is a summation of complex unit-magnitude vectors

(e−j2πnm/N ) times the original data (x(n)). Dividing the DFT result by
N returns the signal to its real magnitude.
• The DFT generates a vector of N complex points as outputs. These

correspond to the frequency range
2π m − 1
ωm = , (218)
dt N
i.e., the first frequency is 0, and the last frequency is slightly less than
the sampling rate.
• The sampling theorem limits our usable frequency range to only one-half
of the sampling rate, called the Nyquist rate. The points higher than
π/dt correspond to negative frequencies, and a complex-conjugacy holds:
X(2) = X ∗ (m), X(3) = X ∗ (m − 1), and so on. What happens near the
Nyquist rate depends on whether N is odd or even.
An additional multiplication of the DFT result by 2 will give peaks which
are of about the right magnitude to be compared with the time-domain
signal. With this scaling, the value X(1) is twice the true DC value.
• The Nyquist rate depends only on the sampling time step dt, but the
frequency vector accompanying the DFT can be made arbitrarily long
by increasing the number of data points. To improve the frequency
resolution of the DFT, a common approach is to zero-pad the end of the
real data,
• The data x(k) should be multiplied by a smoothing window w(k), for

two reasons. First, if the window goes to a small value near its tails,
Gibb’s effect (the spectral signature of a discontinuity between x(1) and
x(N )) can be minimized. The second rationale for windowing is based
on the discrete nature of the transform. Because the DFT provides
frequency data at only N/2 unique frequencies, there is the possibility
that a component in the real data lies in between one of these DFT
frequencies. The DFT magnitude at a specific ω(m) is in fact an average
of continuous frequencies from the neighborhood of the point. Smoothing
windows are generally chosen to achieve a tradeoff of two conflicting
properties relating to the average: the width of the primary lobe (wide
primary lobes bring in frequencies that belong to the adjacent bins ω(m−
1) and ω(m + 1)), and the magnitude of the side lobes (tall side lobes
bring in frequencies that are far away from ω(m)).
There are many smoothing windows to choose from; no windowing at
all is usually referred to as applying a rectangular window. One of the
simplest is the Hann, or cosine, window:
" !#
1 2πk
w(k) = 1 − cos
2 N
• The DFT of a given signal is subject to bias and variance. These can
be reduced by taking separate DFT’s of sub-sections of data (perhaps
zero-padded to maintain frequency resolution), and then averaging them.
It is common to use segments which do not overlap, but it is not a
necessity. This averaging idea is especially important in transfer function
verification; see below.
• If possible, the DFT should be run on a number of samples N which is
a multiple of a large power of two. If N is prime, the DFT will take a
long time to execute.
The above steps, all of which are available through the Matlab function
spectrum() for example, will ensure a fair spectral analysis of a time series.
21.4.2 Estimating the Transfer Function

In continuous time, when the plant input y(t) and output u(t) are transformed
into frequency space, the estimated transfer function follows from
21.4 Transfer Function Estimation – Broadband Input 149
y(s)
P̂ (s) = . (219)
u(s)
As noted above, spectral analysis of a signal benefits by using multiple seg-
ments and averaging; we now want to include the same approach in our esti-
mation of P (s). The procedure is:
• For the p’th segment of data, compute the transfer function estimate
Yp (m)Yp∗ (m)
P̂p (m) = .
Up (m)Up∗ (m)
• For the p’th segment of data, compute two covariances and a cross-
covariance:
Γpuu (m) = Up (m)Up∗ (m)

Γpyy (m) = Yp (m)Yp∗ (m)
Γpyu (m) = Yp (m)Up∗ (m).
• Construct average values of the transfer function estimate and the other
quantities:
P̂ (m) = avgp (P̂p (m))

Γuu (m) = avgp (Γpuu (m))
Γyy (m) = avgp (Γpyy (m))
Γyu (m) = avgp (Γpyu (m))
• Compute the coherence function, which assesses the quality of the final
estimate P̂ (m):
Γyu (m)Γ∗yu (m)

Coh(m) = .
Γyy (m)Γuu (m)
If the coherence is near zero, then the segmental cross-covariances Γpyu (m)
are sporadic and have cancelled out; there is no clear relation between
the input and the output. This result could be caused by either dis-
turbances or sensor noise, both of which are reasonably assumed to be
random processes, and uncoupled to the input signal. Alternatively, if
the coherence is near one, then the cross-covariances are in agreement
and a real input-output relationship exists. With real data, the coher-
ence will deteriorate at high frequencies and also at any frequency where
disturbances or noise occur.
21.5 Time-Domain Simulation

The time-domain simulation approach tweaks the parameters of a simulation
so that its output matches the observed output. The method has its main
strength in the fact that it applies to any model that can be simulated, in-
cluding those of high order and with significant nonlinearities. On the other
hand, the method is computationally expensive and gives no guarantee of a
useful solution, or even of convergence.
At the outset, we need to come up with some structure of the plant model.
This can be based on physics in many cases. Consider, for example, the case
of a mass mounted on a spring and a dashpot, driven by the input force u(t).
A fair guess for the actual dynamics has the form my 00 + by 0 + ky = u(t), and
we plan to look through the three-dimensional parameter space θ~ = [m, b, k].
The simulation operation can be written this way: ŷ(t) = G(θ, ~ u(t)), and the
system identification problem is to minimize ||ŷ(t) − yobs (t)||, say, where || · ||
here indicates the Euclidean norm. For a given parameter vector θ, ~ running the
simulation generates a new ŷ, and computing the norm gives a scalar measure
of goodness.
Since the normed error is a complicated function of both u(t) and θ, ~ the mini-
mization must proceed iteratively. The Nelder-Meade simplex method is easy
to use, and can be invoked with the Matlab function fmins. As an example,
the three programs listed comprise a working Matlab set for identification of
a first-order, nonlinear system. Some notes on use:
• In this example, the same simulation generates the “observed” data and
the simulated response. Since the program simulate always uses the
global variable theta as the parameters, we must be careful about setting
theta in the calling programs.
21.5 Time-Domain Simulation 151
• After a simulation run is complete, the data is interpolated to the same

time scale as the observed data, in order to compute the error.
• The initial guess for theta is a random vector; the Simplex method will
take over from this point. In many instances, however, theta is roughly
known, and a better starting value can be given.
• The Simplex method may head into invalid parameter space, e.g., nega-
tive mass. The error calculation, however, can be easily augmented by
a term which penalizes invalid parameters, e.g.,
err = err + 1000*(1-sign(mass)).
• There is no guarantee that a global minimum will be found or even exists.

Starting from different initial guesses for θ may help find better results,
but we are still at the mercy of the minimization algorithm, and a very
complicated function.
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
clear all ; clear global ;

global u_obs t_obs y_obs dt theta ;
dt = .3 ; % time step
theta = [1 2] ; % true parameter vector
t_obs = 0:dt:20*dt ; % observed time vector
u_obs = ones(length(t_obs),1) ; % observed input
[t_raw, y_raw] = ode45(’simulate’, [0 max(t_obs)], 0) ;
y_obs = spline(t_raw, y_raw, t_obs) ; % observed output
[theta_final] = fmins(’get_err’, randn(2,1)) ;
disp(sprintf(’final theta(1): %g.’, theta_final(1))) ;

disp(sprintf(’final theta(2): %g.’, theta_final(2))) ;
theta = theta_final ;
y_sim = spline(t_raw, y_raw, t_obs) ;
figure(1) ; clf ; hold off ;
plot(t_obs, y_obs, t_obs, y_sim, t_obs, u_obs) ;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [err] = get_err(theta_arg) ;

global y_obs t_obs theta ;
theta = theta_arg ;
y_sim = spline(t_raw, y_raw, t_obs) ;
err = norm(y_sim - y_obs) ;
disp(sprintf(’error: %f.’, err)) ;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function [ydot] = simulate(t,y) ;
global u_obs dt theta ;
k = theta(1) ;
tau = theta(2) ;
ind = floor(t/dt) + 1 ; % we have to choose which u_obs to use:

% a zero-order hold as implemented.
ydot = ( k*u_obs(ind) - y^3 ) / tau ;
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
153
22 CARTESIAN NAVIGATION
The bulk of our discussion on maneuvering and control has assumed that the
necessary system states can be measured. The marine engineer is in fact faced
with choices between many different basic sensor packages, notably compasses,
paddle wheels, inertial navigation units, rate gyros, and depth guages, for ex-
ample. These listed sensors are self-contained and rely primarily on the phys-
ical properties of the natural environment. There is also a class of distributed
sensor systems; these generally involve an array of communicating elements, lo-
cated remotely from the vehicle. We present the fundamental concepts behind
two methodologies in this second class: the global positioning system (GPS)
and acoustic navigation, both of which can provide high-accuracy absolute
Cartesian navigation.
22.1 Acoustic Navigation

Consider a transponder A, which can transmit an acoustic signal, and also
measure, with microsecond accuracy, the time to receive a reply. Next, place
a responder B at a distance R away from A; the job of B is just to transmit
a signal whenever it receives one, with a (short) predictable response time Tt .
Thus, the elapsed time T between a tranmission and consequent reception at
A is
T = Tt + 2R/cw ,
where cw is the speed of sound in water, about 1450m/s. The range R follows
by inversion:
(T − Tt )c2
R= .
2
Suppose that the location of B is known; then a given measurement of R
places A on a sphere around B. It is a case of three unknowns (x, y, z) and one
equation:
(xA − xB )2 + (yA − yB )2 + (zA − zB )2 = R2 .

154 22 CARTESIAN NAVIGATION
The introduction of a another responder C (typically listening and responding

at a different frequency than B) places A on the intersection of two spheres,
i.e., a circle. There are two equations, but still three unknowns. When A,
B, and C lie in a nearly horizontal plane, then the intersection circle lies in a
vertical plane; the addition of a depth sensor to our suite would allow us to pin
A’s location at one of two points on the circle. Finally, if we know which side A
is on, and do not allow for abrupt crossovers, then we have a functional set of
measurements for acoustic navigation with just two responders. The baseline is
the line connecting responders B and C; when A is near this baseline, positional
accuracy will be very poor since the two solution spheres are tangent.
Better and better performance can be obtained by increasing the number of
responders, and consequently of the baselines and spheres. With three re-
sponders, for example, the intersection of a sphere (responder D) and a circle
(responders B and C) is two points. Here there are three equations and three
unknowns, but the nonlinearity of the equations leads to the non-uniqueness
in the solution. A fourth responder or a depth transducer would be needed to
completely constrain the solution.
The above discussion is a minimum conceptual explanation of acoustic naviga-
tion. There are many other pieces to the approach, including an account of the
variation of sound speed cw with water depth, obtaining the Cartesian loca-
tions of the responder network, and handling various geometric configurations
that give rise to poor or degenerate solutions.
There are two common configurations used for acoustic navigation, named for
the length of the baselines relative to the target (transponder A) range.
(Ultra) Short-Baseline The geometry of SBL or USBL puts very short

baselines between the responders compared to the target distance. For in-
stance, the fixed net is often attached to a vessel or other structure, with base-
line lengths on the order of 10-100m. Typical frequencies in use are around
100kHz, with a working range of 100-500m to the target. The wavelength of a
100kHz signal is about 1cm and 5cm is a reasonable estimate of the accuracy
for these systems.
Long-Baseline Long-baseline systems typically involve a larger responder

net, and the target ranges are similar to the baseline lengths. Very large
systems utilize frequencies of 10-15kHz (for a placement accuracy around 2-
5m), and may have ten-kilometer baselines. Frequent sources of error in long-
baseline systems are variations of cw , and also in multipath. In the latter
22.2 Global Positioning System (GPS) 155
B C
A3
A2
B C
A1
A*2
A*3
Figure 6: Left: General configuration of a long-baseline acoustic system, with

target transponder A and fixed responders B and C. The position A1 is on
the baseline and has poor accuracy in the direction normal to the baseline; in
contrast, A3 is well-posed. A∗3 is an additional solution to the two-responder
problem. Right: General configuration of a short-baseline system.
condition, false signals can be caused by reflections off the seafloor and the
surface; these signals are typically eliminated by rejecting those receptions
which are outside a very tight and slowly moving window on travel time T .
Sometimes, bottom topography can shield a direct acoustic path, and the only
receptions available are via multipath!
22.2 Global Positioning System (GPS)

The GPS system is the most powerful system publicly available for absolute
position reference above water. It is similar in concept to the acoustic naviga-
tion systems described above, but with one fundamental difference: only the
one-way travel time from each satellite is measured at the target. The accu-
racies achieved with GPS are therefore strongly dependent on the accuracy of
time-keeping.
The core of the system is a network of 24 satellites arranged in six planes (4

per plane) inclined at 55 degrees from the equatorial plane, at an altitude of
about 20,000km. The altitude is chosen carefully to correspond with a 12-hour
orbit; geosynchronous orbit is much higher, at around 40,000km. The orbit
lines over a non-rotating earth are thus near-sinusoids that reach lattitudes
of 55◦ N and 55◦ S, and the six lines are spaced 60◦ apart in longitude. With
reference to a rotating Earth, the spatial frequency is doubled, so that the
longitudinal distance between tracks is 30 degrees, or 1800 nautical miles.
Each satellite transmits a regular signal which contains, among other items, the
time of its transmission, and the three-dimensional location of the satellite (the
ephemeris). It is from these two pieces of information, from many satellites,
that triangulation and navigation are performed at the target.
For the ideal case that the time bases of the target and the satellites are
exactly aligned, triangulation of the type described for acoustic navigation is
possible. In practical terms, the speed of light c = 3 × 108 m/s implies that
a 1ns timing error will cause a 0.3m range error. For this reason, each of
the satellites carries four atomic clocks on board, good to 1ns per day. The
time base is continually monitored and updated from a ground station, and
accurate to within 1ns. The ephemeris of each satellite is also monitored and
updated from the ground station, using a combination of least-squares analysis
of past data (1 week), and a Kalman filter to predict the future ephemeris.
The accuracy of satellite position is better than 10m.
We now come to the last thorny issue: how to make target receivers that don’t
require atomic clocks! The solution is very clever, and illustrated for the case
of two-dimensions; the arguments for three dimensions are the same. In two
dimensions, the range measurements from two sources locate the target on one
of two points (the intersection of two circles). Suppose that the target clock
is too slow by an amount t̃, so that the range estimate is
r + r̃ = c(t + t̃).
The estimated range circles are too large, and the estimated location of the
target too far from the baseline. Introduce a third satellite now so that the
intersections of the circle pairs now occur at six points. Three of these are
very close, and indicate the approximate true solution. The trick is to find the
correction for t̃ that puts the three close solutions onto a single point. This
single point is near the centroid of the three approximate points, and represents
the best position solution. The target time base can be kept up to date by
22.2 Global Positioning System (GPS) 157
performing this check for every set of signals. Hence, the two-dimensional
timing problem is solved with three satellites; the three-dimensional timing
problem is solved with four satellites.
The position specification for GPS is 25m, at the 95’th percentile. This is a re-
markable feat, given that it is an absolute measure over the entire surface and
atmosphere of our 6000km-radius Earth. Major sources of error include: clock
base and satellite navigation (≈ 2.5m), ionosphere and troposphere electro-
magnetic variations (≈ 2.5m), and inaccuracies in the receiver and multipath
(≈ 2m). GPS is also subject to selective availability or S/A, the addition of
a slow-varying random component in the satellite ephemeris data. Selective
availability is controlled by the United States Department of Defense, and de-
grades the position specification to 100m (typical). The advent of differential
GPS solved most of the S/A concerns of American allies, by providing high-
accuracy navigation in local areas. The idea here is that a stationary target
can detect the effects of S/A (since it is not moving) and then transmit the
corrections over a small geographical area. Many current GPS receivers are
capable of decoding these local corrections, which are then applied to their
own satellite navigation processing. Differential GPS typically provides 1-2m
accuracy.
Finally, as with the acoustic navigation systems described, an independent
altitude measurement will enhance the accuracy of GPS, essentially reducing
a three-dimensional to a two-dimensional problem.
159
23 REFERENCES
Short notes indicate how the reference is related to the text.
Dynamics of Mechanical and Electromechanical Systems, S.H. Crandall, D.C.

Karnopp, E.F. Kurtz, Jr., and D.C. Pridmore-Brown, Krieger: Malabar, FL,
1985. Coordinate systems and rigid-body dynamics.
Optimal Spacecraft Rotational Maneuvers, J.L. Junkins and J.D. Turner, El-
sevier: New York, 1986. Euler angles and quaternions.
Principles of Naval Architecture, 2nd Revision, E.V. Lewis, ed., Society of

Naval Architects and Marine Engineers: Jersey City, NJ, 1988. Classical treat-
ment of stability, propulsion, and lifting surfaces.
Marine Hydrodynamics, J.N. Newman, MIT Press: Cambridge, 1997. General

reference.
Guidance and Control of Ocean Vehicles, T.I. Fossen, Wiley: New York, 1994.
Comprehensive modern treatment of marine dynamics and control.
Formulas for Natural Frequency and Mode Shape, R.D. Blevins, Krieger: Mal-
abar, FL, 1984. Extensive added mass formulas.
A method for estimating static aerodynamic characteristics for slender bodies

of circular and noncircular cross section alone and with lifting surfaces at
angles of attack from 0◦ to 90◦ , L. Jorgensen, NASA Technical Note D-7228,
April 1973.
Fluid Dynamic Lift, S.F. Hoerner and H.V. Borst, Hoerner Fluid Dynamics:
Bakersfield, CA, 1975. Extensive data on lift of bodies and wings.
Fluid Dynamic Drag, S.F. Hoerner and H.V. Borst, Hoerner Fluid Dynamics:
Bakersfield, CA, 1975. Data on drag of bodies and wings.
Marine Propellers and Propulsion, J.S. Carleton, Butterworth-Heinemann:

Boston, 1994. Includes a chapter on transverse and azimuthing thrusters.
The influence of thruster dynamics on underwater vehicle behavior and their

160 23 REFERENCES
incorporation into control system design, D. R. Yoerger, J.G. Cooke, and J.-
J. E. Slotine, IEEE J. Oceanic Engineering, 15:167-178, 1990. First-order
nonlinear thruster model.
Toward an improved understanding of thruster dynamics for underwater ve-

hicles, A.J. Healey, S.M. Rock, S. Cody, D. Miles, and J.P. Brown, Proc.
IEEE Symposium on Autonomous Underwater Vehicle Technology, pp 340-
352. 1994. Second-order nonlinear thruster model.
Accurate four-quadrant nonlinear dynamical model for marine thrusters: The-

ory and experimental validation, R. Bachmayer, L.L. Whitcomb, and M.A.
Grosenbaugh, IEEE J. Oceanic Engineering, 25:146-159, 2000. State-of-the-
art in thruster modeling.
Acceleration and steady-state propulsion dynamics of a gas turbine ship with

controllable-pitch propeller, C.J. Rubis, Proc. SNAME Annual Meeting, No.
10, 1972.
Governing ship propulsion gas turbine engines, C.J. Rubis and T.R. Harper,
SNAME Trans., 94:283-308, 1986.
Flow-Induced Vibrations, 2nd Edition, R.D. Blevins, Van Nostrand Reinhold,

New York, 1990. Cable strumming.
Modern Control Engineering, K. Ogata, Prentice-Hall: Englewood Cliffs, NJ,

1970. SISO linear control.
Circuits, Signals, and Systems, W.M. Siebert, MIT Press: Cambridge, MA,
1986. General reference on transforms.
Applied Nonlinear Control, J.J.-E. Slotine and W. Li, Prentice-Hall: Engle-

wood Cliffs, NJ, 1991.
Feedback Control Theory, J.C. Doyle, B.A. Francis, and A.R. Tannenbaum,
Macmillan: New York, 1992. Loopshaping concepts.
Control Engineering: A Modern Approach, P.R. Bélanger, Sauders College

Publishing, Philadelphia, 1995. General control reference.
Introduction to Dynamic Systems, D.G. Luenberger, Wiley: New York, 1979.

161
Dynamic programming and LQR derivation.
Applied Optimal Estimation, A. Gelb, ed., MIT Press: Cambridge, 1988.

Kalman filter practical aspects.
Linear Optimal Control Systems, H. Kwakernaak and R. Sivan, Wiley: New

York, 1972. Kalman filter derivation and properties.
Multivariable feedback design: concepts for a classical/modern synthesis, J.C.

Doyle and G. Stein, IEEE Trans. Automatic Control, 26:4-16, 1981. Seminal
paper on LTR.
A tutorial on the LQG/LTR method, M. Athans, Proc. American Control

Conference, pp 1289-1296, 1986.
Multivariable control of a submersible using the LQG/LTR design methodol-

ogy, R. Martin, L. Valavani, and M. Athans, Proc. American Control Confer-
ence, pp 1313-1324, 1986. More tutorial LTR.
Optimal Control, F.L. Lewis and V.L. Syrmos, Wiley: New York, 1995. Gen-
eral reference on LTR.
Numerical Recipes in C, 2nd Edition, W.H. Press, S.A. Teukolsky, W.T. Vet-
terling, and B.P. Flannery, Cambridge University Press: Cambridge, UK,
1992. Function minimization.
Spectral Analysis in Engineering, G.E. Hearn and A.V. Metcalfe, Halsted:

New York, 1996. Transfer functions from spectral analysis.
NAVSTAR: Global positioning system – Ten years later, B.W. Parkinson, S.W.
Gilbert, Proc. IEEE, 71:1177-1186, 1983. The whole issue is in fact dedicated
to global positioning systems. There is another such issue: IEEE Proc. Vol.
87, No. 1, 1999.
162 23 REFERENCES
163
24 PROBLEMS
Many of the following problems use data presented by Fossen and in Principles
of Naval Architecture.
1. Inertial dynamics. The gas generator turbine of a LM-2500 marine

engine rotates at p = 10, 000rpm (revolutions per minute), and has a
mass of approximately 200kg. The following question pertains to the
inertial forces experienced by this spinning rigid body, with Ixy = Ixz =
Iyz = xg = zg = 0.
Because of high rotational speed, the turbine is exceedingly sensitive to
lateral balancing. Estimate the net force that the radial bearings would
have to support for a 1mm mass imbalance, say, yg = 1mm with respect
to the spin axis.
2. Kinematics. Although dead-reckoning is not recommended as a method

for long-term navigation, it can be quite useful for pre-planning trajec-
tories. For example, consider a surface vessel heading profile as follows:
t=0 φ(t) = 0 rad

0 < t <= 60s φ(t) = t/60 × π/4 rad
60s < t < 240s φ(t) = π/4 rad
240s < t < 300s φ(t) = (300 − t)/60 × π/4 rad
t = 300s φ(t) = 0 rad
φ
y
π/4 φ
t = 0s t = 60s t = 240s t = 300s
What is the total distance traveled in the Cartesian x direction during

these five minutes? You may assume that the forward speed of the vessel
is u = 1m/s, and that there is no significant sideslip. Hint: Break the
trajectory up into three parts, which can be evaluated separately.
164 24 PROBLEMS
3. Coefficients and system modeling. Consider a weather vane in a

wind of velocity Uo . If θ is the angle of the vane with respect to the wind
direction,
(a) Write the single-degree of freedom (N ) linearized equations of mo-

tion about the fixed axis 0.
(b) Write Nθ , Nθ̇ , and Nθ̈ in terms of Nv , Nr , Nṙ , etc..
(c) If we consider the differential equation
Aÿ(t) + B ẏ(t) + Cy(t) = 0,
the condition for stability is that A, B, and C must have the same
sign. Express this requirement in terms of the derivatives in the
previous question. Give physical interpretations for what would
make such a device stable or unstable.
(d) Create a numerical model of this system, using the MATLAB ODE
solver ode45. The system equation can be written as two first-order
equations:
( ) " #( )
d θ̇ −B/A −C/A θ̇
= .
dt θ 1 0 θ
Simulate the system response to nonzero initial conditions (e.g.,

θ(0) = 1, θ̇(0) = 0). Discuss, using several examples, response
sensitivity to B and C, which are related to the aerodynamic coef-
ficients. For example, look at the range {A, B, C} = {1, ±3, ±3}.
O θ
x
Uο
165
4. Hydrodynamics. The figure below shows some characteristic fluid

force curves versus a motion parameter. Give the linear hydrodynamic
coefficient at two different operating conditions, origin O and A: is it
zero, small, finite positive, finite negative?
F F F
O O
u u u
O A A A
5. Second-order system. Consider the vibration of a mass-spring-dashpot

system:
mẍ + bẋ + kx = 0,
wherein all the coefficients are positive and real (i.e., physical).
(a) Set x(t) = est , and find two solutions for s in terms of m, b, and k,
using the quadratic formula. The solution pairs depend on whether
b exceeds a critical value: what is the critical value and how do the
solutions change?
(b) From the initial condition x(0) = xo , and ẋ(0) = 0, find the coeffi-
cients for the general solution
x(t) = c1 es1 t + c2 es2 .
in terms of xo , s1 , and s2 . What do you think will happen when

s1 = s2 ?
(c) Write this x(t) for sub-critical b (that is, complex s) in terms of the
standard parameters
166 24 PROBLEMS
s
k
ωn = (undamped natural frequency)
m
b
ζ = √ (damping ratio)
2 km
q
ωd = ωn 1 − ζ 2 (damped natural frequency).
Give your answer in terms of a real exponential multiplied by a

single sine or cosine, and make a sketch.
(d) Sketch the response x(t) for supercritical b, i.e., both s are real, for
the same initial conditions. What happens when the roots si are
far apart vs. when they are close together?
(e) Consider now the case where an input acts to drive the mass from
zero initial conditions:
mẍ + bẋ + kx = u.
This equation defines a system, with input u, and output x. Using

the Laplace transform, write the transfer function for this system,
i.e., the impulse response, both in frequency (Laplace) space, and
in the time-domain. Make a sketch of the time-domain result. Be
sure to cover both the sub-critical and super-critical damping cases.
(f) What is the step response of this system, in both the subcritical
and super-critical damping cases? Include sketches.
6. Submarine roll dynamics and control. A submarine has weight

1200t (tons) and the center of gravity is 0.5m above the center of buoy-
ancy (What can you conclude?). The rolling motion can be assumed
to be decoupled from the other motions. This submarine has anti-
rolling fins to ensure stability. The control hydrodynamic derivative is
Kδ = −2.8tm per degree of fin rotation δ, at a forward speed of 5m/s.
(a) Write the equation of motion for roll (K).

(b) If the automatic control law δ = k1 ψ is used, where ψ is the roll
angle, what range of k1 ensures stability?
167
(c) If the speed suddenly drops to 2.5m/s, how does the range of sta-
bilizing k1 change?
7. Steering. Draw curves of rudder angle δ vs. yaw rate r for a dynamically
stable surface vessel, and then for an unstable vessel. Indicate areas
where the vessel turns against the rudder action.
8. Submarine pitch/heave dynamics. Consider a submarine moving

forward at speed U , and restricted to small motions in the vertical (x−z)
plane. Assume that:
• The submarine is symmetric in the x − y plane and yG = 0. The

submarine is not symmetric in the x − z plane due to the sail, etc.,
but assume it is nearly symmetric, so that you can omit certain
hydrodynamic terms.
• At rest, the submarine is neutrally buoyant and stable with the x
and y axes horizontal. What does this tell us about the magnitude
of the buoyant force and where it acts with respect to the center of
weight?
(a) Derive the equations of motion for the submarine moving at speed
U.
(b) Derive the hydrostatic, restoring pitch moment for small pitch θ.
(c) Linearize the inertial terms in the equations of motion.
(d) Expand the fluid forces and moments in terms of the motions. Omit
nonlinear and memory effects, and be sure to include the hydrostatic
moment. Explain your choices.
(e) Write out the complete linearized equations of motion. Does surge
decouple from pitch and heave? Write out the coupled equations in
matrix form.
(f) Can the submarine be stable in pitch without feedback control?
Can the submarine be stable in depth without feedback control?
9. Submarine stability via slender body theory.
(a) An underwater vehicle hull has the shape of a circular cylinder,

except at the very ends, where it rapidly tapers down to a point. If
168 24 PROBLEMS
the radius is 0.5m, length is 17m, and speed is 2m/s, use slender-
body theory to express the following coefficients: Zẇ , Mq̇ , and Mw ,
all referenced to the body midpoint.
(b) It turns out that in order to match the lift and moment of the hull
in experiments, a flat tail of radius 0.4m has to be included in the
slender-body estimate. Determine the corrected value for Mw , and
in addition compute Zw . Where is the aerodynamic center with
respect to the center of the hull, and is this hull stable in pitch?
Recall that the aerodynamic center is the equivalent location of the
observed lift force that creates the observed moment.
(c) A fin with pre-determined lift slope of 3.6/rad is to be applied at
the tail to make the device just stable in pitch. How big is the
required fin area that brings the net Mw to zero, as desired?
10. Lifting surfaces. Using the lifting surface formulas, estimate the coef-
ficient Kv of a submarine sail as shown below, for U = 10m/s.
7m
10m
15m
11. Slender body theory. Consider a long ellipsoidal body of length l and
diameter d.
(a) With l/d = 7.0, approximate the cross-body added mass, using
slender-body theory, and compare it with the exact results from
the table below (Blevins).
(b) Perform the slender body calculation also for a sphere, and compare
again to the exact result.
(c) What is the added mass in the in-line direction?
169
Sphere added mass: 2ρπa3 /3

Ellipsoid cross-body added mass: 4αρπab2 /3
where a/b = 0.1: α = 0.075
0.2: 0.143
0.6: 0.355
1.0: 0.500
2.0: 0.704
5.0: 0.894
7.0: 0.933
10.0: 0.960
∞: 1.000
Ellipsoid longitudinal added mass: 4αρπab2 /3
where a/b = 0.1: α = 6.148
0.2: 3.008
0.6: 0.908
1.0: 0.500
2.0: 0.210
5.0: 0.059
7.0: 0.036
10.0: 0.021
∞: 0.000
2b
2a
12. Submarine pitch stability with various methods. We will calculate

the pitch derivative due to an angle of attack, Mw , for a small submarine,
using several different methods.
The submarine is a rotationally symmetric ellipsoid with L/D = 7, L =
35m. We fix the body origin to the midpoint as shown below. The body
170 24 PROBLEMS
is appended with two fins as shown in the figure: each has an area of
1m2 , geometric aspect ratio of 3, and has its longitudinal center of action
1m forward of the body stern.
(a) Apply wing theory to characterize Mw due to the fins alone.
(b) Apply slender body theory to estimate Mw for the body alone.
(c) Linearize the Munk moment, and give a corrected slender-body

value for Mw of the body.
(d) Apply Jorgensen’s approximate formulas to estimate Mw for the

body; make linearizations where necessary.
(e) Use the experimental data from Hoerner (p. 13.2) below to estimate
Mw for the body.
(f) Use the Hoerner result and your result from wing theory to write a
net Mw for the body plus fins.
(g) What is the location of the net aerodynamic center? If it has an

unstable location forward of the midpoint, what increase in fin area
would bring it back to the midpoint?
(h) What is the diameter of a flat stern that will allow the slender body
theory to give the same moment as the Hoerner (experimental)
data, for the body alone?
(i) What is the diameter of a flat stern that allows the approximate
formulas of Jorgensen to give the same moment as the Hoerner data,
for the body alone?
Hoerner p. 13.2:
Symmetric body of revolution with L/D = 6.7.
Force and moment referenced to body midpoint:
Z = −0.5ρU 2 D2 Cydb tan(w/u). Cy = 1.20.
M = 0.5ρU 2 D2 LCnb tan(w/u). Cm = 0.53. DESTABILIZING
171
Each fin: 1m 2 area

AR = 3
1m
L/D = 7
x y
L = 35m
13. Cable dynamics. A towing cable is steel-jacketed, with diameter 3cm,

an effective extensibility of E = 100 × 109 P a, and a density of ρc =
5000kg/m3 . The vehicle it tows is streamlined, and has negligible mass
(M ). If we are towing at low speed and L = 4000m depth,
(a) Calculate ω1 , the first undamped natural frequency in heave, with

mL >> M .
(b) If the vessel heaves with a P = 2m amplitude and a 4.5-second
period, what is the dynamic tension amplitude at the top (s = L) of
the cable? The formula for this dynamic tension (with mL >> M )
is T̃ = EAP k sin(ks)/cos(kL),
q where A is the cross-sectional area
of the cable, k = ω m/EA, and m is the cable mass per unit
length.
(c) Taking into account the static tension induced by in-water weight,
does the cable unload at the surface, for the heaving conditions
above?
14. Cable mechanics (hard!). You are asked to assess the operational
envelope of a cable/vehicle system which has been installed on a vessel.
The cable is steel-jacketed, with diameter 3cm, an effective extensibility
of 100 × 109 P a, and a density of 5000kg/m. The vehicle is streamlined,
and has a mass (material plus added) of 100kg.
(a) The towing angle at the vessel cannot exceed 25deg from vertical,
for reasons of deck and crane geometry. If this angle is considered
equal to the critical angle φc , what is the maximum speed for the
system? Assume the normal drag coefficient of the cable is Cn =
1.2.
172 24 PROBLEMS
(b) Considering the undamped, uncoupled axial dynamics, express the

natural frequency as a function of the vehicle depth, and sketch the
curve. If the fastest waves that the surface vessel responds to have
period 4s, what is the “threshold” operating depth?
(c) The undamped, uncoupled lateral dynamics follow the equation
∂2q ∂2q ∂q
(m + ma ) = T̄ + wn sin φ̄ ,
∂t2 ∂s ∂s
where ma is the added mass of the cable per unit length, and wn is
the in-water weight of the cable. Separation of variables q(s, t) =
q̃(s) cos ωt gives
∂ 2 q̃ ∂ q̃
T̄ + wn sin φ̄ + wn γ q̃ = 0,
∂s2 ∂s
where γ = (m + ma )ω 2 /wn . Simplify this expression by writing
T̄ = wn s (the vehicle weight effect is small), sin φ̄ ' 1 (the cable is
√
nearly vertical), and then use the substitution z = 2 γs to arrive
at the Bessel equation
∂ 2 q̃ ∂ q̃
z2 2
+ z + z 2 q̃ = 0.
∂z ∂z
The derivatives can be transformed from s to z coordinates via the
chain rule:
∂ q̃ ∂ q̃ ∂z
= ·
ds ∂z ∂s
!2
∂ 2 q̃ ∂ 2 q̃ ∂z ∂ q̃ ∂ 2 z ∂z
= + · · .
ds2 ∂z 2 ∂s ∂z ∂z∂s ∂s
(d) The equation above has a solution of the form q̃ = cJ0 (z), where
c is a constant to be found, and J0 (z) is the zero’th order Bessel
function of the first kind. If q̃(s = L) = Q, that is, we impose a
harmonic input at the top, find c in terms of Q and z(s = L). What
condition makes q̃ blow up, indicating a resonance frequency ωn ?
173
(e) Recalling that strumming occurs when ωn ' 0.2U/d, construct a

graph of towing velocities vs. deployment depths for which the
strumming would be centered. Include at least the three lowest
modes in your sketch, and discuss the trends for the higher modes.
J0 (x) = 0 for for x ' π(n − 0.25), n = 0, 1, 2, · · ·, especially when x
is large. You should find that strumming is hard to avoid in deep
towing applications!
15. Vessel heading control. We consider the yaw/sway dynamics of a

high-speed container ship. All the calculations below are to be made in
nondimensional coordinates, so you do not need to perform any trans-
formations in this question.
(a) The nondimensional system with states ~x0 = [v 0 , r0 ] evolves accord-

ing to d~x0 /dt0 = A~x0 + Bδ, where
" # ( )
−0.90 −0.42 −0.13
A= , B= .
−4.8 −2.3 1.4
If the output is yaw rate r0 , i.e., C = [0 1], write the transfer

function r0 (s)/δ(s) = C(sI − A)−1 B. Use matrix inversion for the
2 × 2 matrix (sI − A), perform the matrix multiplications, and then
use the quadratic formula if necessary to express your answer as
r0 (s) s + z1
=K ,
δ(s) (s + p1 )(s + p2 )
where K is a constant gain, −z1 is the zero (there is one real zero),
and [−p1 , −p2 ] are the two poles of the second-order system. Are
both poles stable?
(b) For the purposes of autopilot design, the transfer function φ(s)/δ(s)
is crucial. Fortunately, all you have to do is divide r0 (s)/δ(s) by an-
other factor of s to account for the fact that dφ/dt0 = r0 . Sketch all
the poles and zeros of the plant transfer function P (s) = φ(s)/δ(s)
on a root locus plot.
174 24 PROBLEMS
(c) A PID controller C(s) will work to stabilize the heading angle,
and brings to P (s)C(s) one additional pole at the origin, and two
arbitrary zeros. Demonstrate a rough PID design by devising two
controller zero locations that will attract the closed-loop poles into
the left-half plane. Draw the path that you expect each closed-loop
pole to follow, as the control gain increases from zero to very large
values.
You do not need to calculate the PID gains that go with your pro-
posed zero locations.
16. Full-state feedback of a submarine. The coefficients governing the

pitch/heave dynamics of a Deep Submergence Rescue Vehicle (DSRV)
are given below. They are nondimensionalized as in the lecture notes,
using the factors ρ/2, U , and L.
0
Ixx = 0.000118 Mq0 = −0.0113 Zq0 = −0.0175
0
Izz = 0.00193 Mq̇0 = −0.00157 Zq̇0 = −0.000130
x0G = 0 Mw0 = 0.0112 Zw0 = −0.0439
m0 = 0.0364 Mẇ0 = −0.000146 Zẇ0 = −0.0315
U = 2.0m/s Mδ0 = −0.0128 Zδ0 = −0.0277
L = 15.0m Mθ0 = −0.156/U 2
(a) Write the linearized (nondimensional) dynamics in the matrix form

ẋ = Ax + Bu, where the input u = δ, and the state vector is
x = [w0 , q 0 , θ, Z 0 ]. Z 0 here is the elevation of the vehicle in inertial
coordinates; your approximation for Ż 0 should include both w0 and
θ.
(b) Is the DSRV vehicle open-loop stable, that is, without control action
δ? You can assess this either by finding the eigenvalues of your A-
matrix above, or by computing the stability parameter C 0 .
(c) In preparation for controller design, create a simulation using Mat-
lab. Make a graph of the open-loop step response, showing all the
dimensional state variables versus dimensional time.
(d) Assuming that all the states can be measured accurately, a full-
state controller δ = −Kx can be used, where K is a 1 × 4 gain
matrix. Under this control, the system dynamics are governed by
175
ẋ = (A − BK)x, so that if A − BK has all negative eigenvalues,

the system is stable.
Use the Matlab function place to put the four closed-loop poles at
g(−0.95 ± 0.31i) and g(−0.59 ± 0.81i), g = 1. Demonstrate that
your closed-loop design is stable against nonzero initial conditions.
What are the effects of increasing or decreasing g?
Note that the pole locations suggested above are with respect to
the dimensional system, i.e., using dimensional time.
17. Nyquist stability. The inverted pendulum shown below is often used
as a simple model for rocket flight, and can also illustrate the dynamic
behavior of an unstable ocean vessel which is propelled from the stern,
e.g., a barge being pushed by a tugboat.
For this problem, we assume that all of the mass (m) is concentrated
at the distal end of an arm of length l; a high-performance positioning
system sets the horizontal position of the cart uo .
g l
uo
A Lagrangian derivation of the dynamics gives:
g 1
φ̈ = sin(φ) − cos(φ)üo .
l l
Note that the acceleration of the cart is now considered to be the input
to the plant, e.g., u = üo .
(a) If the observation is taken to be the angle of the bar φ, i.e., y =

φ, write the state-space representation of the dynamics linearized
about the point φ = 0.
176 24 PROBLEMS
(b) Show that this plant is fully state observable and controllable.
(c) Now make the assumption that l = 1, and g = 1 (it can be achieved
through a proper nondimensionalization) What is the transfer func-
tion of the plant, P (s)? Use this plant model for the rest of this
question.
(d) Where are the plant poles and zeros, if any?
(e) A stabilizing controller is not hard to find, and a suggested one is:
−4(2s + 1)
C(s) = .
s+2
Create a rough Nyquist plot of the loci of P (s)C(s), for the frequen-
cies ω = [0, 2.2, ∞]rad/s. These three frequencies are all that you
need to sketch the overall shape; there are no hidden loops or other
features. In the case of ω = 2.2, for which you will need to make
explicit calculations, you may find the following identity useful:
a + jb (a + jb) (c − jd) (ac + bd) + j(bc − ad)

= = .
c + jd (c + jd) (c − jd) c2 + d2
Be sure to include the complex-conjugate points for −ω on your

plot.
(f) Invoke the Nyquist stability criterion to confirm that the closed-
loop system is stable.
(g) About how many degrees of phase margin does this design provide?
What reduction in low-frequency gain can be tolerated? Is it a
reasonable design?
18. Root locus and loopshaping. The parameters governing the surface
maneuvering of a high-speed container ship are given below for reference:
m0 0.00792 0
Izz 0.000456 x0G −0.05
L 175.0m U 8m/s Yv̇0 −0.00705
Yṙ0 0.0000 Yv0 −0.0116 Yr0 0.00242
Yδ0 −0.00258 Nv̇0 0.0000 Nṙ0 −0.000419
Nv0 −0.00385 Nr0 −0.00222 Nδ0 0.00126
177
Note that the center of vessel mass is located aft of the origin; for this
model, the origin coincides with the center of added mass, so that Yṙ =
Nv̇ = 0. The nondimensional system with states ~x0 = [v 0 , r0 ] evolves
according to d~x0 /dt0 = A~x0 + Bδ, where
" # ( )
−0.90 −0.42 −0.13
A= , B= .
−4.8 −2.3 1.4
The relevant output is yaw rate: C = [0 1] and D = 0. For the purposes

of autopilot design, however, the transfer function φ(s)/δ(s) is needed.
The following steps create two heading autopilots, using the root-locus
and loopshaping techniques. In addition to the Matlab commands listed
below, you will find very useful the convolution function conv() which
can be used to combine systems, e.g., for numerators,
numPC = conv(numP,numC);. Also, be sure that you equalize axis scal-
ing for your plots in the complex-plane, by using axis(’equal’);.
(a) Use the Matlab command tf(), or ss(), to create a system model
of the open-loop transfer function P (s)C(s), using the plant above
and a PID-type controller:
1

C(s) = kp 1 + τd s + .
τi s
The actual numerical values for kp , τd , and τi are to be found in the

next step.
(b) Using τd = 2 and τi = 6 as suggested values, use the Matlab com-
mand rlocus() and then rlocfind() to select a controller gain
kp , that puts the three slow poles in the following sector: 1) min-
imum undamped frequency (nondimensional) of 0.3, 2) maximum
frequency of 0.5, and 3) minimum damping ratio 0.7. Give a root
locus plot, with your pole locations clearly marked on top of the
trajectories taken as kp varies. You don’t need to show the fourth,
fast pole, which will be quite far to the left.
(c) Apply the kp you selected to P (s)C(s), and then use the Matlab
functions feedback() to create the resulting feedback system, and
178 24 PROBLEMS
step() to plot the closed-loop system response to a step input in

desired heading.
(d) Use the Matlab command nyquist() to make a Nyquist plot of
P (s)C(s) for your design. Make a visual estimate of the gain and
phase margins.
(e) An alternate approach for controller design of this stable plant
is loopshaping: For the open-loop function L(s) = ωc /s, where
ωc = 2.0, invert the plant to come up with a compensator: C(s) =
L(s)/P (s). This design has infinite gain margin and 90 degrees
phase margin.
(f) As above, create the feedback system, and plot the closed-loop step
response.
(g) The loopshaping control is not quite a PID-controller; how does it
differ, and what would L(s) have to contain to make it a PID?
(h) The controllers you just designed are in nondimensional time coor-
dinates; give the P,I, and D gains for use on a real time scale, for
the root-locus design.
19. LQR. Consider the state-space system and LQR design:
" #
0 −1
A = ,
1 0
B = [1 0]T
C = [0 1]
D = 0
Q = CT C
R = ρ.
The plant is an undamped oscillator with undamped poles at ±j. Note

that the plant output is position for this problem.
(a) What is the control gain K in terms of ρ? Hints: There are two
solutions for p12 ; choose the positive one. Also, the expression for
p22 is messy; luckily, you won’t need to use it.
179
(b) Determine the limiting approximations for K with ρ very small

and very large – these are the cheap control and expensive control
problems.
(c) Derive the limiting closed-loop pole locations for ρ −→ 0, giving the
frequency and damping ratio of the Butterworth pattern in terms of
ρ. You can get the characteristic equation for the poles as det(sI −
(A − BK)) = 0, and then make it fit the form s2 + 2ζωn + ωn2 = 0.
20. LQR. Consider the first-order, unstable system governed by ẋ = x + u,

for which the output is y = x. The state-space matrices are A = B =
C = 1.
(a) Solve the LQR Riccati equation for P , for the case Q = C T C and
R = ρ. Your answer should give a positive P as a function of ρ
alone.
(b) Consider the cheap-control problem, with ρ → 0. Write the leading-
term approximations for P and then K, and compute the eigenvalue
(there is only one) of the closed-loop system. Note that eig(M ) =
M , if M is a scalar instead of a matrix.
(c) Now look at the expensive-control problem, with ρ → ∞. Again,
write the approximations for P and K, and find the closed-loop
system eigenvalue.
21. LQG/LTR. For the inverted pendulum plant (see previous question)
with state-space matrices
" #
0 1
A= , C = [0 1] ,
1 0
the KF Riccati equation yields H = [100 14]T , for the choices V1 =

[100 0 ; 0 0], and V2 = 0.01.
(a) Compute the open-loop transfer function for the Kalman filter loop:
L(s) = C(sI − A)−1 H.
180 24 PROBLEMS
(b) Even though this L(s) is unstable, its magnitude plot confirms that
it is a reasonable loopshaping design. What is the low-frequency
gain for this L(s), and what kind of tracking performance should
we expect from the associated LTR design?
(c) Now consider the closed-loop transfer function, which you can write
as S(s) = L(s)/(1 + L(s)) (the sensitivity). What is the character-
istic equation for the closed-loop system, and about what damping
ratio has the Kalman filter provided?
22. LQG/LTR design. The parameters for the linearized sway/yaw mo-
tions of a swimmer delivery vehicle are given below.
% Parameters, all nondimensional except [U,L]

U = 4.0 ; % m/s
L = 5.3 ; % m
Izz = 0.006326 ;
m = 0.1415 ;
xg = 0. ;
Yv = -0.1 ;
Yr = 0.03 ;
Nv = -0.0074 ;
Nr = -0.016 ;
Ydelta = 0.027 ;
Yvdot = -0.055 ;
Yrdot = 0. ;
Nvdot = 0. ;
Nrdot = -0.0034 ;
Ndelta = -0.013 ;
You are asked to develop an LQG/LTR controller for this plant, and it is
suggested that you compose a single Matlab script to perform the steps
in sequence. Please make sure you answer all the questions, and include
a listing of your code. This entire design is made in nondimensional
coordinates.
(a) Plant Modeling and Characteristics

181
i. Construct a state-space plant model, to take rudder angle δ as

an input and give heading angle φ as an output. Please provide
the numerical values for the A, B, C matrices. There should be
three states in your model, with one input channel and one
output channel.
ii. Compute and list the controllability and observability matrices;
is the plant state-controllable and state-observable?
iii. Where are the poles of your plant model? Is this model stable?
iv. Show a plot of your plant’s step response.
(b) LQR and KF Designs
i. Using the Matlab command lqr(), you can compute the LQR
feedback gain K, for given A, B, Q, and R matrices. With the
choices Q = C T C, and R = ρ, list K and plot the closed-loop
step responses for the choices ρ = [0.1, 0.001, 0.00001]. How do
the gains and step responses change as you make ρ smaller and
smaller?
Note that the fundamental closed-loop LQR system is
~x˙ = (A − BK)~x + BK~xdesired

y = ~x,
i.e., the input to the closed-loop system is ~xdesired and the out-
put is ~x. Your plot should show specifically the output φ, for an
input of ~xdesired = [vdesired = 0, rdesired = 0, φdesired = 1]. This
compression can be achieved in one step by premultiplying the
system by C T , and post-multiplying it by C:
~x˙ = (A − BK)~x + BKC T ydesired

y = C~x,
ii. The Matlab command lqe() can be used to generate the Kal-
man filter gain H, given design matrices A, C, V1 , and V2 . For
the choices V1 = I3×3 and V2 = 0.01, compute H, and make
a plot of the closed-loop step response. Be sure to give the
numerical values of H.
182 24 PROBLEMS
Note that the lqe() command asks for a disturbance gain ma-
trix G; you should set this to I3×3 . The closed-loop KF system
is as follows:
x̂˙ = (A − HC)x̂ + Hy
ŷ = C x̂,
i.e., the input is the measurement y and the output is an esti-
mated version of it, ŷ.
(c) Loop Transfer Recovery
The LQG compensator is a combination of the KF and LQR designs
above. With normal negative feedback, the compensator C(s) has
the following state space representation:
~z˙ = (A − BK − HC)~z + He
u = K~z,
so that the input to the compensator is the tracking error e =
r − y, and its output u is the control action to be applied as
input to the plant. The total open-loop transfer function is the
P (s)C(s); in Matlab, you may simply multiply the systems, e.g.,
sysPC = sysP * sysC ;.
i. Make a log(magnitude) plot of the KF open-loop transfer func-
tion L(s) = C(sI − A)−1 H, versus log(frequency). You may
find the Matlab command freqresp() helpful. |L(s)| should
be large at low frequencies, and small at high frequencies, con-
sistent with the rules of loopshaping.
ii. As ρ → 0, the product P (s)C(s) → L(s). Demonstrate this by
computing P (s)C(s) for the three different values of ρ above,
and overlaying the respective |P (s)C(s)| over the plot of part
3a).
iii. Make a closed-loop step response plot for the smallest value of
ρ. How does it compare with the KF step response of part 2b)?
In real LTR applications, the particular values of V1 and V2 can be picked

to control the low-frequency gain, and crossover frequency of the open-
loop KF system L(s) = C(sI − A)−1 H.
183
23. Inertial navigation. A triaxial accelerometer package is aligned with

the [x,y,z] (fwd, port, up) axes of an underwater vehicle. At a particular
instant in time, the three raw accelerations from the strain guages are
[ẍ = −0.96, ÿ = 0.80, z̈ = 9.73] meters/second. Note that under the
influence of gravity alone, this sensor will always report that the vehicle
is accelerating upwards.
(a) Ignoring the coupling of Euler small angles, what estimates can
you give for the roll and pitch angles? I suggest that you draw the
measurement vector in the coordinates of the vehicle frame, and
then consider the orientation of the body which would lead to it.
(b) Given that the acceleration due to gravity is 9.81m/s, how do we
know that these are reasonable (but not foolproof!) angle calcula-
tions to make?
184 24 PROBLEMS
24. LQG/LTR design: term project. You are asked to carry out mod-
eling and control system design for the surge dynamics of an oceanic
cable-laying vessel. This vessel needs to have very good speed control
so that the cable, which is being paid out constantly and sinks slowly
under its own weight, lies properly on the seafloor. The bathymetry
for the path is well-known but variable, so the vessel will need to slow
down and speed up frequently. Unexpected deviations in speed from the
calculated trajectory are likely to create loops or kinks in the cable, or
induce large tensions.
vessel length L 145m

draft h 9m
beam b 22m
displacement ∇ 17750m3
surge added mass ma 0.07ρ∇
thrust reduction factor t 0.19
wetted surface area Aw 5585m3
towed resistance coefficient Cr 0.0025
wake fraction at propeller w 0.22
propeller diameter D 7.21m
pitch/diameter ratio P/D 1.1
rotative efficiency ηr 1.025
zero-speed thrust coefficient kt (J = 0) 0.6
slope of kt with J -0.522
zero-speed torque coefficient kq (J = 0) 0.1
slope of kq with J -0.0833
rotational moment of inertia Ip 2.03 × 105 kgm2
engine parameter a 0.876
b 0.208
c 2.173
d -.0939
maximum fuel rate fm 1.5kg/s
maximum torque Qm 70000N m
maximum speed nm 60Hz
gearbox reduction ratio λ 32
efficiency ηg 0.97
185
It is desired to create a linear controller for the nominal surge motion,

and then demonstrate that it will work with a simulation of the real
nonlinear system, within a neighborhood of the nominal speed.
The vessel is powered by a single gas turbine engine, driving one pro-
peller. The parameters for the vessel, propeller, and engine system are
as given in the table.
The gas turbine torque-speed characteristic fits the relation:
! !
Qe f ne f
=− a +b + c +d ,
Qm fm nm fm
where Qe represents the engine torque in N m, f the fuel rate in kg/s,

and ne the rotation rate of the engine in Hz.
(a) Make a map of the gas turbine characteristics. For instance, make
a contour plot which gives Qe /Qm (the fraction of maximum torque
developed) as a function of f /fm (the fraction of the maximum fuel
rate) and ne /nm (the fraction of the maximum rotation rate of the
engine).
Be sure to include clipping on your contours, where the calculated
torque exceeds Qm . Similarly, the engine cannot develop negative
torque.
(b) Make a table of some steady operating conditions for fo /fm in the
range of 0.05–0.95, in which you show: advance ratio seen by the
propeller, engine torque, propeller rotational speed, engine rota-
tional speed, and vessel linear speed.
Plot the pairs [fo /fm , neo /nm ] on the characteristic plot from Part
1, and assess whether the vessel, engine, and propeller are well
matched to operate over a range of fo .
(c) Construct a linear approximation for the plant dynamics around
the operating condition fo /fm = 0.8. If the plant input is δf , and
the output is δu, list the A, B, C matrices for your plant, and the
eigenvalues. δ here indicates the perturbed value from the steady
state, e.g., f = fo + δf .
186 24 PROBLEMS
(d) Create an LTR-type controller for the plant. First, choose the filter
gains V1 and V2 to make the crossover frequency (where the magni-
tude of L(s) = C(sI −A)−1 H passes through unity) about 0.6rad/s.
This choice corresponds with the closed-loop system being able to
overcome waves with period 10seconds or longer. Note that for this
scalar design, you will be able to move the curve |L(s)| up or down
on the Bode plot, but you cannot easily change its shape, or move
it side-to-side.
Secondly, recover this loop shape with an LQG-type controller, us-
ing a small control penalty.
You should prepare for this part: a plot of log(|L(s)|) versus log(ω),
a listing of your V1 and V2 choices, and the A, B, C matrices of your
LQG compensator.
(e) In the event of a current or wind disturbance, the actual vessel
speed will vary from the nominal value. This is not acceptable in
the long term, given that the vessel is laying out a cable.
An integral action can be added quite easily in the LTR design
technique. First, add an integrator to the input of the plant, so
that an “augmented plant” is created. This augmented plant has
one more state than the original plant:
" #
A B
Aaug =
0 0
Baug = [0 1]T
Caug = [C 0]
The idea is to then carry out the KF design and LTR as before, with
the augmented plant model. When this is done, move the integrator
from the augmented plant into the compensator. The addition of an
integrator channel to the plant or compensator can be accomplished
with the command sysPaug = sysP * tf([0 1],[1 0]) ;, for ex-
ample.
Present a figure of |L(s)| and the other quantities as you gave them
in Part 4. This is to be kept as a separate controller design.
(f) For the purpose of demonstrating your controllers, construct a non-
linear simulation of the plant system, with input f . You don’t
187
need to show anything for this task; it is mostly to verify that

the steady conditions you calculated above, with f = 0.8fm , are
actually steady conditions from the point of view of the whole sim-
ulation. You may wish to use global variables in Matlab so that
you don’t have to type everything in more than once.
(g) Augment your simulation above with the compensator dynamics. If
the compensator has states z, and your simulation has states [u, np ],
make a new state vector [u, np , z]. You will propagate the first two
states as usual with the nonlinear equations and the fuel rate f as
input. The compensator states z are propagated by a set of system
matrices [Ac , Bc , Cc ], which you just designed. Note that the input
to the compensator is an error signal: e = r − u, where u is the
speed of the vessel and r is a reference speed. The output of the
compensator is then fuel rate f .
We are interested mainly in disturbance rejection, for which you
will just set r = uo , the steady condition speed. Recall that you are
implementing a controller for operation about a nominal condition;
whenever your compensator generates f from the error signal, be
sure to then add on 0.8fm , the part needed to keep us near the
steady condition.
Demonstrate four properties, showing time-domain simulations for
each:
• The first controller, without integral action, does not com-
pletely reject steady disturbances. I suggest you implement
a disturbance as an acceleration:
uprime = uprime + k. In the absence of a controller, this
line would drive the system with an additional acceleration of
km/s2 , thus modelling the effects of steady wind or current.
This line is to be added after computing the usual parts of
uprime.
• The first controller does not reject high frequency disturbances.
A suggestion is to let
uprime = uprime + k*sin(10*omega_c*t), where omega_c is
the crossover frequency. This adds an oscillating acceleration
at ten times the crossover frequency.
• The first controller does, however, reject some lower frequency
disturbances. For example, try
188 24 PROBLEMS
uprime = uprime + k*sin(omega_c/10*t)/100. Note the ex-

tra factor of 100 is included so that the result can be compared
directly with the high-frequency disturbance case above. The
effect of an acceleration disturbance onto the observed output,
velocity u, scales with the inverse of the frequency.
• The second controller, with integral action, rejects steady dis-
turbances. Use the same approach as above for steady distur-
bances
(h) What is the propeller response during the high-frequency distur-
bances? Is it reasonable or do we need to slow down the controller
bandwidth?

Maneuvering and Control of Marine Vehicles

Uploaded by

Copyright:

Available Formats

Maneuvering and Control of Marine Vehicles

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Maneuvering and Control of Marine Vehicles

Uploaded by

Copyright:

Available Formats

MANEUVERING AND CONTROL

Maneuvering and Control of Marine Vehicles

2 KINEMATICS OF MOVING FRAMES 14

3 VESSEL INERTIAL DYNAMICS 20

7 STANDARD MANEUVERING TESTS 52

10 PRACTICAL LIFT CALCULATIONS 68

11 FINS AND LIFTING SURFACES 74

12 PROPELLERS AND PROPULSION 77

14 TRANSFER FUNCTIONS & STABILITY 96

15 CONTROL FUNDAMENTALS 100

16 MODAL ANALYSIS 109

17 CONTROL SYSTEMS – LOOPSHAPING 113

18 LINEAR QUADRATIC REGULATOR 121

20 LOOP TRANSFER RECOVERY 137

21 SYSTEM IDENTIFICATION 143

22 CARTESIAN NAVIGATION 153

The elements of a vector have a graphical interpretation, which is particularly

1. Vector addition is pointwise.

Graphically, addition is stringing the vectors together head to tail.

2. Scalar multiplication is pointwise.

1.1.2 Vector Magnitude

this scalar is commonly used to normalize a vector to length one.

1.1.3 Vector Dot Product

The dot product also satisfies

where θ is the angle between the vectors.

1.1.4 Vector Cross Product

1. direction is normal to the plane formed by the two vectors,

2. direction is given by the right-hand rule, rotating from ~x to ~y ,

3. magnitude is the area of the parallelogram formed by the two vectors –

4. (signed) magnitude is equal to ||~x||||~y || sin θ, where θ is the angle between

The schoolbook formula is

1.2.2 Multiplying a Vector by a Matrix

where n is the number of columns in A. ~c will have as many columns as

as many rows as A has columns; they have consistent inner dimension n.

~c = A~v = A:1 v1 + A:2 v2 + · · · + A:n vn .

We are multiplying column vectors of A by the scalar elements of ~v .

1.2.3 Multiplying a Matrix by a Matrix

C = AB = [AB:1 AB:2 · · · AB:k ],

where k is the number of columns in matrix B. The same inner dimension

1. Associative. (AB)C = A(BC).

2. Distributive. A(B + C) = AB + AC, (B + C)A = BA + CA.

3. NOT Commutative. AB 6= BA, except in special cases.

1.2.4 Common Matrices

The identity always satisfies AIn×n = Im×m A = A.

det(A) = A11 A22 − A21 A12

In higher dimensions, the determinant is more complicated to compute. The

If the determinant of a matrix is zero, then the matrix is said to be singular –

Once this computation is made, the inverse follows from

1.2.9 Eigenvalues and Eigenvectors

Thus, A has two eigenvalues, λ1 = −1 and λ2 = 2. Each is associated with a

1.2.10 Modal Decomposition

X = [~x1 · · · ~xn ] , and

Next, construct a diagonal matrix of eigenvalues:

Hence A can be written as a sum of modal components.1

1.2.11 Singular Value

where U is m × m, Σ is m × n, and V is n × n. The star notation indicates a

Other properties of the singular value include:

1.3 Laplace Transform

where s is an unspecified complex number; Y (s) is considered to be complex as

1.3.3 Convolution Theorem