1

Unified Approach to Specifying the Psi-angle Error Equation in Strapdown

Inertial Navigation Systems

Yong Li
University of New South Wales, Sydney 2052, Australia
Email: [email protected]


Abstract

Error equations for inertial navigation systems (INSs) are typically derived using either the phi-angle
() approach or the psi-angle () approach. These two approaches have been shown to be
theoretically equivalent. The psi-angle error model is widely used in modern INS-multisensor
integrated navigation systems because the translatory error is isolated from the psi-angle error
equation. However, the misalignment vector takes on a different meaning in different
implementations which has resulted in two forms of the psi-angle model being referred to in the
literature, which have opposite signs [1]. To resolve this problem, this paper proves that the two forms
of the psi-angle model are indeed identical through the use of explicit notations of the vector psi. The
derivation proves that the sign change problem is due to the application of the psi vectors in opposite
directions. In addition, this paper derives the psi-angle models for all four forms of the vector psi. All
of these forms can be used in INS-multisensor integrated systems. The corresponding psi-angle
corrections are given as well. Upon different forms of vector psi, the velocity and position error
equations in the psi-angle model are given too.


Key words: inertial navigation system, error equation, psi-angle model, sign change problem, vector
psi


Notations and Definitions

The following coordinate systems are used to derive the psi-angle model:

Body frame (b-frame): frame fixed to host vehicle on which INS sensors are mounted.
Computer frame (c-frame): local-level local-north coordinate system for the INS-derived position.
Inertial frame (i-frame): inertial coordinate system.
Platform frame (p-frame): analytical platform into which the computed direction cosine matrix
transforms vectors from the b-frame.
True frame (t-frame): local-level local-north coordinate system with true geo-position.
DCM (C): direction cosine matrix which transforms vectors from one coordinate system to another.

Except when specifically stated, all symbols are vectors (italic lower case letters) or matrices (italic
capital letters), with appropriate subscripts and superscripts. Vectors and matrices include:

: angular velocity of host platform.
: angular difference between p-frame and c-frame.
: angular difference between p-frame and t-frame.

u
u
: angular velocity of a-frame with respect to r-frame, projected into the a-frame, where a-frame
and r-frame are arbitrary frames (can be b-,c-, i-, p-,or t- frame).
C
u

: DCM from a-frame to r-frame.

(:): cross-product matrix of column vector v (where v is an arbitrary vector).

b
b
: gyro-sensed angular velocity of b-frame with respect to i-frame, projected into the b-frame.
e
b
: error of gyros in b-frame.

2

Introduction

The combination of Micro-Electro-Mechanical inertial sensors (MEMS) with Global Navigation
Satellite Systems (GNSS) can potentially result in an ideal navigation system characterized by
comparatively low-cost, light-weight and low power consumption, and long-term stable accuracy
compared with a standalone precise inertial navigation system (INS), which may be quite expensive
and heavy, and subject to accuracy drift. The psi-angle error model is widely used in modern
INS/GNSS multisensor integrated navigation systems because the translatory error is isolated from
the psi-angle error equation. However, as indicated in [1] investigators develop different equations for
psi, appropriate for their own coordinate systems and definition of the misalignment vector. Two
forms of the psi-angle model have been proposed, for example:

= -
c
c
+ e
p
(1)

This is the psi-angle model in [2, 3, 4], originally derived for gimballed INSs and also widely used in
strapdown INSs, and

= -
c
c
- e
p
(2)

This is the psi-angle model referred to in [1, 5, 6] and originally derived for strapdown INSs.

The two forms of the psi-angle model have opposite signs for the gyro error term, or equivalently the
two forms signify opposite sign of psi. However in both forms psi is denoted by the same symbol,
which can confuse readers and those implementing INS algorithms in software. The opposite signs of
psi will lead to opposite corrections of DCM C
b
p
in multisensor integrated systems.

In the literature, psi is usually defined as the angular misalignment between the p-platform and the c-
platform. Few authors fully describe the misalignment - psi as a vector implies a direction and needs
to be projected into a certain coordinate system to facilitate computation. The following four vector
forms of psi are introduced:

(1)
pc
p
: angular difference of c-frame with respect to the p-frame, projected into the p-frame.
(2)
pc
c
: angular difference of c-frame with respect to the p-frame, projected into the c-frame.
(3)
cp
p
: angular difference of p-frame with respect to the c-frame, projected into the p-frame.
(4)
cp
c
: angular difference of p-frame with respect to the c-frame, projected into the c-frame.

The time-derivative of psi is the relative angular velocity between the p- and c- frames. All four forms
of psi and their associated differential equations are summarized in Table 1.

Table 1. Vector forms of psi angle
Index definition
1

pc
p
=
pc
p

Angular difference of c-frame with respect to
p-frame, projected into p-frame
2

pc
c
=
pc
c

Angular difference of c-frame with respect to
p-frame, projected into c-frame
3

cp
p
=
cp
p

Angular difference of p-frame with respect to
c-frame, projected into p-frame
4

cp
c
=
cp
c

Angular difference of p-frame with respect to
c-frame, projected into c-frame



3

Derivation of the Psi-angle Error Equation

The derivation of the psi-model in [2] refers to gimballed INSs. The author in [1] proved that this psi-
model is also valid for strapdown INSs, apart from a sign change.











Figure 1. Geometry of psi and phi angles [1]

The inertial navigation systems basically solve Newtons force equations from measurements
(accelerometer outputs) of specific force (non-gravitational accelerations) within a frame whose
orientation with respect to an inertial frame is controlled or known via gyroscopes [2]. The INS
angular errors are attributable to two sources: (1) translatory error caused by an error in the INS
position; and (2) angular error from the gyro errors.

Fig.1 shows the geometry of the psi and phi angles, which are the angular difference between the p-
and c- frames, and the angular difference between the p- and t- frames, respectively. As illustrated in
Fig. 1, the INS-derived position error causes the t-frame to vary to the c-frame. Because of gyro errors,
the DCM- or attitude-updating calculation has an error which in turn results in the DCM transforming
a vector in the b-frame into the p-frame, instead of into the c-frame. In principle, the t-, c-, and p-
frames are the same frame (local-level local-north). The INS errors cause them to be different from
each other. For a gimballed INS, the b-frame is always controlled to track the c-frame. However,
because of the gyro drift and calculation error the b-frame actually tracks the p-frame.

Ideally, the DCM, in the error-free case, transforms vectors in b-frame to t-frame, the differential
equation is:

C

b
t
= -(
bt
t
)C
b
t
(3)

The angular velocity
bt
t
can be written as:

bt
t
=
t
t
-
b
t
(4)

Substituting Eq. (4) into (3) and noting that (
bt
t
) represents the vector cross-product, one can
obtain:

C

b
t
= -(
t
t
-
b
t
)C
b
t
= (
b
t
)C
b
t
- (
t
t
)C
b
t
(5)

Using the equation:

(
b
t
)C
b
t
= C
b
t
(
b
b
) (6)

Eq. (3) can be rewritten as:

C

b
t
= C
b
t
(
b
b
)-(
t
t
)C
b
t
(7)
x
t
z
t

x
c
x
p
z
c
z
p

Where
b
b
is the angular velocity of the b-frame with respect to the i-frame, projected into the b-
frame. It can be sensed by the gyros.
t
t
is the angular velocity of the t-frame with respect to the i-
frame, projected into the t-frame.

In Eq. (7), the true value of
t
t
is calculated from the true geo-position. However, because of the
translatory error, the computed
t
t
using the INS-derived position becomes
c
c
. The local-level local-
north frame with the INS-derived position is then the c-frame. The DCM becomes C
b
c
and Eq. (7)
becomes:

C

b
c
= C
b
c
(
b
b
)-(
c
c
)C
b
c
(8)

The real gyro outputs have errors:

b
b
=
b
b
+ e
b
(9)

where e
b
is the lumped error of the gyros.

As a result of the gyro errors, the local-level local-north frame is further perturbed to become the p-
frame. Then Eq. (8) in the INS computation is actually:

C

b
p
= C
b
p
(
b
b
)-(
c
c
)C
b
p
(10)

where the calculated DCM is C
b
p
instead of C
b
c
in Eq. (8), which means that the INS-computed DCM
transforms a vector in the b-frame into the p-frame instead of the c-frame. In other words, the
difference between the p-frame and the c-frame is only contributed to by the gyro errors. The angular
difference between the p-frame and the c-frame is defined as the psi-angle, which links the difference
of the two DCMs:

C = C
b
p
- C
b
c
(11)

Substituting Eqs. (9) and (11) into Eq. (8) yields:

C

b
c
= (C
b
p
- C)(
b
b
- e
b
) - (
c
c
)(C
b
p
- C) (12)

Now differentiating both sides of Eq. (11) and further substituting Eqs. (10) and (12) and neglecting
the 2
nd
order term of Ce
b
, one obtains:

C

= C

b
p
- C

b
c
= C
b
p
(e
b
) + C(
b
b
) - (
c
c
)C (13)

C can also be derived using another approach. In fact, assuming the psi-angle is so small that the p-
frame is very close to the c-frame, the transformation matrix from the p-frame to the c-frame can be
written as:

C
p
c
= I - (
pc
p
) (14)

where
pc
p
is the angular difference of the c-frame from the p-frame, projected into the p-frame.

Substituting Eq. (14) into (11), one obtains:

C = C
b
p
- |I - (
pc
p
)]C
b
p
= (
pc
p
)C
b
p
(15)

5

Now differentiating both sides of Eq. (15) yields another form of the differential equation of C:

C

= (

pc
p
)C
b
p
+ (
pc
p
)C

b
p
(16)

Substituting Eq. (10) into (16) yields:

C

= (

pc
p
)C
b
p
+ (
pc
p
)|C
b
p
(
b
b
) - (
c
c
)C
b
p
] (17)

Substituting Eq. (15) into (13), one obtains:

C

= C
b
p
(e
b
) + (
pc
p
)C
b
p
(
b
b
) - (
c
c
)(
pc
p
)C
b
p
(18)

Comparing Eqs. (17) and (18):

(

pc
p
)C
b
p
+ (
pc
p
)|C
b
p
(
b
b
) - (
c
c
)C
b
p
] = C
b
p
(e
b
) + (
pc
p
)C
b
p
(
b
b
) - (
c
c
)(
pc
p
)C
b
p

(19)

Right-multiplying C
p
b
on both sides, and moving the 2
nd
term on the left-hand side of the equation to
the right-hand side, one obtains:

(

pc
p
) = C
b
p
(e
b
)C
p
b
+ |(
pc
p
)(
c
c
) - (
c
c
)(
pc
p
)] (20)

Note that following two expressions:

(
c
c

pc
p
) = (
c
c
)(
pc
p
) - (
pc
p
)(
c
c
) (21)
and
(e
p
) = (C
b
p
e
b
) = C
b
p
(e
b
)C
p
b
(22)

Applying them to Eq. (20), one obtains:

(

pc
p
) = (e
p
) - (
c
c

pc
p
) = (e
p
-
c
c

pc
p
) (23)

Note that () is the cross-product operator of the vector, thus its vector form can be written as:

pc
p
= -
c
c

pc
p
+e
p
(24)

This is the psi-angle model, which is the same as in Eq. (13) in [2] and Eq. (38c) in [4].

Similarly, using
cp
p
= -
pc
p
in Eq. (24), one obtains:

cp
p
= -
c
c

cp
p
- e
p
(25)

This is the psi-angle model represented by Eq. (28) in [1], Eq. (552a) in [5], and Eq. (33) in [6].

Using a similar derivation, the psi-angle model for
pc
c
can be derived:

pc
c
= -
c
c

pc
c
+e
p
(26)

and the psi-angle model for
cp
c
can be derived as:

cp
c
= -
c
c

cp
c
- e
p
(27)
6

It can be seen that the four forms given above can be categorized into two forms: Eqs. (24) and (26)
correspond to Eq. (1), and Eqs. (25) and (27) correspond to Eq. (2). Therefore one can conclude that
the current two forms of the psi-angle model are theoretically identical. However explicit notation for
the vector psi is needed to accurately describe the psi-angle model as presented in Eqs.(24) - (27).
Upon using different forms of vector psi, corresponding correction of C
b
p
should be applied in the
multisensor integrated system, as summarized in Table 2. After correction, the DCM transforms the
vectors from the b-frame into the c-frame, such as:

C
b
c
= C
b
p
C
p
c
(28)

Table 2. Different forms of the psi-angle model and corresponding corrections of C
p
c

Form C
p
c

1

pc
p
= -
c
c

pc
p
+ e
p
C
p
c
= I - (
pc
p
)
2

pc
c
= -
c
c

pc
c
+ e
p
C
p
c
= I - (
pc
c
)
3

cp
p
= -
c
c

cp
p
-e
p
C
p
c
= I + (
cp
p
)
4

cp
c
= -
c
c

cp
c
-e
p
C
p
c
= I + (
cp
c
)

Note that coordinates mis-matching occurs in all forms of the psi-angle model, such as the term

c
c

pc
p
in forms 1 and 3, and the term e
p
with
pc
c
in form 2 and a similar one in form 4. This mis-
matching results in higher order errors that can be neglected.

Although the psi-angle in [1] is defined as the angular difference between the c-frame and the p-frame

pc
p
(corresponding the angular vector from c-frame to b-frame), upon which the psi-angle model
should have the form of Eq. (1) as proven above, [1] derived the psi-angle model of the form of Eq.
(2). It has been found that this mistake is caused by wrongly applying Eq.(8) in [1]. According to Eq.
(8) in [1], the psi-angle in [1] should be the angular difference between the p-frame and the c-frame

cp
p
or
cp
c
(corresponding the angular vector from b-frame to c-frame), upon whichso that the psi-
angle model in [1] is consistent with the results of this paper.

The widely-used velocity error equation in the psi-angle model is with
pc
c
[2][3]:

o:
c
= -(
c
c
+
c
c
) o:
c
+ v
c
-
pc
c

c
+ og
c
(29)

It should be the same for
pc
p
, with which the velocity errors are resolved in the p-frame. For
cp
c
it
should be:

o:
c
= -(
c
c
+
c
c
) o:
c
+ v
c
+
cp
c

c
+ og
c
(30)

It should be the same for
cp
p
, with which the velocity errors are resolved in the p-frame. Note that the
term has an opposite sign as that in Eq. (29).

The position error model has no difference for all the forms of vector psi [2][3]:

or
c
= -
cc
c
or
c
+ o:
c
(31)


Concluding Remarks

Through the use of explicit notations of the vector psi, this paper has theoretically defined all four
forms of the psi-angle model, which can be categorized into the two forms currently found in the
7

literature. Using this approach it is relatively easy to identify the two forms of the psi-angle model
because they can be derived from one to another. The analysis has demonstrated that the sign change
problem is caused by application of the psi vectors of opposite directions, which is either the angular
difference from the c-platform to the p-platform, or visa versa. All forms of the psi-angle model can
be equivalently used in INS/GNSS multisensor integrated systems, and the corresponding psi-angle
corrections are given in Table 2. The widely-used velocity error equation in psi-angle model is upon

pc
p
or
pc
c
. For
cp
p
or
cp
c
, the term in the velocity error equation of the psi-angle model
should have an opposite sign. The position error model has no difference for all the forms of vector
psi.

Finally, the psi-angle in [1] should be the angular difference from the p-frame to the c-frame, ensuring
that the psi-angle model in [1] is then consistent with the result of this paper.

References

[1]. A. Weinred and I.Y. Bar-Itzhack, Psi-angle error equation in strapdown inertial navigation
systems, IEEE Trans. Aerospace and Electronic Systems, Vol. AES-14, No. 3, 539-542, May
1978.
[2]. D.O. Benson, A comparison of two approaches to pure-inertial and Doppler-inertial error
analysis, IEEE Trans. Aerospace and Electronic Systems, Vol. AES-11, No. 4, 447-455, July
1975.
[3]. I.Y. Bar-Itzhack and N. Berman, Control technical approach to inertial navigation systems,
Journal of Guidance, Vol. 11, No. 3, 237-245, May-July 1988.
[4]. D.G. Meskin and I.Y. Bar-Itzhack, Unified approach to inertial navigation system error
modeling, Journal of Guidance, Control and Dynamics, Vol. 15, No. 3, 648-653, May-June
1992.
[5]. G. Arshal, Error equations of inertial navigation, Journal of Guidance, Vol. 10, No. 4, 351-358,
July-August 1987.
[6]. X. Kong, E.M. Nebot, and H. Durrant-Whyte, Development of a non-linear psi-angle model for
large misalignment errors and its application in INS alignment and calibration, Proceedings of
the IEEE Inernational Conference on Robotics & Automation, Detroit, Michigan, May 1999,
1430-1435.