IOP Conference Series: Materials Science and Engineering

PAPER • OPEN ACCESS Related content

- Quadrotor trajectory tracking using PID
Leader-Follower Formation Control for Quadrotors cascade control
M Idres, O Mustapha and M Okasha

To cite this article: Falin Wu et al 2017 IOP Conf. Ser.: Mater. Sci. Eng. 187 012016 - Intelligent passively stabilized quadrotor
D Sayfeddine, A G Bulgakov and T N
Kruglova

- Trajectory tracking in quadrotor platform

by using PD controller and LQR control
View the article online for updates and enhancements. approach
Maidul Islam, Mohamed Okasha and
Moumen Mohammad Idres

Recent citations
- Backstepping Based Formation Control of
Quadrotors with the State Transformation
Technique
Keun Lee et al

This content was downloaded from IP address 129.113.130.47 on 15/11/2019 at 21:44

 ICMAE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 187 (2017) 012016 doi:10.1088/1757-899X/187/1/012016

International Conference on Recent Trends in Physics 2016 (ICRTP2016) IOP Publishing

Journal of Physics: Conference Series 755 (2016) 011001 doi:10.1088/1742-6596/755/1/011001

Leader-Follower Formation Control for Quadrotors

Falin Wua, Jiemin Chenb and Yuan Liangc

School of Instrumentation Science and Opto-electronics Engineering, Beihang
University, No.37 XueYuan Road, HaiDian District, Beijing, China
a
[email protected], b [email protected], c [email protected]

Abstract. Quadrotors are gaining an increasing interest in public and extensively explored in
recent years. In many situations, a team of quadrotors is desired to operate in a certain shape,
which is also called formation. In this paper, a linear PID controller is used to control each
single quadrotor and a slide mode controller is adopted to solve the formation flying problem
which employs the leader-follower structure. The formation simulations are run in the
Matlab/Simulink environment to evaluate the performance of control laws.

1. Introduction
Quadrotor, or commonly known as drone and quadcopter, is a simple aerial vehicle which can operate
in the narrow space with four propellers. With the pace of electronic device industry accelerating,
quadrotors are being extensively explored and studied by an increasing number of scholars over the
past few years. Carrying with sensors and other necessary devices, they make a difference to
numerous civilian and military applications, including disaster mitigation, environment preservation,
reconnaissance and academic research [1].
However, in many application scenarios, a team of unmanned aerial vehicles (UAVs) needs to
follow the preset trajectory while maintaining a specific geometric shape. Compared with the
conventional system, formation flight can increase the anti-interference performance and efficiency,
improve the probability of success in search tasks, expand the region of surveillance and reduce the
expense of military missions. As a result, the cooperation of multiple quadrotors is an active and
popular research topic as formation has broad applications in accomplishing complicated tasks in real-
world domains [2-4].
There are various methods to realize formation cooperation, such as leader-follower strategy [3; 5],
virtual structure approach [6] and behavior-based method [7].
The leader-follower method is the most common hierarchical structure used in the formation flight.
Each follower aircraft is controlled to maintain its position and velocity to a designated leader aircraft
which flies along some predefined trajectory. This structure is characterized by its easy
reconfiguration and simple expandability to new objects, for its function can be quickly taken over by
other UAVs [8]. However, this strategy does not use direct feedback from the followers in general.
Virtual architecture approach enables quadrotors to fly as if they were embedded in a rigid
structure. The formation has a so-called geometry center, which is a virtual point determined by shape,
flying speed, orientation of the formation or some other factors. If one of the UAVs in the team loses
its position, the overall flock can change the flight trajectory to help the lost one. The literature [9]
describes this approach applied in formation of space crafts. However, this strategy is relatively
difficult to be implemented in practical applications.

Content from this work may be used under the terms of the Creative Commons Attribution 3.0 licence. Any further distribution
of this work must maintain attribution to the author(s) and the title of the work, journal citation and DOI.
Published under licence by IOP Publishing Ltd 1
 ICMAE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 187 (2017) 012016 doi:10.1088/1757-899X/187/1/012016

The behavior-based approach prescribes the desired behaviors and generates a vector based on a
weighted average of sensory inputs. Possible behaviors include formation keeping [7], trajectory
tracking and obstacle avoidance. This strategy is suitable for applications with multiple agents and has
an advantage in communication among agents. However, it is difficult to assure some of the
characteristics of the formation.
In this paper we adopt a slide mode controller [10] to manipulate formation flight of three UAVs,
which takes the leader-follower structure. PID control laws are used and preformed for each of the
three UAVs. To verify the control effect, simulations with six-degrees-of-freedom state space UAV
model are presented, which are run in Matlab/Simulink software.
The outline for this paper is as follows: Section 2 introduces the dynamic model of the quadrotor.
Section 3 lays out the controller strategy, including PID controller to manipulate a single quadrotor
and a formation controller to guide the whole team of UAVs. The simulation results are summarized
in Section 4 which evaluates the performance of the strategy. Section 5 is the conclusions.

2. Dynamic model of single quadrotor

The kinematic of quadrotor is formalized in earth frame E and body-fixed frame B . The position and
Euler angles (roll, pitch and yaw) of the vehicle are defined as r   x y z  and      
T T

respectively, with mass m and inertia matrix I . The angular velocity    p q r  is defined in the
body-fixed frame B .
Some reasonable assumptions need to be considered to simplify the design which can be referred in
literature [11]. Firstly, the quadrotor is a symmetric rigid body with the center of mass and body-fixed
frame origin coinciding. So the inertia matrix I can be simplified as I  diag  I xx I yy I zz  .
Secondly, the quadrotor operates in the hover state where roll and pitch are small ( c  1 , c  1 ,
s   , s   ), so the angular velocity  can be seen equal to  and the gyroscopic torque can be
neglected.
The translational and rotational dynamics equations are as follows:
mr  mgez  TB RBE ez (1)
Iω  ω  Iω   (2)
where g is the gravity acceleration; ez  0 0 1 is the unit vector w.r.t. earth frame E ; TB is the
T

total force generated by the four rotors w.r.t. body-fixed frame B ; RBE  SO  3 is the rotation matrix
from body-fixed frame to earth frame;  is the control torque in three dimensions obtained by
changing the rotor speeds. The speeds of four rotors are manipulated by adjusting the total force TB
and torque  , both of which actually are the components of inputs u of the system. The relationship
between inputs and the speed of rotors is as follows:
 u1  TB   k k k k  12 
u     0 lk 0 lk   2 
u   2     
  2  (3)
u3      lk 0 lk 0  32 
      
u4     d d d d  24 
where k is the thrust factor; l is the distance from the center of mass to any one of the rotors; d is
the drag coefficient; i is the speed of the rotor i .

3. Control strategy
In this section, a PID law is utilized to make the single UAV move along the predefined trajectory.
Then we focus on leader-follower formation control according to the geometry of formation and the
relative dynamics.

2
 ICMAE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 187 (2017) 012016 doi:10.1088/1757-899X/187/1/012016

3.1. PID control

The controller is designed to calculate the attitude and velocity needed for UAVs to follow the desired
trajectory. Each quadrotor is controlled independently by the nested feedback loops shown in figure 1.
The outer and inner loops control position and attitude respectively. The reference signals for velocity
control are the sums of the reference values given by reference trajectory generator and correction
signals calculated by rigid body dynamics. The superscript d represents the corresponding value is
desired.

Figure 1. The nested control strategy block diagram

3.1.1. Attitude control. The attitude controller is designed at a point where roll and pitch are small.
According to the assumptions, the products of inertia are zero and I xx  I yy because of the symmetry
then:
I xx p  u2  qr  I zz  I yy  (4)
I yy q  u3  pr  I xx  I zz  (5)
I zz r  u4 (6)
The component of the angular velocity r is very small. So the rightmost terms in (4) and (5) can
be ignored compared with other terms. The system is presented in the hover state, so    . For those
reasons, simple proportional derivative control laws can be employed as follows:
u2d =k p, ( d   )  kd , ( d   ) (7)
u3d  k p, ( d   )  kd , ( d   ) (8)
u4d  k p, ( d  )  kd , ( d  ) (9)

3.1.2. Position control. We use pitch and roll to control position in XY plane, u4 to control yaw angle,
and u1 to control height. The desired position is denoted as ri d , which can be calculated from PID
controller
 ri,T  ri）0
(ri ,T  ri d )  kd ,i (ri ,T  ri )  k p ,i (ri ,T  ri )  ki ,i （ = (10)
where ri ,T  ri ,T  0 for hover.
By linearizing equation (1), the relationship between the desired acceleration and angles (roll and
pitch) is obtained according to following assumptions:
a. The roll and pitch are small, so c  1 , c  1 , s   , s   .
b. The yaw keeps unchanged, which means  = T = 0 , where  0 is the initial yaw and  T is the
desired yaw.
c. In hover state TB  mg .
After linearization, we can get
1 d
d  (r1 sin T  r2d cos T ) (11)
g
1
 d  (r1d cos T  r2d sin T ) (12)
g

3
 ICMAE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 187 (2017) 012016 doi:10.1088/1757-899X/187/1/012016

u1d  mr3d (13)

3.2. Leader-follower formation control

The process of formation flight is as follows: the leader flies along a predefined trajectory which is
stipulated by the trajectory tracking controller. Then the formation controller is designed to keep a
specific shape in XY plane (e.g. the shape shown in figure 3), according to the relative kinematics
between the leader and the follower. After the formation controller generating the desired velocity for
the follower, the follower can track this velocity so as to keep the relative distance  and orientation
angle  constant at the same height. The leader tracks the predefined trajectory of the formation while
the follower tracks the desired velocity to keep the shape, but the same tracking control strategy can be
used for both UAVs. The formation flying strategy is shown in figure 2.

Figure 2. Control strategy block diagram

Figure 3. Leader-follower structure Figure 4. Quadrotors formation in XY plane

The leader-follower structure for three UAVs is shown in figure 3, with a geometry called ‘V’
shape. The position of UAV can be determined by velocity vi and angle velocity i for the yaw angle,
which are the reference inputs to the trajectory tracking controller of the follower. More specifically,
vix and viy are the velocity components in X and Y direction respectively w.r.t. body-fixed frame
coordinate B , the subscript i denotes the leader ( i  L ) or the follower ( i  F ).  i is the angle
between the heading direction and the x axis of earth frame E .
Considering the dynamic characteristics of the quadrotor, it can be found that its vertical subsystem
and horizontal subsystem are completely decoupling. Therefore, it is possible to design these two
subsystems separately. For vertical subsystem,
zi  viz , i  F , L (14)

4
 ICMAE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 187 (2017) 012016 doi:10.1088/1757-899X/187/1/012016

where viz is the velocity component in z axis w.r.t. body-fixed frame B and zi is the velocity
component w.r.t. earth frame E .
Then we mainly focus on the relative translational kinematics in XY plane at the same altitude,
which is demonstrated in figure 4. xi and yi are the velocity components w.r.t. earth frame E , while
vix , viy are the velocity components w.r.t. body-fixed frame B . The translational dynamics in the XY
plane can be described as
xi  vix cos i  viy sin i (15)
yi  vix sin i  viy cos i (16)
 L  L (17)
Similarly, xi , yi w.r.t. E can be transformed to vix , viy w.r.t. B .
vix  xi cos i  yi sin i (18)
viy   xi sin i  yi cos i (19)
As shown in figure 4,  is the distance from the mass center of the leader to one of the followers,
x ,  y are the x , y coordinates in the leader’s body-fixed frame( BL ). Then,
x    xL  xF  cos L   yL  yF  sin L (20)
y   xL  xF  sin L   yL  yF  cos L (21)
Differentiating (20) w.r.t. time and using (15), (16), (17), (18) and (19), we can obtain
x  yL  xF cos L  yF sin L  vLx (22)
By defining e   F  L , using the equation (15), (16) with the trigonometric identities, one gets
x  yL  vFx cos e  vFy sin e  vLx (23)
Similarly,  y can be written as
y  yL  vFx sin e  vFy cos e  vLy (24)
The desired distance  and angle  are constant, so     0 . Define the formation error
d d d
x
d
y

ex  xd  x (25)
ey    y
d
y (26)
e   F  L (27)
Differentiating (25), (26), (27) w.r.t. time, one gets
ex    yd  ey  L  vFx cos e  vFy sin e  vLx (28)
ey   xd  ex  L  vFx sin e  vFy cos e  vLy (29)
e  F  L (30)
From the formation dynamics (28), (29) and (30), a control law is designed to make the errors
converge to zero. For doing this, vFx , vFy and F are considered as the formation control inputs and
the dynamics of the formation error is written as
  F     G    (31)
where
T
  ex ey e  (32)
T
  vFx vFy F  (33)

5
 ICMAE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 187 (2017) 012016 doi:10.1088/1757-899X/187/1/012016

 eyL   1 
F      exL   2  (34)
 L 
 ce se 0
G       se ce 0 (35)
 0 0 1 
with
 1  vLx  L yd (36)
 2  vLy  L  d
x (37)
The slide mode technique is used to design a robust controller. The switching function is
 2    k f   dt (38)
where k f is a constant matrix. The dynamics of the formation error are expected to keep on a surface
defined by  2 =0 through choosing k f appropriately, on which the errors converge to zero. On this
surface  2 =0 , so we can get
 2    k f   F     G   eq  k f   0 (39)
 eq can be obtained by (39) as
eq =G 1     F     k f   (40)
To make sure the system keep on the surface  2 =0 considering perturbations, it can be obtained
 2    k f   F     G   eq  k f    Lsgn( 2 ) (41)
where L is a positive constant. Then the control input can be obtained by
eq =G 1     F     k f   Lsgn( 2 )  (42)
The sgn( x) is defined as
1 x0
sgn( x)   (43)
1 x  0
and the vector
sgn(1 )   sgn(11 ) sgn(12 ) sgn(13 )
T
(44)

4. Simulation and test results

The main aim of the paper is a numerical simulation which is run in Matlab/Simulink software. The
controller system is highly sensitive to parameters. Improper parameters result in large overshoot, as a
consequence, the expected distance among UAVs cannot be maintained. So the gains for PID
controller are tuned experimentally to provide as short as possible convergence time and non-
oscillatory character of the system response.
The quadrotors used in the simulation are all the same, with mass m = 0.58 kg, and moments of
inertia Ixx = Iyy = 0.1147 kg∙m2 and Izz = 0.0522 kg∙m2. The products of inertia are assumed to be zero.
These are typical values for small UAVs. In the simulation, the leader and the followers are set to
follow the similar trajectory which is an s-shape curve actually. The velocity of the leader is provided
to the followers. The initial position of quadrotor formation is 0 0 0 and the selected distance
T

x  y  5 m . The simulation time is 250 seconds. The red line is the leader’s trajectory while the
black and blue line describe the trajectory of Follower 1 and Follower 2 respectively.

6
 ICMAE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 187 (2017) 012016 doi:10.1088/1757-899X/187/1/012016

Simulation results of the quadrotor formation flight trajectory tracking in three dimensions and XY
plane are presented in figure 5 and figure 6 respectively, indicating that the agents try to follow the
leader’s trajectory fairly well.
The position errors of Follower 1 in X and Y direction in the first 10 seconds are shown in figure 7
and figure 8. Due to the fact that the three quadrotors start from the same position, the initial position
errors in X and Y direction are both 5 m according to the predefined x and  y . Then the Follower 1
moves to (5, 5) while the Follower 2 moves to (-5, -5) in the XY plane w.r.t. body-fixed frame B of
the leader. So ex and e y converge to zero during the first 5 seconds. As can be seen from the figures,
the errors converge to zero at a fast rate and fluctuate around zero after 5 seconds. The position errors
of Follower 2 in X and Y direction are similar to Follower 1. In addition, the error of yaw also
oscillates between -0.04 rad and 0.08 rad in the first 5 seconds and then fluctuates around zero after 5
seconds, which is not illustrated in the figure. The error is relatively small without considering the
communication delay and disturbance. The analysis of error demonstrates that the performance of the
formation control is quite satisfactory under the circumstance where the roll and pitch are small.
10

25 0

20
-10
15
z(m)

10
y(m)

-20
5

0
Leader -30

-5 Follower 1
20 Follower 2
0 30 -40

-20 20 Leader
10 Follower 1
y(m) -40 Follower 2
0 x(m) -50
-60 -10 -10 -5 0 5 10 15 20 25 30
x(m)

Figure 5. Quadrotors formation flight Figure 6. Quadrotors trajectory tracking in

trajectory in three dimensions XY plane
ey(m)
ex(m)

time(s) time(s)

Figure 7. Position error in X direction Figure 8. Position error in Y direction

5. Conclusions
The paper focuses on formation control with a leader-follower structure designed for three quadrotors.
PID control law is presented to manipulate a single vehicle and a slide mode controller is adopted for
formation control. The experiments illustrate the effective performance of control laws. However, this
research is restricted to simulations without considering the practical noise and communication delay.
Future research will focus on the hardware platform for formation flight experiments.

6. References

7
 ICMAE IOP Publishing
IOP Conf. Series: Materials Science and Engineering 187 (2017) 012016 doi:10.1088/1757-899X/187/1/012016

[1] Floreano D and Wood RJ 2015 Science, technology and the future of small autonomous drones
Nature 521 460-6
[2] Kendoul F 2012 Survey of advances in guidance, navigation, and control of unmanned
rotorcraft systems J. Field Robot 29 315-78
[3] Yun B, Chen BM, Kai YL and Tong HL 2010 Design and implementation of a leader-follower
cooperative control system for unmanned helicopters J. Control Theory Appl. 8 61-8
[4] Nathan PT, Almurib HA et al 2011 A review of autonomous multi-agent quadrotor control
techniques and applications. 4th Int. Conf. on Mechatronics (Kuala Lumpur) pp 1-7
[5] Roldão V, Cunha R, Cabecinhas D, Silvestre C and Oliveira P 2014 A leader-following
trajectory generator with application to quadrotor formation flight Robot. Auton. Syst. 62
1597-609
[6] Lewis MA and Tan K-H 1997 High precision formation control of mobile robots using virtual
structure Autonomous Robot 4 387-403
[7] Balch T and Arkin RC 1998 Behavior-based formation control for multirobot teams IEEE Trans.
Robot. Automa. 14 926-39
[8] Weitz LA, Hurtado JE and Sinclair AJ 2008 Decentralized cooperative-control design for
multivehicle formations J. Guid., Control, Dyna. 31 970-9
[9] Ren W and Beard R 2004 Decentralized scheme for spacecraft formation flying via the virtual
structure approach J. Guid., Control, Dyna. 27 73-82
[10] Mercado DA, Castro R et al 2013 Quadrotors flight formation control using a leader-follower
approach Proc. European control conf. (Zurich) pp 3858-63
[11] Basri MAM, Husain AR and Danapalasingam KA 2015 Stabilization and trajectory tracking
control for underactuated quadrotor helicopter subject to wind-gust disturbance Sadhana 40
1531-53

Acknowledgements
This work was supported by Postgraduate Innovation Research Fund of Beihang University under
Grant YCSJ-01-2016-10.