Download as pptx, pdf, or txt
Download as pptx, pdf, or txt
You are on page 1of 36

“DESIGN FORMATION CONTROL SYSTEM

WITH COLLISION AVOIDANCE FOR


WHEELED MOBILE ROBOTS”
Student Name : DANG DUY TAN
Student Code : 20192067
Major : Control Engineering and Automation
Supervisor : Assoc. Prof. NGUYEN HOAI NAM
Semester : 2023.1
Table of Contents

1. Introduction

2. Preliminaries

3. Control Design

4. Simulation and Experiment Results

5. Conclusions and Future Work


1. Introduction
1.1 Motivation

• In recent years, the field of multi-agent systems control is receiving


increasing attention. Formation control of mobile agents is one of
the main problems in this context

 There are numerous applications related to formation control such Figure 1.1 – A Kangaroo - inspired Swarm of Mobile
Robots (Source: Prof. Trung Dung Ngo, 2024/01,
as control of vehicle platoons, wheeled robots, satellites,….
HUST)

 The advancement in technologies make swarm robot


applications feasible.

Design formation controllers for a multi-robot


system
Figure 1.2 - Swarm of combat UAVs (Source:
Breaking Defense)
1
1. Introduction
1.2 Results

 Proposing a formation scheme for a 2-wheeled multi-robot system with collision avoidance on theory,
comparing with existing controllers.

 Contribution:

• Design of distributed controllers using only neighbors’ information which guarantee formation
control and avoid collisions between agents and obstacles for a group of agents with non-holonomic
dynamics.

3
2. Preliminaries
2.1 Graph Theory
 A directed graph is the set of nodes and x is the set of edges - A representation for a network of
agents.
 The adjacency matrix : if
(1)
otherwise

 The Laplacian matrix :


* : The set of neighbors of node .
(2)

 is an undirected graph if it satisfies: For every edge , there correspondingly exists an edge .
* The adjacency matrix is symmetric, consequently.

 is a connected graph if it satisfies: For each pair of vertices, there exists at least one single path joining
them.

 is an unweighted graph if it has each element . If not, is a weighted graph.

4
2. Preliminaries
2.1 Graph Theory
 Leader-follower case: A system includes + 1 agents with 1 leader and followers
 follower agents - Represented by graph .
 The leader agent - Represented by node 0 and labeled “”.

 An augmented graph – A representation for the leader-follower system:


 The node set is added with the leader’s presence.
 The edge set ) is added with the leader’s relations to followers.

: if follower has a direct connection with the leader,


(3)
otherwise

 Pinning matrix = + (4)

Assumption 1: Undirected graph is connected and at least 1 follower has a direct connection with the leader.

Lemma 1: [29]. If is an undirected and connected graph, and at least one follower has direct a connection
with the leader, such that = 1, then matrix will be symmetric positive definite.

5
2. Preliminaries
2.2 Mobile Robot’s Model
 A multi-robot system consists of +1 () self-balancing mobile
robot 2-wheeled mobile robots.

 The leader robot’s kinematic model – [19]:


 x L t   vL (t ) cos  L t  ,

 y L t   vL (t ) sin  L t  , (5)
 t   w t 
 L L

 The follower robot ’s kinematic model – [19]:


 x i t   vi (t ) cos i t  ,

 y i t   vi (t ) sin i t  , (6)
 t   w t 
 i i
Figure 2.1 - Global and local coordinates of
multi-robot system  ,,,,,

Assumption 2: The signals , are bounded, there exist constant and > 0 such that and .

6
2. Preliminaries
2.3 Artificial Potential Field (APF)
 Basic idea: Construct a virtual potential energy field.

 Components: Attractive APF& Repulsive APF

 Mathematics: Function .

 Utilization: ∇
(7)

 Collision avoidance problem Repulsive APF.

 Common way: Select a distance based repulsive APF function =

(8)

* : Distance between object and object ,


Figure 2.2 - Three-dimensional APF : The collision range,
Visualization [22] : The detection range

7
2. Preliminaries
2.4 Control Objective

 Designing a distributed control law and for all follower mobile robots () so that the formation pattern of the
multi-robot system can be achieved without collisions among agents and obstacles:

1) for

where : Relative position between follower robot and the leader.

2) for , is the index of the object around the agent and .

8
3. Control Design
3.1 Distributed Observers
 Adopting and reforming from [19].
 Neighborhood positions, orientation, linear velocity  The estimation algorithms for robot :
and angular velocity errors of the follower :

(9) (14)

(15)
(10)
(16)

(11)

(12) (17)

(13)
(18)

• = and , , , , , , are positive constants.

9
3. Control Design
3.1 Distributed Observers
Theorem 1: Consider the neighborhood positions, orientation, linear velocity, angular velocity errors (9) - (13),
and the estimation algorithms (14) - (18) with Assumption 1 and Assumption 2. Then, , , , , will converge to the
leader’s states for .

 Proof: Select a candidate Lyapunov function:

1 T T 1 1 1 1
V1  H H x  yT H T H y  T H T H   vT H v  wT H w,
x (19)
2 2 2 2 2

• where: ; ; (: the mutual denotation for , , , , ).

• Further analysis is illustrated in pages 13-15 of the thesis report.

10
3. Control Design
3.2 Repulsive Artificial Potential Field Function 2
    R   2
aij
2

 Select repulsive APF function: Vaij   min 0,  ij 2 2 


 , (20)
   aij  r  
 
• , : Distance and APF function’s gain between object and
object , respectively.

 Repulsive APF function’s gradient components:


0 if  a  R

Vaij  
 R 2  r 2   a2ij  R 2 y  y 
* y  4 ij if r   a  R (21)
 
3 i j
i   2aij  r 2

0 if  a  r
Figure 3.1 - Visualization of detection
0 if  a  R
range and collision range 
Vaij  
 R 2  r 2   a2ij  R 2  x  x 
* x  4 ij if r   a  R (22)
 
3 i j
i   2aij  r 2

0 if  a  r

11
3. Control Design
3.2 Repulsive Artificial Potential Field Function

Assumption 3: The initial configuration of the multi-robot system is set in a way that every robot does not
violate the collision range of any other objects and the setup of environment is unlike cases in the thesis report
(pages 8-11).

Assumption 4: There exists an optimal bound for the gain component such that the value of partial
derivations of repulsive APF function with respect to vertical and horizontal coordinates will be bounded to
generate enough push-back effects, rather than reaching to infinite.

12
3. Control Design
3.3 Distributed Controller Design
 Define the local coordinate tracking errors:
(23)

(24)

(25)

 Integrate with repulsive APF function, the errors in positions:

(26)

(27)

 where is the set of all objects around the robot .


 Using the transformation matrix:
 xEi   cos  sin i 0   E xi 
   i
 
 yEi     sin i cos i 0   E yi  (28)
   0 0 1   E 
  Ei    i

13
3. Control Design
3.3 Distributed Controller Design
 The structure of the proposed controller:
 
vi  vˆLi cos  Ei   i xEi , (29)

wi  wˆ Li  i Ei 
  .y
tanh  Ei
.vˆLi , (30)
E i
Ei

• where , are positive constants.

 The update law for the dynamic gain:

4 ( 𝑅2 −𝑟 2) ( Δ 2𝑎 − 𝑅2 )
𝜍˙ 𝑖𝑗 =
𝑖𝑗

2 3 [𝐸 (𝑥 𝑥𝑗 𝐸𝑗 − 𝑥𝐸 ) + 𝐸𝑦 ( 𝑦 𝐸 − 𝑦 𝐸 ) ] (31)
(Δ 2
)
𝑖 𝑗 𝑗 𝑖

𝑎𝑖𝑗 −𝑟

14
3. Control Design
3.4 Closed-loop Stability Analysis
Theorem 2: Consider the system of robots with models described in (5), (6) with control laws (29), (30),
update laws (14)-(18), (31), and Assumption 1 – 4 are hold, then the closed-loop system errors in (28) are
exponentially bounded and control objective can be achieved.

 Proof: Select the following candidate Lyapunov function:


1 N 2 1 N 2 1 N 2 1 T T 1 1
V2   xEi   yEi    Ei  xH Hx  y T H T Hy  T H T H 
2 i 1 2 i 1 2 i 1 2 2 2 (32)
1 T 1 T 1 N
 vHv  w Hw    ij2 ,
2 2 2 i 1 jSi
• where: , , , .

• Further analysis is illustrated in pages 18-20 of the thesis report.

15
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.1 Formation Control Scenario
 All parameters are in SI units.

Parameters Values
Formation pattern , ,

Leader’s signals ,

Observers ,,

Controller ,
Repulsive APF ,,
function

Figure 4.1 - Connection graph for formation Table 1 – Parameters for formation control simulation
control simulation

16
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.1 Formation Control Scenario

Figure 4.2 - Trajectories of multi-robot system with different control


schemes in formation control simulation

17
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.1 Formation Control Scenario
 Define - [19]
(33)

Figure 4.3 - Comparison results of different controllers in formation control simulation

18
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.1 Formation Control Scenario

Figure 4.4 - Estimated leader’s states in Figure 4.5 - Estimated leader’s inputs in
formation control simulation formation control simulation

19
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.1 Formation Control Scenario
 Define: ; ; ; .

Figure 4.6 - Formation state errors in Figure 4.7 - Formation input errors
formation control simulation in formation control simulation

20
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.2 Formation Control with Collision Avoidance Scenario

Parameters Values

Formation pattern ,,

Leader’s signals ,
Observers ,,

Controller ,

Repulsive APF function ,,


Obstacles' positions ,
Obstacles’ radius

Figure 4.8 - Connection graph for formation Table 2 – Parameters for formation control with
control with collision avoidance simulation collision avoidance simulation

21
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.1 Formation Control Scenario

22
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.2 Formation Control with Collision Avoidance Scenario

Figure 4.9 - Trajectories of multi-robot system in Figure 4.10 - Distances of all pairs of objects in
formation control with collision avoidance simulation formation control with collision avoidance simulation

23
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.2 Formation Control with Collision Avoidance Scenario

Figure 4.11 - Formation state errors in formation Figure 4.12 - Formation input errors in formation
control with collision avoidance simulation control with collision avoidance simulation

24
4. Simulation and Experiment Results
4.1 Simulation Results
4.1.2 Formation Control with Collision Avoidance Scenario

Figure 4.13 - Comparison of static gain and dynamic gain for the
repulsive APF function in formation control with collision avoidance

25
4. Simulation and Experiment Results
4.2 Experiment Results
4.2.1 Experiment Workspace

Figure 4.16 - Main components


of control system

Figure 4.15 - QBot2e

Figure 4.14 - Workspace for experiment

26
4. Simulation and Experiment Results
4.2 Experiment Results
4.2.2 Formation Control with Collision Avoidance Scenario

Parameters Values

Formation pattern ,
Leader’s signals ,
Observers ,,
Controller ,.

Repulsive APF ,,
function
Obstacles’ radius

Figure 4.17 - Connection graph for Table 3 – System’s parameters for experiment
experiment

27
4. Simulation and Experiment Results
4.2 Experiment Results
4.2.2 Formation Control with Collision Avoidance Scenario

28
4. Simulation and Experiment Results
4.2 Experiment Results
4.2.2 Formation Control with Collision Avoidance Scenario

Figure 4.18 -Trajectories of multi-robot Figure 4.19 - Distances of all pairs of objects
system in experiment in experiment

29
4. Simulation and Experiment Results
4.2 Experiment Results
4.2.2 Formation Control with Collision Avoidance Scenario

Figure 4.20 - Estimated leader’s state Figure 4.21 - Estimated leader’s state
errors in experiment errors in experiment

30
4. Simulation and Experiment Results
4.2 Experiment Results
4.2.2 Formation Control with Collision Avoidance Scenario

Figure 4.22 - Formation state errors Figure 4.23 - Velocities of all agents
in experiment in experiment

31
5. Conclusions and Future Work
 Conclusions:
 Studying the problem of Formation control with Collision avoidance for a class of multi-robot system.

 Proposing an integration of Leader-Follower method with a dynamic gain for repulsive APF function.

 Employing Lyapunov theory to prove that the overall system is input-to-state stable.

 Simulating and conducting experiments to verify the proposed control scheme.

 Future Work:
 Extending to different types of mobile robots: Three-wheeled mobile robots; four-wheeled mobile robots,...

 Practical issues: Communication delay, dynamic/ time varying communication graph, presence of
disturbances , input/state saturation, …

 Associating the overall stability of kinematic model control with dynamic model control.

32
References

[4] Gottlieb, Yoav, and Tal Shima, “UAVs task and motion planning in the Presence of Obstacles and Prioritized Targets,”
Sensors 2015, 15, 29734-29764.
[7] Luca, A.D., Oriolo, G. (1995), “Modelling and control of nonholonomic mechanical systems,” Angeles, J., Kecskeméthy,
A. (eds) Kinematics and Dynamics of Multi-Body Systems, CISM International Centre for Mechanical Sciences, vol 360.
Springer, Vienna.
[8] W. Kowalczyk and K. Kozlowski, "Leader-follower control and collision avoidance for the formation of differentially-driven
mobile robots," 23rd International Conference on Methods & Models in Automation & Robotics (MMAR), Miedzyzdroje,
Poland, 2018, pp. 132-137.
[9] Mastellone, S., Stipanović, D. M., Graunke, C. R., Intlekofer, K. A., & Spong, M. W. (2008), “Formation control and
collision avoidance for multi-agent non-holonomic systems: Theory and experiments,” International Journal of Robotics
Research, 27(1), 107-126.
[10] X. Yu and L. Liu, "Distributed formation control of nonholonomic vehicles subject to velocity constraints," IEEE
Transactions on Industrial Electronics, vol. 63, no. 2, pp. 1289-1298, Feb. 2016.
[18] Kwang-Kyo Oh, Myoung-Chul Park, Hyo-Sung Ahn, "A survey of multi-agent formation control," Automatica, Volume 53,
2015, Pages 424-440, ISSN 0005-1098.
[19] Z. Miao, Y. -H. Liu, Y. Wang, G. Yi and R. Fierro, "Distributed estimation and control for leader-following formations of
nonholonomic mobile robots," IEEE Transactions on Automation Science and Engineering, vol. 15, no. 4, pp. 1946-1954,
Oct. 2018.
References

[20] P. Lu, H. Wang, F. Zhang, W. Yu and G. Chen, "Formation control of nonholonomic mobile robots using distributed
estimators,” IEEE Transactions on Circuits and Systems II: Express Briefs, vol. 67, no. 12, pp. 3162-3166, Dec. 2020.
[21] Qun Lu, Zhiqiang Miao, Dan Zhang, Li Yu, Wenjun Ye, Simon X. Yang, Chun-Yi Su, "Distributed leader-follower
formation control of nonholonomic mobile robots," IFAC-PapersOnLine, Volume 52, Issue 15, 2019, Pages 67-72, ISSN
2405-8963.
[22] H. Choset, K. M. Lynch, S. Hutchinson, G. Kantor, W. Burgard, L. E. Kavraki and S. Thrun, “Principles of robot motion:
Theory, algorithms, and implementations,” MIT Press, Boston, 2005.
[26] C. Kuang, L. Li and Q. Geng, "Formation control and collision avoidance of nonholonomic mobile robots," 5th CAA
International Conference on Vehicular Control and Intelligence (CVCI), Tianjin, China, 2021, pp. 1-6.
[29] Lewis, F.L., Zhang, H., Hengster-Movric, K., and Das, A. (2014). “Cooperative control systems: Optimal and adaptive
design approaches”. London, U.K.: SpringerVerlag.
[30] Stipanovic, D. M., Hokayem, P. F., Spong, M. W. and Šiljak, ´ D. D. (2007), “Cooperative avoidance control for
multiagent systems,” Journal of Dynamic Systems, Measurement and Control 129(5): 699–707.
Thank you for listening!
Q&A ?

You might also like