Coordinative Behavior in Evolutionary Multi-agent System by

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
1

International Graduate

School of Dynamic

Intelligent Systems


Coordinative Behavior in
Evolutionary Multi
-
agent System by
Genetic Algorithm

(T. Shibata and T. Fukuda)

Chuan
-
Kang Ting

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
2

International Graduate

School of Dynamic

Intelligent Systems


Path Planning for Multiple Robots

1

2

3

4

5

6

7

8

9

10

2

1

4

3

6

5

8

7

10

9

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
3

International Graduate

School of Dynamic

Intelligent Systems


Selfish
-
planning

1

2

3

4

5

6

7

8

9

10

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
4

International Graduate

School of Dynamic

Intelligent Systems


Collision & Deadlock

1

2

3

4

5

6

7

8

9

10

Selfish

Coordination

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
5

International Graduate

School of Dynamic

Intelligent Systems


Setting

•
Subject:
path planning of multiple robots

•
Purpose:

optimize the path of each robot to avoid
collisions, reduce waiting time, and decrease tour
length

•
Conditions:

no global communication between
robots but local sense to surrounding

•
Method:

apply genetic algorithm to optimize
motion planning of each robot

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
6

International Graduate

School of Dynamic

Intelligent Systems


Evolution in Nature

Chromosomes

Mutation

Crossover

Mating

Survival of the Fittest

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
7

International Graduate

School of Dynamic

Intelligent Systems


Genetic Algorithms

Genetic Algorithm

()

{


t := 0;


init_population

P(t);


evaluate

P(t);



do {



P’(t) :=
select_parents

P(t);



crossover

P’(t);




mutate

P’(t);



evaluate

P’(t);



P(t+1) :=
survive

P(t), P’(t)



t := t + 1;


} while not terminated;

}

Initialization

Mutation

Evaluate

& Survive

Crossover

Selection

Terminated?

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
8

International Graduate

School of Dynamic

Intelligent Systems


Chromosome
-

background

•
MAKLINK

–
Construct the available free space between obstacles

–
Based on free links: whose two ends are either corners of polygonal
obstacles or a point on the working
-
space boundary

–
Every node can allow only one robot at the same time

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
9

International Graduate

School of Dynamic

Intelligent Systems


Chromosome
-

representation

•
Chromosome

–
An encoded expression of
potential solutions. (usually
in the form of string)

•
Path representation

–
Encode the order of visiting
points as a string

•
Example

–
0
-
1
-
2
-
6
-
9
-
11
-
15
-
17
-
22
-
25
-
26

–
0
-
4
-
8
-
10
-
15
-
20
-
23
-
26

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
10

International Graduate

School of Dynamic

Intelligent Systems


Chromosome
-

population

•
Population
–

a set of chromosomes

•
Chromosomes are generated randomly. i.e. the
population contains a set of random solutions

•
Each robot is equipped with one population


0
-
1
-
2
-
6
-
9
-
11
-
15
-
17
-
22
-
25
-
26

0
-
4
-
8
-
10
-
15
-
20
-
23
-
26

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
11

International Graduate

School of Dynamic

Intelligent Systems


Operator
-

Crossover

•
Exchange information from two selected
chromosomes

–
Crossover point: the first identical node in both parents

–
Example:

Parent1:
0
-
1
-
2
-
6
-
9
-
11
-
15
-
17
-
22
-
25
-
26

Parent2:
0
-
4
-
8
-
10
-
15
-
20
-
23
-
26


Child1:
0
-
1
-
2
-
6
-
9
-
11
-
15
-
20
-
23
-
26

Child2:
0
-
4
-
8
-
10
-
15
-
17
-
22
-
25
-
26

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
12

International Graduate

School of Dynamic

Intelligent Systems


Operator
-

Mutation

•
Mutation rate is low (for
escaping local optimum)

1.
Randomly selects one gene
(node) to mutate

2.
Its following nodes are
picked randomly from
sequentially connected
nodes

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
13

International Graduate

School of Dynamic

Intelligent Systems


Evaluation
-

Fitness

•
Fitness = Tour Length + Waiting Time





(Static)


(Dynamic)

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
14

International Graduate

School of Dynamic

Intelligent Systems


Evolution Flowchart

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
15

International Graduate

School of Dynamic

Intelligent Systems


Experimental Results (1)

•
Optimize the path for single robot (selfish
-
planning)

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
16

International Graduate

School of Dynamic

Intelligent Systems


Experimental Results (2)

•
Fitness variation of multiple robots

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
17

International Graduate

School of Dynamic

Intelligent Systems


Experimental Results (3)

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
18

International Graduate

School of Dynamic

Intelligent Systems


Experimental Results (4)

1

2

3

4

5

6

7

8

9

10

1

2

3

4

5

6

7

8

9

10

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
19

International Graduate

School of Dynamic

Intelligent Systems


Summary

•
The paper proposes a strategy of coordinative
motion planning for multi
-
agent systems.

•
Without global communication, the robot uses
Genetic Algorithm (GA) to search the feasible
solutions, which consider the tour distance and
waiting time at the same time.

•
Simulation results demonstrate this strategy can
avoid collisions and achieve coordination of
multiple robots.

Coordinative Behavior in Evolutionary Multi
-
agent System
by Genetic Algorithm

Chuan
-
Kang Ting

–

Page:
20

International Graduate

School of Dynamic

Intelligent Systems


Thank you