A Distributed Cooperative
Multi–Robot Path Planning
Using Differential Evolution
PRESENTED BY
BRAJA GOPAL PATRA
[Link] COMPUTER SC. & ENGG.
NIT,AGARTALA
Index
Introduction
What is Multi-Robot Motion Planning Problem
Constrains to be satisfied for the problem
Comparison between Centralized & Distributed multi-robot path-
planning
Detection of Next position
Formulation of the objective function
Representing the first constraint as a penalty
Possible Approaches
Differential Evolution Algorithm
Pseudo code
Some of experimental results by researchers
Application
Conclusion
Bibliography
Introduction
The late 1990s have seen a significant
progress in mobile robotics. Path planning is
regarded as a fundamental problem in mobile
robotics.
The path planning problem attempts to
determine a trajectory of motion for the
robots from a predefined starting point to a
given goal point without colliding with any
obstacle as well as the other robots in the
map.
The path-planning problem in the given
context attempts to minimize the total
distance traveled by the robots in the given
workspace, subject to the constraint that the
robots do not hit each other or the static
obstacles.
The multi-robot path planning is an NP-
complete problem, as no polynomial time
algorithm to solve the problem is known at
this time.
What is Multi-Robot Motion Planning Problem
Constrains to be satisfied for the problem
Collision
Collision
Collision between two robots Collision with obstacle
Comparison between Centralized & Distributed
multi-robot path-planning
In the centralized approach the objective functions
and the constraints for path/motion planning of
all the robots are considered together, where as
the distributed approach divides the complexity of
centralized path planning into problems of small
complexity to be shared by the robots.
Consequently, in distributed planning, each robot
attempts to construct its path almost
independently, avoiding collision with static
obstacles or teammates engaged in path planning.
Detection of Next position
Goal (xig, yig)
Next Position (xi’, yi’)
Initial Position (xi, yi)
Velocity (Vi) Steering Angle
Formulation of the objective function
• Objective Function (F) is the total path length traversed
by the robot to reach its goal position.
(xig , yig)
Optimum
Path
t acle
Next Position s
(xi , yi) Ob
(xi , yi)
f = Distance (current-next) + Distance (next-Goal)
Re-formulation of the objective
function using steering angle
Evaluation of the next position of the robot from its current position so that
the total path should be minimum.
Next Position
yi (xi , yi)
Vi
(xi , yi) i
yi x′i = xi + vi cos i t
Current Position
y′i = yi + vi sin i t
x
xi xi
Representing the first constraint as a
penalty
Robot 1 Robot 2
r r
d=
>2r
When (d = 2r) collision occurs
Penalty Function =
Possible Approaches
1. Centralized Planning: A manager globally planes the on-line
trajectories for robots.
2. Distributed Planning: n manager, one for each robot plane the
on-line trajectories for robots.
Differential Evolution Algorithm
In 1995, Storn and Price proposed an alternative
form of EA that includes selection, differential
mutation and recombination to generate trial
vectors for the next iteration.
It begins with a randomly initialized population
of D-dimensional real-valued parameter
vectors. Each vector forms a candidate solution
to the multi-dimensional optimization problem.
Position of all robots and obstacle
Where rand (0, 1) is a uniformly distributed
random number lying between 0 and 1.
Where and are the lower and upper
bound respectively.
Now mutant vector
Trial vector is
(CR=0.5x(1+rand(0,1))
DE uses the Darwinian principle of “survival of
the fittest” in its selection process which may
be expressed as
Pseudo Code
Input: Initial position ,goal position and
velocity Vi for n robots where 1 ≤ i ≤ n and a threshold
value ε.
Output: Trajectory of motion P for each robot R from
i i
to
Begin
Set for all robot i
and
For robot i = 1to n
Repeat
Call DE ( , pos-vector)
// pos-vector denotes current position of all robots and
obstacle//
Move-to( );
Until
// curr _ i =( ), Gi = //
End for;
End.
Procedure DE ( , pos-vector)
Begin
Initialize population.
For Iter = 1 to Max iter do
Begin
Create trial vector using equation.
Evaluate fitness.
If the trial vector is better than its target vector
Then replace the target by trial in the next
generation;
End If;
End for;
Update:
;
;
Return;
End.
Some of experimental results by
researchers
Fig: Comparison of training in Distributed and centralized approach
• Researchers have seen that Distributed approach
converges faster than the centralised approach.
In case of centralized approach converges very
poorly.
Application
Use of a team of robots can help in monitoring,
surveillance, and search and rescue operations, thus
removing the need for human intervention in
dangerous areas.
A simple example is exploration and search of an
earthquake-hit building where each robot has a
sensor(s) that can detect heat, light, sound, or
other, and communicate wirelessly with other
robots. Material handling and bomb detections are
several other such aspects where multiple robots
can co-ordinate among themselves to achieve
required goal.
Conclusion
In this topic I have covered the issue of multi-robot
motion planning by DE, and solved the problem by
both centralized and distributed approaches.
Since n parallel DEs have been employed to handle
the distributed motion-planning problem of n
robots, the run time required for the proposed
distributed realization is small in comparison with
other classical algorithms.
The distributed DE-based algorithm was also
found to be comparable to or better than PSO-
based path planning.
BIBLIOGRAPHY
[1] Jayasree Chakraborty, Amit Konar, Uday K. Chakraborty
and L.C. Jain “Distr ibuted Cooperative Multi–Robot Path
Planning Using Differential Evolution”, 2009 IEEE
Congress on Evolutionary Computation (CEC 2009)
[2] Yuan-qing qin, De-Bao Sun, Ning li, Yi-Gang Cen,”Path
planning for mobile robot using the Particle swarm
optimization with Mutation operator” Proceedings of the
Third International Conference on Machine Learning and
Cybemeacs, Shanghai, 26-29 August 2009
[3] LI Gao-yang, LIU Ming-guang,” The Summary of
Differential Evolution Algorithm and its Improvements”,
2010 3rd international Conference on Advanced
Computer Theory and Engineering (iCACTE)
