Sunteți pe pagina 1din 53

Optimal Path Planning Of

Mobile Robots Using


Computational Algorithms
Outline
Introduction
Problem Statement
Objectives
Literature Review
Theory
Algorithms
Research Methodology
Research Utilization
Results
Recommended Further Work




Introduction
Optimization methods in Mathematics (both stochastic and
deterministic) are used to find optimal strategies in complex
problems.
Strategies which minimize or maximize functions or
quantities such as trajectory f(x,y,z,t), distance travelled, time
or energy consumed require Variational Calculus in
Deterministic Methods or Stochastic Methods based on
random search techniques.
Contd. Introduction(2)
Optimal strategies for mobile robots are computed by fields
extending into Mathematics (optimization) and Robotics.
Mobile Robots (MR) are electromechanical systems controlled
either externally or are autonomous based on embedded
systems in onboard computers.
The field of Optimal Path Planning in mobile robots is an
active interdisciplinary research area covering Mathematics,
Engineering and Computer Science.
Problem Statement
Generally optimal path planning of mobile robots is achieved by
two methodologies viz Stochastic and Deterministic.
Stochastic Methods are based on laws of probability for sampling
of random events and have the advantage that they can handle
large problems.
Deterministic Methods are based on Variational Calculus for
determining optimal trajectories and optimization techniques for
Multivariable Calculus.
Contd. Problem Statement
Mobile robots move in
environments which can be
static or dynamic.
In the static environment
shown in the Fig., the MR is
required to move from the
start point to the end point
(Goal) on permissible paths
while avoiding collisions with
the obstacles and minimizing
the total distance travelled,
time taken or energy
consumed.

Contd. Problem Statement
The mathematical formalism is encoded into a
computational algorithm which is embedded or burnt
into the onboard computer system of the MR as shown in
the Fig
http://www.arrownac.com/solutions/beinmotion/
Objectives
This research aims at computing the optimal trajectory of a
MR in a static or dynamic environment.
The methodologies used are:
A Stochastic Evolutionary technique viz Genetic Algorithms (GA),
and The Random Particle Optimization (RPO) based on
interaction potentials .

Contd. Objectives
The GA method is applicable to both static and dynamic
environments. In static environments all coordinates (MR,
obstacles, goal) are input into the onboard computer system
of MR while in dynamic environment sensors need to be used.
The RPO method is also applicable to both static and dynamic
environments as it is a sensor- based approach in which the
positions of obstacles and/or target may vary with time.
Contd. Objectives
The algorithms developed are intended for implementation,
by engineers, in two mobile robots.
Literature Review
The literature reviewed consisted of
(i) Books
(ii) Research Papers
(iii) Dissertations (MS)

Contd. Literature Review
Books
[B1] J C Latombe Robot Motion Planning 1991 (soft copy available)
[B2] J.-P. Laumond, Editor, Robot Motion Planning and Control,
Springer-Verlag London Limited 1998. (soft copy available)
[B3] M Thoma, Editor, Lecture Notes in Control and Information
Sciences (soft copy available)
[B4] Steven M. LaValle, Planning Algorithms, Cambridge University
Press, 2006 (soft copy available)
[B5]J.H.Holland. 1975. Adaptation in Natural and Artificial Systems,
Ann Arbor: University of Michigan Press.
[B6] Randy L. Haupt and Sue Ellen Haupt, Practical Genetic
Algorithms, Second Edition, John Wiley & Sons Inc., 2004.
[B7] Ulrich Bodenhofer, Genetic Algorithms: Theory and
Applications, Lecture Notes, Second Edition WS 2001/2002,
Institut fr Algebra, Stochastik und wissensbasierte mathematische
Systeme, Johanne Kepler Universitt, A-4040 Linz, Austria


Contd. Literature Review
Research papers
[P1] Hagen Burchardt, Ralf Salomon, Implementation of Path Planning using Genetic Algorithms on Mobile
Robots
[P2] I. I. Mahmoud, M. Salama and A, Abd El Tawab, Implementation of Hardware Genetic Algorithm,
Proceedings of ICEENG, 2008, Cairo, Egypt.
[P2] Tamilselvi, Mercy Shlinie, Hariharasudan, Optimal Path Selection for Mobile Robot Navigation Using
Genetic Algorithm, IJCSI International Journal of Computer Science Issues, Vol. 8, Issue 4, No 1, July 2011
ISSN (Online): 1694-0814 www.IJCSI.org
[P3] P. Raja and S. Pugazhenthi, Optimal path planning of mobile robots: A review, International Journal of
Physical Sciences, Vol. 7(9), pp. 1314-1320, 23 February, 2012.
[P4] Nagib, G. and Gharieb, W., 2004, Path planning for a mobile robot using genetic algorithms,
Proceedings of the International Conference on Electrical, Electronic and Computer Engineering
(ICEEC04), Cairo, Egypt, pp.185-189.
[P5] C. Liu, H. Liu and J. Yang, A Path Planning Method Based on Adaptive Genetic Algorithm for Mobile
Robot, Journal of Information & Computational Science 8: 5 (2011), pp 808-814. (Available at
www.joics.com)
[P6] Sedighi, K.H., Ashenayi, K. ; Manikas, T.W. ; Wainwright, R.L. and
Heng-Ming Tai, Autonomous local path planning for a mobile robot using a genetic algorithm, Volume 2,
pp 1338-1345.
B. Mohajer, K. Kiani, E. Samiei and M. Sharifi, A New Online Random Particles Optimization Algorithm for
Mobile Robot Path Planning in Dynamic Environments, Mathematical Problems in Engineering, Volume
2013, Article ID 491346, January 2013
Contd. Literature Review
Dissertation
MS (Mobile robot path planning optimization in static and
dynamic environment By Ahmed Elshami)
Theory
Theory review of

Genetic Algorithm Method
Random Particle Optimization Method
Theory: Genetic Algorithms
Genetics
Gene
Chromosome
Fitness
Selection: Next
Generation
Cross-over
Mutation
Convergence
Optimization In Path Planning
Path Point
Path
Objective function
Selection: Best Paths
Cross-over of paths
Mutation of Path point
Convergence
Algorithm used: GA
Create Environment
Input Start Point, End-
Point & Obstacle Data
Generate Population
Calculate fitness of
Each chromosome
Select best
chromosomes
Cross-over
Mutation
Obtain next
Generation
Print Optimal
Solution
Converged ?
No
Yes
GA Program

A chromosome
consists of path
points from Start to
End

Each path has an
associated length
Or distance

Objective is to find
the optimal path
i.e. the shortest
path
GA: Create Environment
The Environment is created by defining the workspace i.e. the 2D min
and max of the coordinates (x,y); there are 7 obstacles and path points
labeled 0-15 i.e. 16 path points; Starting Position is 0 and End-Point is 15
Nagib, G. and Gharieb, W., 2004, Path planning for a mobile robot using genetic algorithms, Proceedings of the
International Conference on Electrical, Electronic and Computer Engineering (ICEEC04), Cairo, Egypt, pp.185-189.
Input Data
NPTS=16; NOBS=7; % Number of obstacles
i x(i) y(i) --Link Points-- sizeL
1 1.00 7.00 2 4 5 3
2 1.00 11.00 1 3 2
3 3.00 14.00 2 6 12 3
4 3.00 1.00 1 7 9 3
5 5.00 8.00 1 6 7 3
6 6.00 11.00 3 5 11 3
7 6.00 4.00 4 5 8 3
8 8.00 4.00 7 10 14 3
9 10.00 1.00 4 14 2
10 10.00 7.00 8 11 16 3
11 10.00 11.00 6 10 12 3
12 11.00 14.00 3 11 13 3
13 13.00 12.00 12 16 2
14 12.00 20.00 8 9 15 16 4
15 14.00 3.00 14 16 2
16 14.00 8.00 10 13 14 15 4

Fitness of Each Chromosome
The fitness of a path is the inverse of the length
of the path from starting point to end-point

Fitness(iPath)=1.0/DistanceInPath(iPath)
Chromosome Length
No. of chromosomes: NC=20
For each iteration select 40% of these i.e. 8 chromosomes

No. of Bits NBITS = log NPTS /log 2 = log 16 / log 2 = 4
If there are 1024 points, then NBITS = log 1024/log 2 = 10

Chromosome Length
CHR_LEN=(NOBS+2)*NBITS = (7+2)*4 = 36 for 7 obstacles and 4 bits
CHR_LEN= . . . = (7+2)*10=900 for 7 obstacles and 10 bits

Selection of Path Points
A Path is generated by selected points from a specified start-
point to a specified end-point

Starting Point
Consider each link point
Find distance from each link point to end-point
Calculate the Probability of moving to each link point (PDF and
CDF)(means moving of obstacle)
Generate a random number in (0,1)
Select the next point in path (based on the random number
generated)
Continue till end point is reached

Generate Initial Population
I PP Distance Fitness Path Points
1 7 50.7594 1.9701e-002 1- 5- 7- 4- 9- 14- 16-
2 8 25.9620 3.8518e-002 1- 2- 3- 6- 11- 12- 13- 16-
3 6 22.5571 4.4332e-002 1- 2- 3- 12- 13- 16-
4 7 56.6506 1.7652e-002 1- 4- 9- 14- 8- 10- 16-
5 6 17.9749 5.5633e-002 1- 5- 7- 8- 10- 16-
: : : :
: : : :
: : : :
15 7 23.9713 4.1717e-002 1- 2- 3- 6- 11- 10- 16-
16 10 41.9620 2.3831e-002 1- 2- 3- 12- 11- 6- 3- 12- 13- 16-
17 5 44.5951 2.2424e-002 1- 4- 9- 14- 16-
18 6 54.5468 1.8333e-002 1- 4- 9- 14- 15- 16-
19 6 17.9749 5.5633e-002 1- 5- 7- 8- 10- 16-
20 6 22.5571 4.4332e-002 1- 2- 3- 12- 13- 16-

PP = Number of Path Points
Select Best Chromosomes
---- Iteration No. 1 ---------------

NPth SelPth Distance Fitness . . . . . P a t h . . . .

1 5 17.9749 5.5633e-002 0000 0100 0110 0111 1001 1111 0-4-6-7-9-15-
2 6 17.9749 5.5633e-002 0000 0100 0110 0111 1001 1111 0-4-6-7-9-15-
3 14 17.9749 5.5633e-002 0000 0100 0110 0111 1001 1111 0-4-6-7-9-15-
4 19 17.9749 5.5633e-002 0000 0100 0110 0111 1001 1111 0-4-6-7-9-15-
5 7 19.4085 5.1524e-002 0000 0100 0101 1010 1001 1111 0-4-5-10-9-15-
6 9 19.4085 5.1524e-002 0000 0100 0101 1010 1001 1111 0-4-5-10-9-15-
7 10 19.4085 5.1524e-002 0000 0100 0101 1010 1001 1111 0-4-5-10-9-15-
8 13 20.2959 4.9271e-002 0000 0011 0110 0111 1001 1111 0-3-6-7-9-15-

Note: The maximum chromosome size was calculated as 36; the above paths have 6
path points, and each path point is represented by 4 bits; therefore length of each
chromosome is 24
Cross-over
Consider two chromosomes from paths 4 and 5

4 19 17.9749 5.5633e-002 0000 0100 0110 0111 1001 1111 0-4-6- 7-9-15-
5 7 19.4085 5.1524e-002 0000 0100 0101 1010 1001 1111 0-4-5-10-9-15-

0000 0100 0110 0111 1001 1111 0 4 6 7 9 -- 15
0000 0100 0101 1010 1001 1111 0 4 5 10 9 -- 15

After cross-over
0000 0100 0110 1010 1001 1111 0 4 6 10 9 15
0000 0100 0101 0111 1001 1111 0 4 5 7 -- 9 15

Note: The cross-over has to be checked against the permissible list of link points and
new paths are required to be feasible i.e. connected.
Mutation
Consider the path point

0000 0100 0101 0111 1001 1111

0 4 5 7 -- 9 15

If a bit between the start-point and an
end-point is chosen at random e.g.
the 10
th
bit point 1 and it is mutated
to bit 0, then the path reads

0000 0100 0001 0111 1001 1111

And the path has now become

0 4 1 7 9 15
If this is feasible i.e. connected,
then it can be chosen
Obtain Next Generation
Two methods are used for selection of parents

i- Random selection
ma = ceil(NCsel*rand(1,NCsel)); % mothers
pa = ceil(NCsel*rand(1,NCsel)); % fathers

e.g. The command: ma = ceil(8*rand(1,8)) gives
6 4 2 4 2 1 7 5
Ceil is a matlab command ceil(0.9)=1, ceil(7.2)=8

ii- Rank Weighting
Selection probability of a parent is in proportion to the fitness; thus a PDF and
a CDF is constructed and a random number (0,1) is used to select a
parent

Convergence and Optimal Solution
When the paths are all the same
and no further improvement is
possible by cross-over or
mutation, the solution is said to
have converged

Here, the converged solution
which is the optimal path is

0 4 6 7 9 15

with a distance of 17.9749


Summary
Genetic Algorithm is an Evolutionary Method

An optimal solution is obtained by generating an initial population of
chromosomes (paths consisting of path points), computing their fitness to
select best parents for producing the next generation of chromosomes by
carrying out evolutionary procedures of cross-over and mutation. The
procedure is continued until no further improvement is possible; this is called
convergence which corresponds to an optimal solution.

The optimal solution in the case of path planning of mobile robots can be
refined by specifying more path points i.e taking longer chromosomes to get
better accuracy.
GA: Work carried out
A GA algorithm was developed and validated against a
paper with 7 obstacles and 16 path points given in this
presentation.

The computational efficiency of the GA method for
obtaining an optimal solution was investigated.
Research Utilization: GA
The GA algorithm developed in this research can be used for
implementation in a Mobile Robot to steer through an
environment, both static and dynamic, consisting of
obstacles.

Possible environments are:

i- warehouse movement of goods
ii- automated car parking system
iii- route-planning for large problems
Theory: Random Particle
Optimization
Random Particle Optimization (RPO) is an online optimal
local path planning of mobile robots in dynamic
environments sensor-based updates

Environment consists of barriers representing obstacles
as repulsive Coulomb barriers expressed as Gaussian
potentials with specified intensity and width. Similarly,
the target is modeled as an attractive potential.

An on-line decision for the mobile robot involves
evaluation of cost functions which are strongly
dependent on the potential parameters specified.
(imp)Proposed Algorithm: RPO
Create Environment
Input Start Point, End-
Point & Obstacle Data
Set cost function params
Detect Obstacles
Gen. particles
Evaluate Cost Fns
Select best point
Move MR to this point
Obstacles
found ?
Yes
Target
Reached ?
Glide to next position
No
No
Yes
End
Start
Attractant-Repulsive Potentials
Obstacles are represented by repulsive potentials, which must be avoided, and
the target is represented by an attractive potential.

Fig. shows the four obstacles with = 1, = 1 in the potential

()
=

exp (

2
)
and

()
=

exp (

2
)
Continued Attractant-Repulsive Potentials
Obstacles (4) and the target with
()
= 1,

= 1 and
()
= 1,

= 1
MR Trajectory
The Glide Phase
Initially, MR will move in a straight line from start-point to End-point, unless and
until the sensor detects an obstacle

The Maneuver Phase
The MR enters the maneuver phase
to go around the obstacle, as
shown in the Fig. below with the
robot (R), obstacle (O) and target(T)
in a line.
2 2.5 3 3.5 4 4.5 5 5.5 6
2
2.5
3
3.5
4
4.5
5
5.5
6
1
2 3
4
5 6
R
O
T
2-D grid with robot position R a
t=0 and six artificial generated
points at 60 deg each, on a circle of radius Ct.
Maneuver Phase
Artificial particles are placed, at equal angle intervals, on a circle of radius



Selection Criteria for point
Points = 1,2,3,
Cost functions:
distance error

+ =

() < 0 for particle closer to target


potential error

+ = (

+ ) (

()) < 0 for particle in attractive


zone
where
+ refers to the artificial particles
refers to the present position of the robot

+ the distance from the artificial particle to the target



cost function is a repellant-attractive Gaussian function for the obstacle
()
and
target
()
respectively, and =
()
+
()
.

The procedure is then to move the robot to the best point found viz one for
which the distance error is smallest and negative and the cost function is
negative.
Algorithm Development
for Maneuver Phase
Step 1: The x,y position of the robot, obstacle and target are specified:
x y
ROBOT 4.00 3.00
OBS 4.50 4.00
TARGET 5.00 5.00

Step 2: The step size is: Ct = 0.4472. (0.4*Dist between MR and Obst)

Step 3:
% Cost function parameter
alpObs = 1.0; % height of repellant
muObs = 4.0; % width of repellant
alpGoal = 1.0; % depth of attractant
muGoal = 4.0; % width of attractant
Alg. Dev. For Mnvr. Phase (2)
Step 4:
NPTS=6; stepDeg=360/NPTS;

Step 5:
JobstT = exp(-4.0*( (xc-xo(1))^2+ (yc-yo(1))^2 ) );
JgoalT =-exp(-4.0*( (xc-xT )^2+ (yc-yT )^2 ) );
JT = JobstT + JgoalT;

Step 6:
DTG = sqrt( (xc-xT)^2 + (yc-yT)^2 );

Step 7:
% generate bacteria at time t+ delta T
theta(1)=0; % starting angle for first point on circle (in degrees)
for ib=1:NPTS
% x, y points of artificial points generated on circle of radius
x(ib) = xc + Ct* cos(pi*theta(ib)/180);
y(ib) = yc + Ct* sin(pi*theta(ib)/180);

Alg. Dev. For Mnvr. Phase (3)
% cost function for artificial point
Jobst(ib) = exp(-4.0*( (x(ib)-xo(1))^2 + (y(ib)-yo(1))^2 ) );
Jgoal(ib) =-exp(-4.0*( (x(ib)-xT )^2 + (y(ib)-yT )^2 ) );
% total cost function
J(ib) = Jobst(ib) + Jgoal(ib);

% error of cost function (difference between J for artificial point and present point of
robot location
errJ(ib) = J(ib) - JT;

% error of distance (distance from artificial point distance from present point of robot
location)
DTGb(ib) = sqrt((x(ib)-xT)^2 + (y(ib)-yT)^2 );
errD(ib) = DTGb(ib) - DTG;

% go for the next point
theta(ib+1) = theta(ib)+stepDeg;
end
Alg. Dev. For Mnvr. Phase (4)
i x y Jobst Jgoal J errJ errD
1 0 4.4472 3.0000 1.8113e-002 -3.3148e-008 1.8113e-002 1.1375e-002 -0.16
2 60 4.2236 3.3873 1.6412e-001 -2.7214e-006 1.6411e-001 1.5737e-001 -0.45
3 120 3.7764 3.3873 2.7432e-002 -7.6034e-008 2.7432e-002 2.0694e-002 -0.21
4 180 3.5528 3.0000 5.0606e-004 -2.5876e-011 5.0606e-004 -6.2319e-003 0.23
5 240 3.7764 2.6127 5.5852e-005 -3.1518e-013 5.5852e-005 -6.6821e-003 0.45
6 300 4.2236 2.6127 3.3414e-004 -1.1281e-011 3.3414e-004 -6.4038e-003 0.27
2 2.5 3 3.5 4 4.5 5 5.5 6
2
2.5
3
3.5
4
4.5
5
5.5
6
1
2 3
4
5 6
R
O
T
The best point out of 6 points above
is the one with errD<0 and errJ<0

Since none of the 6 points meets
these conditions, more points are
needed

Alg. Dev. For Mnvr. Phase (5)
i theta x y errJ errD
1 66.00 4.1819 3.4085 1.5790e-001 -0.45
2 60.00 4.2236 3.3873 1.5737e-001 -0.45
3 72.00 4.1382 3.4253 1.5135e-001 -0.44
4 54.00 4.2629 3.3618 1.4985e-001 -0.44
: : : : : :
25 138.00 3.6677 3.2992 2.0408e-003 -0.08
26 348.00 4.4374 2.9070 1.5406e-003 -0.07
27 144.00 3.6382 3.2629 -9.0533e-004 -0.03
28 342.00 4.4253 2.8618 -1.2449e-003 -0.02
29 150.00 3.6127 3.2236 -2.8905e-003 0.02
30 336.00 4.4085 2.8181 -3.1169e-003 0.02
: : : : : :
59 240.00 3.7764 2.6127 -6.6821e-003 0.45
60 246.00 3.8181 2.5915 -6.6823e-003 0.45

New point found:
Index= 27 theta = 144.00 x = 3.64 y = 3.26
theta x y DistanceToNextPt Distance2 TotalDistance
144.00 3.64 3.26 0.4472 2.2073 2.6545
Alg. Dev. For Mnvr. Phase (6)
2 2.5 3 3.5 4 4.5 5 5.5 6
2
2.5
3
3.5
4
4.5
5
5.5
6
R
O
T
d
1
d
2
X
Y
Maneuver with 60 points on circle
Alg. Dev. For Mnvr. Phase (7)
Table 1: Effect of sample size on the best point selected


10 144.00 3.64 3.26 0.4472 2.2073 2.6545
20 144.00 3.64 3.26 0.4472 2.2073 2.6545
30 144.00 3.64 3.26 0.4472 2.2073 2.6545
40 144.00 3.64 3.26 0.4472 2.2073 2.6545
50 144.00 3.64 3.26 0.4472 2.2073 2.6545
100 144.00 3.64 3.26 0.4472 2.2073 2.6545
500 344.88 4.43 2.88 0.4472 2.1916 2.6388
1000 344.88 4.43 2.88 0.4472 2.1916 2.6388
5000 141.91 3.65 3.28 0.4472 2.1910 2.6382
10000 141.91 3.65 3.28 0.4472 2.1910 2.6382

The minimum total distance travelled by the MR is thus 2.6382 units
Arbitrary Maneuver
0 2 4 6 8 10
0
2
4
6
8
10
1
2
3
4
5
6
7
*
*
x
y
Determination of Optimal Path
-2 0 2 4 6 8 10 12
-2
0
2
4
6
8
10
12
1
2
3
4
5
6
7
*
*
x
y
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
1
2
3
4
5
6
7
*
*
Start
Target
Obstacle
Swing and RPO from Start to Target
Determination of Optimal Path (2)
A close-up view of the trajectory near obstacles 5 and 1 is shown in Fig below
where the curvature of the trajectory is evident in the RPO. This can be
compared with the larger distance travelled in the swing phase.
2 2.2 2.4 2.6 2.8 3 3.2 3.4 3.6 3.8
1
1.5
2
2.5
3
3.5
4
4.5
1
2
3
4
5
6
7
*
*
x
y


RPO
Swing Trajectory
Obstacle
Determination of Optimal Path (3)
The final result of this analysis is that the straight-line
distance of 14.1421 without obstacles is 16.5820 in the
arbitrary swing strategy and 14.4214 in the optimal
RPO trajectory.
Research Summary
This research focused on two methods for optimal path planning of mobile
robots in both static and dynamic environments viz (i) Genetic Algorithms
applied to static environments, and (ii) Random Particle Optimization applied
to both static and dynamic environments.

The limitations of research carried out in this dissertation are:

A mobile robot has been assumed to be a point robot i.e. its size has not been
considered, and
A holonomic robot has been considered i.e. one which has all degrees of
freedom and can turn to any direction, unlike a car, from its present position.
CONCLUSIONS (GA)
Efficiency of the algorithm, i.e. less computational effort is dependent on the
choice of model parameters such as the sample size, length of chromosome, and
the probability distribution specified for determination of parents as well as in
the offspring.
Genetic Algorithms for path planning was found to provide a converged optimal
path fast enough for it to be used efficiently for movement of a robot;

One notable difference observed between GA for path planning and GA for
optimization of functions is that schemes such as cross-over and mutation can be
carried out readily for the case of variables provided they remain within the
specified constraints for the optimization problem.
In path planning however, cross-over and mutation are applicable also subject to
the constraint but a more restrictive one in which the connectibility of the paths
will be the deciding factor e.g. a robot can not fly over an obstacle.

GA, being an evolutionary method, is essentially based on stochastic procedures
and should be favored when the problem is very large. This remains its best
advantage.
CONCLUSIONS (RPO)
The sample size N should be greater than ~ 20 for
sufficient accuracy
The magnitude of the repulsive barrier
()
does not
play a significant role as the robot is already in close
proximity.
The width of the repulsive barrier
()
can be of use
for obstacles of large size.
The magnitude of the attractive potential
()
should be
orders of magnitude higher than that for the repulsive
barrier due to the Gaussian exponential, especially for
large distance of the robot from the target.
Recommended Further Work
The GA code developed and investigated in this research is
ready to be implemented in a static environment. In that case
the computational efficiency of the hardware needs to be
established before it is applied to a dynamic environment.
Similarly, the RPO code developed and investigated in this
research is ready to be implemented in both static and
dynamic environments.
Thank You

S-ar putea să vă placă și