Professional Documents
Culture Documents
Fyp Robotic Poster
Fyp Robotic Poster
Faculty of Engineering
Performance Comparison Of Multi-query Sampling Based Path
Planning Algorithms In Simulated Environments
GOH WEI YANG
Abstract
163271
Results
Several multi-query sampling based algorithms had been introduced over past
decade but there is no complete comparison about the performance for those
algorithms with a clear classification. This thesis is focus on the comparison for
performance of 6 algorithms in 5 different kinds of environment. To do so,
algorithms were converted into MATLAB codes and put into environments for
simulation. A scheme was made to judge the performances. The scheme are biased
toward the success rate and the rest are the same percentage. The results were
defined separately based on the types of environment. Obstacle based probabilistic
roadmap obtained the best performance in simple environments (77.5%). Low
dispersion probabilistic roadmap obtained highest of success rate in 14 out of 20
environments (70%) and low deviation of path length in 12 out of 20 environments
(60%). This results the algorithm had the best performance in maze environments
(67.75%) and mixed environments (69.00%). Meanwhile, optimal probabilistic
roadmap had the best performance in bug trap environment (69.25%). Gaussian
sampling had the best performance in narrow passage environment (64.50%).
Introduction
Gaussian sampling
Problem Statements
No research make simulations by using several multi-query sampling based
algorithms in different kinds of environments.
No standardize variables for comparing performance of algorithms
Objectives
Simulate 6 multi-query sampling based algorithms in 20 environments.
Failure Case
Legend
Starting point
Ending point
Node
Edges
Shortest Distance
Scope of study
2 Dimension, point robot, static obstacles
Obstacle
Methodology
Gaussian Sampling
PRM
70.00
LDPRM
60.00
OBPRM
50.00
40.00
Gaussian
Sampling
30.00
RBB
20.00
10.00
0.00
PRM*
Simple
Maze
Narrow Passage
Bug Trap
Mixed
Types of Environment
Conclusion
Simple environments
OBPRM (77.50%)
Maze environments
LDPRM (67.75%)
Narrow passage environments Gaussian sampling (64.50%)
Bug trap environments
PRM* (69.25%)
Mixed environments
LDPRM (69%)
Benefits
Provide considerable suggestion for user to choose a better algorithms
Provide a guideline for others who want to study sampling based algorithm.
Supervisory Committee
Supervisor
: Associate Professor Dr. Tang Sai Hong (UPM)
Co-supervisor 1 : Dr. Weria Khaksar (UNITEN)
Co-supervisor 2 : Professor Yoshio Yamamoto (Tokai University, Japan)
Examiner 1
: Associate Professor Ir. Dr Nawal Aswan bin Abdul Jalil (UPM)
Examiner 2
: Ir. Razali bin Samin (UPM)