Development of a Bioinspired optimization algorithm for the automatic generation of multiple distinct behaviors in simulated mobile robots
In this study, the utilization of a multi-objective approach in evolving artificial neural networks (ANNs) for an autonomous mobile robot is investigated. The ANN acts as a controller for radio frequency (RF)-Iocalization behavior of a Khepera robot simulated in a 3D physics-based environment. A fi...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Research Report |
| Language: | English |
| Published: |
Universiti Malaysia Sabah
2006
|
| Subjects: | |
| Online Access: | https://eprints.ums.edu.my/id/eprint/23200/1/Development%20of%20a%20Bioinspired%20optimization%20algorithm%20for%20the%20automatic%20generation%20of%20multiple%20distinct%20behaviors%20in%20simulated%20mobile%20robots.pdf https://eprints.ums.edu.my/id/eprint/23200/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | In this study, the utilization of a multi-objective approach in evolving artificial neural networks (ANNs) for an autonomous mobile robot is investigated. The ANN acts as a
controller for radio frequency (RF)-Iocalization behavior of a Khepera robot simulated in a 3D physics-based environment. A fitness function has been created from the preliminary experiment with optimized two conflicting objectives: (1) maximize the virtual Khepera robot's behavior for homing towards a RF signal source and (2) minimize the number of hidden neurons used in its ANNs controller by the utilization of Pareto-frontier Differential Evolutionary Multi-objective (POE) algorithm. Bootstrap problem found during the fitness function optimization. Thus, another component has
been included into the fitness function in order to overcome the bootstrap problem, so called obstacle avoidance component. Furthermore, the performance of the generated
robot controll~r has been improved with integrated three extra components; average wheels speed component, maximize wheels speed component and shortest time component. The testing results showed the controller performed better after the hybridization of the mentioned components. Further, a comparison between elitism and non-elitism used has been conducted attributable to no study has been conducted yet
in comparing the application of elitism and non-elitism. As a result, this study has thus shown that the multi-objective approach to evolutionary robotics in the form of the
elitist PDE-EMO algorithm can be practically used to automatically generate controllers for RF-Iocalization behavior in autonomous mobile robots. |
|---|