Abstract: |
Autonomous mobile robot navigation is a complex problem and evolutionary principles applied to such problems provide good solutions with relatively less computational effort. This also allows an automatic evolution of such systems. We describe how to evolve a neural network control system for a mobile robot using a simulator applying concepts of multi-objective optimization. Sometimes a single objective may not be adequate to describe the desired performance of the robot. In such cases, typecasting the problem as a multi-objective problem becomes necessary. In this paper we investigate the possibility of using evolutionary algorithms to evolve a controller for a mobile robot with multiple objectives to be satisfied simultaneously. Such behavior includes obstacle avoidance, smooth motion and target acquisition. The novelty of this method lies in the evolution of different navigational behaviors simultaneously using concepts of Pareto-optimality and evolutionary algorithms. A neural network is utilized to provide the control structure for the navigation of the mobile robot. A multi-objective evolutionary algorithm (FSGA) is utilized to identify the optimal neural network weights. The simulation results show that the proposed methodology is efficient and robust for evolving different behaviors simultaneously. Simulation results are provided and discussed. |
