Autonomous robot navigation using adaptive potential fields

In this paper is presented a new scheme for autonomous navigation of a mobile robot, based on improved artificial potential fields and a genetic algorithm. In conventional artificial potential field methods, the robot is attracted by the goal position only, and rejected by several obstacles. Use of a single attraction point can lead to trap situations where the method is unable to produce the resultant force needed to avoid large obstacles. In the scheme presented here, multiple auxiliary attraction points have been used to allow the robot to avoid large, or closely spaced, obstacles. The configuration of the optimum potential field is automatically determined by a genetic algorithm. Simulation experiments performed with three different obstacle configurations, and ten different routes, showed that the scheme reported has a good performance in environments with high obstacle densities, achieving a success rate of 93 per cent.