Path planning algorithm and its application to a car-like robot

The problem of path planning is studied for the case for a mobile robot moving in an environment filled with obstacles which shapes and positions are known. An aggressive algorithm using a description of the obstacles based on a neural network is proposed. The algorithm allows to construct an almost optimal path. The path is piecewise linear with changing directions at the corners of the obstacles. The proposed algorithm is a significant improvement of the potential field algorithms because it finds an optimal path without being trapped in local minimums and the calculation speed for the proposed algorithm is comparatively fast. The algorithm solves the navigation problem in very complex environments such as polygonal makes. Simulation results and an application to a car-like robot "Khepera" show the effectiveness of the proposed algorithm.