A genetic algorithm for nonholonomic motion planning

Abstract

The paper presents a genetic algorithm to find and optimize solutions for nonholonomic motion planning problems. Mainly focusing on mobile robots, the algorithm uses present randomized algorithms to come up with suboptimal paths and iteratively optimizes them according to a fitness function which includes domain specific knowledge. The major advantages of this method include being an any-time algorithm, and improving the quality of the solution throughout the evolutionary process. An extensive experimental analysis comparing our results with state of the art algorithms outline the effectiveness of the proposed methodology.