An evolutionary method for active learning of mobile robot path planning

Several evolutionary algorithms have been proposed for robot path planning. Most existing methods for evolutionary path planning require a number of generations for finding a satisfactory trajectory and thus are not efficient enough for real-time applications. la this paper we present a new method for evolutionary path planning which can be used on-line in real-time. We use an evolutionary algorithm as a means for active learning of a route mop for the path planner. Given a source-destination pair, the path planner searches the map for a best matching route. If an acceptable match is not found, the planner uses another evolutionary algorithm to generate on-line a path for the source-destination pair. The overall system is an incremental learning planner that gradually expands its own knowledge suitable for path planning in real-time. Simulations have been performed in the domain of robotic soccer to demonstrate the effectiveness of the presented method.