Hybrid evolutionary motion planning via visibility-based repair

Abstract

Introduces a hybrid evolutionary system for global motion planning within unstructured environments. This hybrid system combines a novel obstacle representation scheme, the concept of evolutionary searching and a new concept that we refer to as 'visibility-based repair' to form a hybrid which quickly transforms infeasible motions into feasible ones. Our hybrid evolutionary system differs from other evolutionary motion planners in that (1) more emphasis is placed on repairing infeasible motions to develop feasible ones rather than using simulated evolution exclusively as a means of discovering feasible motions, (2) a continuous map of the environment is used rather than a discretized map, and (3) it develops motion plans for moving or mobile destinations. In this paper, we demonstrate the effectiveness of this new hybrid by using two challenging motion planning problems where the target tries to move away from our point robot.