Multiobjective motion planning for a nonholonomic vehicle

Abstract

A technique is proposed for integrating a probabilistic graph construction algorithm with an evolutionary multiobjective optimizer. A hybrid planner (EvoVBPR) for nonholonic robotic vehicle is presented. It integrates a probabilistic roadmap construction method (VBPR) with the SPEA2 evolutionary multiobjective algorithm and an additional deterministic graph pruning step. The result is a Pareto set of roadmaps that represent different tradeoffs between length of path and obstacle clearance.