A hybrid sampling strategy with optimized Probabilistic Roadmap Method

Probabilistic Roadmap Method(PRM) is sampling-based techniques being extensively used for virtual humans field. In this paper, we present a hybrid sampling strategy with PRM for multi-agent path planning in a complex environment. The two aspects are optimized: first, we propose a hybrid sampling strategy which is composed of bridge test sampling and non-uniform sampling to enhance milestones in narrow passages and boundary regions; second, we propose a optimized A-star algorithm which is able to remove redundant milestones to plan a proper path. Our planner is tested on five agents in complex environment. Preliminary experiments show that the hybrid sampling strategy enables effectively increase the number of milestones in crucial space, and the optimized A-star algorithm is able to availably shorten the length of path.