Adaptive and relaxed visibility-based PRM

In this paper we introduce a new sampler for robotic motion planning using the probabilistic roadmap approach. We improve the previously-introduced visibility-based sampler with two heuristics. First, we adopt a space decomposition approach can guide the sampler by identifying promising areas of the configuration space where milestones can be efficiently generated. While the original visibility-based sampler uses a very selective criterion to determine whether a sampled configuration point is to be accepted as a new milestone, we use a relaxed criterion that can significantly reduce the computational cost in the roadmap construction phase while increasing the size of the roadmap only mildly