Collision-probability constrained PRM for a manipulator with base pose uncertainty

摘要: We address the motion planning problem for a manipulator system with base pose uncertainty, e.g., when is mounted on mobile base. Using particle based representation we extend PRM (probabilistic roadmap) approach to deal this uncertainty. Because of path associated probability being collision-free, which fundamentally changes nature PRM's query phase. plan shortest such that collision-free higher than user defined threshold, were follow path. The becomes collision constrained (CP-CSPP), and shown as NP-hard w.r.t. number particles [1]. then present lazy algorithm, called Lazy-CPC-PRM (collision LazyPRM), k-shortest algorithm in conjunction labeling algorithm. exploits key insight if portion considered by invalid (the it less threshold) or dominated another sub-path, all longer paths containing can not be solution This leads significant efficiency gains practice. Although, worst case complexity exponential particles, empirically show effectiveness our 30 simulated 3-dof