Multi-query motion planners based on sparse roadmaps for changing and difficult environments

Abstract

We investigate multi-query motion planners that use sparse roadmaps, and extend their domain of applicability to nonreversible systems and path-existence LTL trajectory specifications; we prove probabilistic completeness for the extensions. We offer simulation and experimental evidence that such planners are competitive with single-query planners for changing environments. We present a planner system for intricate manipulation tasks that would be too difficult for a single-query planner to handle in a practical amount of time. We propose a data structure to guide planning in highly constrained environments and provide an object classification criterion that captures kinematic interactions between objects and is suitable for motion planning.