Efficient Motion Planning Based on Disassembly

摘要

Disassembly-based motion planning (DBMP) is a novel and efficient single-query, sampling-based motion planning approach for free-flying robots. Disassembly-based motion planning uses workspace information to determine the workspace volume of a potential solution path and uses this information to exclude large portions of configuration space from exploration. It also identifies the most constrained placements of the robot along the potential solution path. These placements are referred to as assemblies because they are highly constrained by the environment, much like parts in an assembly are constrained. The constraints limit the possible motions of the robot and thus can be exploited to further limit configuration space exploration. The use of these two sources of workspace information permits the solution of many practical problems with very limited configuration space exploration. This reduction in configuration space exploration results in performance improvements of several orders of magnitude, compared to state-of-the-art motion planning methods. For non-free-flying robots, disassembly-based motion planning performs at least as well as the sampling-based motion planning method it is based on.