Human computer cooperation in interactive motion planning

摘要

This paper presents an interactive motion planning system that generates collision-free motions of robots or objects. The problem of planning collision-free motions has a high complexity, and most motion-planning programs often take a long time to plan motions when the environment is cluttered. Humans on the other use a set of learned heuristics to effortlessly plan collision-free motions, but do not have as good of an accuracy in geometric collision checking as the motion-planning programs. The proposed system provides the user with a facility to utilize the heuristic power of humans along with the algorithmic power and the geometric accuracy of motion-planning programs. More specifically, the user performs a global analysis of the environment and specifies robot's configurations (called subgoals) critical to finding a collision-free path, while a motion-planning program performs collision checking and finds a collision-free paths between subgoals. It has resulted in a system that is much more powerful and efficient than either a human or a computer algorithm in motion planning tasks. This system is expected to reduce motion planning time drastically for tele-operated robot manipulators and for verification of feasible part motions in mechanical assembly.