A recursive sampling based method for path planning

This paper presents a new technique of path planning in synthetic 3D environment. The environment may involve any number of obstacles. The algorithm precomputes a global roadmap of the environment by using a variant of randomized motion planning algorithm. The local planner used to connect two samples is a recursive one and always finds a path between the two samples avoiding obstacles. Finally, it performs graph searching and automatically computes a collision-tree shortest path between two user specified locations. The method is applied to a two-joint manipulator. Experimental results show that the probability of finding a path is high and planning can be done in a few seconds.