Octree-based approach to real-time collision-free path planning for robot manipulator

This paper presents a real-time calculation method to generate the trajectory of robot manipulator for the purpose of avoiding collision. In order to model 3-D workspace, we use an octree which has been used for fast collision detection. Path planning is done at two levels. First, a global planner finds a sub-goal in the 3-D workspace using the structure of the octree. Then a local planner searches the path to the sub-goal in the joint space. If the local planner fails, a new sub-goal is selected by the global planner and subsequently examined by the local planner. In the local planner, the levels of the octree are used as the cost function to represent the distance between the manipulator and the obstacles. This criterion is not exact, but, due to this, we can obtain the approximate feasible trajectory extremely quickly. We show the effectiveness of our method with some simulation examples.