Image-based path-planning algorithm on the joint space

Abstract

To design an image-based path-planning algorithm, we should decrease the sum of moving distances of degrees-of-freedom of a manipulator. The algorithm A* selects the shortest path in the joint space by hard search. The combination search requires long distance moving. In a computer (virtual) world, the moving is not always time consuming. However in a real world, the moving is extremely time consuming. To overcome this drawback, we are seeking for a one-way search. For this reason, the sensor-based path-planning algorithms are preferred to an image-based path-planning algorithm. On the observation, we adopt a three or more dimensional sensor-based path-planning algorithm as the image-based path-planning algorithm. Its heuristics is defined by a difference between present and objective images. If the heuristics amounts to zero, the image-based algorithm stops. Therefore, convergence of a camera on a manipulator tip to an objective state is kept, and it is not influenced at all by uncertainties of robot and camera parameters.