NEAR-TIME OPTIMAL ROBOT MOTION PLANNING FOR ONLINE APPLICATIONS

Solving current formulations of the time-optimal point-to-point motion problem for robotic manipulators is a computationally intensive task. Thus, mast existing solutions are not suitable for on-line motion pla nning applications, such as the interception of moving targets, where time-optimality of the motion is advantageous. A novel technique is pr oposed in this article that separates the time-optimal point-to-point motion problem into the following two sub-problems: (1) selection of a near-time-optimal path between the two endpoints, and (2) generation of time-optimal motion along the selected path (i.e., constrained cont inuous path motion). Although our approach uses known path-constrained time-optimal-motion algorithms for the second sub-problem, a new meth od is proposed for the selection of near-time-optimal paths. Based on a study of the characteristics of global-time-optimal paths, the near- optimal path is selected as a minimum-curvature joint spline, tangent to one of the manipulator's acceleration directions at the start point , and tangent to the required manipulator velocity direction at the en d point. The algorithm for determining the overall near-optimal path i s described herein, along with an example. Simulation test results and computation-time studies indicate that the proposed method is suitabl e for on-line motion planning applications. (C) 1995 John Wiley & Sons , Inc.