We have developed direct numerical methods of solving optimal control problems that are sufficiently fast that it is becoming reasonable to use them to perform time optimal path planning. An overview of the nature of time optimal robot motions has been developed for elbow, polar, and SCARA robots, and it has been shown that for some maneuvers a very large savings in time is possible. This offers the opportunity to increase productivity on assembly lines by making the robots more efficient in executing their point-to-point motions. This paper takes one step toward converting these time optimal results into ones that can be used in routine robot operation. Methods are developed that allow one to make the existing robot feedback controllers produce the optimal torque histories needed for time optimal operation. It is shown that this is easiest accomplished by back-computing the commands to give the robot links from the desired torque histories. (Author)
AbstractThe authors and co-workers have developed direct numerical methods of solving optimal control problems that are sufficiently fast that it is becoming reasonable to use them to perform time optimal path planning. In a series of publication, a good overview of the nature of time optimal robot motions has been developed for elbow, polar, and SCARA robots, and it was shown that for some maneuvers a very large savings in time is possible. This offers the opportunity to increase productivity on assembly lines by making the robots more efficient in executing their point-to-point motions. This paper takes one step toward converting these time optimal results into ones that can be used in routine robot operation. Methods are developed that allow one to make the existing robot feedback controllers produce the optimal torque histories needed for time optimal operation. It is shown that this is easiest accomplished by back computing the commands to give the robot links from the desired torque histories.