Online improvement of speed and tracking performance on repetitive paths

Abstract

When a mechanism such as a robot executes a trajectory, the tracking error increases as the trajectory speed is increased. This paper reports the experimental evaluation of an algorithm which locally adjusts the speed of a repetitive trajectory to achieve a specified level of tracking error. In regions of the trajectory where error is too high, the trajectory is slowed down, in regions where the error is below the specification, the trajectory is speeded up. The algorithm was experimentally evaluated on a five-axis mini direct-drive robot and it stably converged to a satisfactory trajectory for a range of error levels and speeds in spite of wide variations in the key algorithm parameters. The method is independent of the splining method used to generate the trajectory and of the feedback control law, and no model of the system is required.