This paper addresses the problem of determining an optimal set of gains for a controller for a teleoperated arm. Specifically, an automatic tuning technique was applied and investigated for tuning an independent-joint proportional-derivative controller for a teleoperated manipulator. The Hooke and Jeeves method is used in conjunction with a one-dimensional search routine in the tuning algorithm. The algorithm was used to optimize gains for a two-link teleoperator simulation and the results of several optimizations were used to determine the best form for an input trajectory and cost function. The desired joint angle trajectory is taken from low-pass filtered step inputs with randomly generated magnitudes, which vary at a predetermined interval. Both positive and negative angles are generated, but they are constrained to lie within the manipulator work space. It was determined that the cost function should be based on tracking error, peak position error over the entire desired path, overshoot, actuator torque bounds, and gain limits. The optimized gains obtained from the simulation were applied to an actual teleoperator and some improvement was seen.
展开▼