An experimental adaptive position and force controller for an electrohydraulic manipulator

During a changeout test using a dexterous electrohydraulic manipulator on a type-6B On-orbit Replacement Unit, two major shortcomings of using a teleoperation-type servo controller for a telerobotic task were noticed. The first shortcoming was the application of a large force to the environment when transitioning from force control to position control caused by inaccurate servo control. The second shortcoming was the force controller's inability to track step commands caused by improper model structure and ad hoc gain tuning methods. The focus of this dissertation has been to reduce or eliminate the effects of these shortcomings.;Due to difficulties modeling and calculating a dexterous manipulator's rigid body dynamics, the manipulator is modeled as independent servovalves. Coupling between the joints is considered unmodeled disturbances. A non-constant auxiliary signal applied to the servovalve's input ensures steady state tracking. In an experimental study, adaptive position control was shown to follow a model in both joint space and Cartesian space more accurately than a fixed gain controller. A key finding from the study was the RLS estimates the proper auxiliary signal without persistent excitation.;For the adaptive force controller, contact was modeled as a linear spring. Results of an experimental study show that in force directions, the adaptive force controller adapts to various contact stiffnesses, showing great improvements over a fixed gain force controller. A key factor in the amount of improvement over the fixed gain controller is the manipulator's servo stiffness in the force controlled direction. Due to difficulties in accurately measuring torque, adaptive force control does not improve the force controller's performance in torque directions.;As a solution to both shortcomings, adaptive position and force controllers have been developed and experimentally tested for the electrohydraulically driven Kraft manipulator. The indirect adaptive controllers use the recursive least squares (RLS) method to explicitly estimate the dynamic models. The controller is continuously redesigned based upon the current estimate of the dynamic models. To update the force gains, the adaptive controller uses the criteria that the manipulator's action (motion or force application) follows the response of a pre-determined model.