| In this dissertation, three problems have been primarily studied in the area of robot control: (1) characterization of robot nonlinear dynamics, (2) force control of robot manipulators with unknown dynamics and disturbances, and (3) nonlinear impact force control.; In order to identify the robot control problems, we studied the effects of plant dynamics, focusing on robot joint dynamics such as joint flexibility and friction. Extensive experimental studies have been done on robot transmissions. The experiments have included: transmission linearity, backlash, static and dynamic friction, and forward efficiency. In order to understand the influence of transmission properties on overall system performance, a comparative evaluation was performed on three competing transmission types: worm-gear drive, cone-drive and traction drive transmission. The results of experiments performed on worm-gear drive validated a load-dependent friction model, which was derived for feed-forward friction compensation in feedback control.; With understanding of how transmission nonlinearities influence closed-loop and controlled behavior, a new controller has been developed by combining Natural Admittance Control with Time Delay Control. The proposed nonlinear controller is not model based control. The only system parameter that must be estimated is inertia, rendering it easy to implement. It rejects unmodeled dynamics, nonlinearities, and disturbances without a difficult characterization process while preserving desired dynamics. The simulation results demonstrate not only good external disturbance rejection, robustness to parameter changes and insensitivity to noise, but also demonstrate good trajectory tracking providing good rejection of internal Coulomb friction.; For stabilization of a robot manipulator upon collision with a stiff environment, we proposed a novel impact force control strategy which is developed based on the observation of human interactive behavior. It uses a robust natural admittance/time-delay control with an added negative force feedback to absorb impact force and stabilize the system. During the impact phase, this control input alternates with zero control input when no environment force is sensed. Simulation results show that this simple bang-bang control approach produces a stable interaction with a very stiff environment and its performance is comparable to the other existing impact force control techniques. |
