Design and control of direct-drive systems with applications to robotics

Abstract

This thesis presents advances in the design and control of direct drive systems. The McGill/MIT synchronous direct drive motor, a new high performance torque sensor, and a complete hydraulic dynamometer testbed are designed and built. In the area of motor torque control, a new commutation strategy with respect to the motor's phase torques and the amplifier's frequency response characteristics is presented. It minimizes the torque ripple over a specific velocity range while maintaining minimum power losses. The high performance of the controller at low and high velocities is demonstrated by experiments and simulations, respectively. In order to re-tune the open-loop controller continuously, an on-line estimator based on the motor's voltage equation is developed. Simulations and experimental results have demonstrated convergence of the estimated parameters to their true values. Positive joint torque feedback is nested inside the motion control to compensate the effect of the load dynamics on position tracking performance. A concise explicit model of the remaining dynamics as well as the condition on the robot kinematics for linearity are derived. In the case of non-colocated actuator and torque sensor, the effect of a torque sensor's compliance on stability is investigated using a singular perturbation method. Finally, an H∞ joint torque feedback is proposed which takes the actuator's dynamics and uncertainty into account and minimizes the system's sensitivity to the load torque disturbances or load dynamics. On our experimental direct-drive motor, the additional H∞ positive torque feedback remarkably improved the disturbance attenuation (25dB) and load decoupling properties of a simple PID motion controller.