Autonomous torque sensor calibration and gravity compensation for robot manipulators.

摘要

This thesis addresses calibration of joint torque sensors and identification of a gravity compensation (GComp) model. The two problems are related: calibrated torque information is prerequisite for the GComp model identification; the identified GComp model makes on-line automatic torque sensor calibration possible.; In the first part, we propose an autonomous joint torque sensor calibration method, which utilizes combinations of single-joint rotations and an arm's own gravity load. The method determines not only joint torque sensor gains and offsets, but also those of the joint angle sensors. When one joint of a manipulator is rotated, the gravity torque exerted on the joint varies sinusoidally with rotation angle. From the fitted parameters one may extract sensor gains and offsets. A custom designed calibration load is used to provide a reference torque for torque sensor gain calibration. A key feature of the proposed method is that nothing is assumed known about the arm's inertial parameters or of the location of the reference load in the grasp. Experiments are conducted on the Sarcos Dextrous Arm.; The second part of the thesis presents a new and simple procedure to identify link mass parameters used for gravity compensation and on-line automatic torque sensor calibration for a robot manipulator. The approach employs a single-joint rotation and a recursive procedure that proceeds distally to proximally. Composite mass moments are identified for each link, and are further divided into configuration dependent and independent terms. It is shown that the configuration-independent terms may actually combine contributions from several links. Simulation and experimental results on the Sarcos Dextrous Arm verify the approach.