Torque Estimation in a Wrist Rehabilitation Robot Using a Nonlinear Disturbance Observer

In this paper, we present implementation of a nonlinear disturbance observer algorithm to estimate disturbance torques on a wrist rehabilitation robot. The ultimate goal is to enable accurate estimation of user interaction torques without force/torque sensors. A dynamic model of the Wrist Gimbal two degree-of-freedom forearm and wrist exoskeleton is developed. A nonlinear disturbance observer (NDO) algorithm that incorporates the robot dynamics is implemented. Friction models for both joints of the robot are identified. The robot dynamic model and the accuracy of the disturbance torque estimations are experimentally verified under closed-loop trajectory tracking scenarios.