An interaction torque control improving human force estimation of the rehab-exos exoskeleton

This paper describes the interaction torque control of the Rehab-Exos, an upper-limb robotic exoskeleton with direct torque joint sensors for interaction in Virtual Environments and rehabilitation. The control architecture consists in a centralized torque control and separated optimal torque observers for each joint of the exoskeleton. The optimal observer is a full-state Kalman filter providing the estimates of both internal and external torques acting on the joints and overcoming most of the issues due to the noise in the torque sensor signals. The centralized torque control is based on a full dynamics model of the exoskeleton, calculates the kinematics and dynamics of the system and estimates the feed-forward contribution for the compensation of dynamic loads measured by joint torque sensors. Experimental tests have been carried out to validate the desired torque tracking in haptic interaction tasks.