Smith-Predictor-Based Torque Control of a Rolling Diaphragm Hydrostatic Transmission

Abstract

Rolling Diaphragm Hydrostatic Transmissions (RDHT) are high-performance low-cost solutions to delocalize heavy actuators away from the joints of robotic systems. Exploiting a low-cost pressure-based sensing technique, we propose here a Smith-predictor-based joint torque control of an RDHT-based actuation system. We also use a load-cell sensor for ground truth validation. The developed feedback controller is conveniently tuned based on an arbitrary pre-specified closed-loop natural frequency and damping ratio. This preserves the open-loop bandwidth while removing the intrinsic oscillations of the lightly damped highly transparent open-loop plant. Experimental tests validate the proposed control strategy, both in a stand-alone torque setpoint configuration and in a series of Human-Robot-Interaction tests confirming the significant advantages of the closed-loop control architecture.