Auto Cable Pretension Method for Soft Exosuit Based on Gait Trajectory Prediction Network

Abstract

Cable-driven systems are widely used in various kinds of robots. And this actuation method is also applied in SIAT Soft Exosuit (SSEX). In practice, due to the softness of cable and necessity of not restricting wearer's mobility, the cable is designed to be a little loose and will be looser in some particular time. Consequently, force will not be transmitted through the cable until it is stretched, which will increase system latency. To this end, we propose an auto pretension method to reduce system latency in cable-driven soft exosuit system. Our approach achieves latency decreasing by pre-tightening the cable before the time point we need it to transmit force. And thus, we have to predict the gait trajectory to know when to start pretension procedure in advance. To solve this problem, we propose a Gait Trajectory Prediction Network (GTPN) which can predict gait trajectory by analyzing wearer's body parameters. Experimental results demonstrate the obvious decreasing of system lag and better wearing experience.