Design of Adaptive Gravity Compensation Controller for Upper Limb Exosuit: The Concurrent Learning-based Approach

This paper deals with the design of an adaptive gravity compensator (AGC) for an upper limb soft exosuit. An exosuit is an assistive device for a wearer that supports locomotion to reduce human effort. The human upper limb with the exosuit can be modelled as an Euler-Lagrange system actuated by the human torque and assistive torque generated using the DC motor. The gravity compensator design aims to develop an adaptive control law that drives the assistive device's actuation, enabling the wearer to lift additional payloads with reduced effort. The adaptive gravity compensator is based on an estimation algorithm which estimates the unknown parameters. Often these algorithms require the signal to be persistently exciting to ensure exactness in estimation. To relax the per-sistence of excitation conditions, a concurrent learning-based estimation algorithm is introduced with the aim of exponential convergence of the parameter estimation error. It is shown that the concurrent learning-based adaptive gravity compensator can improve the response by reducing human effort.