Iterative Learning Control of Impedance Parameters for a Soft Exosuit

Abstract

In this paper, the human ankle impedance information will be added into the human-exosuit dynamic model, and then a human-in-the-loop control of soft exosuit is designed to provide plantar flexion assistance. The gradient-following and betterment schemes are employed to obtain a desired impedance model. The scheme can provide an auxiliary force for the ankle to push off the ground in the variable human-exsosuit dynamic interaction. When the subject wears the exosuit and walks on the ground, this iterative learning method can be used to give the exoskeleton the human-like process learning skills, so that it can automatically adapt to the forces and instabilities of the surrounding environment. The effectiveness and stability of the control scheme are verified by experiments on different subjects. Results show that a desired interaction performance can be achieved by learning impedance parameters, indicating our proposed method has potential to facilitating exosuit control.