Artificially induced joint movement control with musculoskeletal model-integrated iterative learning algorithm

Neuromuscular electrical stimulation (NMES) delivers tiny electrical impulses to artificially induce muscle contraction, which can be used for the purpose of neurological rehabilitation or muscle strength enhancement. For neurological patients, NMES is not only able to artificially induce movements but also to improve physiological functions and proprioception. To make the induced movements functional, achieving accurate and persistent movement is important but difficult due to highly nonlinear time-variant properties of human musculoskeletal system. This paper proposes a musculoskeletal model-integrated iterative learning control (MMILC) strategy where the musculoskeletal model accelerates the self-learning of iterative learning control (ILC) to accomplish repetitive joint movements. Taking advantage of the feedforward information from the musculoskeletal dynamic model and the self-learning ability of ILC, fast response of the controller and accurate tracking performance can be achieved simultaneously. Both simulation and experiment research are conducted to evaluate the performance of the proposed MMILC in NMES-induced joint movement control. Comparing with traditional ILC, the proposed MMILC illustrates faster and better tracking performance in the control of typical knee joint movements. From the statistic results on six able-bodied subjects, the proposed MMILC can compensate the subject-specific differences in musculoskeletal characteristics responding to the artificial electrical stimulation and represent feasible, effective control of NMES to achieve different motor control purposes.