Iterative learning control of FES with embedded simultaneous volitional EMG

Seventeen million people are left with limb impairment following stroke, and there is an urgent need for new effective rehabilitation technology. Functional electrical stimulation (FES) has been shown to facilitate motor re-learning by artificially activating muscles during practice of motor tasks, however clinical outcomes depend on how closely FES matches the intended motion. Electromyography (EMG) signals recorded from muscle can capture voluntary intention, and have successfully been employed in openloop FES controllers. Likewise, model-based controllers using force and/or position data have yielded accurate movement control in clinical trials. Iterative learning control (ILC) in particular has been successful in clinical tests, since the process of rehabilitationis inherently iterative. This paper develops a new control strategy to combine both EMG and model-based control. It begins by developing a novel hybrid model of muscle dynamics incorporating both EMG and FES. ILC is then employed to enable precise control over both variables, thereby offering substantial improvements over existing control schemes in the domain of stroke rehabilitation. Experimental results confirm efficacy of the hybrid control scheme, as well as its suitability for clinical application