Electrical Stimulation and Iterative Learning Control for Functional Recovery in the Upper Limb Post-Stroke Katie Meadmore, Timothy Exell, Christopher Freeman, Mustafa Kutlu and Eric Rogers Electronics and Computer Science University of Southampton Southampton, UK SO17 1BJ Email: klm@ecs.soton.ac.uk Ann-Marie Hughes, Emma Hallewell and Jane Burridge Faculty of Health Science University of Southampton Southampton, UK SO17 1BJ Abstract—Therapies using functional electrical stimulation (FES) in conjunction with practice of everyday tasks have proven effective in facilitating recovery of upper limb function following stroke. The aim of the current study is to develop a multi-channel electrical stimulation system that precisely con- trols the assistance provided in goal-orientated tasks through use of advanced model-based ‘iterative learning control’ (ILC) algorithms to facilitate functional motor recovery of the upper limb post-stroke. FES was applied to three muscle groups in the upper limb (the anterior deltoid, triceps and wrist extensors) to assist hemiparetic, chronic stroke participants to perform a series of functional tasks with real objects, including closing a drawer, turning on a light switch and repositioning an object. Position data from the participants’ impaired upper limb was collected using a Microsoft Kinect® and was compared to an ideal reference. ILC used data from previous attempts at the task to moderate the FES signals applied to each muscle group on a trial by trial basis to reduce performance error whilst supporting voluntary effort by the participant. The clinical trial is on-going. Preliminary results show improvements in performance accuracy for each muscle group, as well as improvements in clinical outcome measures pre and post 18 training sessions. Thus, the feasibility of applying precisely controlled FES to three muscle groups in the upper limb to facilitate functional reach and grasp movements post stroke has been demonstrated. Keywords: Functional electrical stimulation; Iterative learning control; Stroke rehabilitation; Technology; Upper limb; Wrist. I. I NTRODUCTION Stroke is a leading cause of death and disability world- wide, leaving many stroke survivors dependent on others for activities of daily living [1], [2]. Motor dysfunction is one of the main outcomes from stroke. It has been estimated that about 60% of patients have mobility problems one year post-stroke [2] and up to 85% are left with impairment of the upper extremity [3]. Of particular importance are upper limb impairments which cause many stroke patients to have difficulty in performing everyday tasks that involve reaching and grasping. This impacts both daily living and well-being [4]. Thus, it is important that rehabilitation systems which facilitate recovery of functional movement in the upper limb are developed. Research has shown that repetitive, goal-orientated prac- tice of movement is vital for recovery of upper limb function following stroke (see [5]). In addition, voluntary effort is also associated with increased positive therapeutic effects [6], [7]. As such, it is important that rehabilitation technologies incorporate and maximise these aspects of rehabilitation. Functional electrical stimulation (FES) is a promising reha- bilitation therapy, as it allows repetitive training of precise movements despite muscle weakness and paralysis often found post-stroke [5]. Indeed, FES has proved effective in improving motor function in the upper limb (e.g., [7], [8], [9], [10], [11], [12]. However, functional movements involve the coordination of multiple muscle groups in the impaired limb, and to date, most rehabilitation systems have applied FES to only one or two muscles in the upper limb [8]. In addition, even though systems employing FES technologies may allow patients to practice for longer, it has been suggested that assistive devices used in rehabilitation, such as robotic devices and FES, may reduce the voluntary effort patients exert during training [13]. To address these issues, a multi-channel stimulation reha- bilitation system for the upper limb that precisely controls applied FES through advanced iterative learning control (ILC) algorithms has been developed. This system, termed GO-SAIL: Goal Oriented, Stimulation Assistance through It- erative Learning, builds on previous work that has controlled FES signals using ILC algorithms [14], [15], [16], [17], [18]. ILC operates by using data from previous attempts at the task to update the FES control signal that is applied on the subsequent task attempt independently to each muscle group, with the objective of sequentially increasing tracking performance. This data may comprise kinematic, kinetic and stimulation signals, which are used in combination with an underlying bio-mechanical dynamic model of the arm [17], [18]. In this way, for each trial, the amount of stimulation applied provides a minimal level of assistance to correct