Iterative Learning Control of Upper Limb Reaching using Functional Electrical Stimulation ⋆ C. T. Freeman, I. L. Davies, P. L. Lewin and E. Rogers School of Electronics and Computer Science, University of Southampton, Southampton, SO17 1BJ, UK (e-mail: cf@ecs.soton.ac.uk) Abstract: An iterative learning control scheme for the application of functional electrical stimulation to the human arm is designed and implemented. The task is to track trajectories in the horizontal plane and stimulation is applied to the triceps muscle. A model of the arm is first derived which includes assistive torque about the shoulder provided by a robotic arm. A linearising controller is then designed and a linear ILC algorithm is applied to the resulting system. Experimental results show that a high level of performance can be achieved in practice, and provide justification for the system to be subsequently used by stroke patients for rehabilitation. 1. INTRODUCTION Strokes affect between 174 and 216 people per 100,000 pop- ulation in the UK each year, and half of all acute stroke pa- tients starting rehabilitation will have a marked impairment of function in one arm (Mant et al. [2004]). Functional electrical stimulation (FES) can provide the experience of moving for the patient, which is necessary if sensory-motor function is to be regained. Recent studies have shown that when stimulation is associated with a voluntary attempt to move the limb, improve- ment is enhanced (Burridge and Ladouceur [2001]). Open-loop methods for the control of FES (see, for example, Davoodi and Andrews [2004]) have not provided the high level of per- formance necessary to fully promote this association. Closed- loop and model-based schemes, however, have overwhelmingly concentrated on the lower rather than the upper limb. Neural networks are one of the few approaches that have successfully been used to control FES applied to the arm, but these require extensive training and have unresolved stability issues due to their black-box structure (see Lan et al. [1994]). An experimental test facility incorporating a five-link planar robotic arm and an overhead trajectory projection system (see Freeman et al. [2007] for details) has been developed in order to provide a controlled environment in which to apply electrical stimulation to stroke patients. The subject is seated with their arm strapped to the robot, and the task presented to them is to repeatedly track a number of reaching trajectories using a combination of voluntary control and surface FES applied to muscles in their impaired shoulder and arm. The electrical stimulation is mediated using iterative learning control (ILC), a technique that is applicable to systems operating in a cyclical mode. This is one of the few advanced control techniques which has previously been applied to stimulation of the upper limb, although a high level of performance has not been achieved in practice (Dou et al. [1999]). Given the nature of the task, ILC is an obvious choice but also provides a framework which is capable of producing accurate tracking provided that the differ- ence in voluntary effort (interpreted as an external disturbance ⋆ This work is supported by the Engineering and Physical Sciences Research Council (EPSRC). Grant no. EP/C51873X/1. applied to the system) is sufficiently small from one trial to the next (see, for example, Moore [1992]). When accurate tracking of the trajectory is achieved, the stimulation will be reduced in order to promote sustained voluntary effort by the subject. In this paper ILC is used to control the FES applied to the triceps of an unimpaired subject who provides no voluntary effort. The robot supplies an assistive torque about the shoulder to allow full reaching tasks to be accomplished that are driven by the stimulation. The results show that high performance can be achieved, and confirm the efficacy of the system prior to its use by stroke patients. 2. WORKSTATION DESCRIPTION The robotic workstation consists of a five-link planar robotic arm rigidly connected to an overhead projection system, and is shown in Figure 1. A subject is strapped to the extreme link and Fig. 1. Unimpaired subject using the robotic workstation. a 6 axis force/torque sensor records the force they apply to the robotic end effector. The robotic arm is used to constrain the subject’s arm, to impose forces on the end-effector that make the task feel ‘natural’ to the subject, and to apply assistance during the performance of tracking tasks. The stroke patient’s Proceedings of the 17th World Congress The International Federation of Automatic Control Seoul, Korea, July 6-11, 2008 978-1-1234-7890-2/08/$20.00 © 2008 IFAC 13444 10.3182/20080706-5-KR-1001.0202