Iterative learning control of the redundant upper limb for rehabilitation

A non-linear iterative learning control approach is developed for application to stroke rehabilitation. The subject is seated in a robotic workstation and electrical stimulation is applied to their triceps muscle to assist the tracking of trajectories in a horizontal plane. In addition to rotation about vertical axes through the shoulder and elbow joints, the forearm is also permitted to elevate in order to provide full arm extension. A dynamic model of the human arm is first developed, and then constraints are introduced to overcome kinematic redundancy. The expressions necessary to implement the control law are derived, and experimental results confirm its ability to achieve a high level of performance in practice.