Cascade based iterative learning control of robotic-assisted upper extremity stroke rehabilitation

This paper develops a combined cascade and iterative learning control based scheme for regulating the assistive stimulation applied to the human muscle in robotic-assisted upper-limb stroke rehabilitation. Guided by a robot the patient makes repeated attempts at a finite duration task with assistive stimulation applied to the relevant muscles. Once each attempt is complete, the arm is reset to the starting location and an iterative learning control algorithm is used to compute the stimulation to be applied on the next attempt, where if the patient is improving with each attempt the level of assistive stimulation required should decrease and the voluntary effort increase. This property has been observed in clinical trials where a critical problem is the response of the muscles to electrical stimulation and, in particular, fatigue. The new results in this paper relate to the addition of a cascade controller in an inner feedback loop around the muscle model to counter the onset of fatigue. Results from a simulation based assessment of the final design using patient data are given, where such a study is a prerequisite for obtaining ethical approval to conduct a clinical trial.