Iterative Learning Control of Gravity Compensation for Upper-Arm Robot-Assisted Rehabilitation

Robot-assisted rehabilitation allows patients e.g. suffering from a stroke to practice without continuous supervision from a therapist. To activate neuroplasticity, the patient has to actively participate in the rehabilitation therapy and the robot should only provide as much assistance as required based on the patient's needs and abilities. For this purpose, gravity compensation is a promising approach as simplifying movements enables the patient to increase the training's intensity and number of repetitions. Thus, the aim of this paper is the application and implementation of an iterative learning control scheme to adjust the gravity compensation during therapy based on the patient's abilities. For this purpose, a norm-optimal iterative learning control scheme and an optimization-based proportional-type iterative learning control algorithm are used. To validate and compare them, an experiment with a linear and a second one with a circular motion trajectory is done, while a slowly changing repetitive disturbance in form of an artificial force is applied to imitate the patient. In this case, the measured number of samples per cycle differs due to the underlying control scheme of the robot. For this reason, a mapping process based on the Dijkstra method is done. The results illustrate that both algorithms are robust against disturbances and yield good tracking performance. Thus, also other factors such as the computation effort of both algorithms should be considered in future research.