Assistive Control of Soft-Robots for Orthopedic and Neurological Rehabilitation using Adaptive Model-Based Approaches

Abstract This paper presents two improved adaptive model-based assistive controllers for therapeutic robotic devises based on direct rotary pneumatic soft-actuators for orthopedic and neurological rehabilitation. The controllers are based on separated models of the soft-robot and of the patient’s extremity, in order to take into account the individual patient’s behavior, effort and ability during control, what is essential to facilitate muscle reconstruction in orthopedic and relearn lost motor functions in neurological rehabilitation. The high inherent compliance of soft-actuators allows for a general human-robot interaction and provides the base for effective and dependable assistive control. An inverse quasi-static model of the soft-robot with estimated parameters is used to achieve transparency of the robot during treatment and adaptive inverse models of the individual patient’s extremity allow the controllers to learn on-line the individual patient’s behavior and effort and react in a way that assist the patient only as much as needed. Due to the use of the models the high compliance of soft-actuators can remain during control. The effectiveness of the controllers is evaluated using a first prototype of an assistive soft elbow trainer.