Adaptive optimal control for rehabilitation systems

This paper describes the design of a control strategy for robotic rehabilitation devices, aimed at recovering the motor skills of patients affected by particular diseases that damage the neuromuscular apparatus of the human body. There are two different but related activities in the project here described. At first, a non-linear parameter identification procedure, based on the Recursive Least Square (RLS) method, has been used to identify the impedance at the end-point of the human arm. To obtain more accurate estimation values, the RLS algorithm uses the state values (position and velocity) provided by a state observer based on acceleration-aided Kalman Filter (aaKF). Then, the design of a Non-Linear Adaptive Controller (based on the arm's characteristic) is described. The controller aims at helping the patient during the rehabilitation exercises, by using a level of assistance that varies according to the patient's performance. In this way it is possible to emulate the help of the physiotherapist, teaching the patient the correct execution of the exercise only if necessary, thus implementing an assist-as-needed policy.