SRACO Based Co-Adaptive Control for a Class of Upper-Limb Robotic Rehabilitation Training Systems

The goal of this paper is to develop a SRACO (i.e., saturated robust adaptive critic-based optimal control) based coadaptive control system for a class of upper-limb rehabilitation training systems. The control system should be able to combine the rehabilitation robot and the patient to learn to accomplish goal-directed training tasks using the reward-punishment method, rather than clear training signals. Specifically, if the control behavior generated by the control system is good for rehabilitation training and enable the coadaptation between the patient and the rehabilitation robot, then the critic will give a reward, otherwise, the critic will punish the developed controller to modify its control behavior, till the controller finds the optimal control policy for the everyday functional rehabilitation training through the natural decision mechanism. Finally, the developed control approach (through Lyapunov stability analysis) is proved to achieve a robust optimal tracking, provided the control gains are selected appropriately. Simulation studies also show that the improved approach has robust optimal tracking than the existing controllers even with limiting control input.