Modified adaptive critic-based control for a class of robotic rehabilitation systems

A modified adaptive fuzzy actor-neural network critic (FANC)-based controller is proposed for a class of robotic rehabilitation system(RRS). Based on the modified FANC control approach, the developed RRS can utilize different kinds of prior knowledge and measured information to initialize its fuzzy actor, critic network, and ensure the RRS stability during the whole control process with the help of a PD-based supervisory controller. In addition, to attenuate the effects of the approximation error and the uncertainty in the system, robust terms are also integrated into the proposed FANC agent. Lyapunov stability analysis shows the proposed FANC can yield a semi-global asymptotic result. Simulation experiments showed that the feasibility of the developed FANC for robotic rehabilitation system.