Adaptive Optimal Control of Unknown Constrained-Input Systems Using Policy Iteration and Neural Networks
A robotic exoskeleton with adjustable robot behavior is proposed in this paper. The assistive exoskeleton helps the people to execute the task and optimizes the system's performance. We propose a novel adaptive impedance control for the robotic exoskeleton, whose end-effector's motions are prescribed by the relative trajectory and constrained by the physical limits. With the purpose of shaping the dynamics of the robotic exoskeleton to minimize motion tracking errors and human effort, the linear quadratic regulation (LQR) approach is employed to acquire an optimal impedance model. Integral reinforcement learning (IRL) is employed to deal with the given LQR design without the human model's information. The impedance control incorporating adaptive parameter learning technique makes tracking errors convergence while the constrained region is not transgressed by the end-effectors. Experiment results show that the proposed controller is effective in tracking the desired trajectories.