Completely model-free approximate optimal tracking control for continuous-time nonlinear systems

This paper studies the optimal tracking control problem for continuous-time affine nonlinear systems and proposes a completely model-free approximate optimal tracking control design approach. This approach only uses measurement data collected from the trajectories of the system in real time to learn the optimal tacking control. At first, a new tracking policy iteration algorithm is developed based on the integral reinforcement learning technique. Then, the algorithm is implemented based on the actor-critic structure, where the critic neural network and the actor neural network are updated iteratively. Finally, simulation results are provided to show the efficiency of the method.