Output Feedback Adaptive Dynamic Programming for Linear H∞ Tracking Control

An online model-free integral reinforcement learning (IRL) output-feedback algorithm is developed to solve the H
∞
tracking control problem for the linear continuous-time system. The difficulty is due to the fact that access to system state information is not always available, and the proposed algorithm tackles this by presenting an output-feedback algorithm that solves the model-free H
∞
tracking problem. We rebuild the system states using a limited set of measured system outputs over the trajectory using the state reconstruction idea, and the output feedback version of the IRL Bellman equation is built to provide the updated control policy and disturbance policy simultaneously only by the measured system outputs. An online IRL output-feedback algorithm is then developed with no prior system dynamics or system state knowledge. The proposed scheme converges to the GARE solution and an application for the F-16 aircraft is demonstrated to illustrate our main result.