Finite-horizon optimal control of discrete-time linear systems with completely unknown dynamics using Q-learning

This paper investigates finite-horizon optimal control problem of completely unknown discrete-time linear systems. The here refers to that the system dynamics are unknown. Compared with infinite-horizon control, Riccati equation (RE) is time-dependent and must meet certain terminal boundary constraints, which brings greater challenges. Meanwhile, have also caused additional main innovation this developed cyclic fixed-finite-horizon-based Q-learning algorithm approximate input without requiring dynamics. consists two phases: data collection phase over a fixed-finite-horizon parameters update phase. A least-squares method used correlate phases obtain by cyclic. Finally, simulation results given verify effectiveness proposed algorithm.