Approximate dynamic programming for output feedback control

This paper studies the adaptive and optimal output feedback control problem using approximate dynamic programming. It is shown that, under the recursive algorithm, the control policy converges to its optimal value, up to a constant proportional to the magnitude of the inaccuracy caused by observation errors. On the basis of this result, direct adaptive output feedback strategies are developed for solving both discrete-time and continuous-time LQR problems with uncertain parameters. Finally, numerical examples are given to demonstrate the efficiency of the proposed control schemes.