Revisiting LQR Control From the Perspective of Receding-Horizon Policy Gradient

We revisit in this paper the discrete-time linear quadratic regulator (LQR) problem from perspective of receding-horizon policy gradient (RHPG), a newly developed model-free learning framework for control applications. provide fine-grained sample complexity analysis RHPG to learn that is both stabilizing and -close optimal LQR solution, our algorithm does not require knowing initialization. Combined with recent application Kalman filter, we demonstrate general applicability estimation streamlined analyses.