Approximate Midpoint Policy Iteration for Linear Quadratic Control
This work provides a faster converging algorithm for solving linear quadratic optimal control problems, which could benefit researchers and practitioners working with these control systems.
This paper introduces a midpoint policy iteration algorithm for linear quadratic optimal control. In a model-based setting, it achieves cubic convergence, outperforming standard policy iteration (quadratic) and policy gradient (linear) algorithms. The algorithm can also be approximately implemented model-free using least-squares estimates, achieving cubic convergence to approximately optimal policies with the same sample budget as approximate standard policy iteration.
We present a midpoint policy iteration algorithm to solve linear quadratic optimal control problems in both model-based and model-free settings. The algorithm is a variation of Newton's method, and we show that in the model-based setting it achieves cubic convergence, which is superior to standard policy iteration and policy gradient algorithms that achieve quadratic and linear convergence, respectively. We also demonstrate that the algorithm can be approximately implemented without knowledge of the dynamics model by using least-squares estimates of the state-action value function from trajectory data, from which policy improvements can be obtained. With sufficient trajectory data, the policy iterates converge cubically to approximately optimal policies, and this occurs with the same available sample budget as the approximate standard policy iteration. Numerical experiments demonstrate effectiveness of the proposed algorithms.