Fast Policy Learning for Linear Quadratic Control with Entropy Regularization
This work addresses policy learning efficiency for control problems, offering incremental improvements in convergence rates for specific LQC settings.
The paper tackles the problem of policy learning in linear-quadratic control with entropy regularization by proposing two new methods, RPG and IPO, which are proven to converge linearly to optimal policies, with IPO achieving super-linear convergence near the optimum and in transfer learning scenarios.
This paper proposes and analyzes two new policy learning methods: regularized policy gradient (RPG) and iterative policy optimization (IPO), for a class of discounted linear-quadratic control (LQC) problems over an infinite time horizon with entropy regularization. Assuming access to the exact policy evaluation, both proposed approaches are proven to converge linearly in finding optimal policies of the regularized LQC. Moreover, the IPO method can achieve a super-linear convergence rate once it enters a local region around the optimal policy. Finally, when the optimal policy for an RL problem with a known environment is appropriately transferred as the initial policy to an RL problem with an unknown environment, the IPO method is shown to enable a super-linear convergence rate if the two environments are sufficiently close. Performances of these proposed algorithms are supported by numerical examples.