Policy Gradient Converges to the Globally Optimal Policy for Nearly Linear-Quadratic Regulators

Open in new window