Convergent Policy Optimization for Safe Reinforcement Learning
This addresses safety constraints in reinforcement learning for applications like control and multi-agent systems, but it appears incremental as it builds on existing optimization methods.
The paper tackles safe reinforcement learning with nonlinear function approximation by formulating policy optimization as a constrained nonconvex problem and constructing surrogate convex problems using policy gradient estimators, proving convergence to a stationary point and applying it to optimal control and multi-agent examples.
We study the safe reinforcement learning problem with nonlinear function approximation, where policy optimization is formulated as a constrained optimization problem with both the objective and the constraint being nonconvex functions. For such a problem, we construct a sequence of surrogate convex constrained optimization problems by replacing the nonconvex functions locally with convex quadratic functions obtained from policy gradient estimators. We prove that the solutions to these surrogate problems converge to a stationary point of the original nonconvex problem. Furthermore, to extend our theoretical results, we apply our algorithm to examples of optimal control and multi-agent reinforcement learning with safety constraints.