Accelerating Safe Reinforcement Learning with Constraint-mismatched Baseline Policies

Tsung-Yen Yang, Justinian Rosca, Karthik Narasimhan, Peter J. Ramadge