Probabilistic Guarantees for Safe Deep Reinforcement Learning

Edoardo Bacci, David Parker

Deep reinforcement learning has been successfully applied to many control tasks, but the application of such agents in safety-critical scenarios has been limited due to safety conc…