Off-Policy Safe Reinforcement Learning with Constrained Optimistic Exploration

Off-Policy Safe Reinforcement Learning with Constrained Optimistic Exploration explores A novel off-policy safe RL algorithm that enables cost-bounded exploration and stable value learning for safety-critical applications.. Commercial viability score: 7/10 in Safe Reinforcement Learning.