Tag
OpenAI proposes standardizing constrained RL as the formalism for safe exploration and introduces Safety Gym, a benchmark suite for evaluating safe deep RL algorithms in high-dimensional continuous control tasks with safety constraints.