Hindsight Experience Replay capability added to reinforcement learning toolkit

AI Impact Summary

Change Type CAPABILITY indicates a new capability in the RL toolkit, introducing Hindsight Experience Replay. This method reinterprets past experiences by aligning them with alternative goals, which can dramatically improve data efficiency in sparse-reward environments. Engineering teams should plan to update training pipelines to enable HER, which may entail adjustments to the replay buffer, goal-conditioned policies, and environment interfaces to support goal specifications.