RL Unplugged

RL Unplugged是由DeepMind创建的一套用于离线强化学习方法评估的基准套件。该数据集涵盖了从Atari游戏到模拟运动控制问题等多个领域的数据，包括部分或完全可观察的领域、连续或离散的动作空间以及随机与确定性的动态。数据集旨在通过详细的评估协议，促进离线RL方法的研究和比较，提高实验的可重复性，并使RL研究对更广泛的社区更加系统和可访问。

RL Unplugged is a benchmark suite developed by DeepMind for evaluating offline reinforcement learning (RL) methods. The datasets contained in this suite cover multiple domains ranging from Atari games to simulated motor control tasks, including partially or fully observable environments, continuous or discrete action spaces, as well as stochastic and deterministic dynamics. This suite aims to facilitate the research and comparison of offline RL methods through detailed evaluation protocols, improve the reproducibility of experiments, and make RL research more systematic and accessible to the broader community.