D4RL Offline Dataset

该数据集专为单智能体离线任务设计，涵盖了健身房领域、Adroit任务以及AntMaze挑战。在Adroit任务中，需要控制一个具有24个自由度的机械手，而在AntMaze任务中，则需要通过组合次优轨迹来实现最优策略。这些任务都属于单智能体强化学习范畴。

This dataset is specifically designed for single-agent offline tasks, encompassing the Gym domain, Adroit tasks, and the AntMaze challenge. In Adroit tasks, the agent is required to control a robotic manipulator with 24 degrees of freedom, while in the AntMaze task, the optimal policy must be achieved by combining suboptimal trajectories. All these tasks fall under the scope of single-agent reinforcement learning.