Adroit Expert Demonstration Dataset

该数据集包含了大量的演示数据，每个任务包含5000个轨迹，总计有100万个来自训练有素的强化学习策略的转换。这些数据被用于评估从专家演示中进行的模仿学习。规模上，数据集达到了100万个转换，任务专注于模仿学习。

This dataset contains a large corpus of demonstration data, with 5000 trajectories per task, totaling 1 million transitions generated by well-trained reinforcement learning policies. These data are utilized to evaluate imitation learning from expert demonstrations. In terms of scale, the dataset consists of 1 million transitions, and all its tasks are focused on imitation learning.