Mujoco locomotion tasks

该数据集包含了三个连续控制机器人任务（Hopper、Walker和AntPusher）的专家演示数据，这些数据是从经过良好训练的PPO策略中生成的。该数据集被用于评估在机器人控制领域中，层次化策略学习方法的有效性。数据集涵盖了三个任务，这些任务具有不同的状态和动作维度（Hopper任务具有11维状态和3维动作；Walker任务具有17维状态和6维动作；AntPusher任务具有107维状态和8维动作）。这些数据集主要应用于机器人控制及模仿学习任务中。

This dataset contains expert demonstration data for three continuous control robotic tasks (Hopper, Walker, and AntPusher), generated from well-trained PPO policies. It is utilized to evaluate the effectiveness of hierarchical policy learning methods in the field of robotic control. The dataset covers three tasks with distinct state and action dimensions: the Hopper task has an 11-dimensional state space and 3-dimensional action space; the Walker task has a 17-dimensional state space and 6-dimensional action space; the AntPusher task has a 107-dimensional state space and 8-dimensional action space. This dataset is primarily applied to robotic control and imitation learning tasks.