MuJoCo Robotic Locomotion and Manipulation Environments

该数据集包含了多种MuJoCo机器人运动（如蚂蚁、人形、半猎豹）和环境操作（如FetchPush、FetchSlide、FetchPickAndPlace）场景，用于评估不同的技能发现方法。这些环境旨在促进对通过无监督学习方法所掌握技能的评价，并且该数据集被用来比较所提出的LSD方法与几种基线方法的表现。该数据集覆盖了运动和操作任务，规模广泛，任务涉及无监督技能发现和目标跟踪任务。

This dataset contains a range of MuJoCo robot locomotion scenarios (e.g., Ant, Humanoid, HalfCheetah) and manipulation environments (e.g., FetchPush, FetchSlide, FetchPickAndPlace), tailored for evaluating diverse skill discovery methods. These environments are designed to facilitate the assessment of skills mastered through unsupervised learning approaches, and the dataset is employed to compare the performance of the proposed LSD method against several baseline methods. The dataset covers a broad spectrum of locomotion and manipulation tasks, including both unsupervised skill discovery and object tracking tasks.