arth-shukla/MS-HAB-PrepareGroceries

ManiSkill-HAB PrepareGroceries数据集是一个用于机器人操作和低级别控制的演示数据集。每个子任务/对象组合（例如pick 002_master_chef_can）包含1000个成功的episode，每个episode包含200个样本/演示。数据是通过RL策略生成的，并通过基于规则的事件标签系统过滤以确保机器人行为的安全性。数据集包含Pick和Place两个子任务，相对于其他MS-HAB长期任务（如TidyHouse、SetTable），PrepareGroceries Pick和PrepareGroceries Place属于高难度任务。数据集的结构包括每个子任务/对象组合的JSON和HDF5文件，分别包含元数据和演示数据。数据集是纯合成的，使用ManiSkill的默认渲染生成数据。

The ManiSkill-HAB PrepareGroceries Dataset is a whole-body, low-level control/manipulation demonstration dataset for ManiSkill-HAB PrepareGroceries. Each subtask/object combination (e.g., pick 002_master_chef_can) has 1000 successful episodes (200 samples/demonstration) gathered using RL policies filtered for safe robot behavior with a rule-based event labeling system. PrepareGroceries contains the Pick and Place subtasks. Relative to other MS-HAB long-horizon tasks (such as TidyHouse, SetTable), PrepareGroceries Pick and PrepareGroceries Place are of hard difficulty (on a scale of easy-medium-hard). The dataset can be used to train vision-based learning from demonstrations and imitation learning methods, and it may also serve as synthetic data for computer vision tasks.