Marcovldn/maniskill_distracting_arms

该数据集是一个用于机器人操作任务的数据集，专注于拉动立方体工具手臂场景。它包含四个配置：高难度动态环境、高难度静态环境、低难度动态环境和低难度静态环境。每个配置都包含训练集和测试集，特征包括图像观察、图像掩码、状态（浮点数组）、动作（8维浮点数组）、奖励（浮点值）、终止标志和截断标志。数据集适用于强化学习、机器人控制或模拟环境中的任务学习，旨在支持算法在复杂操作任务中的训练和评估。

This dataset is designed for robotic manipulation tasks, focusing on the pull cube tool arms scenario. It includes four configurations: hard dynamic environment, hard static environment, low dynamic environment, and low static environment. Each configuration contains train and test splits, with features such as image observation, image mask, state (float array), action (8-dimensional float array), reward (float value), terminated flag, and truncated flag. The dataset is suitable for reinforcement learning, robotic control, or task learning in simulated environments, aiming to support algorithm training and evaluation in complex manipulation tasks.