sanjar-normuradov-agile-robots/austin-sailor

该数据集是一个用于机器人技能模仿学习的开源数据集，包含240个episodes和353094帧数据，涵盖4个任务。数据集详细记录了机器人的状态（包括关节位置、末端执行器位置和夹爪宽度）、动作指令、图像观察（普通图像和腕部图像）以及时间戳等信息。数据预处理包括平滑零状态样本、过滤初始帧等步骤。数据集结构清晰，特征描述详尽，适用于机器人学习和控制算法的研究与开发。

This dataset is an open-source dataset for skill-based imitation learning in robotics, containing 240 episodes and 353094 frames across 4 tasks. It includes detailed records of robot states (such as joint positions, end-effector poses, and gripper width), action commands, image observations (regular images and wrist images), and timestamps. Pre-processing steps include smoothing zero-state samples and filtering initial frames. The dataset has a clear structure and detailed feature descriptions, making it suitable for research and development in robot learning and control algorithms.