SLMC_poke_sponge_nudge_car_v1

该数据集使用LeRobot工具创建，专为机器人学任务设计，特别是针对so101_follower类型的机器人。数据集包含20个完整的情节（episodes），总计1150帧数据，覆盖2个不同任务，采样频率为10 FPS。数据以Parquet格式存储，分为训练集（全部20个情节）。数据集结构丰富，包括动作特征（如末端执行器的位置、旋转和夹爪位置）、观测状态（与动作特征类似但顺序不同）、来自腕部和前部摄像头的图像观测（256x256 RGB视频，编码为AV1格式），以及高级特征如VFM（视觉基础模型）特征、VLM（视觉语言模型）特征、技能特征、物体特征和全文本特征。这些特征用于支持机器人控制、感知和决策任务，适用于机器学习和人工智能研究。

This dataset is created using the LeRobot tool and is specifically designed for robotics tasks, particularly for the so101_follower type of robot. It contains 20 complete episodes, totaling 1150 frames of data, covering 2 different tasks with a sampling frequency of 10 FPS. The data is stored in Parquet format and is divided into a training set (all 20 episodes). The dataset has a rich structure, including action features (such as end-effector position, rotation, and gripper position), observation states (similar to action features but in a different order), image observations from wrist and front cameras (256x256 RGB videos, encoded in AV1 format), as well as advanced features like VFM (Visual Foundation Model) features, VLM (Visual Language Model) features, skill features, object features, and full-text features. These features are used to support robot control, perception, and decision-making tasks, making the dataset suitable for machine learning and artificial intelligence research.