ivogeorg/aic-cable-insertion-gazebo-v3

AIC灵巧电缆插入——Gazebo (LeRobot v3)数据集是一个模拟演示数据集，针对AI for Industry Challenge (AIC) 2026资格赛阶段的灵巧电缆插入任务。该数据集在Gazebo AIC模拟器中记录，来自自主演示策略，并以LeRobot v3格式打包。数据集包含274个片段、289,753帧，帧率为20 FPS，任务为将电缆连接器插入正确的目标端口。特征包括三个手腕摄像头图像（左、中、右）、臂关节位置（6自由度）、手腕力/扭矩扳手和TCP姿态目标。数据集专为模仿学习（如ACT）设计，但仅限于特定任务和模拟器（Gazebo、AIC任务板），且所有演示都是合成/自主生成的，非人类远程操作。The AIC Dexterous Cable Insertion — Gazebo (LeRobot v3) dataset is a simulated demonstration dataset for the dexterous cable-insertion task from the AI for Industry Challenge (AIC) 2026, Qualification phase. Recorded in the Gazebo AIC simulator from an autonomous demonstrator policy and packaged in the LeRobot v3 format. The dataset includes 274 episodes, 289,753 frames, at 20 FPS, with the task being Insert the cable connector into the correct target port. Features consist of three wrist-mounted camera images (left, center, right), arm joint positions (6 DoF), wrist force/torque wrench, and TCP pose target. It is intended for imitation learning (e.g., ACT) but is task- and simulator-specific (Gazebo, AIC task board), with all demonstrations being synthetic/autonomously generated, not human teleoperation.