ckx666/Robotwin_v2_eef_lerobot_v3
RoboTwin V2 数据说明(本目录) 总结:为什么要单独做 EE(末端位姿)数据集? RoboTwin 官方提供的 LeRobot 转换链路默认读 joint_action,因此常见的 robotwin_official_* 里 observation.state / action 是 14 维关节角 + 夹爪,适合复现官方 pi0、在关节空间做模仿学习。 但在统一动作空间 的训练与部署里,更自然的是用 末端执行器绝对位姿(位置 + 四元数 + 夹爪):与 end.position / effector.position 配置对齐、便于和真机 EE 控制接口一致,也避免在策略头里再隐式学一遍「关节 → 末端」的映射。 原始 HDF5 里其实一直带有 endpose/,只是官方脚本没有把它写进 LeRobot。本目录在保留原始 hdf5/ 与官方 joint 版 lerobot/robotwin_official_* 的前提下,额外提供了从 endpose 重转 的 lerobot/robotwin_ee_{clean… See the full description on the dataset page: https://huggingface.co/datasets/ckx666/Robotwin_v2_eef_lerobot_v3.
| Author / organization | ckx666 |
|---|---|
| License (as declared) | mit |
| Downloads | 209 (as reported by Hugging Face) |
| Likes | 0 |
| Created | 9 October 2026 |
| Last updated at source | 12 October 2026 |
Popularity, last 3 days
209 now (+5 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.