Neural Actuation Dataset (NAD)

面向低成本执行器动力学、外力估计和状态诊断的真机遥测数据；固定修订公开非gated，覆盖OpenManipulator-X、SO-101和Franka，但Franka只支撑未来状态条件离线评测。

开放状态: 已核验开放
许可证: MIT declared by dataset card
商业使用: MIT permits commercial use; verify upstream robot logs and privacy obligations
人工评审结论: verified_public_nongated_fixed_revision
机器人: ROBOTIS OpenManipulator-X、SO-101、Franka Emika Panda
任务: joint sweeps、pick and place、lift and hold、directional force contact、force-gauge pushing、motor-condition classification
模态: joint position、joint velocity、command、motor current/load、temperature、force/torque reference、RGB video for documentation
格式: CSV、video、Parquet mirror
采集/生成方式: hardware execution and platform-specific data-collection tools
已知问题: Franka uses a future-state proxy offline; torque head is a surrogate, not calibrated physical torque; full download and independent replay not completed.