zijian2022
ACT sim eval: FANUC ER-4iA single-screw pick
Isaac Sim evaluation rollouts of an ACT policy trained on real FANUC ER-4iA screw insertion data, tested on unseen object poses with 10/11 success rate across seeds 100-110.
Downloads13
Episodes11
Why This Matters for Physical AI
This evaluation dataset demonstrates sim-to-real transfer of visuomotor policies for precise robotic manipulation, validating ACT policies trained on real robot data when executed in simulation under distribution-shifted object poses.
Technical Profile
- Modalities
- rgb
- Robot Embodiments
- FANUC ER-4iA
- Action Space
- end_effector_delta
- Environment
- simulation
- Task Types
- pick_and_placeinsertion
- Episodes
- 11
- Data Format
- LeRobot
- Annotation Types
- action_labelsreward_labels
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack