insagur2026apache-2.0
PointBridge HA2 demos (orange pick-and-place, OpenArm + RH56F1)
Success demonstrations of a contact-aware Point-Bridge / VITRA base policy in MuJoCo for offline imitation learning with residual TD3-BC, collected in a precisely-specified simulation environment with an orange pick-and-place task into a bowl.
Downloads0
Episodes300
Why This Matters for Physical AI
This dataset provides contact-aware demonstration data for training residual reinforcement learning policies that combine imitation learning with model-free RL, enabling sample-efficient learning of dexterous manipulation tasks in simulation.
Technical Profile
- Modalities
- proprioceptioncontact
- Robot Embodiments
- OpenArm + RH56F1 handGR1 (Inspire-mappable)
- Action Space
- joint_positions
- Environment
- simulation
- Task Types
- pick_and_placegraspingmanipulation
- Episodes
- 300
- Data Format
- HDF5
- Annotation Types
- contact_labelssuccess_labels
- License
- apache-2.0
Access
Need custom proprioception data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack