DecentVLA2026apache-2.0
moplan_Cube_Blue_on_Red_franka_left
A simulated Franka Panda robotic arm performing cube stacking tasks in Isaac Sim, where actions are generated by a motion planner discovered through LLM-driven program search (GEPA), not teleoperated demonstrations.
Downloads52
Episodes99
Hours0.5746
Why This Matters for Physical AI
This dataset provides a controlled evaluation of learned motion planning behaviors for manipulation tasks with consistent embodiment and action space, useful for studying how models generalize across different scene configurations with the same underlying policy.
Technical Profile
- Modalities
- rgbproprioception
- Robot Embodiments
- Franka Panda
- Action Space
- 8-DoF: panda_joint1, panda_joint2, panda_joint3, panda_joint4, panda_joint5, panda_joint6, panda_joint7, gripper
- Environment
- simulation
- Task Types
- manipulationpick_and_placecube_stacking
- Episodes
- 99
- Total Hours
- 0.5746
- Data Format
- LeRobot
- Annotation Types
- environment_state
- License
- apache-2.0
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack