nurtayev-dmit
seasondish-try-1-raw
Raw recorded trajectories of MikasaSeasonDish-v0 task with 1000 episodes at 10 Hz. Robot must remember a marked condiment cue, pick it up, and hold it over a bowl while leaving the other condiment untouched.
Downloads8
Episodes1000
Why This Matters for Physical AI
This dataset demonstrates partial observability and memory requirements in manipulation tasks, where robots must reason about past observations to solve complex multi-step object interaction problems.
Technical Profile
- Modalities
- rgbproprioception
- Robot Embodiments
- Fetchds_fetch
- Action Space
- joint_positions
- Environment
- simulation
- Task Types
- manipulationpick_and_placememory
- Episodes
- 1000
- Data Format
- HDF5
- Annotation Types
- reward_labelsaction_labels
- License
- mit
Access
Need custom rgb data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack