mulligan2026mit
sim-square-broad-c01-dagger-mixed
A robotic manipulation dataset from the Mulligan project containing 284 episodes of a sim-square-broad task collected through mixed sources including policy rollouts and raw collections. The dataset includes multi-view observations from an agent view camera and robot eye-in-hand camera.
Downloads72
Episodes284
Why This Matters for Physical AI
This dataset contributes to physical AI research by providing multi-view robotic manipulation trajectories from simulated environments that can be used for training and evaluating robot learning policies through imitation and reinforcement learning approaches.
Technical Profile
- Modalities
- rgb
- Robot Embodiments
- robot0
- Environment
- simulation
- Task Types
- manipulation
- Episodes
- 284
- Data Format
- parquet
- Annotation Types
- reward_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