mulligan2026MIT
sim-square-broad-c01-dagger-sobol-no-cf
Part of the Mulligan project, this dataset contains 100 episodes of simulated square manipulation tasks using DAGGER training with Sobol sampling and no counterfactual augmentation. Data is provided in parquet format with state-only observations.
Downloads75
Episodes100
Why This Matters for Physical AI
This dataset contributes to physical AI research by providing simulated manipulation trajectories generated through DAGGER imitation learning with controlled sampling strategies, enabling study of policy learning methodologies and data augmentation effects.
Technical Profile
- Modalities
- proprioception
- Environment
- simulation
- Task Types
- manipulation
- Episodes
- 100
- Data Format
- parquet
- License
- MIT
Access
Need custom proprioception data?
Claru builds purpose-built datasets for simulation applications with dense human annotations and quality assurance.
Request a Sample Pack