Real-deployment anchor data, shared benchmarks, and reproducibility standards for embodied intelligence in operational environments. LeRobot-native. Community-run. Open.
Physical AI needs its own data layer, benchmarks, and reproducible baselines. Smiling Buddha builds and publishes them openly, one area at a time.
Structured, versioned, LeRobot v2.1 native. Real-deployment anchor data at V1–V3 fidelity.
Standardised evaluation suites for perception, planning, and long-horizon manipulation in physical environments.
A published framework for extending narrow real-world captures into rich simulation training data with V0-V3 validation scoring.
CLIs, converters, and validation tools for physical AI data engineering. LeRobot / ROS 2 / PyTorch compatible.
Our flagship methodology publication. Describes how narrow real-world captures can be systematically extended into rich simulation training data, with a V0 → V3 validation taxonomy.
Frontier robotics models converge on the same bottleneck: narrow real-world data is expensive to capture but generalises poorly. Pure simulation is cheap but suffers from the sim-to-real gap. We propose Real-to-Sim Enrichment (RSE), a systematic method for turning a narrow real capture into a distribution of validated simulation variants, scored across four validation levels (V0 → V3) representing physical fidelity and behavioural transferability.
Freelancers, students, PhD candidates, engineers between roles, working professionals — the Edge Case Library is being built by a distributed community. Named recognition. Remote-first.
Join us →