Skip to content
Embodied AI · Open research stack

From first principles to physical action.

Learn the foundations, execute complete pipelines, and judge every result by its evidence. One bilingual system for understanding and building embodied intelligence.

Start learning View field map Explore pipelines 中文
01ObserveSensors and state
→
02UnderstandGoals and world models
→
03ActPolicy, control and safety
→
04LearnFeedback and evidence
14Foundation lessons
10Engineering pipelines
5Evidence levels
EN · 中文Bilingual entry

Choose your route

01 · LEARN

Build the mental model

Math, learning, frames, kinematics, sensing, control, systems, safety and evaluation.

 02 · ORIENT

Choose a research direction

See capabilities, prerequisites, pipeline contracts, evidence and deliberate non-claims in one map.

 03 · BUILD

Run one complete system

Start from inputs, execute every stage, retain artifacts, and check an explicit promotion gate.

 04 · MEASURE

Compare with context

Read protocol, data budget, episode count, negative results and raw-artifact boundaries together.

System coverage today

SMOKESimulation & dataTask → expert → episodes → QA SMOKEVLA policyMultimodal data → policy → closed loop SMOKEWorld modelsTransitions → prediction → rollout → planning SMOKERL post-trainingMDP → PPO → evaluation → regression INTERFACERFM / cross-embodimentCanonical schema → adapter → safety INTERFACEEmbodied reasoningInstruction → subgoals → skills → replan SMOKEDexterous retargetingLandmarks → geometry → optimization → time DOCSim-to-RealReplay → HIL → shadow → guarded rollout DOCPerception & stateCalibration → sync → fusion → uncertainty DOCNavigation & locomotionState → planning → control → recovery

Evidence before claims

Execution is not performance, and simulation is not hardware validation.

DoF separates import, smoke execution, deterministic tests, benchmark evidence, and hardware validation. A lower level never implies a higher one.

Read the validation policy →

Verify the repository

python scripts/run_pipeline.py --validate
python scripts/run_pipeline.py --list
python scripts/audit_repository.py
python -m pytest tests/ -q

The root English README and 中文 README are the concise product landing pages. This site is the deeper learning and engineering layer.