Deployment¶
A hands-on playbook for moving a UniLab policy across hardware, simulation backends, and source frameworks. Each tutorial follows the same shape:
What you start with — the trained artefact and config.
What changes — the minimal set of edits in code, YAML, and assets.
How you validate — concrete commands and checkpoints.
Choose your journey¶
Prepare a trained policy for G1 / Go2 / Allegro bring-up with ONNX exports and deploy-side contract checks.
Switch the same task between MuJoCo and Motrix without retraining from scratch.
Bring tasks over from Isaac Lab / Legged Gym / rsl_rl / skrl.
🤖 Sim → Real¶
End-to-end pipeline + go/no-go checklist.
29-DoF humanoid; motion-tracking deploy.
Joystick, rough terrain, Go2W wheels.
Cube rotation; friction + vision.
Training playback exports, ONNX Runtime checks, and deploy prototype inputs.
Priority-ordered DR recipes.
Soft limits, EMA, e-stop, watchdog.
Training-side latency knobs and deploy-side measurement checks.
Symptom → cause → fix cookbook.
🔀 Sim → Sim (MuJoCo ↔ Motrix)¶
🔁 Framework Migration¶
GPU-resident → CPU + shared-mem.
Class-based env → NpEnv.
Trainer split: collector + learner.
Algo coverage and trade-offs.
Side-by-side field map.
Common reward terms in UniLab style.