microduck_rl
microduck_rl: laboratorio sim-to-real reproducible
Task manifest, policy card, telemetría y puertas de hardware.
Qué aprenderás
- Map environment, simulator, and runner
- Run a reproducible simulation
- Identify sim-to-real gates
Antes de empezar
- Basic Git and command-line usage
- Comfort reading a project README
You can explain the project, run its documented first step, and decide what to verify before adopting it.
Conclusiones clave
- Policy cards connect training to hardware review.
- Replayable fixtures make regression visible.
- Hardware gates should be enforced progressively.
Project brief
La policy card conserva revisiones, reward, semilla, checkpoint y evaluaciones y evoluciona de simulación a gates de export y firma humana.
Use the pinned source evidence, record versions and assumptions, and ask a qualified reviewer to validate the result before adopting it.
Policy card
Record repository/simulator revisions, assets, seed, observation/action schema, reward and termination, training budget, checkpoint hash, evaluation seeds, telemetry, and reviewer. Keep raw data outside Git; commit sanitized fixtures and hashes.
Staged roadmap
Begin trace-only in simulation, add advisory regression and domain-randomization checks, then enforce export gates for timing, saturation, sensor dropout, watchdog, and operator sign-off. Accessible Markdown/SVG remains canonical; 3D views are optional.
Use the pinned source evidence, record versions and assumptions, and ask a qualified reviewer to validate the result before adopting it.
Cómo elegir
| Criterio | Opción A | Opción B |
|---|---|---|
| Best when | You need predictable behavior and easy auditing | You need adaptive optimization and have reliable telemetry |
| Main risk | May leave performance on the table | Can become difficult to explain or debug |
Pasos de implementación
- 1
Define task, policy, telemetry, and review fields.
- 2
Add deterministic simulation fixtures.
- 3
Build disturbance and export checks.
- 4
Require qualified operator approval before hardware.
Ejemplo para copiar
policy:
task: microduck_walk
simulator_revision: <sha>
seed: 42
reward_hash: <hash>
checkpoint_sha256: <hash>
eval: { success: <value>, disturbances: <suite> }
hardware_gate: holdPreguntas frecuentes
Should Three.js be part of the lab?
Only as an optional exploration layer; text, metrics, and local SVG must remain accessible and complete.
Can a policy card replace operator review?
No. It organizes evidence but does not replace qualified safety approval.
Fuentes
- microduck_rl repositoryFuente verificada 2026-08-31