microduck_rl
microduck_rl Comparison and Selection: mjlab, Other Simulators, or Real-World Data
Choose a robotics RL stack by fidelity, throughput, assets, hardware path, reproducibility, and safety burden.
What you will learn
- Map environment, simulator, and runner
- Run a reproducible simulation
- Identify sim-to-real gates
Before you start
- 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.
Key takeaways
- Simulator choice follows fidelity and handoff needs.
- Equal tasks prevent misleading throughput comparisons.
- Safety and operator cost belong in selection.
Match the research question
microduck_rl is a focused Microduck/mjlab environment surface. A general simulator may offer more robots and sensors; real-world data may capture reality better but costs hardware time and carries safety risk.
Use the pinned source evidence, record versions and assumptions, and ask a qualified reviewer to validate the result before adopting it.
Run an equal task
Compare the same observation/action contract, reward, disturbances, seeds, rollout length, success metric, and policy budget. Include environment setup, debugging, export, and operator training costs.
Choose a handoff
Prefer the stack with the clearest asset license, telemetry, reset semantics, and sim-to-real plan. Document when a checkpoint must be discarded or retrained.
Use the pinned source evidence, record versions and assumptions, and ask a qualified reviewer to validate the result before adopting it.
Decision guide
| Criterion | Option A | Option 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 |
Implementation steps
- 1
Define task, fidelity, data, and safety constraints.
- 2
Benchmark candidate stacks equally.
- 3
Compare tooling, assets, and export paths.
- 4
Document choice and reevaluation trigger.
Copy-ready example
requirements -> mjlab | other simulator | real data
contract + task fixture -> quality/throughput/safety
operator + export cost -> decision + exitFrequently asked questions
Is a general simulator automatically better?
No. A focused environment can be easier to reproduce and debug for the target robot.
When should real-world data enter?
After simulation baselines and safety instrumentation are ready, with qualified operators and a bounded data plan.
Sources
- microduck_rl repositorySource checked 2026-08-31