microduck_rl
microduck_rl 快速上手:运行一次有边界的机器人 RL 回合
固定 mjlab 环境并检查观察、奖励、终止和轨迹。

你将学会
- Map environment, simulator, and runner
- Run a reproducible simulation
- Identify sim-to-real gates
开始前需要
- 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.
先看结论
- A small episode reveals the environment contract.
- Reward and termination must be visible.
- Simulation scores are not hardware validation.
Pin the experiment
固定代码、仿真器、CUDA/PyTorch、种子和 rollout,从最小环境开始,保存观察/动作形状、奖励、终止原因和视频。
Use the pinned source evidence, record versions and assumptions, and ask a qualified reviewer to validate the result before adopting it.
Inspect one episode
Before training, print observation/action shapes, reward terms, termination causes, and contact or sensor telemetry. Save a video or trajectory so a score has visible context.
Scale deliberately
Increase parallel environments and rollout length only after a baseline is reproducible. Keep simulation success separate from any claim about physical-robot safety.
Use the pinned source evidence, record versions and assumptions, and ask a qualified reviewer to validate the result before adopting it.
如何选择
| 比较维度 | 方案 A | 方案 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 |
实施步骤
- 1
Pin code, simulator, seed, and config.
- 2
Run one evaluation episode.
- 3
Inspect shapes, rewards, and telemetry.
- 4
Scale parallelism after parity.
可复制示例
git clone https://github.com/pollen-robotics/microduck_rl.git
cd microduck_rl
# follow the repository's pinned mjlab/PyTorch setup
# run the smallest environment and save a rollout常见问题
Can I start with hardware?
No. Begin in simulation and complete qualified sim-to-real and operator checks before hardware.
What should a bug report include?
Revision, seed, simulator stack, config, checkpoint, telemetry, and rollout artifact.
资料来源
- microduck_rl repository来源核查 2026-08-31