MoonRLLab

MoonRLLab is a small reinforcement-learning lab written in MoonBit for the 2026 MoonBit Software Synthesis Challenge.

Repository links:

• GitLink: https://gitlink.org.cn/lzylzy78/MoonRLLab
• GitHub: https://github.com/liuzhiyug/MoonRLLab

Project policy:

• default branch for submission: master
• primary contributor: 刘智宇 <2579597201@qq.com>
• no external contributor names are used in the history

It focuses on a practical but expandable slice of the RL stack:

• Environment: finite, discrete environments with reset/step/render support
• Policy: epsilon-greedy action selection
• Agent: Q-learning and SARSA tabular control
• Trainer: episode loops and run summaries
• Logger: readable progress output

What this project tries to solve

The MoonBit ecosystem has many language and systems building blocks, but fewer end-to-end examples for experimentation workflows. MoonRLLab fills that gap by offering a compact lab for:

• trying reinforcement-learning ideas in MoonBit
• studying a clear environment/agent separation
• extending the framework toward other discrete control tasks

Default demo

The repository ships with a 4x4 GridWorld demo.

The demo is intentionally small, but the structure is designed to extend to:

• other discrete environments
• different exploration strategies
• n-step or eligibility-trace learners
• richer reporting and visualization

Build and run

moon check
moon test
moon run cmd/main

Source notes

This project is newly authored for the competition. It does not copy upstream RL code. The design is informed by standard tabular reinforcement-learning references and the MoonBit textbook and toolchain docs.

License

Apache-2.0