Romulan War (Dodger) · Arcade Lab

Dodge-survival game with a polished desktop build (Python/pygame), a reliable web hub (HTML5 canvas — no WebAssembly) including Neon Circuit (car arena with human / AI / versus + engine audio), and an ML / RL stack: tabular Double Q on dodge, DQN on a discrete car-ring env, documented in ml/ALGORITHMS.md.

The site is served from the web/ folder (plain HTML + Canvas + JavaScript). Pygbag is no longer required — deployment is a simple static upload.

1. Push main and wait for Actions → Deploy to GitHub Pages to finish.
2. Settings → Pages → source Deploy from a branch → gh-pages / (root).
3. Open https://<user>.github.io/<repo>/ (e.g. https://pat749.github.io/dodger-game/).

Hard-refresh after updates (Ctrl+Shift+R). Use a Chromium-based browser for best results.

python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
cd game
python main.py

Legacy launcher: python dodger.py in Dodger-game python/Dodger-master/ (forwards to game/main.py).

See ml/README.md and ml/ALGORITHMS.md. Summary:

• ml/dodge_env.py + ml/train_q.py — tabular Q / Double Q → ml/checkpoints/q_table.json.
• ml/car_env.py + ml/train_dqn_car.py — DQN (PyTorch) → ml/checkpoints/dqn_car.pt.

pip install -r requirements-ml.txt
python3 -m ml.train_q --seed 0

pip install -r requirements-ml-rl.txt
python3 -m ml.train_dqn_car --steps 20000
python3 -m ml.eval_dqn_car --episodes 300
python3 -m ml.export_dqn_car_web   # optional: Neon Circuit AI in the browser (see web/dqn_car_policy.json)

CITATION.cff is included for GitHub’s “Cite this repository” widget.

• Old Jekyll / docs errors: use gh-pages + / (root), not /docs, for the live game (see docs/ note in earlier commits if applicable).
• Pygbag black screen: use the web/ build instead; it is the supported browser target now.

Extended from the classic Dodger-style pygame tutorial; web and RL layers added for teaching and research demos.