RustyRunways

RustyRunways is a deterministic airline logistics simulation built in Rust with a CLI, GUI, and Python APIs. It’s designed for AI/ML/RL: train agents, prototype policies, and run fast, reproducible experiments in a realistic, constraint‑driven game world.

Keywords: AI, machine learning, reinforcement learning, RL environment, Gymnasium, Stable‑Baselines3, vectorized environments, simulation game, Rust game engine, Python bindings, deterministic seeds, parallel stepping.

---

Install

• Python (PyPI):

pip install rusty-runways
# with Gym wrappers
pip install 'rusty-runways[gym]'

• Build from source (Rust):

cargo build --release

Docs for CLI, GUI, and Python: https://dennislent.github.io/RustyRunways

---

GUI Frontends

• egui (native): crates/gui — current Rust native UI using eframe/egui.
• Tauri + React: apps/tauri — desktop shell (src-tauri) with a React UI (ui).

Dev notes (Tauri UI):

• Requires Tauri CLI v2: cargo install tauri-cli --locked --version ^2
• Quick start (one command): scripts/dev_tauri.sh
  • Starts the UI dev server on http://localhost:5173 and launches the desktop app via cargo tauri dev.
  • Note: Do not open the UI in a regular browser; use the desktop window.
• Manual (two terminals):
  • Terminal A: cd apps/tauri/ui && npm install && npm run dev
  • Terminal B: cd apps/tauri/src-tauri && cargo tauri dev
• Production build:
  • cd apps/tauri/ui && npm run build
  • cd apps/tauri/src-tauri && cargo tauri build

Both frontends can coexist; the Rust core lives in crates/core and is shared.

---

Local Dev & Checks

Run the full set of linters, formatters, Rust tests, and Python dev tests in one go:

scripts/check.sh
# or auto-fix Python lint/format issues
scripts/check.sh --fix

This will:

• Lint/format Python under crates/py with Ruff/Black
• Build and test all Rust crates (cargo fmt, clippy, test)
• Build and install the Python package (maturin develop) and run Python tests (pytest)

Gym wrappers live under the pure‑Python package rusty_runways and are tested if gymnasium is available (the check script installs it for you).

---

Quick Start (Python)

Engine bindings (single + vector envs):

from rusty_runways_py import GameEnv, VectorGameEnv

g = GameEnv(seed=1, num_airports=5)
g.step(1)
print(g.time(), g.cash())

venv = VectorGameEnv(4, seed=1)
venv.step_all(1, parallel=True)
print(venv.times())

Gym wrappers (optional, install extra):

from stable_baselines3 import PPO
from stable_baselines3.common.vec_env import DummyVecEnv
from rusty_runways import RustyRunwaysGymEnv, make_sb3_envs

vec_env = DummyVecEnv(make_sb3_envs(4, seed=1, num_airports=5))
model = PPO("MlpPolicy", vec_env, verbose=1)
model.learn(total_timesteps=10_000)

See Python docs for observation/action spaces, reward shaping, and vectorized stepping.

---

Why RustyRunways for RL/AI

• Deterministic seeds: reproducible training and evaluation.
• Vectorized environments: fast rollouts with optional parallel stepping.
• Compact action surface (refuel/load/unload/depart/advance/maintenance) and rich JSON observations.
• Real‑world constraints (runway length, fuel costs, payload, deadlines) ideal for decision‑making under constraints.
• Multi‑frontend: CLI for scripting, GUI for inspection, Python for training and research.

---

Links

• Docs: https://dennislent.github.io/RustyRunways
• PyPI: https://pypi.org/project/rusty-runways/
• Issues: https://github.com/DennisLent/RustyRunways/issues

Contributions welcome! Open issues and PRs for bug fixes, features, or RL tooling.