duae26/-traffic-signal-optimization-Blida

Hybrid LSTM + Meta-Heuristic Traffic Signal Optimization — Blida, Algeria README.md # Hybrid Predictive Meta-Heuristic Framework for Traffic Signal Optimization **M2 Thesis — University of Yahia Fares Medea — 2025-2026** A hybrid framework combining LSTM-based traffic prediction with meta-heuristic optimization (WOA, HHO, GA) for dynamic traffic signal control applied to a real 9-intersection network in Blida, Algeria. --- ## Key Results | Method | Low | Medium | Peak | | ------------- | ------ | ------ | ------ | | GA (reactive) | -11.1% | -7.7% | -19.1% | | WOA + LSTM | +18.8% | +27.0% | +18.8% | | HHO + LSTM | +27.9% | +38.2% | +34.6% | Improvement relative to fixed-time baseline (mean over 5 runs). --- ## Requirements - Python 3.10+ - SUMO 1.25.0 - PyTorch 2.0 - See requirements.txt for full list Install dependencies: ```bash pip install -r requirements.txt ``` --- ## How to Run **1. Run all experiments (baseline + WOA + HHO + GA):** ```bash python run_experiment.py ``` **2. Run statistical validation (5 independent runs):** ```bash python run_5x.py ``` **3. Generate all figures and Excel table:** ```bash python generate_results.py ``` --- ## Project Structure ├── src/ │ ├── utils/ # TraCI interface, fitness function, TLS IDs │ ├── lstm/ # LSTM predictor module │ ├── woa/ # Whale Optimization Algorithm │ ├── hho/ # Harris Hawks Optimization (with signed fix) │ └── ga/ # Genetic Algorithm (reactive benchmark) ├── models/ # Trained LSTM weights and scalers ├── sumo_network/ # Blida road network and scenario configs ├── results/ # Figures and statistical summary ├── controller.py # Rolling horizon control loop ├── run_experiment.py # Main experiment runner ├── run_5x.py # Statistical validation └── generate_results.py # Results visualization --- ## Network The road network covers the Bab Dzair district of Blida city center (36.4670°N–36.4745°N, 2.8250°E–2.8360°E), extracted from OpenStree …