Traffic Flow Optimizer

Description

This project is a Python-based framework designed to automate the optimization of urban traffic flow. The system integrates three major components:

• Perception – Detects and tracks vehicles from video data.
• Flow Prediction – Computes real-time traffic flow metrics and predicts future trends using machine learning models.
• Flow Optimization – Determines network flow rate for different routes to maximize throughput and regulate congestion.

Repository Structure

.
├── dataset
├── optimization
│   ├── linear-program/
│   │   ├── network_model/  # OOP classes for building traffic network
│   │   ├── lp_solver.py    # Optimization solver via linear program
│   │   └── main.py
│   └── README.md
├── perception
│   ├── model/              # Trained model weights, not tracked
│   ├── results/
│   └── src
│       ├── utils/ 
│       └── main.py
├── prediction
│   ├── results
│   │   ├── analysis/ 
│   │   └── dataframe/ 
│   └── src
│       ├── random_forest/
│       ├── utils/ 
│       └── main.py
├── README.md
└── requirements.txt

Further details about each functional module are described below.

Functional Overview

1. Perception

The perception portion of the framework handles the following tasks:

• Trains YOLOv8 model for vehicle detection.
• Tracks detected objects using DeepSort.
• Iterates through .mp4 videos in dataset/vid (local only, not committed).
• Exports frame-by-frame detection and tracking results to perception/results in the following JSON structure:

{
    "vid_id": "vid_id",
    "detections": [
        [
            {
                "TRACK_ID": [X_1, Y_1, X_2, Y_2]
            },
            ...
        ],
        ...
    ]
}

2. Flow Prediction

The flow prediction module is responsible for:

• Processes each JSON file in perception/results to compute traffic metrics: average speed, density, and flow.
• Uses these features to train a Random Forest regression model for traffic flow prediction.
• Evaluates performance using standard metrics such as MAE and RMSE.

3. Signal Timing Optimization

The optimization module includes:

• Simulating traffic networks using directed graphs.
• Determining capacities, thresholds, and flow constraints for each edge and path.
• Formulating graph-based Linear Programming (LP) problems to optimize network flow and regulate congestion.
• Detailed mathematical setup is documented in ./optimization/README.md.