Deep Reinforcement Learning Approach for Urban Traffic Signal Control

Overview

This repository presents a centralized deep reinforcement learning (DRL) framework for adaptive traffic signal control in urban road networks.
The system learns optimal traffic light phase decisions by interacting with a traffic simulation environment, with the objective of minimizing congestion, vehicle waiting time, and queue lengths.

The project is designed as a research-oriented prototype, suitable for academic evaluation, experimentation, and further extension.

---

Key Objectives

• Replace static traffic light timing with adaptive, learning-based control
• Optimize traffic flow using Deep Reinforcement Learning
• Evaluate system performance using a simulated urban traffic environment
• Provide a modular and extensible codebase for experimentation

---

System Architecture

The system follows a standard reinforcement learning pipeline:

1. Traffic Simulator

   • Simulates urban traffic dynamics
   • Provides environment state (vehicle density, queue lengths, etc.)

2. RL Environment Wrapper

   • Converts simulation state into an RL-compatible format
   • Computes reward signals based on traffic efficiency metrics

3. Deep RL Agent

   • Learns optimal traffic signal actions
   • Uses neural networks to approximate decision policies

4. Training & Evaluation Pipeline

   • Training loop for policy optimization
   • Evaluation module for performance assessment

---

Reinforcement Learning Formulation

• State Space

  • Encodes traffic conditions such as vehicle counts, waiting times, or lane occupancy

• Action Space

  • Traffic signal phase selection or phase switching decisions

• Reward Function

  • Designed to penalize congestion and delay
  • Encourages smoother traffic flow and reduced waiting time

---

Installation

Prerequisites

• Python 3.7+
• Traffic simulator (e.g., SUMO)
• Required Python packages

Setup

git clone https://github.com/cosmicbit/Deep-Reinforcement-Approach-on-Traffic-System.git
cd Deep-Reinforcement-Approach-on-Traffic-System
pip install -r requirements.txt