Chuyển tới nội dung chính

Introduction

The Problem​

Current traffic control systems (fixed-timer or semi-smart) are designed with a single goal: optimize traffic flow (reduce wait times, reduce congestion). However, they are completely "blind" to a critical factor: localized environmental impact. Continuous stop-and-go traffic or prolonged idling at an intersection creates "pollution hotspots".

Example: A system might optimize for fast vehicle flow but inadvertently push a large amount of PM2.5 and CO emissions into an intersection near a school or hospital. This is a failure in terms of public health optimization.

The Core Solution​

GreenWave builds a real-time adaptive traffic signal control system using AI (Reinforcement Learning) to simultaneously optimize two objectives (multi-objective):

  1. Traffic Goal: Minimize average wait time and queue length.
  2. Environment Goal: Minimize estimated emissions and/or observed air pollution indices at the site.

The system is no longer "blind"; it makes trade-offs between these two goals.

Architecture & Data Flow​

Architecture

The system revolves around Smart Data Models:

  1. Input Layer:
    • Traffic: AI Cameras (counting, classification), Induction Loops.
    • Environment: Air Quality Sensors (PM2.5, CO, NO2) at intersections feeding AirQualityObserved.
  2. Processing Layer:
    • Camera data -> TrafficEnvironmentImpact (Estimated impact).
    • Sensor data -> AirQualityObserved (Real-time impact).
  3. Decision Layer (AI Agent):
    • State: Traffic queues, current light phase, AirQualityObserved, TrafficEnvironmentImpact.
    • Action: Extend Green, Red Now, Priority Phase.

Goals​

Intelligent Control

  • AI-Driven Traffic Coordination - DQN-based adaptive signal control
  • Multi-Objective Optimization - Balance traffic flow & air quality
  • Real-Time Decision Making - Sub-second response to traffic changes

Environmental Awareness

  • Air Quality Monitoring - PM2.5, CO, NO2 sensors at intersections
  • Emission Estimation - Traffic-based pollution prediction
  • Pollution Hotspot Prevention - Protect sensitive areas

Monitoring & Visualization

Admin Dashboard - Comprehensive control panel with AI/manual modes

  • Public Air Quality Portal - User-facing environmental metrics
  • Historical Analytics - Time-series data visualization
  • Real-Time Monitoring - Live dashboards and analytics for instant insights
  • Control Panel By Hand/AI - Manual and AI modes for traffic control
  • Manage Sensors - Add, remove, and configure sensors
  • Manage Subscriptions - Subscribe to real-time data streams

DevOps & Standards

  • CI/CD Pipeline - Automated testing & deployment
  • OpenAPI Documentation - Comprehensive API specs
  • Open Source Compliance - MIT License, Contributing Guidelines

Quick Started​

Prerequisites​

  • Docker 28.3.2: To run the Orion-LD Context Broker. Download at Docker
  • Node 24.6.0: To run the frontend. Download at NodeJS
  • SUMO 1.25.0: To run the SUMO Traffic Simulation. Download at SUMO

Start your site​

Run the following command to start all services (Context Broker, Backend, Frontend, and Documentation):

cp .env.example .env
cd src/backend
cp .env.example .env
docker compose up -d

Once the services are up and running, you can access them at the following addresses: