3D Immersive Traffic Simulation

Inspiration# Multi-Agent Traffic Simulation using Google ADK

Inspiration

Traffic congestion and road safety remain major challenges in urban environments. We wanted to explore how Agentic AI could improve decision-making by simulating the interactions between pedestrians, drivers, traffic signals, and safety systems. Our goal was to demonstrate how multiple AI agents can collaborate to make intelligent traffic management decisions in real time.

What We Built

We developed a Multi-Agent Traffic Simulation Platform powered by Google ADK and Gemini 2.5 Flash. The system consists of specialized agents responsible for traffic supervision, pedestrian behavior, driver actions, traffic signal control, and safety assessment. These agents collaborate to evaluate traffic conditions and generate coordinated recommendations for safe road usage.

The application is deployed using Google Cloud Run for the AI backend and Firebase Hosting for the frontend experience.

How We Built It

Our architecture combines:

Google ADK for agent orchestration
Gemini 2.5 Flash for reasoning and decision-making
Flask-based backend APIs
Firebase Hosting for frontend deployment
Cloud Run for scalable backend hosting
Three.js-based traffic simulation interface

The decision logic evaluates traffic density, weather conditions, and signal status to determine pedestrian crossing permissions, vehicle movement, and overall safety risk.

A simplified risk model is:

$$ Risk = f(TrafficDensity, Weather, SignalState) $$

where higher traffic density increases risk and affects agent decisions.

Challenges We Faced

One of the biggest challenges was orchestrating multiple agents while ensuring consistent decision outputs. We also faced deployment challenges related to containerization, API configuration, Cloud Run deployment, and frontend-backend integration. Managing secure API access while maintaining a smooth user experience required careful architecture decisions.

What We Learned

This project provided valuable experience with Agentic AI design patterns, multi-agent orchestration, cloud deployment, and real-time AI-powered decision systems. We learned how specialized agents can collaborate effectively to solve complex problems and how Google ADK simplifies the development of intelligent multi-agent workflows.