AI & Smart City Hackathon

Traffi

An intelligent traffic light signal optimization model designed to reduce congestion and vehicle idle times in the city of Trnava.

DomainDeep Reinforcement Learning
Target AreaTrnava Municipality

The Challenge

Fixed-timer signal light cycles struggle to adapt during peak transit hours in Trnava, leading to bottlenecked intersections, unnecessary idle emissions, and delayed public transport.

  • Inflexible timer intervals during unpredictable traffic shifts
  • Elevated CO2 emissions caused by excessive stop-and-go idling
  • Lack of real-time adaptive flow management across intersections

AI-Driven Solution

Traffi models key municipal intersections as a multi-agent reinforcement learning environment, dynamically tuning signal lengths based on live simulated traffic density.

Adaptive Signal Timing

Dynamically alters green light durations to maximize vehicle throughput.

Emission Reduction

Minimizes stop times to directly decrease urban carbon footprint.

Technical Architecture

Core Language

Python 3

Primary language powering data processing, simulation logic, and model training.

ML Framework

PyTorch

Deep Neural Networks and Reinforcement Learning agents driving decision policies.

Network Simulation

Traffic Flow Modeling

Simulated traffic topology modeled on real-world Trnava road networks.

View on GitHub

Check out the model implementation, source code, and project documentation on GitHub.