AI-driven Smart Urban Traffic Management System

01 / PROBLEM

Urban congestion can increase travel delay, vehicle idle time, fuel use, and vehicle-related environmental impact. Static dashboards alone do not show how a traffic prediction can inform a decision or how that decision might perform under controlled conditions.

02 / APPROACH

An academic decision-support prototype that follows an Observe → Predict → Decide → Simulate → Measure workflow. It turns predicted congestion and demand into bounded signal-timing recommendations, then compares an AI-assisted strategy with a fixed-time baseline in a simulated traffic network.

03 / KEY FEATURES

• Multi-junction command center for three simulated urban junctions
• Random Forest congestion classification with Low, Moderate, and High classes and prediction probabilities
• Short-horizon traffic-demand forecasting
• Adaptive green-time recommendations constrained to 20–90 seconds with an exact 180-second cycle
• Digital-twin-style simulation of arrivals, queues, signal states, vehicle movement, delay, throughput, and idle time
• Fixed-time versus AI-assisted strategy comparison under the same simulated conditions
• Model-based fuel, CO₂, and PM2.5 estimates derived from simulation outputs
• Simulated environmental signals and bounded manual signal overrides

TECHNOLOGY

ARCHITECTURE

Simulated traffic inputs → feature preparation → Random Forest congestion classifier and demand forecaster → constraint-aware signal optimizer → traffic simulation → environmental estimates → command-center dashboard.

IMPLEMENTATION

The system separates model predictions from signal recommendations. A bounded optimization layer applies the 20–90 second green-phase limits and exact 180-second cycle before recommendations are evaluated in a queue-based simulation. The dashboard presents the prediction, forecast, signal plan, simulation comparison, and environmental interpretation as one workflow.

OUTCOMES & NOTES

The repository documents an end-to-end prototype that connects ML prediction to constrained signal planning and simulation-based evaluation. Add measured experiment results here only after confirming them from the project outputs; no real-world congestion reduction or emissions reduction is claimed.

Academic prototype using synthetic/simulated traffic data. It does not control real traffic signals, connect to municipal infrastructure, or measure real-world emissions. Environmental values are model-based estimates, not direct sensor measurements.