AI That Fixes Broken Traffic Lights Could Save Your Commute

Traffic lights rely on sensors to count cars and adjust timings, but those sensors often break or get blocked. When that happens, lights use guesses instead of real data, leading to jams and longer waits. Researchers created an AI called LMP-GNN that fills in those missing numbers by predicting how many cars are likely on each road.

The AI doesn’t change how traffic lights work—it just provides better guesses when sensors fail. In tests on five city traffic networks, it improved travel times by up to 13.74% when sensors were missing data in a specific pattern. The AI is also lightweight, running 81-95% faster than other methods while using far fewer computing resources.

Unlike other AI systems that require retraining or complex changes, this one works alongside existing traffic light rules. It’s designed to be transparent, so engineers can understand and trust its predictions. The goal isn’t to replace human decisions but to make traffic lights smarter when they’re flying blind.

Think of it like a GPS rerouting you when your usual route is blocked—except this AI reroutes entire city traffic flows to keep things moving smoothly.