A traffic light on a fixed timer is one of the dumbest pieces of infrastructure you interact with daily. It holds you at a red for ninety seconds while the cross street sits empty, because it's following a schedule written years ago that has no idea what's actually happening at the intersection right now. It's open-loop control — acting without ever looking at the result. Fixing that is the point of my Adaptive Traffic Light Controller.
Open-loop control acts blind
A fixed-timer light runs the same cycle regardless of reality. Rush hour or 3am, packed queue or empty road — same timing. That's the signature of open-loop control: it executes a predetermined plan and never measures whether the plan is working. It can't adapt, because it never looks.
The failure is baked in. The moment conditions differ from the assumptions the schedule was built on — which is almost always — the system is doing the wrong thing with total confidence. No amount of tuning the fixed schedule fixes a system that fundamentally isn't allowed to react.
Close the loop: measure, then decide
The adaptive controller replaces the schedule with a feedback loop. It senses the actual vehicle throughput at the intersection and adjusts signal timing to match real demand — giving green time to the direction that needs it instead of the one the clock says.
- Sense before you act. The controller measures what's actually there — real vehicle presence and flow — and lets that drive the decision, so an empty street doesn't get a long green while a queue waits.
- Timing follows demand, not a clock. Green duration adapts to throughput in real time. The intersection responds to the traffic it has, not the traffic someone predicted it would have.
- Feedback beats prediction. You don't need a perfect model of traffic if you can measure it and react quickly. A tight sense-decide-act loop handles the messy real world better than the smartest fixed schedule ever could.
The takeaway
The jump from open-loop to closed-loop control is one of the highest-leverage ideas in all of engineering, and it's everywhere: a thermostat, cruise control, a well-run system of any kind. Anything following a fixed plan while ignoring its own results is leaving enormous improvement on the table. Measure reality, feed it back, and adapt.
Building this controller taught me to spot open-loop thinking and ask "why aren't we reacting to what's actually happening?" The full design is on the project page.
👉 See it: www.divyakush.com/projects/adaptive-traffic-controller
Divyakush Punjabi — Full-Stack & AI Systems Engineer
🌐 https://www.divyakush.com · 💼 LinkedIn · 💻 GitHub
Top comments (0)