Signal controller
Green goes where the traffic is.
Lane data becomes recommended green, yellow and red timings that clear the heaviest lanes first. A language model never drives the signal.
Who does what
| Role | What it does |
|---|---|
| Gemini | Reads the scene and explains each recommendation in plain language |
| The optimizer | Calculates the timings with a deterministic algorithm, inside the safety limits |
| A city engineer | Reviews the plan and its explanation, then approves or rejects it |
How the optimizer decides
| Condition detected | Action recommended |
|---|---|
| High density on an approach | Extend green for that approach, up to a maximum green cap |
| Low density on an approach | Reduce green and reallocate the time to busier approaches |
| Emergency vehicle detected | Priority green corridor along its route, then restore the normal cycle |
Limits on every plan
- Minimum green time on every approach
- Fixed yellow and all-red intervals
- Enough time for pedestrians to cross
- A maximum wait, so no lane gets starved
If a camera feed fails or confidence is low, the system reverts to the existing fixed-time plan.
One step at a time
| Mode | What happens | Status |
|---|---|---|
| Advisory | The system recommends, and a city engineer decides | Pilot |
| Simulated control | Approved plans run in the sandbox so effects can be tested | Demo |
| Live control | Approved plans are sent to a signal controller | Future phase. Needs city authorization and hardware integration |
Today nothing touches a real signal.
Keep reading
See it for yourself.
The app runs on sample data. Open the demo account and try the dashboard, the sandbox and the assistant.