PATC

๐ŸŸข Every green light works for you.
You Fixed lights PATC release
โ›ถ Full simulation

Problem first

Bengaluru's signals are still solving the previous queue.

Better roads help, but coordination decides whether each green clears pressure or simply moves the jam to the next signal.

0 lost per driver ยท TomTom 2025
3 J corridor PATC coordinates 3 signalized junctions under one adaptive timing plan

Safety boundary

No black-box control. No direct live writes in year one.

Shadow first. PATC recommends and records decisions while the real signal runs normally.

Hard rules. Minimum greens, amber/all-red, pedestrian windows, emergency priority, stale data, and manual override remain non-negotiable.

Privacy by design. The pilot defaults to aggregate labels; any QA clips are short-lived, access-logged, and not used for identity or enforcement.

Product

A decision layer for connected signal sectors.

PATC starts with a selected corridor, maps real signal geometry and sensor positions, builds a trusted replay of how traffic actually moves, and only then recommends timing changes.

01 โ€” Map

Map the sector

Cameras, signals, stop lines, bus stops, turns, blind spots, and sensor points are marked before data collection.

02 โ€” Measure

Measure the wave

Phase notes, weather, events, and labels estimate queues, discharge, startup loss, and downstream block risk.

03 โ€” Replay

Replay safely

The model compares observed behavior with candidate timing choices before recommending anything operational.

04 โ€” Coordinate

Coordinate

Operators see an action, reason, confidence score, safety constraints, and a replay trace for audit.

Why coordination

One green can make the next junction worse.

In corridors like HSR, Silk Board, and Bommanahalli, one signal releases vehicles into a downstream link that is already blocked. The queue moves backward, two-wheelers fill gaps, buses slow discharge, and the next signal inherits a worse problem.

PATC treats traffic like flow: density, discharge, startup loss, spillback, weather, and event waves. The product is not an uncontrolled AI switch. It is a measured operating layer that proves when an adaptive action helps, why it helps, and when it must fall back.