PhaseGrid reads real traffic at the signal head — queues, pedestrians, cyclists, emergency vehicles — and adjusts the phase plan in real time. No more sitting at a red for cars that left ten minutes ago.
Just an idea right now — no product, no pilots yet. Building in the open.
The majority of signals still run on fixed timing plans written years ago, or react only to a single loop detector buried in the pavement. They can't tell a full queue from an empty lane, or a pedestrian waiting from an empty crosswalk — so they hold phases too long, cut others too short, and leave the timing plan to go stale as traffic patterns change.
PhaseGrid slots into the signal cabinet next to the existing controller. It doesn't replace how signals work — it changes what they're allowed to decide, phase by phase.
Camera and radar units read every approach — vehicle queues, pedestrians, cyclists, and emergency vehicle signals — continuously, not just at the stop line.
An on-site model weighs every approach against the others each cycle and picks the phase sequence that clears the most people safely, within engineer-set guardrails.
The signal extends a green for a trailing queue, shortens one that's already empty, and logs every decision so traffic engineers can review and override it.
No forklift upgrade, no new intersection hardware to trench in. PhaseGrid adds one unit next to the equipment that's already there.
Camera and radar units mounted at the signal head read every approach continuously — vehicle queues, pedestrians, cyclists, and emergency vehicle signals.
PhaseGrid's unit sits next to the existing signal controller, reads the sensor data, and proposes phase timing every cycle — no cloud round-trip for a decision that has to happen in milliseconds.
Every decision is stored with the sensor data that triggered it. Engineers set the guardrails up front, review the log, and can override or roll back any change instantly.
Signal cabinets are being upgraded on their own replacement cycle regardless — this is a case for putting real intelligence in them while that happens, not for ripping out infrastructure that works.
There's no product yet — this is the starting point. If you work in traffic operations, a DOT, or signal engineering, we'd genuinely like to hear whether this is worth building and what would make it useful to you.