How end-of-queue warning systems work

When a lane closure or a flagging operation backs traffic up, the most dangerous place in the work zone is the back of the queue. Drivers arrive at highway speed and meet stopped vehicles, and rear-end crashes are among the most frequent work zone crashes. A queue warning system finds the back of the queue as it moves and tells drivers before they reach it.

Radar sensors are placed upstream of the work zone through the whole distance a queue can reach. Guidance commonly calls for about half a mile between sensors, and seldom more than two miles. Each sensor reports the average speed per direction, usually in 60-second bins, and cameras can confirm what the radar sees.

The tail of the queue lies between the last sensor that reads slow traffic and the first that reads free flow. Closer spacing gives a more precise tail, which matters most where the tail is hard to see.

The thresholds belong in the system plan; 45 and 20 mph are common choices. Messages follow the MUTCD's rules for portable changeable message signs: no more than two phases, each of up to three lines of eight characters.

To keep signs from flickering between messages, a good system confirms a change before acting on it and waits longer before relaxing. LaneWarden starts a warning after two consecutive 60-second bins agree and ends it only after three calmer bins.

The right sign is not simply the nearest one. It has to be far enough upstream that drivers can read it and slow down before they reach the tail. LaneWarden uses the nearest sign at least half a mile before the tail. As the queue grows the warning moves upstream to the next sign, and as it shrinks it moves back.

A queue just past a hill crest is invisible until drivers reach the top. Stopping sight distance, the distance a driver needs to notice a hazard and brake to a stop, sets how far past a crest a queue becomes dangerous. With AASHTO's design values of 2.5 seconds to react and 11.2 ft/s² of braking, it grows quickly with speed:

When the tail is that close past a crest, the warning has to be posted before the crest. LaneWarden only lets a sign at least 1,000 ft before the crest carry it, and the operating plan lists every crest and curve where this applies.

Sensors lose power and modems drop out. A sign that loses the detection it depends on must not keep showing a travel time or nothing at all, because that tells drivers the road ahead is clear. It should post a general caution such as WORK ZONE, then SLOW DOWN, and alert a technician. This logic belongs on the sign itself, so it works when the server or the network does not.

• Which thresholds trigger each message, and how long before a warning comes down?
• How is the tail located, and how quickly does the sign react?
• Which sign carries the warning, and how are crests and curves handled?
• What does each sign show when its data stops?
• Who approves messages, and is every change logged?
• Who is alerted, and when, for example when delay passes 15 or 20 minutes?
• How long is the data kept, which reports come out of it, and is there a WZDx feed?

The LaneWarden live demo runs these rules on a simulated corridor, including a queue that grows over a crest. You can scrub through the morning and watch the warning move from sign to sign.