About Fleet 247

At a glance

Corridors

2

Miles

2,739

Segments

13,662

Bridges, each its own

2,339

Instruments read

505

Hours

Now, +1 h, +2 h

What Fleet 247 does

Fleet 247 reads every instrument within 15 miles of each quarter mile of I-45 and I-10: road stations with their pavement sensors, airfields, the state and federal weather networks, fire weather stations and the citizen stations that pass the checks. From them it reads the surface the road most likely has, dry, wet, slush, snow or ice, with how likely each is and how confident the answer is, and turns that into an action for a high profile truck, empty or loaded, and for a robotaxi able to drive in snow: proceed, lengthen the gap, slow, route around, or exit at the next safe stop. Every alert in force on the road is given in full, to everyone.

Dry only on a positive observation

A road is dry only where a road station within 10 miles reported a dry surface within the hour, or where a forecast hour is sure of itself (confidence 0.5 or more), above 37 F, with nothing falling. An airfield that sees nothing falling says nothing of a road surface miles away, and a missing report says less: either leaves the road not known. Not known is never shown as clear and never told to proceed. At or below 37 F a not known road slows every vehicle.

Every mile says why

Each segment carries the rule that fired, and a road that is not known says why in one clause: a gap with no instrument within 15 miles, a report older than 90 minutes, a temperature only, nothing falling with no road surface reported, a bridge with no measured temperature, or a reading unsure. The reason travels with the action into the map, the exceptions, the segment drawer and the API.

Bridges

A bridge deck freezes before the road beside it. At or below 37 F a bridge is read icier than the road, but only when something measured that temperature: a road station's pavement, or an instrument's air. A bridge with no measured temperature is not known.

The exceptions first

Each corridor opens on its exceptions, the stretches that ask a vehicle to do anything but proceed, worst first, each with its reason, its alert and the vehicles that reported on it, and the miles of each action beneath them.

Scored in public

Every half hour the answers for both corridors and all three hours are kept. The next day each answer an hour or two ahead is set mile by mile against the road when its hour came, and each road station that reported its surface is held out of its own segment and the segment read again without it. The score page publishes, per corridor and day, wet called dry, ice called wet, not known that was ice, false exits, and the miles where the action changed and the road agreed.

Vehicles as instruments

A truck, a plow or a roadside device can send its own snapshot of the road as a Road Weather Message, the message of SAE J2945/3, in JSON Encoding Rules: time, place, wipers and a traction or brake flag are enough. Each snapshot is kept whole, snapped to its segment and to the bridge when it is on one, and checked for range, a stuck value, a jump, and agreement with a ranked road station or another vehicle, before it votes. Two vehicles with wipers on and no traction lost, over a pavement measured above 35 F, move a not known segment to wet; one traction flag never calls ice. Every sending vehicle is ranked as a station is, and one failing for a week stops voting. With no vehicle on a segment, nothing changes: silence never turns a gap green. The profile, with curl examples.

The rule hash

Every answer carries a hash of the rules that made it. A fleet that sets its own rules beside the defaults can tell, for any past action, which rules decided it, and a later change never rewrites what was said.

The API

/api/corridor gives every stretch of a corridor with its action and reasons, and every alert, free; with a key, every segment with its figures, the instruments that decided it and the stops along the way. /wdms/rwm/download gives the segment a vehicle is on, and /wdms/rwm/upload takes its snapshot. The developer documentation.

Questions

What is Fleet 247?

Road weather for trucks and robotaxis, every quarter mile of a corridor: the surface the road most likely has, how sure that is, and what each vehicle should do about it, now and one and two hours ahead.

Which roads does it cover?

I-45 from Galveston to Dallas, and I-10 from Santa Monica to Jacksonville. Each is cut into quarter mile segments, and each bridge is a segment of its own.

When does Fleet 247 call a road dry?

Only on a positive observation: a road station within 10 miles that reported a dry surface within the hour, or a forecast hour sure of itself, above 37 F, with nothing falling. Anything less is not known.

What does not known mean?

Nothing near reads the road surface closely enough to say. A not known road is never shown as clear and never told to proceed, and it always says why: a gap, a report older than 90 minutes, a temperature only, or a bridge with no measured temperature.

How is it scored?

Every half hour the answers are kept. The next day each answer an hour or two ahead is set against the road when its hour came, and road stations held out of their own segments check the surface. Wet called dry, ice called wet, not known that was ice and false exits are published per corridor and day.

Can a vehicle send its own reading?

Yes. A fleet, plow or device registers once and uploads a Road Weather Message snapshot: time, place, wipers and a traction or brake flag are enough. Each snapshot is checked before it votes, and every sending vehicle is ranked as a station is.