Use real vehicule count data for traffic simulation

[XioNoX]

Hi,

My city publishes counting data for both motorized vehicles and bicycles on many roads (less for bikes...).
You can see them by zooming in on both:

• https://geo.brest-metropole.fr/portal/apps/mapviewer/index.html?layers=c42197968f52452da692df690475b18a
• https://geo.brest-metropole.fr/portal/apps/mapviewer/index.html?layers=c8ec1f97a9974e96bf4ef882bad71841

Most of them are "one off" measurements, so some roads have been measured in 2017, others in 2020, etc... I think some (and bikes) are supposed to be more or less real time (I asked, waiting for their answer).

The data at least have:

• Speed (average, 85%ile, for each direction, etc)
• Count (obviously), average, morning peak time count, evening peak time count, trucks vs. cars, day average, night average, and this for both directions
• The actual Geographical segment the measurement has been done on
• Link to a PDF with per hour data: (eg. https://echanges.brest-metropole.fr/VIPDU72/dep/Comptage_vehicules/BR-Moulin_15-avr19.pdf)
• Road width
• ...

Both are also available from the ArcGis API:

• https://geo.brest-metropole.fr/arcgis/rest/services/public/GPB_DEP/MapServer/1717006
• https://geo.brest-metropole.fr/arcgis/rest/services/public/GPB_DEP/MapServer/1710005

Even though I don't think the per-hour is available from the database (I asked, waiting for their answer) and the data is many years old in some locations, it seems like a good idea to use it to simulate the city more accurately. It doesn't seem trivial to implement though but could be a useful long plan goal. At least in France it's common for cities to publish such data (or maybe even mandatory).

[dabreegster]

Hi, awesome that you could track down this data! In short, I don't really know how to make any use of it. An agent-based simulator like A/B Street ultimately needs trip requests as input -- start at building 1 at 8am, drive to building 2. There's not a clear way I know of to generate that from counts along a road. The best I know how to do is synthesize this "travel demand model" from various other datasets or guesses, calculate the likely route people would take, and then compare to real count data. https://a-b-street.github.io/docs/tech/trafficsim/travel_demand.html and https://dabreegster.github.io/talks/tds_seminar_synthpop/slides.html is a summary of my understanding of this general problem. It's something I'm partly working on as part of my job right now, but I don't expect to have any decent results usable in A/B Street anytime soon.

A related effort is finding census data in different places, and using that to synthesize travel demand models -- at least knowing how many people live in different parts of a city is a start to generating realistic trips. Just finding the data sets and cleaning them up is quite a chore, and one that has to be done over and over for every country / region. There is an open source project at my work that's starting to be worked on now: https://github.com/dabreegster/popgetter/. Do you happen to know much about French census data, how to find shapefiles for the finest/smallest resolution of zones where there's data available, and how to find datasets describing how many people live in each zone, what their age ranges are, worker / student / retired / etc status, etc?

[XioNoX]

I read your presentations and investigated a bit this topic as I don't know anything about it.
So France splits its whole territory into IRIS areas.
Those areas are of 3 types:

• habitat (housing) - 1800 to 5000 inhabitants
• activity - more than 1000 employees and have twice more employees than inhabitants
• divers (various) - parks, harbors, etc.

You can see them live by zooming in on https://www.geoportail.gouv.fr/donnees/iris and clicking on any zone to see their ID and type.
They can be downloaded as SHP file on https://geoservices.ign.fr/irisge
For the whole of France 2023 or smaller parts (eg. mine)

Then for the population data, the last census was in 2019, and there you can find a spreadsheet split by IRIS areas:
https://www.insee.fr/fr/statistiques/6543200 (direct csv)
With total population per area, then split by sex, age, work category, etc.
Actually that data is split based on the IRIS areas from Jan 1st 2021.

So I guess that's everything you're asking for? :)

About travel demands, there is a bit more!
From Oct. 2017 to Nov. 2018 the city did a travel survey across the metropolitan area, asking people about a typical work day (Mon-Fri).
They collected about 30k "travels" across ~7500 people, including transportation method, goals, time, sex, worker type, if they have a car/bike and use it for their commute. I couldn't find the exact legend but that can be solved if deemed useful.
The start/stop points seem to match the IRIS areas.
https://opendata.brest-metropole.fr/explore/dataset/deplacements-des-menages-2018-les-deplacements/table/