Hi, I am exploring use of vehicle probe data which provides average speeds for short links (HERE, TomTom, Google, etc.). Has anyone done validations/calibration studies against conventionally collected speed point-speed data from tube counters, pavement loops, Trtl or speed cameras? The two methods of measuring speed are very different. I hope we can show the link between them in order to ease the industry into using probe-speed data. Many thanks.
Hi Chris, good point :-) We have recently started using floating car data (FCD) and were interested in comparison with traditional profile data. I know of Belgian study - http://www.sciencedirect.com/science/article/pii/S2352146516301120 - where FCD was approx. by 10 km/h higher compared to radar-speed. In our study we did only limited comparison (7 sites) and found the same trend, with difference on average 2 km/h. I believe there are several explanations to this excess:
- different drivers (FCD = usually commercial/professional, i.e. maybe faster drivers vs all drivers detected by radar)
- different definitions of free-flow speed
- different days, times, seasons, etc. (radar measurements are usually under controlled conditions, FCD are not)
Nevertheless this field is fresh so let's start discussion, other references, etc. :-)
Thank you. I had a quick look and passed it onto our experts in this area. It is interesting that you are referring to this as FCD from commercial drivers. Our data is much more varied, and comes from a range of sources including monitored heavy vehicles scheme (very accurate and biased), corporate fleets (probably more like what you have), GPS navigation units, and yes - personal phones (least accurate by the way). The data is 'cleaned' by the supplier to exclude a whole range of common errors (e.g. reversing, pedestrians, ferries, drivers on parallel or private roads). We are about to launch into a validation/calibration exercise against tube/loop point-speed data.
Any other research links would be welcome.
Thanks for explanation, your approach is wide, I see... TomTom has done several studies, but I could not find precise numbers in them - worth googling up and checking out for yourself :-)
Yes that seems to be the problem. The data providers don't publish academically, and what they publish often falls short of the answers I seek. I googled...
PS: https://trid.trb.org/view.aspx?id=1361860 useful, does not answer how well the data fits
A Czech 2015 study - Comparative analysis for floating car and loop detectors data - under the code "EU - ITS-1302" on page https://proceedings.itsworldcongress.com/bestpapers
Great study, Jiri. I like the 'virtual traffic detectors' based on the FCD data. The speed relationship was not developed.
- Badges
- Science topic