I’m interested in implementing a traffic/speed data service for OSM-based applications.
I’ve written the following post about it. Please let me know what you think.
1 Like
Introduction can benefit from clarification of speed limit data as maximum speed limits vs current live actual speeds.
I think you mean second one, but it is not entirely clear from the initial introduction.
1 Like
Hi @contrapunctus - I think you and I are singing from the same songsheet!
This just recently released: https://play.google.com/store/apps/details?id=com.speedlimit.app
iOS version coming soon.
Quality control IS an issue which I want to improve, and am actively seeking community input on this.
I’m very happy to compare notes with you.
Thanks, I’ve updated the post.
Just keeping up with construction mapping and making sure regular congestion points have the correct hazard tag, and using a router that adds a small penalty to construction=minor and hazard=queues_likely tends to handle all but the edge cases, without introducing the problems realtime feedback creates (namely false-positives and distracted driving).
not really
it does not allow to distinguish roads which are passable without problem and ones with monstrous traffic jams (and when one want to know when traffic jams are likely and when not…)
the same for traffic intensity on roads, which is relevant to cyclists
still does not seem really clear - and I am now more confused by this average speed. So it is not about live data? But rather about being able to guess/predict which speed is likely for given road at given time?
Well enough in most situations, it does. It just doesn’t let you know when that hazard doesn’t apply, a negligible distinction, given that congestion is usually caused by too many cars and a sign of impending gridlock. Being able to generally avoid known congestion hotspots and taking routes that avoid construction is more likely to produce consistent results than attempting to optimize for edge cases. Plus as we’ve seen with Waze in the real world, this information often draws traffic to temporarily decongested routes, making overall congestion worse. Great if you’re part of the vanguard that beat the rush, bad for everyone else.
Taking traffic into account also doesn’t necessarily need realtime data so much as enough data in general to infer a pattern. This could be abstracted from the two decades of GPX data already uploaded.
Pretty easy to infer though. If you got a bunch of motorists rat-running a bike route through a neighborhood, odds are the nearest one or two major roads on either side are congested and going to be worse. My brouter profile for bicycles has a penalty for hazard=queues_likely and construction=minor precisely because high traffic routes suck, and construction zones with their lane drops and often only open 1 lane in a direction with no passing allowed, is straight up dangerous to bicycles (and just plain sucks to endure when uphill, pushing hard with someone right on your rear wheel while stuck sucking congestion exhaust).
not really
it is hard for me to do as a human looking at map, after living for decades in general area
no idea how to do it automatically
and what worse it dramatically changes over time, both based on day of week, hour, holidays, weather and more
1 Like
No, this is indeed about live speed data. But - and please correct me if I’m wrong here - even when we say live speed/traffic data, don’t we really want the median speed over (say) the last 15 minutes, rather than the speed of the last user traveling that stretch?
If one or more users have gone through a spot and we have the live median speed, great. And if there’s insufficient live data, we can look into the *historical* median speeds at a given time of day…
1 Like
That’d be fair. But based on articles I’ve seen over the years, and some casual A|B testing on my own between Osmand and Magic Earth (and Magic Earth with the live traffic and feedback on and off), the live traffic and feedback is generally not a value gain.
My husband (who still uses Google for reasons I can’t fathom given it’s glaringly bad in Oklahoma) has tried similar things with that service and had similar results. The feedback portion is basically useless and the live traffic data just isn’t bringing anything to the table that historical average traffic data wouldn’t have been able to tell you, barring extreme edge cases that are going to appear on active traffic management signage and announced on the radio anyway.
We’ve tried this in a variety of locations around the country as we travel, too, and this seems to be fairly consistent over time, location and traffic data provider. AFAICT, they work largely on the placebo effect. “I’m doing something, so it must be working.” Kinda like when you give a toddler a busybox when you’re working on a project so they feel like they’re included and doing something, except the toddler’s an adult operating heavy machinery.
yes, though weighted toward recent info if you are lucky enough to have data
if you are google they you can quite quickly detect say accident on motorway, without waiting for 15 minutes average to drop to 0
My experience with googles floating car data is quite the opposite. As a tendency, if anything, it is too sensitive; a couple of cars waiting for a red light will already turn up. Obviously how a routing engine reacts to that is a different question.
1 Like
The sensitivity isn’t the problem, it’s that when it suggests an alternative, it’s not like other routes are appreciably faster at scales smaller than “multi-state drive” or couldn’t have been predicted in advance from historical averages.
This would be very interesting once there are map apps that take advantage of live traffic feeds from e.g., road authorities. Then it would be easier to build on top of that, as starting such a service from scratch would not be super useful.
Comaps is doing that, it’s at a state where you can compile the traffic branch and have it shown on the map.
1 Like
Hello,
I think this is something a lot of people would like, basically as far as I understand you intend to create a service that serves as an alternative to Waze with some extra features.
In the french forum there’s been recent talk of reviving an old project called OpenEventDatabase that could serve that same idea (for roadworks and traffic jams in particular), with a greater idea of all kinds of events in the database, such as cultural events too. Not traffic specific in a sense.
In case you find this interesting, here are two threads you can check out: Open Event Database : le retour? and OpenEventDatabase reboot?
1 Like