Keepright, Routing vs Printing

Posted by Dion Dock on 3 April 2010 in English (English)

I've been cleaning up a bunch of maplint reported by Keepright,, which rocks!

(All comments based on USA data.)

Unfortunately, I'm starting to get to the conclusion that routing OSM data just won't work for any road smaller than highway=tertiary. It seems that the TIGER import didn't connect a bunch of roads near county lines that are away from the cities. That's not too hard to fix if mappers spend some effort.

There is also a fair amount of bad data due to edits. For example, streets are split into parallel one-way roads, but the new road doesn't get an intersection at every road it crosses. Or the new driveway almost touches the main road.

The trouble is things look good in the web. Intersections are rendered the same as non-intersecting roads. Visually, the map is good enough. However, from a routing perspective, the roads never meet, so you'll never get to make that right turn. This is going to be a serious problem for anyone who wants to use OSM data for routing. Without going to external websites, the map appears just fine, possibly leading to more incomplete edits.

Until this data gets beat into submission, I think OSM maps are going to be good for printing and viewing but risky for door-to-door routing.

So how important is routing?

Comment from wallclimber21 on 3 April 2010 at 01:25

I share your skepticism.

TIGER data is a gift in that it kick starts that database with roads that cover the whole US. But it also means that bad data (and there's a TON of it) will never see any correction or review. Which makes it very tenuous to rely on it for routing. It's simply not going to work unless you're sure that a particular area has been worked on extensively.

It's more than just county lines or missing intersections. You have zillions of miles of impassable fire roads marked as 'residential'. You have residential streets intersecting with interstates. There are many roads with a location that's off by more than half a mile.


I *love* OSM because it allows me to add the features I care about on a map (basically: hiking and mountain biking trails). But for routing use, your only real option is a commercial map. I don't see this changing anytime soon.

Comment from Nightdive on 3 April 2010 at 05:19

It's more fun to see you mapped streets coming live in OSM than correcting errors in the data. The tiger import is probably one reason why there are not that many mappers in the US like in germany.
Less mappers means that there are not many people fixing errors.
Germany data is good enough for routing with a few issues but that should get better.

Comment from amm on 3 April 2010 at 06:25

Although the US might have disproportionate number of problems, road connectivity for routing is definitely an issue in other places too, including Germany ( as e.g. shown by the OSM Inspector ). Once people start using osm more for (hobby) routing, things will improve, as more people notice the problems and fix them. But it will be a bit painful at the beginning. Never-the-less, I do think it is one of the important use cases for OSM data.

In the future even more emphasis will thus have to be put on quality assurance to fix these sorts of problems. But our tools for that, such as keepright, are slowly improving, they just need to become more widely known and used.

Comment from Godzookee on 3 April 2010 at 16:44

I think routing would be very important there seems to be no point if we can't youed the maps in the GPS's

Comment from Ethan O'Connor on 3 April 2010 at 20:49

I think that the solution to this lies in QA efforts that focus on the end product - routes - and not the data. The 250 Cities Project was an example of this sort of approach:

But this is of course just the start of the possibilities. 250 Cities used a blanket approach looking for high ratios of Route Length to Geographic Distance; altering this just slightly to take into account scaling effects, terrain effects, population density effects, and street-layout style effects and adjust the ratio thresholds should make it possible to extend the 250 Cities effort to a 1000 Cities, 1000 Routes project -- fix the intercity grid for the 1000 largest cities in the US and fix 1000 Routes within each city prioritized by population density. This is a big project, but is tractable if an effort and be coordinated and sustained for a decent amount of time.

Another approach that should come online is comparing calculated routes to the OSM .gpx database with a blended heuristic/trained probabilistic filter interposed. This could be a significant source of routing data, especially if we can port the Waze source ( to draw from and report to OSM rather than Waze's databases. This approach can be especially helpful in detecting plausibly-short but illegal routes (missed turn restrictions, one-ways, etc.) and in detecting new routes as they are created through construction.

Further, the existing QA tools such as KeepRight should incorporate or be integrated into planning/coordination/progress tracking tools that allow estimation of work remaining and visualization of progress made on definable sub-tasks (no duplicate nodes within a certain county, for example). This has featured prominently in the 250 cities project and the various duplicate nodes projects, for example, and really helps drive participation.

So -- let's do this! :)

Comment from Andy Allan on 5 April 2010 at 09:48

Don't be discouraged by the poor quality of the data - just get on and fix your area and encourage others to do the same! There's no magic pill to sort out all the routing problems, but there's no reason to not fix the problems at hand either. I spent an ungodly amount of time fixing interstates when I started the 250 cities project - hundreds and hundreds of hours. When other people saw the amount of work being done, they got stuck in too.

Make the area that you're interested in an exemplar of how to do things, and other people will raise things to that standard. But when everyone accepts the current state of play, then few people are motivated to improve things.

Login to leave a comment