Since we took CycleStreets to public ‘beta’ launch three weeks ago, we’ve had lots of feedback, all of which is extremely useful to us. We’ve been hard at work fixing various bugs and adding improvements.
We now have a weekly refresh of mapping data taking place. This means that we can now more effectively work through cases where errors in the underlying mapping data cause routing problems.
The ‘cycle’ of improvement works like this:
- We receive feedback that a route seems to have an error in it. (Just hit the ‘feedback’ button on the page, and add a sentence or two describing a problem you’ve noticed.)
- We check out the route. Where it turns out this is an error in the map data, e.g. a one-way street marked the wrong direction, our OpenStreetMap contacts will fix this where possible.
- Once a week, probably at the weekend, we import the OpenStreetMap data and put this through a very extensive set of optimisations that are used to speed up the routing performance when you use the site.
- We can check that the route problem has been fixed, by going to the original page and using ‘Change’ to re-plan the route from the same start/end points.
As an example, this journey (#50547) was suggesting a route through a one-way street marked in the wrong direction. The user noticed this and kindly sent us a feedback note pointing this out. (Such occasional errors are to be expected in a data set as large as OpenStreetMap’s excellent data.) Here’s the extract concerned:
Shaun from OSM kindly fixed the direction. On Friday night we did our new weekly refresh, which therefore included this fix. (Don’t rely on it always being on Friday yet, though!)
The result is now this new journey (#52023) which includes this amended extract:
Of course, in other cases, the routing engine may be the source of the problem. Most of the errors we’ve so far heard about are due to things we haven’t yet implemented in this beta release, namely hills and taking traffic lights / turns into account more strongly in the routing. There are also problems relating to bridleways at present and a few other ‘provision types’, as we call them. These issues are amongst the things we are working on in coming weeks as we move towards a public release – stay tuned to this blog for more news.
We have quite a backlog of feedback messages to work through, but we aim to respond to every one of these in due course. Please do keep feedback, however small, coming through. Please be assured that we read every one, and each report will help get both the mapping data, and our routing engine, improved.
2 thoughts to “Weekly refreshes of mapping data”
How are you passing errors to your OSM ‘contacts’? It would be good if the OSM community could have a look at a list of problems reported (maybe with an rss feed) and then could pick ones for them to check.
In fact it might be good just to put errors on Openstreetbugs.org, I’ve started viewing that a lot more. It’s very helpful.
Shaun’s good, but he can’t be everywhere the UK has a problem!
We have at present a couple of OSMers who receive the feedback reports when the ‘route feedback’ radio button is used. Some have already been fixed up, but it’s early days yet, and we have a backlog of apparent errors we’d like to get moved quickly.
So if other OSMers can help handle the route feedback problems that result from data errors, please do get in touch via the contact page – this would be very useful, both for CycleStreets and presumably OSM as a project! Now that we have a weekly refresh this should result in quick routing quality improvements.
Effectively we’d like people either to fix problems directly, or act as a proxy between the feedback reporter and Openstreetbugs, and then respond to the user in clear and untechnical terms. Do get in touch if you’re an OSMer that can help!