We have a new off-road cycle track with a good surface in Durham, mostly shared with pedestrians but some segregation. It is now taken account of in your routing data because I was able to plan a route which uses it:
https://cycle.travel/map/journey/190535
However, cycle.travel does seem to prefer using the main roads, which have no cycle infrastructure to speak of, and which, despite being 30mph limits, have speeding drivers on. What can I do in OpenStreetMap to improve the tagging in a way which will get this cycle track to be recommended more? Or perhaps the roads need tagging somehow to indicate how hostile they are?
Here are two examples, still going to the same destination, where the new cycle track is not used:
https://cycle.travel/map/journey/190534
In this one, Mill Hill Lane is reasonably quiet, but the other roads are not. The turning from Elvet Hill Road into Potters Bank is a nasty junction on a blind bend. I'd have thought for most people the new cycle track would be better.
https://cycle.travel/map/journey/190532
In this route there is perhaps more of an distance penalty to using the new cycle track. The first stretch is on a narrow shared-use pavement, and then Potters Bank is a road where drivers frequently exceed 30mph because it has a rural appearance. Again, I would have thought that the new cycle track would be better for most users.
How should I be improving the tagging? Does the surface, width, etc. have an impact? Or is it just that the cycle track is shared with pedestrians, crosses lots of roads and wiggles about that makes the routing engine so reluctant to use it?
Comments
Couple of things here!
The extra distance plays a part, but I think the main thing affecting the cycle track's low weighting is the amount of wiggling/turns. cycle.travel tries to choose 'unfussy' routes where possible, and the cycle track has quite a lot of junctions and crossings. Unfortunately there isn’t any traffic data for Elvet Hill Road or Quarryheads Lane, so cycle.travel has to rely on guesswork and is probably understating the amount of traffic on them.
If it’s a signposted route in real life, one thing you could do is to create a cycle route relation in OSM (network=lcn) with this route on. This would reduce the penalty given at each turn.
Separately I’ll have a look at the turn calculations as the route stands – I suspect there’s a bit of room for improvement in how cycle.travel calculates them. (I currently have a test process for road/path weightings but not for turns, so that’s something to work on!)