[Talk-GB] barrier=kerb on highways may be blocking OSRM (Car) routing

[Edward Catmur]

I think your new mapping is correct, as it reflects reality - which is what
most matters. You could improve it by curving the cycle path/sidewalk up to
the crossing and aligning the kerb nodes to the aerial photo, but that's
nitpicking.

That said, there's still a problem in that at least one routing
engine (OSRM in bicycle mode) still sees kerbs as impassable to bicycles:
https://www.openstreetmap.org/directions?engine=fossgis_osrm_bike&route=55.0123%2C-1.4837%3B55.0130%2C-1.4810
but
that's something the routing engine devs should fix (if you feel like
contacting them); compare
https://github.com/fossgis-routing-server/cbf-routing-profiles/blob/master/foot.lua#L35

https://github.com/fossgis-routing-server/cbf-routing-profiles/blob/master/bike.lua#L40


Even a non-lowered kerb is an obstacle to cycling, but not an
insurmountable one, at least to able-bodied cyclists who can raise their
bike over the kerb, so it should perhaps have a penalty in routing but not
enough to make a 2km detour preferable. That said, the same goes for cars -
other than the lowest bodied sports cars, pretty much all motor vehicles
are capable of taking a kerb at low speed.

On Wed, Dec 18, 2019 at 12:35 PM James Derrick <lists at jamesderrick.org>
wrote:

> Hi,
>
> After investigating two reports of OSRM routing failures around North
> Tyneside, the common factor I can see is barrier=kerb tags added to
> highway=crossing nodes intersecting highway=tertiary and
> highway=cycleway/ footway ways.
>
> Here are links to the two map note reports:
>
> https://www.openstreetmap.org/note/2030228
> https://www.openstreetmap.org/note/2030238
>
> To investigate the report, I entered the postcodes given into the
> default routing engine on the OSM map and found VERY odd routes going
> 10x the direct distance, and avoiding very obvious direct paths:
>
> https://www.openstreetmap.org/directions?engine=fossgis_osrm_car&route=55.0659%2C-1.4624%3B55.0511%2C-1.4530#map=14/55.0590/-1.4747&layers=N
>
> Personally, I'd not noticed the OSM main map had added several routing
> engines as I use separate tools, so have no idea how often the routing
> engines update their database extracts but expect the issue to be
> visible for a few days.
>
>
> After two examples of bad routing, I checked the paths between the
> geolocated points given and found one common factor - barrier=kerb on a
> road / footway highway=crossing node.
>
> https://www.openstreetmap.org/node/4341572135
>
> My hunch is the router isn't familiar with barrier=kerb, so is assuming
> BOTH ways are blocked and using an alternate path.
>
>
> It is debatable how a routing engine should interpret highway=kerb tag,
> however my own thought is the kerb is not on the highway=secondary - it
> is on the highway=footway.
>
> If anywhere, there should be two nodes on the footway separate from the
> secondary to give information to wheelchair accessibility routers.
>
>
> As an experiment, I've removed the barrier=kerb from a highway-crossing
> and added two nodes on the cycleway, with the additional explicit tags of:
>
>    barrier=kerb
>    bicycle=yes
>    foot=yes
>    wheelchair=limited
>    kerb=lowered
>    tactile_paving=yes
>    horse=yes  (ISTR UK law says cycle = horse!)
>
> This is rather cumbersome compared with one barrier=kerb tag on the
> node, but logic suggests this is more consistent with reality and
> current routers.
>
>
> Has any one used the barrier=kerb tag, or is familiar with the inner
> workings of OSRM or similar engines please?
>
> Thanks,
>
>
> James
> --
> James Derrick
>      lists at jamesderrick.org, Cramlington, England
>      I wouldn't be a volunteer if you paid me...
>      https://www.openstreetmap.org/user/James%20Derrick
>
>
> _______________________________________________
> Talk-GB mailing list
> Talk-GB at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk-gb
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk-gb/attachments/20191218/f4517117/attachment-0001.html>