I’m a bit puzzled by how some routers handle railway platforms. And I’m wondering how platforms really should be mapped to facilitate good routing. I personally prefer multipolygons with railway=platform_edge with only short connecting footways to nearest stairs/elevators like they do in Germany for example.
The routers I’ve tried (OpenTripPlanner, OSRM, Graphhopper) don’t seem to support multipolygons leading to bizarre routing with long detours and/or suggesting trespassing over the railway. OpenTripPlanner and OSRM are able to use the edges of railway=platform mapped as a closed way for routing though.
The current solution in Finland is to add footways along the full length of the platform. Island platforms for two tracks are divided to two polygons instead of using railway=platform_edge. This mapping method is for Digitransit project and its OpenTripPlanner router specifically as mentioned here: Digitransit - OpenStreetMap Wiki
But it’s not very clean in my opinion. And if you were to add footways like that in Germany they would be removed very quickly (found this thread auf Deutsch). What should be done?
I was able to do some testing and yes, it does indeed! If the platform is a clean multipolygon with platform_edge parts it works well. But if there are footpaths along the length, then the routing engine gets really confused and insists on walking via the end of the platform… So mapping footpaths over platform areas isn’t quite right.
I’ve recently come across several places where platforms are tagged as a way having both public_transport=platform and highway=footway and also as a closed way with railway=platform. So there is an area and a way splitting the area in half. Example: Way: 28684239 | OpenStreetMap & Way: 1135803011 | OpenStreetMap Why would people do this?
Writing just for the sake of an update. So far I’ve fixed about 200 rail/tram/subway platforms in Finland which had overlapping platform objects mapped over them.
Hopefully in sometime we can be free without footways on platforms where people sometimes get confused and they “correct” the footways by re-tagging them as platforms.
If a router is willing to route around or through a closed way but not around or through an equivalent multipolygon, then that’s likely a bug worth reporting to the developers.
Overall, the router landscape hasn’t been so good for routing optimally through areas, but it’s finally on the cusp of getting better. For platforms in particular, routers are currently so inconsistent that validators have to warn mappers to expect broken routing if they rely on the platform tags alone.
