[Tagging] rail routes: how are platforms and stops associated (rail question 2)

[Tijmen Stam]

On 12-05-17 00:14, Warin wrote:
> On 12-May-17 07:45 AM, Tijmen Stam wrote:
>> On 10-05-17 18:59, Bjoern Hassler wrote:
>>> Hello again,
>>>
>>> In an  osm:relation:route
>>> <https://wiki.openstreetmap.org/wiki/relation:route> (type=route,
>>> route=train/...), you have both platforms and stop positions. How is a
>>> particular platform associated with a stop that serves it?
>>>
>>> E.g. for public transport routing, you'd walk (highway=footway) to a
>>> platform (public_transport=platform), at which point you'd change to a
>>> train stopping at a stop (public_transport=stop_position). How would the
>>> routing algorithm know that the platform is associated with the stop?
>>>
>>> Is there an existing mechanism or convention, e.g. a tag on the platform
>>> that indicates the stop, or both tagged with the same name or similar?
>>>
>>> Thanks!
>>> Bjoern
>>>
>>> PS I've noticed that sometimes the stop position is at the far end of a
>>> platform (i.e. the two stop positions are at opposite ends of the
>>> station). Maybe that's so that an association can be made?
>>
>> Answering your grand question:
>>
>> As I interpret the wiki, it is the route-relation that ties together
>> the stop_position with the platform, by including them as a pair per
>> "halt".
>>
>> Imagine the "renderer" is a transit simulator that simulates a journey
>> from your home to somewhere via "Green line" that stops on track 1
>> (platform A), then the route relation of the green line contains both
>> the stop_position (a node on track 1) and the platform (platform A).
>> It would then plot a walking route to platform A, then transfers you
>> into the train (on track 1) and along the route.
>>
>> I know of people who use a stop_area-relation for each
>> stop_position/platform pair, which then could be used to tie
>> stop_position and platform together, but that is not how I interpret
>> the wiki.
>> I use one stop_area for a whole station.
>
> Here there are different length trains - they usually stop at different
> positions on the platform so the middle of the train is at the middle of
> the platform. And there are short platforms where a full length train is
> too long for the platform - so people wanting to get off must be either
> in the front carriages, the middle carriages or the rear carriages in
> order to get off. (Why the different options? So that the train
> passengers don't all congregate in one portion of the train - different
> platforms have different positions for the train stop) Of course shorter
> length trains can stop with their carriages fully engaged with the
> platform.
> I take the stop position from the train divers point of view - as that
> is what would be designated to be of practical use.

In the Netherlands, most stop_positions for trains were once imported 
automatically.
The railway company has two stop positions per track, one per direction, 
on simple stations.
But on a complex station like Amsterdam, a train on track 7 can stop at:
7 (the full platform length)
7a
7b
7c (this never happens, as the 7c part alone is too short to hold a 
train next to the platform, but as there is a switch between 7b and 7c 
train drivers need to know whether to stop before the switch or that 
they can proceed behind it.)
7ab
7bc
which would require 6 stop_positions per track. But that's IMHO too 
complicated for OSM, where I would suffice with one per "section" and 
just a random choice of 7a or 7b if a train needs to halt at 7ab.

How would one treat a bus or train route where the vehicle will stop at 
one of multiple tracks, undecided by direction? E.G. at each end of 
metro line 53, the train will choose an unoccupied track. Routing each 
combination results in 4 route_relations per directions, which seems 
overkill to me. What I sometimes do is add both tracks up to the 
branching point, but then you get a "route relation contains a 
gap"-warning. Is there a fix or better way for that?

Tijmen/IIVQ