[OSRM-talk] node as waypoint for polygons
frederik at remote.org
Tue Oct 25 20:50:44 UTC 2016
On 10/25/2016 10:14 PM, Jürgen Barthel wrote:
> Problem 1: Airport "nodes" are usually somewhere on the runway. Very
> frequently that results in a navigation for an airport service road
> opposite the terminal. We have identified the street access nodes for
> all airports in Europe and currently add Russia.
It sounds to me as if what you're having is really a geocoding problem.
You say "how long does it take from Munich to Frankfurt airport" and the
error occurs when you translate "Munich" and "Frankfurt airport" to
points on the street network.
>From a purely routing point of view, this could be treated as a m:n
routing problem - take all streets within the Munich polygon, and all
streets withing the Frankfurt Airport polygon, and then find the
shortest trip that connects any Munich road with any Frankfurt airport
road. But that's not what you want, is it?
For towns and cities, I don't think the admin centre is what you should
be looking for; you should use the place node instead. This is not
necessarily contained in the administrative boundary relation, and will
usually be in central location (whereas the admin_centre is just where
the local/regional government resides, which can be quite far away from
the actual centre of something).
The "street access nodes" you're talking about will also likely be
different for different modes of transport - for example a cyclist going
to an airport might want to end their journey elsewhere than a motorist.
And even the motorist will aim for a different part of Frankfurt airport
when they go there for a long-term stay than when they want to drop
someone off, and that again might be different from an ambulance
responding to an emergency.
Your "street access node" might turn out to be "the point where the
average car-driving tourist will want to go by car when visiting the
location" or so.
The points you have collected are certainly not "admin centres" unless
you have actually researched where the local government resides.
> I also addressed before the issue of "nearest suitable road segment"
> (see 17Aug16) also loosely related to this, where issues like "national
> parks" popped up. Shopping centers jump to mind, "car free city
> centers", etc, pp.
I wonder if many of these things might be "the closest car park one can
reach on foot"? But then the question is again, closest to where -
certainly the car park closest to the middle of the runway is not
> I looked for Frankfurt Airport and got
> leading to http://www.openstreetmap.org/relation/5813621 (without any node).
It would probably be too difficult to specify "the" one node that
everyone wants to go to when they say "Franfurt Airport"...
> Looking at Zurich http://www.openstreetmap.org/relation/1682248, there
> is no admin center but "Node Zurich (240025182) as label". Problem is, a
> "label" does not help.
Well the role "label" is a bit funny but the node linked as a label is a
proper place node that is certainly usable as a destination point.
> Question 2: If you look at national parks or other "car free" places,
> they may have multiple waypoints...?
It is easy to identify all crossings where any road leads into the
national park. However, not all of these roads might be suitable for
tourist access. Again the question is, who wants to know - an emergency
vehicle? The park ranger? A cyclist? A lorry driver? A motorcar tourist?
> I think the main question is: Does someone have an idea how to add a
> "node as waypoint" as a "standard" for any polygons into OSM???
It sounds to me as if what we really need is a complicated software
component that executes the steps you are currently executing by hand -
look at the area, determine what objects there are in the area, rate
which ones are the most likely destination(s), and then expose those as
"anchor points" for analysis. That way, the software could be tuned to
support different transport modes and different use cases, rather than
trying to hard-code into OSM the "one centre point to serve everyone".
Do you think something like this is possible, or did your research often
include steps that were so difficult that only a human could do it?
Frederik Ramm ## eMail frederik at remote.org ## N49°00'09" E008°23'33"
More information about the OSRM-talk