[Routing] OpenStreetMap + osm2pgsql + pgRouting into Postgis

[Jaak Laineste (Nutiteq)]

Yes, osm2postgis does not split roads in intersections (or actually it
does vice versa: merges them), so you can technically try pgrouting
for it, but it only uses final road ends, so it works only in some
special cases. Been there, done that :).

 I tried  to"post-fix" the postgis database using some road cutter (I
found best tool from ArcMap and no free alternatives, except maybe
Grass) . The issue is still that it would cut roads in all
intersections, where roads  actually do not intersect (multi-level
intersections). So correct solution would be to use osm2pgrouting tool
instead to convert directly from .osm file to the pgrouting tables.
But last time when I tried it it worked with limited areas (e.g. whole
Germany was too big). Maybe this is fixed now

Jaak

2010/5/18 Tristram Gräbener <tristramg at gmail.com>:
> Hello,
>
> Correct-me if I'm wrong, but osm2postgis is not meant for routing.
> According to http://wiki.openstreetmap.org/wiki/Osm2pgsql
> "this means that you can't tell which linestring is connected to
> which, but for rendering a map that's not important (c.f. routing)"
>
> Generaly OSM data has correct intersections for routing (of course the
> are exceptions).
> So my guess would be that osm2pgsql is not the right tool for routing.
>
> In OSM a way is not an arc in routing graph. You have to split the way
> in multiple arcs
>
> I hope this will help you
>
> On Tue, May 18, 2010 at 1:38 PM, roshni.spain <roshni.aruna at gmail.com> wrote:
>>
>> Hi,
>>    could you please help me into this matter. The scenario is the
>> following:
>>
>>    I have got OSM data that has been inserted into PostGIS using osm2pgsql.
>> I have used MapServer to interact with my map through OpenLayers and this
>> works perfectly fine.
>>    On the other side, I need to use pgRouting to use the algoritms of
>> shortest_path or dijkstra to find a route between A and B and show it with
>> OpenLayers. I have managed to do this with OpenLayers, so i know that if the
>> algorithms return any result, the calculated rout shows properly in my
>> browser. BUT, my problem is that in most of the cases, my algoritms do not
>> return any result... lets say maybe 80-90% of the times there is no route
>> calculated. What can i do ??
>>
>>    My geom table has got the values x1,y1,x2,y2,source,target filled for
>> all the rows.
>>
>>    The columns source and target are filled using this:
>>    SELECT assign_vertex_id('planet_osm_line', 5, 'the_geom', 'gid');
>>
>>    As for the columns x1,y1,x2,y2, these are filled using this:
>>    SELECT astext(StartPoint(way)) as startpoint from planet_osm_line =>
>> used to fill up x1, y1
>>    SELECT astext(EndPoint(way))as endpoint from planet_osm_line => used to
>> fill up x2, y2
>>
>>    I really dont know where the problem could be, as all the rows are
>> filled up and the data seem correct...
>>
>>    Looking further into the matter,
>>    i have noticed that the only cases where i have managed to get some
>> result with the shortest_path function is when there is a straight line
>> between A and B, but if i need to do A and C, and there is no straight line
>> between them (suppose something like A => B => C), then no result will be
>> returned :(
>>
>>    The same applies for other routing algorithms ... and still no clue
>>
>>
>>    I have been told that this is probably because the OSM data is not
>> properly noded at all intersections. If it is so, then how can i fix my
>> data?
>>
>> Thank you all
>>
>> --
>> View this message in context: http://gis.638310.n2.nabble.com/OpenStreetMap-osm2pgsql-pgRouting-into-Postgis-tp5069564p5069564.html
>> Sent from the Routing mailing list archive at Nabble.com.
>>
>> _______________________________________________
>> Routing mailing list
>> Routing at openstreetmap.org
>> http://lists.openstreetmap.org/listinfo/routing
>>
>
> _______________________________________________
> Routing mailing list
> Routing at openstreetmap.org
> http://lists.openstreetmap.org/listinfo/routing
>



-- 
Jaak Laineste