[OSM-talk] Lightning fast car routing built on OpenStreetMap data, with draggable routes
Jean-Marc Liotier
used a routing service that fast. It is apparently tuned for car
routing... And that's all I can say since the Karlsruher Institut für
Technologie whose homepage is linked from the results panel doesn't seem
to say anything about it. If anyone has further information about this
exciting service...
Try Tromso to Cadiz... http://map.project-osrm.org/bz - a 5201k route
generated instantly as far as I can measure. And it crosses ferries too,
so Inverness to Athens works (http://map.project-osrm.org/bA) as does
Moscow to Malta (http://map.project-osrm.org/bB) !
_______________________________________________
talk mailing list
ta...@openstreetmap.org
http://lists.openstreetmap.org/listinfo/talk
Jean-Marc Liotier
> Spotted in @openstreetmap's Twitter feed... I don't remember having ever
> used a routing service that fast. It is apparently tuned for car
> routing... And that's all I can say since the Karlsruher Institut für
> Technologie whose homepage is linked from the results panel doesn't seem
> to say anything about it. If anyone has further information about this
> exciting service...
The service is of course at http://map.project-osrm.org/ as you may have
guessed from the examples in my initial message.
It has been mentioned by about everyone on Twitter
(https://bitly.com/pGcc3J+)... I'm surprised there has been no
conversation about it here.
For news about the project, you may follow
https://twitter.com/ProjectOSRM
Jean-Marc Liotier
> did you see this?
> http://neis-one.org/2011/07/comparison-routing/
> or this
> http://neis-one.org/2011/07/comparison-reloaded/
> but remember it is a few months old and it seems
> that the new version got some improvements too.
I had missed that one - I have only recently become interested in
routing as a user.
For those who had missed it too, this graph shows that OSRM has better
and more consistent performance than any other routing service,
including Mapquest, Bing, Google and Cloudmade :
http://neis-one.org/wp-content/uploads/2011/07/20110717_RoutingEngine_Comparison_Detail.png
I wonder how they do it...
Markus Lindholm
I also came across a situation that it couldn't find a route for, from
Wallingatan 11 to Wallingatan 5.
http://www.openstreetmap.org/?lat=59.336906&lon=18.057388&zoom=18
Wallingatan is a oneway street that crosses a pedestrian street and
about ten meter before the intersection it changes also to pedestrian.
Is this a problem with routing engine or with how the street is tagged?
/Markus
Peter Wendorff
> Had fun testing different routes, and it is fast.
>
> I also came across a situation that it couldn't find a route for, from
> Wallingatan 11 to Wallingatan 5.
> http://www.openstreetmap.org/?lat=59.336906&lon=18.057388&zoom=18
> Wallingatan is a oneway street that crosses a pedestrian street and
> about ten meter before the intersection it changes also to pedestrian.
>
> Is this a problem with routing engine or with how the street is tagged?
is missing - on the two stubs of the Wallingatan and on the small part
of the Drottingatan.
regards
Peter
Philip Barnes
Inverness into one of the boxes and hit show nothing happens.
I can see your routes by clicking the links however. The Inverness to
Athens route does seem a bit bizarre. Not what I would have expected.
Crossing the Pennines on the A66 is strange, continuing on the M6 is the
more normal route. Also crossing from the M2 to M20 to get to Dover is
strange, for a car the M2-A2 directly to Dover is quicker. Most routing
software does seem to prefer the tunnel over the ferry.
http://open.mapquest.co.uk, which also uses OSM mapping provides more
normal routing, certainly for the UK stage. Although it does seem to
struggle with routing in South Eastern Europe.
Google takes you through Italy and across the ferry to Greece, am not
sure of the relative merits of this route, over travelling overland but
it does start to ring alarm bells of real border crossings and the need
for additional insurance.
Lester Caine
> Crossing the Pennines on the A66 is strange, continuing on the M6 is the
> more normal route. Also crossing from the M2 to M20 to get to Dover is
> strange, for a car the M2-A2 directly to Dover is quicker. Most routing
> software does seem to prefer the tunnel over the ferry.
>
> http://open.mapquest.co.uk, which also uses OSM mapping provides more
> normal routing, certainly for the UK stage. Although it does seem to
> struggle with routing in South Eastern Europe.
I must say I'm seeing the same strange effects on routing in the UK. Although
mapquest is a little slower, it does at least pick up the faster roads rather
than routes that are perhaps 0.5km shorter but using roads with many roundabouts
rather than the adjacent motorways or dual carriageways with none.
That and I could not drag the route to use the more practical roads on Seamonkey.
--
Lester Caine - G8HFL
-----------------------------
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk//
Firebird - http://www.firebirdsql.org/index.php
Jean-Marc Liotier
>
> I could not drag the route to use the more practical
> roads on Seamonkey.
Did you manage with another browser ? The method differs from Google :
left click to create a handle, then you can drag it.
Fabrizio Carrai
also almost immediate [1]!
It is pity that the search of the "Start" and "End" places find
strange locations:
"Ragusa, Italy" reports "Malavita, Ragusa". Close, but out of the city.
"Livorno, Italy" reports a natural land that it is in the Livorno
area, but it is a small rock in the middle of the sea!
Compliments for the job!
F.
Jean-Marc Liotier
> The Inverness to Athens route does seem a bit bizarre. Not what I
> would have expected.
>
> Crossing the Pennines on the A66 is strange, continuing on the M6 is
> the more normal route. Also crossing from the M2 to M20 to get to
> Dover is strange, for a car the M2-A2 directly to Dover is quicker.
> Most routing software does seem to prefer the tunnel over the ferry.
>
> http://open.mapquest.co.uk, which also uses OSM mapping provides
> more normal routing, certainly for the UK stage. Although it does
> seem to struggle with routing in South Eastern Europe.
>
> Google takes you through Italy and across the ferry to Greece, am
> overland but it does start to ring alarm bells of real border
> crossings and the need for additional insurance.
I went to Istanbul from France through the Croatian coast and Greece on
the way in and through Italy on the way out. Apart from the touristic
merits of either, Italy is certainly a faster way.
Strangely, OSMR won't give me a route through the Igoumenitsa-Brindisi
ferry though the Mapnik render does show a ferry line between them...
I'll have to take a look at the data. Without that ferry crossing, the
route through Italy is not as short - which might explain the routing
service's preference for the continental route.
Lester Caine
>> > I could not drag the route to use the more practical
>> > roads on Seamonkey.
> Did you manage with another browser ? The method differs from Google :
> left click to create a handle, then you can drag it.
Ok less than intuitive ... all the others you just click and drag
I did try IE7 but that is just a mess ... the route description comes up in the
wrong place and background boxes are missing.
--
Lester Caine - G8HFL
-----------------------------
Contact - http://lsces.co.uk/wiki/?page=contact
L.S.Caine Electronic Services - http://lsces.co.uk
EnquirySolve - http://enquirysolve.com/
Model Engineers Digital Workshop - http://medw.co.uk//
Firebird - http://www.firebirdsql.org/index.php
_______________________________________________
Nick Whitelegg
I tried the route we went on holiday as a family in the 80s... Fernhurst, W Sussex, to Münstertal in Germany. Impressively fast.
The route was much as I remember as far as Reims... but then, rather than routing you along the A4 autoroute to Strasbourg and then down the German autobahn (forget the number) to Freiburg, it routed you much further south along French "N" roads. Is the motorway weighting high enough?
Nick
-----Jean-Marc Liotier <j...@liotier.org> wrote: -----
To: ta...@openstreetmap.org
From: Jean-Marc Liotier <j...@liotier.org>
Date: 17/03/2012 02:01PM
Subject: Re: [OSM-talk] Lightning fast car routing built on OpenStreetMap data, with draggable routes
Steve Doerr
> Also crossing from the M2 to M20 to get to Dover is
> strange, for a car the M2-A2 directly to Dover is quicker.
I don't know. People around here (NW Kent) seem to vary in which way
they prefer to go. TomTom opts for the M20.
I just used project-osrm.org to plan a route from home to Alton Towers
and was surprised to find it took me through central London - over
Westminster Bridge. Most people in their right mind would go through
Dartford Tunnel and round the M25.
--
Steve
John F. Eldredge
From what you and others have said, it sounds like the software is seeking for the shortest distance, rather than the shortest travel time.
--
John F. Eldredge -- jo...@jfeldredge.com
"Reserve your right to think, for even to think wrongly is better than not to think at all." -- Hypatia of Alexandria
Janko Mihelić
Found one error, it doesn't watch for turn restrictions when a way has the "via" role. So, U-turns are not restricted.
Only Mapquest watches for these.
Janko
Michal Migurski
> Spotted in @openstreetmap's Twitter feed... I don't remember having ever
> used a routing service that fast. It is apparently tuned for car
> routing... And that's all I can say since the Karlsruher Institut für
> Technologie whose homepage is linked from the results panel doesn't seem
> to say anything about it. If anyone has further information about this
> exciting service...
Doesn't seem to work at all in the US - I'm getting routing failures between SF and LA when I drag the pins over, and the place search boxes modify my queries after I make them. San Francisco becomes "Santa Rafaela María, Pedro Abad", Los Angeles "Carretera Villaviciosa - Arganda, San Martín de la Vega".
-mike.
----------------------------------------------------------------
michal migurski- mi...@stamen.com
415.558.1610
Eugene Alvin Villar
> It has been mentioned by about everyone on Twitter (https://bitly.com/pGcc3J+)... I'm surprised there has been no conversation about it here.
This had been announced at the dev mailing list:
http://lists.openstreetmap.org/pipermail/dev/2012-March/024558.html
On Michal Migurski wrote:
> Doesn't seem to work at all in the US - I'm getting routing failures between SF and LA when I drag the pins over, and the place search boxes modify my queries after I make them. San Francisco becomes "Santa Rafaela María, Pedro Abad", Los Angeles "Carretera Villaviciosa - Arganda, San Martín de la Vega".
As mentioned at the dev email, the data only contains Europe for now.
Kai Krueger
Lester Caine wrote
>
> I must say I'm seeing the same strange effects on routing in the UK.
> Although
> mapquest is a little slower, it does at least pick up the faster roads
> rather
> than routes that are perhaps 0.5km shorter but using roads with many
> roundabouts
> rather than the adjacent motorways or dual carriageways with none.
>
sometimes seem to generate rather different routes. Either, because they
assume different default speed profiles for various OSM highway classes,
because they add different heuristic penalties (such as for traffic lights
or corners), or because they implement different sub sets of OSM taggings.
Comparing the various routers and where they differ will hopefully help
improve those defaults, as well as identify areas, where the data needs to
be enriched so that the routers have an easier job on selecting the best
route.
In case people are interested, to make this comparison easier
http://apmon.dev.openstreetmap.org/routing/ allows you in a single interface
to select which of the 4 main OSM routing engines one wants to use (OSRM,
MapQuest Open, CloudMade and Gosmore) and allows to quickly switch between
them to compare. One should remember, however, that they all use data
extracts from different times. While OSRM and MapQuest should be pretty up
to date, I am not sure how often CloudMade or Gosmore update.
Unfortunately, given that the dev server, through which the results get
proxied, can be rather slow, one can't really appreciate the wonderful speed
of OSRM.
Kai
P.S. It is really great to see all those improvements flowing into OSRM! It
will hopefully help make OSM data ever more routable. Thanks and
congratulations to Dennis and everyone else who might have been involved!
--
View this message in context: http://gis.19327.n5.nabble.com/Lightning-fast-car-routing-built-on-OpenStreetMap-data-with-draggable-routes-tp5572804p5574723.html
Sent from the General Discussion mailing list archive at Nabble.com.
Nathan Edgars II
> Spotted in @openstreetmap's Twitter feed... I don't remember having ever
> used a routing service that fast. It is apparently tuned for car
> routing... And that's all I can say since the Karlsruher Institut für
> Technologie whose homepage is linked from the results panel doesn't seem
> to say anything about it. If anyone has further information about this
> exciting service...
Seems to be Europe only, despite using a US traffic sign in the logo.