[GraphHopper] Routing through highway=track

[Laurent Bendel]

Yeah, I hadn’t seen the github issue. I’ll reply to it to keep the discussion in one place.

Good example near Kiel where there’s a grade3 track, although IMHO a ‘Betonspurplatten’ should be a grade1..

On 6 janv. 2015, at 12:51, Peter <graphhopper at gmx.de> wrote:

> Hi Laurent,
> 
> thanks for your analysis, very appreciated! Also the knowledge in
> different countries is nice to have :)
> 
> I can remember we excluded those tracks as it was very unusual in most
> countries with grade > 1. So yes, if this is required then it is very
> important to include grade2 and even 3 (I'm not sure if higher), maybe
> we make this configurable in the CarFlagEncoder and per default it is 2?
> Later we make this country-specific (#175)
> 
> (BTW: The only downside will be that more streets are added to the
> network and algorithms without CH will be slightly slower for long routes)
> 
> Also I found this related issue (not yet looked again into it :)):
> https://github.com/graphhopper/graphhopper/issues/253
> 
> Regards,
> Peter
> 
> 
> On 06.01.2015 12:40, Laurent Bendel wrote:
>> OK, the code is actually pretty readable, I did my homework. It turns out the algorithm turns down tracks with tracktype=grade > 1. From CarFlagEncoder.acceptWay:
>> 
>> if (tt != null && !tt.equals("grade1"))
>>                return 0;
>> 
>> by changing this to 
>> 
>> if (tt != null && !tt.equals("grade1") && !tt.equals("grade2"))
>>                return 0;
>> 
>> I get correct results, and given the very low weight given to such tracks, it shouldn’t impact much other routings. Or would it ?
>> 
>> According to http://wiki.openstreetmap.org/wiki/Key:tracktype, grade2 should definitely be acceptable for a general purpose vehicle, grade3 probably too. As for the example outlined below (no man’s land between Mauritania and Morocco), it’s actually more like a grade4, mostly compacted sand, some soft sand. Still crossed by all kind of small cars every day. So if somebody corrects OSM with the more correct grade type, the routing will again fail. But this is unavoidable in the grand scheme of things.
>> 
>> Any thought ? submit a patch ?
>> 
>> Laurent
>> 
>> PS: note that in Australia, some tracks are labeled highway=secondary or even highway=primary, so the problem doesn’t exist.
>> 
>> On 5 janv. 2015, at 19:34, Laurent Bendel <asianrider at gmail.com> wrote:
>> 
>>> Hi there,
>>> I’m new to graph hopper, but I’ve already spotted a problem: it seems the algorithm decides that tracks (highway=track) are impassable by cars, but OK for bikes. E.g.:
>>> 
>>> https://graphhopper.com/maps/?point=Dakhla%20الداخلة%2C%20Morocco&point=Nouadhibou%2C%20Mauritania&layer=MapQuest
>>> 
>>> The problem is, in quite a few countries outside of Europe, main roads can be so bad that they are classified as tracks, but still used by thousands of normal cars, buses and lorries. The border between Mauritania and Morocco is a typical example, it is actually the only link between North and West Africa..!  Maybe they should be weighted « less » than highway=unclassified, but still taken in account. Any reason it is implemented this way ?
>>> 
>>> I’m using the graphhopper.com service, not my own server.
>>> 
>>> Thanks,
>>> Laurent
>>> 
>>> 
>> 
>> _______________________________________________
>> GraphHopper mailing list
>> GraphHopper at openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/graphhopper
> 
> 
> _______________________________________________
> GraphHopper mailing list
> GraphHopper at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/graphhopper