[GraphHopper] GPS-exact routing prioritizes snapping over length

Peter K peathal at yahoo.de
Wed Jan 1 17:51:41 UTC 2014


Hi Joel,

> I've gotten GPS-exact results working thanks to Philipp, and so far
I'm seeing great results.

The latest master has the same feature now.

> Is there any way to add some sort of prioritization to always choose a
shorter path over a closer snap to a path?

If you don't use CH you could indeed implement something
(none-bidirectional) where you only need to modify the finish condition
which checks against a set of nodes instead of one.

But I think what you could need is a proper map matching algorithm (or
also called conflation). Map matching is a research area where where you
take into account the position history, the speed, the vehicle, a
possible previous calculated path, etc

Regards,
Peter.


> Hi, 
> I've gotten GPS-exact results working thanks to Philipp, and so far
> I'm seeing great results. Mostly everything works and the paths are
> looking great.
> I have however run into a slight issue. I suspect it's due to the way
> the routing and snapping is designed. I have several situations where
> I have two parallel paths a and b which eventually converge. My
> departure point is on path a but my destination path is closest to
> path b, but in a place where a and b are still parallel. The routing
> will choose a much longer path which routes me past my destination on
> path a and then switches back/double back on path b. 
> I've attached two screenshots which show an overview and a detail of
> my destination point, in which you can see two paths are parallel. 
>
> Is there any way to add some sort of prioritization to always choose a
> shorter path over a closer snap to a path? 
>
> Joel Haasnoot
>
>
> _______________________________________________
> GraphHopper mailing list
> GraphHopper at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/graphhopper

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/graphhopper/attachments/20140101/ece3ddfb/attachment.html>


More information about the GraphHopper mailing list