[GraphHopper] bidirectional Dijkstra algorithm is not optimal

[Peter K]

Thanks, I'll investigate.

>  
>
> Hi everyone,
>
>  
>
> there seems to be a bug routing with bidirectional algorithms. I've
> got different routing results for same queries. (see attached pictures)
> All bidirectional algos are picking suboptimal pathes sometimes.
>
> For all I know A*-bidirectional is supposed to result in possible
> suboptimal routes, but dijkstrabi and dijkstraNativebi should find the
> shortest/fastest path.
> The finish condition (currFromWeight + currToWeight >=
> nativeBestPath.getWeight())should accept only for the best path or am
> I missing something?
>
>
> Note: to replicate the attached routes you have to disable contraction
> hierarchies
> here are the related URLs:
> http://localhost:8989/?point=48.983206%2C12.140568&point=48.983438%2C12.136411&weighting=shortest&algorithm=dijkstra
>
> http://localhost:8989/?point=48.983206%2C12.140568&point=48.983438%2C12.136411&weighting=shortest&algorithm=dijkstraNative
>
> http://localhost:8989/?point=48.984428%2C12.144517&point=48.983146%2C12.136631&algorithm=dijkstra
>
> http://localhost:8989/?point=48.984428%2C12.144517&point=48.983146%2C12.136631&algorithm=dijkstrabi
>
>  
>
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/graphhopper/attachments/20140212/87841d10/attachment.html>