[GraphHopper] Information about graphhopper algorithms

Peter K peathal at yahoo.de
Fri Jan 17 21:01:52 UTC 2014


Hey Angelo,

> the Dijkstra-based algorithms (let's call them so)

CH is also Dijkstra based. Lets call them none-CH or normal algos ;)


> multiple paths from node A to node B

Do you mean alternative paths? Currently also not possible out of the
box but interesting topic. (Of course with different vehicle or
weighting there is some kind of alternative path possible.)


> multimodal paths from node A to node B

Currently not the possible with GraphHopper. Or you need to give us more
details about your requirements.


> a path composed by some "cars" ways, "pediastran" ways, "cycle" ways
and so on....

The details of which vehicle can follow which one and how often you can
'switch' is highly dependent on your usecase. E.g. taxi? car rent?
normal car? how you would mix bike and car? Or why foot and bike?


> If it's correct.....are you thinking to implement them?

Not on the priority list but this is highly dependent on the community
(pull) requests.

Kind Regards,
Peter.


> Hi
>
> First of all compliments for the GraphHopper Routing Machine...it's
> really amazing
>
> We are thinking to use it in our scenario; since in our case we can't
> use the CH algorithm due to the fact that we have highly dinamyc
> weight values, we are thinking to use the Dijkstra-based algorithms
> (let's call them so) like A*, Bi-directional Dijkstra and so on...
> In order to not use the CH algorithm Peter suggested to us to specify
> prepare.chShortcuts=no in the config.properties file
>
> We did it and we are checking the algorithms...in our case we should
> also provide:
> multiple paths from node A to node B
> multimodal paths from node A to node B
>
> What I mean is that when we calculate the route from point A to point
> B we should be able in providing:
> multiple paths and che client can choose the wished one
> multimodal path: a path composed by some "cars" ways, "pediastran"
> ways, "cycle" ways and so on....
>
> It seems to us that the previously listed cases are not possible in
> GraphHopper (above all the possibility to return a list of Path)....Is
> this correct?
> If it's correct.....are you thinking to implement them? Can we find
> any kind of implementation anywhere?
>
> Thank you
> Angelo
>




More information about the GraphHopper mailing list