[Routing] generalized routing format

[flo Detig]

Robert (Jamie) Munro wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> flo Detig wrote:
>   
>> ps: some random thoughts:
>> What about even more layers?
>> In case of greedy (dijkstra-/a*-style) BFS you could start 'searching'
>> until the first higher-layer-node appears, for instance a 
>> residential-junction.
>> After that stay on that layer and only examine 'arcs' to other junctions
>> until you reach the first 'secondary-layer-junction', after which you 
>> can ignore
>> junctions to residential (lower-level) streets and move on in bigger steps.
>> Once you arrive at the motorway-level you search on that layer / graph
>> that solely consists of arcs between 'motorway-junctions' (german 
>> Autobahnkreuz)..
>> (could be implemented by adding an integer for layer to the nodes table
>> and computing shortest paths between neighbouring nodes in each layer
>> resulting in a edges-/arcs-/pgrouting-table with source-target-cost)
>> But how get 'down' again to find the final destination node?
>>
>> Please excuse if this sounds stupid. I'm just loud thinking..
>> Probably this is this too complicated.
>> Any opinions?
>>     
>
> As said in another post, it might work if combined with a Bidirectional
> search(*), but you'd have to be careful with what you put in each layer.
> For example, a small A road that worked as a short-cut between 2
> motorways would probably have to be on the motorway layer, otherwise you
> might find yourself having to traverse a large section of the motorway
> network to find your way around.
>
> For example, you don't want to go all the way around the M25 to get from
> Dartford to Thurrock, you want to take the A282 over the Dartford
> Crossing (which links the 2 ends of the almost circular motorway)
> http://www.openstreetmap.org/?lat=51.4718&lon=0.2514&zoom=12&layers=B000FTF
>
> Robert (Jamie) Munro
>
>
>   
That is an important point! The reason to stay on 'bigger' roads (e.g. 
motorways) is because it's generally faster. Any shortcut that is faster 
despite being on a lower layer network actually belongs to the higher 
layer network. Computing the layers seems thus more complicated.
Do you think it could suffice to do greedy (dijkstra like) searching in 
all directions, add nodes and edges  of the desired layer (with 
travel-time  as cost) and whenever such a shortcut appears add it as an 
additional edge?

flo detig