I don't know about of any sneaky, optimized algorithms, but graphserver allows the definition and use in a basic dijkatra algorithm of arbitrary weight functions, where the weight function takes as input the state vector at the edge's origin vertex. That'd take I part of the problem. In general it may be simpler to use a stock dijkatra with edge weights found at runtime, than to use a specialized algorithm. Works for me.
-B (206) 992-7567 Sent from some sort of phone On Sep 19, 2008, at 8:34 PM, Stefan Pflumm <[EMAIL PROTECTED]> wrote: > Hello, > > I'm searching for an routing algorithm who can handle this two > properties: > > 1. the edge weight function c(E) of the edge E depends on the costs of > the predecessor edges. A* for example, can't handle this, because it > connects existings paths without updating the successors. If the path > previous to an edge changes and the edge is already expanded the costs > of the edge and all successors will also change. > > 2. the edge weight function c(E) of the node N depends on the > successor > edge. For example: E_1 and E_2 are neighbour edges of E. c(E) to E_1 > must be not equal to c(E) to E_2 > > I hope somebody can give me an advice, thanks > > Stefan > > > > > > _______________________________________________ > Routing mailing list > [email protected] > http://lists.openstreetmap.org/listinfo/routing _______________________________________________ Routing mailing list [email protected] http://lists.openstreetmap.org/listinfo/routing
