From: alexti <al...@sh...> - 2010-03-21 17:47:31
|
Hi Stefan, On Sun, 21 Mar 2010 06:02:15 -0600, Stefan Frey <ste...@we...> wrote: > Some quick questions and comments: > > - I prefer your definition of the graph to the one in tiles.xml > (inherited > from the tiledesigner) as you only have to count edges and you do not > have to > consider that you are not allowed to visit the same side (a node/vertex > in > Rails) twice. Rails defines tile 25 (the green Y) with two edges that > both > start from the same side (see tiles.xml), but only one of each edge can > be > used per revenue turn. I was not clear on distinction between tile definitions and connectivity graph, but I've elaborated on that in another message. Basically, they are separate and the tile definition is similar to what you describe. > > - Is your algorithm implemented in Java and what is the calculation time > for > standard hardware and a reasonable difficult setting? (Your text below > seems > to imply close to zero, if you can use it for AI). It's in C++ and on 2 year old machine I see times < 0.01s, but "reasonable difficult setting" is something rather difficult to quantify. In practice I was finding that the worst case for the performance is not the end game but just before 4s come out, when everyone has 3-4 trains and the track has already been formed to allow multiple ways to run those short trains. > > - From your description I conjecture that you already maintain a > dictionary > that maps tile ids to your vertix/edge defintions. In that case that > would > easy to add, if the data format is used can be converted to xml. Rails > maintains a similar definition file in tiles/tiles.xml. Unfortunately, I don't have such dictionary, because I was entering all tiles in the definition (connectivity) form. However from your description it seems that the tiles in Rails are maintained in a very similar form so the conversion algorithm will be very similar to what I use right now. Alex. -- Using Opera's revolutionary e-mail client: http://www.opera.com/mail/ |