Igor Kondrasovas wrote: > Hello, > > I would like to use the the bb_min /3 predicate to find the lowest > “Cost” solution available in my search / 6 results: > > The search is: search(XYs, 0, first_fail, indomain_split, complete, []), > > XYs is a list of points (X,Y) elements and the cost is equals to the > maximum X value on the entire list, so my goal is to minimize this cost. > > I would like to know if I could use maxlist /2 to define the cost value > or if there is any built-in predicate I could use, since it is not a > simple scalar numeric list (only X value is considered). Your code is perfectly fine (you do need to construct the auxiliary list of Xs). By the way, maxlist(Xs,Cost) is the same as Cost #= max(Xs). > Here is the way I´m defining the predicate. In fact the code hangs when > the bb_min is used. First try doing the search without the bb_min: does it find a solution? If yes, it should at least find that same first solution with the bb_min, print "Found a solution with cost ....", and then go on trying to find a better one. It is also quite likely that you need to consider a search heuristic that is better suited to the geometrical problem you have. What I mean is that your overlap constraints are of the form "no x overlap OR no y overlap". When you have an object at (x1,y1) which must not overlap with (X,Y), then that does not restrict the domains of X or Y, because they are allowed to overlap in one of the dimensions. A 2-D aware search routine could explore the possible relative position of pairs of objects by imposing branching constraints of the form X>x1 or X<x1 or Y>y1 or Y<y1. -- JoachimReceived on Wed May 19 2010 - 05:22:28 CEST
This archive was generated by hypermail 2.3.0 : Wed Sep 25 2024 - 15:13:20 CEST