Hello -
I am trying to do program up a search engine that performs sequential refinement by adding more and more constraints to a model. In the first run, I am happy with a solution at all, but then I want to add increasingly complex constraints until there is no more solution (or the search engine times out). Now, I first run a model with a single constraint. If DFS can find a solution, then I run a completely new model with two constraints and from scratch. If this model is feasible, then I run the model with three constraints, etc. In principle, this solution works, but it is a very inefficient: a lot of time is spent generating the first feasible solution when restarting the search (especially as the problem becomes more constrained). It seems to me that this should be doable with RBS or a variation of the LDS engine, as described in the MPG. But there are very few examples to follow (the closest I found was this question posted on stack overflow http://stackoverflow.com/questions/34182681/how-to-change-the-variable-value-ordering-in-gecodes-restart-based-search but unfortunately it does not contain any answer). I simply cannot understand how to do it right. In particular, where do I post additional constraint on the model (in the master space, the slave space or elsewhere)? Any leads on how to do it, or links to code that adds constraints or changes branching order in RBS would be appreciated! Cheers, Dirk
_______________________________________________ Gecode users mailing list users@gecode.org https://www.gecode.org/mailman/listinfo/gecode-users
