[gecode-users] Symbolic Constraints - fd_relation contraint in Gecode?

Pedro Salgueiro pds at di.uevora.pt
Sat Jun 13 14:47:04 CEST 2009


I guess that now I understood it, I will give it a try and see how that
goes.

Thank you guys, 
Pedro Salgueiro

On Sat, 2009-06-13 at 14:10 +0200, Denys Duchier wrote:
> Pedro Salgueiro <pds at di.uevora.pt> writes:
> 
> > Translating the large domain to a smaller one is a good idea, the
> > problem is that besides I need to use the extensional constraint to make
> > sure that the values belong to some matrix,
> 
> use extensional on the translated domains
> 
> > I also have to apply
> > constraints to some of the values individual values of the tuple, based
> > on the values of the large domain. So, if I translate the domain to a
> > smaller one, then I have no way to apply such constraints.
> 
> yes you do: apply the constraint on de-translated versions of the vars
> 
> > For instance, let's suppose I have the following matrix:
> >
> > matrix={{9900, 1234, 2349} 
> >         {9901, 2345, 5678}}
> >
> > wich will be translated to:
> >
> > translated_matrix={{0,1,2,3}
> > 	           {4,5,6,7}}
> 
> let's say:
> 
> translated_matrix={{0,1,2}
>  	           {3,4,5}}
> 
> so that the number of values actually correspond ;-)
> 
> > Then I will apply an extensional constraint to make sure that the
> > variables C={X1,X2,X3} and C1={Y1,Y2,Y3} belongs to translated_matrix.
> 
> you use extensional on the translated matrix.
> 
> > Then I need to apply constraints that relate individual values on each
> > of the tuple, such as
> >
> > Y1 - X1 < 10
> 
> I write A=<B1,B2,...,Bn>[I] for the element constraint.
> Constrain 2 new variables XX1 and YY1 as follows:
> 
> XX1 = <9900,1234,2349,9901,2345,5678>[X1]
> YY1 = <9900,1234,2349,9901,2345,5678>[Y1]
> 
> XX1 and YY1 are the de-translated versions of X1 and Y1, and you can
> pose the constraint:
> 
> YY1 - XX1 < 10
> 
> > About the element constraint, I didn't got that one.
> 
> I hope the above example clears it up for you.
> 
> Cheers,
> 
> --Denys





More information about the gecode-users mailing list