Re: Using real-valued solver for propogation purposes

Hi Chris,

On Thu, Jan 23, 2003 at 04:57:57PM +0000, Christopher Jefferson wrote:
> I would like to use a real-valued linear constraint solver to perform 
> bounds-tightening on the linear constraints in a CSP, rather than the (I 
> believe) usual method of performing 'bounds consistency' on each of the 
> linear constraints. The CSP will contain other constraints as well of 
> course. I suspect that eclipse already contains some method of doing this 
> but I have been unable to find it...

I don't think ECLiPSe has anything in particular to support this, other than
using eplex's lp_probe predicate to optimise for different objective
functions.

I also asked around here about this kind of technique, and apparently it's
studied in some detail in Susanne Heipcke's PhD thesis:

Heipcke, S.: Combined Modeling and Problem Solving in Mathematical
Programming and Constraint Programming, PhD Thesis, School of Business,
University of Buckingham, 1999

Yves Colombani has also done some work in this area.

Hope that helps.

Cheers,
Warwick