Motivation

Finite Domain Constraint Propagation and Integer Programming are two methods to solve and optimize systems of linear inequations over discrete domains. Experiments show that no one method has a general advantage over the other. Is seems rather that there are problems that are particularly well suited to either one or the other approach, owing to their different characteristics. But even different instances of the same problem can exhibit very different behaviours, which can make it impossible to chose the ``most suitable'' solver for a particular application.

These observations prompted the development of this library: It implements a hybrid solver based on cooperation between the ECLiPSe finite domain solver lib(fd) and the ECLiPSe external Simplex/MIP solver interface lib(eplex). The basic idea is to have the programmable control provided within ECLiPSe, the incremental bound propagation achieved by the finite domain solver, and the global reasoning that is done by the simplex solver.