Inria / Raweb 2004
Project-Team: MODBIO

Search in Activity Report, year 2004:


Project-Team : modbio

Section: New Results

Keywords: constraint programming, integer programming, cooperative solving.

Constraint programming and integer programming

Participant: Alexander Bockmayr.

In a joint work with John N. Hooker (CMU), we present in [12] a state-of-the-art survey of constraint programming (CP), with special emphasis on its relationship to mixed integer programming (MIP). CP methods exhibit several parallels with branch-and-cut methods for MIP. Both generate a branching tree. Both use inference methods that take advantage of problem structure: cutting planes in the case of MIP, and filtering algorithms in the case of CP. A major difference, however, is that CP associates each constraint with an algorithm that operates on the solution space so as to remove infeasible solutions. This allows CP to exploit substructure in the problem in a way that MIP cannot, while MIP benefits from strong continuous relaxations that are unavailable in CP. We overview basic concepts of CP, including consistency, global constraints, constraint propagation, filtering, finite domain modeling, and search techniques. We then indicate how CP may be integrated with MIP to combine their complementary strengths.