A Framework for Integrating Optimization and Constraint Programming
- 439 Downloads
This talk begins with a description of the modeling and computational advantages that can be obtained by combining optimization and constraint programming in a principled way. It then presents a framework for integration based on three elements: a search-infer-and-relax algorithmic paradigm, a unifying theory of duality, and the use of metaconstraints (a generalization of global constraints) for modeling. Inference techniques from constraint programming and relaxation techniques from mathematical programming are combined in both branch-and-relax search and constraint-based (nogood-based) search. The talk illustrates these ideas with examples in freight shipment, employee scheduling, continuous global optimization, airline crew scheduling, the propositional satisfiability problem, and multiple machine scheduling.