A Unified Approach to Optimization

Abstract : The purpose of this research is to identify underlying unity in optimization methods, and to use the resulting insights to combine methods so as to exploit complementary strengths. In particular, we propose strategies for the integration of mathematical programming, constraint programming, dynamic programming, and heuristic methods. We report achievements in seven specific approaches to unification: high-level modeling, finite-domain cuts, bounds from decision diagrams, primal heuristics from decision diagram, decision diagrams and dynamic programming, logic-based Benders decomposition, and unification of exact and heuristic methods.