Abstract:While mixed-integer linear programming and convex programming solvers have advanced significantly over the past several decades, solution technologies for general mixed-integer nonlinear programs (MINLPs) have yet to reach the same level of maturity. Various problem structures across different application domains remain challenging to model and solve using modern global solvers, primarily due to the lack of efficient parsers and convexification routines for their complex algebraic representations. In this paper, we introduce a novel graphical framework for globally solving MINLPs based on decision diagrams (DDs), which enable the modeling of complex problem structures that are intractable for conventional solution techniques. We describe the core components of this framework, including a graphical reformulation of MINLP constraints, convexification techniques derived from the constructed graphs, efficient cutting plane methods to generate linear outer approximations, and a spatial branch-and-bound scheme with convergence guarantees. In addition to providing a global solution method for tackling challenging MINLPs, our framework addresses a longstanding gap in the DD literature by developing a general-purpose DD-based approach for solving general MINLPs. To demonstrate its capabilities, we apply our framework to solve instances from one of the most difficult classes of unsolved test problems in the MINLP Library, which are otherwise inadmissible for state-of-the-art global solvers.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is: although mixed - integer linear programming (MILP) and convex programming solvers have made remarkable progress in the past few decades, solution techniques for general mixed - integer nonlinear programming (MINLP) have not reached the same level of maturity. The complex problem structures in many different application areas are still difficult to model and solve using modern global solvers, mainly due to the lack of efficient parsers and convexification methods to handle their complex algebraic representations. Therefore, this paper introduces a new graphical framework based on decision diagrams (DDs) for globally solving MINLPs. This framework can model complex problem structures that are difficult for traditional solving techniques to handle, provides a new global solving method, and fills a long - standing gap in the decision diagram literature.
Specifically, the main contributions of the paper include:
1. **Contributions to the MINLP literature**: Existing global MINLP solving methods mainly rely on the algebraic representation of constraints. Non - convex nonlinear terms are usually decomposed into simpler forms and convexified separately, which makes these techniques highly dependent on the specific algebraic forms and properties of functions. For example, hyperbolic and trigonometric functions are common in various applications from artificial intelligence to energy systems, but due to the lack of suitable convexification methods, these structures are usually unacceptable in leading global solvers such as BARON and SCIP. In contrast, the decision - diagram - based framework provides significant flexibility in the functional form of the problem. The graphical nature of decision diagrams allows for direct evaluation and effective relaxation of underlying terms without decomposition when constructing decision diagrams. Therefore, this framework can be effectively applied to a wide range of MINLPs containing highly nonlinear, non - convex, non - smooth, and even black - box functions, many of which cannot be handled or are poorly handled by modern global solvers.
2. **Contributions to the decision diagram literature**: Although decision diagrams have been successfully applied to various optimization problems in the past two decades, there are still two main limitations in their applicability: (i) Decision diagrams are mainly applied to problems with specific structures; (ii) Decision diagrams were initially limited to modeling discrete programs. These limitations have hindered the widespread adoption of decision diagrams and highlighted the need to develop general decision - diagram techniques for solving MINLPs. This paper addresses this gap by introducing a new branch - and - cut framework that utilizes the structure of decision diagrams at each stage of the solving process, from constructing effective relaxations, to generating cutting planes and outer approximations, to spatial branch - and - bound methods with convergence guarantees. Therefore, this work establishes the first general - purpose solving method based on decision diagrams for globally solving general MINLPs.
Overall, this paper aims to overcome the limitations of existing global MINLP solving methods by proposing a new graphical framework based on decision diagrams, providing a more flexible and powerful solving tool suitable for a wider range of complex problems.