Global dynamic optimization using edge-concave underestimator (Q2182856)

Global optimization of problems characterized by an embedded system of ordinary differential equations is very important in several engineering areas. Unfortunately, even simple instances of these problems are often nonconvex and exhibit multiple locally optimal solutions. In this article, the authors propose a new method to solve deterministic global optimization problems having an objective function and constraints expressed in integral terms, and an embedded set of nonlinear parametric ordinary differential equations. The method is based on a branch-and-bound algorithm that uses a new class of recently developed underestimators. The performance of the algorithm is compared to several current methods, and the effectiveness of the underestimator is illustrated through numerical examples. The main contributions of the work include proposing a relaxation that departs from traditional convexity-based approaches for global dynamic optimization, and developing a tight edge-concave underestimator for nonconvex problems involving integral terms.