Global optimization of nonconvex bilevel problems: implementation and computational study of the Branch-and-Sandwich algorithm

摘要

We describe BASBL, an implementation of the Branch-and-Sandwich deterministic global optimization algorithm, (Kleniati and Adjiman, J. Glob. Opt. 60, 425–458, 2014), for nonlinear bilevel problems, within the MINOTAUR toolkit. The algorithm is extended to include heuristics for branching and node selection. The computational performance of the bilevel solver is analyzed for different combinations of the heuristics by solving nonconvex bilevel test problems from the literature.