On Robustness of Mixed-Integer reformulations of Generalized Disjunctive Programs

In this work, we compare the performance of the Alternative Mixed-Integer (AMI) reformulation of GDPs against those of the Big-M and Hull Reformulation (HR) in a framework of nonconvex mixed-integer nonlinear models. The robustness and efficiency of the reformulations are tested on a synthetic example using commercially available solvers, namely SBB, DICOPT++, SCIP, and Alpha-ECP. The comparison is carried out on the following key criteria: efficiency (CPU usage, number of iterations), robustness (number of successfully solved problems), quality of the obtained solutions (comparison of the objective function values). Results obtained in this study indicate that the AMI reformulation represents a competitive alternative to Big-M and HR.