Abstract

Abstract. Benders ’ Decomposition (BD) is a prominent technique for tackling large mixed integer programming problems having a certain structure by itera-tively solving a series of smaller master and subproblem instances. We apply a generalization of this technique called Logic-Based BD, which does not restrict the subproblems to have continuous variables only, to a bi-level vehicle rout-ing problem originating in the timely distribution of printed newspapers to sub-scribers. When solving all master and subproblem instances exactly by CPLEX, it turns out that the scalability of the approach is quite limited. The situation can be dramatically improved when using a meaningful metaheuristic – in our case a variable neighborhood search – for approximately solving either only the sub-problems or both, the master as well as the subproblem instances. More generally, it is shown that Logic-Based BD can be a highly promising framework also for hybrid metaheuristics. 1