The capacitated single-allocation p-hub location routing problem: a Lagrangian relaxation and a hyper-heuristic approach

Abstract

A variant of the hub location routing problem studied in this work, which is the problem of locating a set of hub nodes, is establishing the hub-level network and allocating the spoke nodes to the hub nodes. As a particular property of this problem, each cluster of spoke nodes allocated to a hub constitutes a directed route that starts from the hub, visits all the spokes in the same cluster, and terminates to the same hub. We propose a hybrid of hyper-heuristic and a relax-and-cut solution method, which includes cooperation among several low-level heuristics governed and controlled by a learning mechanism. This hybridization provides a mechanism in which the obtained dual information through the Lagrangian relaxation (bundle) method being utilized to guide the local searches for constructing/improving feasible solutions. Several classes of valid inequalities as well as efficient separation routings are also proposed for being used within the relax-and-cut approach. Our extensive computational experiments confirm the efficiency of this solution method in terms of quality as well as computational time.