Solution of the family traveling salesman problem using a hyper-heuristic approach

Abstract

This article is concerned with a recently introduced variant of the generalized traveling salesman problem (GTSP) called the family traveling salesman problem (FTSP). Given a set of nodes partitioned into multiple clusters termed as families, the FTSP consists in finding a tour visiting a pre-specified number of nodes from each of these families in such a manner that the total distance traveled is minimized. FTSP finds application in order picking in modern warehouses, where similar items can be stored at different places as the latest technologies like radio frequency identification (RFID) facilitate item localization. To solve this problem in a manner that is both effective and efficient, a hyper-heuristic approach is presented. This approach makes use of three large neighborhood search methods as low level heuristics. The solution obtained through the hyper-heuristic approach is improved further by using a local search phase. To assess the performance of the proposed approach, computational experiments are performed on the 86 standard benchmark instances of the problem. The proposed approach obtained good-quality solutions in comparison to the state-of-the-art approaches on these instances. Moreover, our approach is several times faster on most of the large instances. We have also reported the performance of our approach on a set of 60 new benchmark instances.