Solving the Clustered Traveling Salesman Problem Using the Lin-Kernighan-Helsgaun Algorithm

The Clustered Traveling Salesman Problem (CTSP) is an extension of the Traveling Salesman Problem (TSP) where the set of cities is partitioned into clusters, and the salesman has to visit the cities of each cluster consecutively. It is well known that any instance of CTSP can be transformed into a standard instance of the Traveling Salesman Problem (TSP), and therefore solved with a TSP solver. This paper evaluates the performance of the state-of-the art TSP solver Lin-Kernighan-Helsgaun (LKH) on transformed CTSP instances. Although LKH is used as a black box, without any modifications, the computational evaluation shows that all instances in a well-known library of benchmark instances, GTSPLIB, could be solved to optimality in a reasonable time. In addition, it was possible to solve a series of new very-large-scale instances with up to 17,180 clusters and 85,900 vertices. Optima for these instances are not known but it is conjectured that LKH has been able to find solutions of a very high quality. The program is free of charge for academic and non-commercial use and can be downloaded in source code.