An improved approximation ratio for the minimum latency problem

Research paper by Michel Goemans, Jon Kleinberg

Indexed on: 01 Jun '98Published on: 01 Jun '98Published in: Mathematical Programming


Given a tour visitingn points in a metric space, thelatency of one of these pointsp is the distance traveled in the tour before reachingp. Theminimum latency problem (MLP) asks for a tour passing throughn given points for which the total latency of then points is minimum; in effect, we are seeking the tour with minimum average “arrival time”. This problem has been studied in the operations research literature, where it has also been termed the “delivery-man problem” and the “traveling repairman problem”. The approximability of the MLP was first considered by Sahni and Gonzalez in 1976; however, unlike the classical traveling salesman problem (TSP), it is not easy to give any constant-factor approximation algorithm for the MLP. Recently, Blum et al. (A. Blum, P. Chalasani, D. Coppersimith, W. Pulleyblank, P. Raghavan, M. Sudan, Proceedings of the 26th ACM Symposium on the Theory of Computing, 1994, pp. 163–171) gave the first such algorithm, obtaining an approximation ratio of 144. In this work, we develop an algorithm which improves this ratio to 21.55; moreover, combining our algorithm with a recent result of Garg (N. Garg, Proceedings of the 37th IEEE Symposium on Foundations of Computer Science, 1996, pp. 302–309) provides an approximation ratio of 10.78. The development of our algorithm involves a number of techniques that seem to be of interest from the perspective of the TSP and its variants more generally. © 1998 The Mathematical Programming Society, Inc. Published by Elsevier Science B.V.