The buy-at-bulk network design problem has been extensively studied in the general graph model. In this paper, we consider geometric versions of the problem, where all points in a Euclidean space are candidates for network nodes, and present the first general approach for solving them. It enables us to obtain quasi-polynomial-time approximation schemes for basic variants of the buy-at-bulk geometric network design problem with polynomial total demand. Then, for instances with a single sink and low capacity links, we design fast polynomial-time, low-constant approximation algorithms.
Approximation Algorithms for a Capacitated Network Design Problem
