Polynomial-time approximability of the k-Sink Location problem

A dynamic network ${\cal N} = (G,c,τ,S)$ where $G=(V,E)$ is a graph, integers $τ(e)$ and $c(e)$ represent, for each edge $e\in E$, the time required to traverse edge $e$ and its nonnegative capacity, and the set $S\subseteq V$ is a set of sources. In the $k$-{\sc Sink Location} problem, one is given as input a dynamic network ${\cal N}$ where every source $u\in S$ is given a nonnegative supply value $σ(u)$. The task is then to find a set of sinks $X = \{x_1,\ldots,x_k\}$ in $G$ that minimizes the routing time of all supply to $X$. Note that, in the case where $G$ is an undirected graph, the optimal position of the sinks in $X$ needs not be at vertices, and can be located along edges. Hoppe and Tardos showed that, given an instance of $k$-{\sc Sink Location} and a set of $k$ vertices $X\subseteq V$, one can find an optimal routing scheme of all the supply in $G$ to $X$ in polynomial time, in the case where graph $G$ is directed. Note that when $G$ is directed, this suffices to obtain polynomial-time solvability of the $k$-{\sc Sink Location} problem, since any optimal position will be located at vertices of $G$. However, the computational complexity of the $k$-{\sc Sink Location} problem on general undirected graphs is still open. In this paper, we show that the $k$-{\sc Sink Location} problem admits a fully polynomial-time approximation scheme (FPTAS) for every fixed $k$, and that the problem is $W[1]$-hard when parameterized by $k$.