On the Parameterized Complexity of the Connected Flow and Many Visits TSP Problem

TL;DR

This paper investigates the parameterized complexity of the Connected Flow problem, a generalization of the Many Visits TSP, providing complexity results, algorithms, and kernelization based on parameters like demand set size, treewidth, and vertex cover.

Contribution

It offers new complexity classifications, fixed-parameter algorithms, and kernelization results for the Connected Flow problem, extending existing approaches and analyzing various graph parameters.

Findings

NP-complete for |D|=2 with unit demands and no costs

FPT algorithm with respect to vertex cover size k

Polynomial kernel for Many Visits TSP case

Abstract

We study a variant of Min Cost Flow in which the flow needs to be connected. Specifically, in the Connected Flow problem one is given a directed graph $G$, along with a set of demand vertices $D \subseteq V(G)$ with demands $\mathsf{dem}: D \rightarrow \mathbb{N}$, and costs and capacities for each edge. The goal is to find a minimum cost flow that satisfies the demands, respects the capacities and induces a (strongly) connected subgraph. This generalizes previously studied problems like the (Many Visits) TSP. We study the parameterized complexity of Connected Flow parameterized by $|D|$, the treewidth $tw$ and by vertex cover size $k$ of $G$ and provide: (i) $\mathsf{NP}$-completeness already for the case $|D|=2$ with only unit demands and capacities and no edge costs, and fixed-parameter tractability if there are no capacities, (ii) a fixed-parameter tractable $\mathcal{O}^{\star}(k^{\mathcal{O}(k)})$ time algorithm for the general case, and a kernel of size polynomial in $k$ for the special case of Many Visits TSP, (iii) a $|V(G)|^{\mathcal{O}(tw)}$ time algorithm and a matching $|V(G)|^{o(tw)}$ time conditional lower bound conditioned on the Exponential Time Hypothesis. To achieve some of our results, we significantly extend an approach by Kowalik et al.~[ESA'20].