Pinning on Tight Cuts: Improved Algorithm and Bounds for Unsplittable Multicommodity Flows in Outerplanar Graphs

2026-06-03Data Structures and Algorithms

Data Structures and AlgorithmsDiscrete Mathematics
AI summary

The authors study a problem about sending multiple flows through a network without splitting any of them, focusing on a type of graph called outerplanar graphs. Previous researchers found some limits on how much capacity might need to be increased to make unsplittable flows possible in simpler graphs like cycles. The authors improve these limits for outerplanar graphs, showing the required capacity increase factor lies between 4/3 and 2. They also introduce a new method that looks at overall properties of the network, which might help in other similar problems.

multicommodity flowunsplittable flowundirected graphedge capacityouterplanar graphcut-conditionfeasible flowcycle graphdemand valueapproximation bounds
Authors
David Alemán Espinosa, Niklas Schlomberg
Abstract
The multicommodity flow problem in an undirected capacitated graph $G$ is specified by a set of source-sink pairs with nonnegative demands. A flow is feasible if it routes all demands without exceeding the edge capacities, and it is unsplittable if it routes each demand along a single path. Let $α$ be the smallest value such that the existence of a feasible flow implies the existence of an unsplittable flow that exceeds the edge capacities by at most $+\,α\,d_{\max}$, where $d_{\max}$ is the maximum demand value. Schrijver, Seymour, and Winkler showed that $α\in\left[1.01,\,1.5\right]$ if $G$ is a cycle. These bounds were ultimately improved to $α\in\left[1.1,\,1.3\right]$ by Skutella and Däubel. Recently, Alemán Espinosa and Kumar extended this constant upper bound to the broader class of outerplanar graphs, and showed that if $G$ is outerplanar then $α\le 3.6$. We show that $α\in\left[\tfrac{4}{3},2\right]$ if $G$ is outerplanar. We introduce a novel technique that considers the global parameters of the instance, and that may be useful in other (more general) settings where the cut-condition is sufficient, or nearly sufficient, for the existence of a feasible flow.