A Localized Method for the Multi-commodity Flow Problem

Abstract: This paper presents a local-control approach to address the multicommodity flow problem. The method involves relaxing both capacity constraints and flow conservation constraints. If the flow exceeds the capacity on an edge, the edge would incur a congestion cost. If the flow into a vertex is not equal to that out of the vertex, the vertex would have a height. Subsequently, a new concept, stable pseudo-flow, is introduced. Potential difference reduction algorithms, which don't rely on any shortest path or augmenting path, are designed to obtain stable pseudo-flow. If the stable pseudo-flow is a nonzero-stable pseudo-flow, there exists no feasible solution for multicommodity flow problem. Conversely, if the stable pseudo-flow is a zero-stable pseudo-flow, the feasible solution exists and the zero-stable pseudo-flow is the feasible solution. Notably, the algorithms work in a localized manner and can be efficiently implemented in parallel, which would further enhance performance.