The Complexity of Flow Expansion and Electrical Flow Expansion.

FLOWEXPANSION is a network design problem, in which the input consists of a flow network and a set of candidate edges, which may be added to the network. Adding a candidate incurs given costs. The goal is to determine the cheapest set of candidate edges that, if added, allow the demands to be satisfied. FLOWEXPANSION is a variant of the MINIMUM-COST FLOW problem with non-linear edge costs. We study FLOWEXPANSION for both graph-theoretical and electrical flow networks. In the latter case this problem is also known as the TRANSMISSION NETWORK EXPANSION PLANNING problem. We give a structured view over the complexity of the variants of FLOWEXPANSION that arise from restricting, e.g., the graph classes, the capacities, or the number of sources and sinks. Our goal is to determine which restrictions have a crucial impact on the computational complexity. The results in this paper range from polynomial-time algorithms for the more restricted variants over NP-hardness proofs to proofs that certain variants are NP-hard to approximate even within a logarithmic factor of the optimal solution.