In this work we focus on an extension of the minimum cost flow problem in layered networks. Feasible arc flows must satisfy a specific compatibility restriction in addition to flow balance and capacity restrictions. Namely, at most one of the crossing arcs is allowed to have positive flow on it. This variant of the minimum cost flow problem, which we call the minimum cost noncrossing flow problem, can frequently be encountered in real life. The determination of optimal temporal quay crane allocations to berthed vessels in container terminals, and optimal train schedules through the stations on the same railroad line are two examples. We first analyze the complexity of the problem and show that the noncrossing flow problem is in fact NP-complete in a layered network. Then, we introduce mixed-integer linear programming formulations and discuss a polynomially solvable special case. Next we show a sufficient condition for the existence of a crossing in an optimal solution, which can be used for preprocessing the arcs in order to reduce the problem size. Our computational experiments on a large test set show that our preprocessing algorithm can significantly reduce the number of arcs.

}, keywords = {integer programming, Layered networks, Network flows, Noncrossing flow}, issn = {0166-218X}, doi = {https://doi.org/10.1016/j.dam.2018.09.016}, url = {https://www.sciencedirect.com/science/article/pii/S0166218X18304815}, author = {I. Kuban Altinel and Necati Aras and Zeynep {\c S}uvak and Z C Ta{\c s}k{\i}n} }