Using shortest paths in some transshipment problems with concave costs

This paper deals with the problem of solving an uncapacitated transshipment problem with either one source and several sinks or one sink and several sources. The cost function of the problem is concave in the amount shipped on each arc and thus local optima are possible. A characterization of adjacent extreme flows in terms of corresponding arborescences is given for this type of networks.This characterization together with shortest path methods is then used to attack the problems of finding local optima and of ranking extreme points.A real-world problem and computational evidence for the usefulness of the method are produced.