Logit network pricing

In this paper, we address a pricing problem defined over a transportation network, and where the underlying flow is assigned according to the logit model. This setting yields a nonlinear optimization problem that yet possesses strong combinatorial features. Taking advantage of the model's analytical properties, we characterize its first-order optimality conditions and introduce rules that simplify the network topology, while leaving the structure of optimal solutions unchanged. Based on these results, a class of unimodal instances is identified. Finally, the connection with the more classical problem known as 'product line pricing' in economics is emphasized.