By Zulqarnain Haider
The speaker-packed session on models and methods for network design was about freight logistics and supply chain problems modelled as service network design problems. The first talk by Mingyao Qi from Tsinghua University, China was about robust service network design under demand uncertainty. The presenter started the talk by mentioning the problem of demand uncertainty for freight forwarding companies followed by some of the approaches, including deterministic and stochastic, used to address the problem. His talk focused on instead using two-stage robust optimization methodology for handling demand uncertainty. He presented the mathematical formulation and a column and constraint generation algorithm to solve the proposed robust models exactly.
The second talk was given by Huan Jin from MIT Global SCALE Network in Ningbo, China. Her presentation was about service network design (SND) with node capacity with application in parcel delivery. She mentioned the case of urban logistics in China where the relatively smaller storage terminals within the city may have capacity constraints in terms of space and parking. After introducing the network formulation and the branch-and-price method used to solve it, she concluded by analyzing the algorithm through comprehensive numerical results.
The third speaker, Ira M. Wheaton, Jr. (photo) from Georgia Tech, talked about integer programming models for freight logistics network design problem with in-tree constraints. The constraints ensure that “at any node in their service network, all shipments that are handled at that node and that have the same destination must be routed via the same outgoing arc.” He presented multiple integer programming models for SND problems with in-tree constraints, compared them and presented the relevant computational results.
The fourth speaker, Qiang Qiang from Penn State, talked about his model for competitive closed-loop supply chain network design problem with equilibrium constraints. The work aims to model the reactions of existing firms when an entering firm enters the closed-loop supply chain network and must compete with the existing supply chain players. He concluded by underlining the complexity of solving the problem and the algorithm, called LQP-ALM, they employed. Numerical examples and results followed.
Julia Y. Yan of MIT then discussed her work on data-driven transit network design problems. She identified the problem of falling ridership throughout U.S. cities and proposed a model to design a transit network that promises higher ridership. She presented a tractable column and constraint generation algorithm to solve the problem of selecting a subset of transit routes that maximizes ridership while satisfying certain coverage constraints. The presented approach “is able to double the transit ridership while satisfying regulatory requirements under tight financial constraints.”
In the final talk, session chair Angelika Leskovskaya, Southern Methodist University, spoke about her work on search heuristics for generalized network flow problems (GNFP). These problems can represent changes of flow on the arcs using multipliers or fixed charges on the arcs. She presented network simplex-based algorithms that exploit the problem structure and heuristic approaches for efficiently solving fixed-charge transportation problems, fixed-charge transshipment problems, and other generalized network problems.