New Aggregation Methods for Multicommodity Network Flow Problems: Theory and Applications to Locomotive Refueling

Large scale network optimization problems appear in a wide variety of logistical and transportation applications. However, such problems are usually difficult to solve using current technology. This thesis is motivated by such a problem in railroad fuel management. It designs mixed-integer programming methods and algorithms to solve the fuel-related challenges of the largest Australian railroad company. Moreover, it extends state-of-the-art optimization techniques to handle large scale network optimization problems, including the fuel management problem that has motivated the thesis.