Bilevel optimization methods to evaluate intermodal freight transportation investments in Colombia

The interaction between a national economy and the transportation system that supports it is dual in nature, and changes in one can have an impact on the other. This is certainly the case in countries with developing economies where the government's efforts to improve transportation infrastructure is driven by the need for economic development and improved market access. Modeling such systems requires accounting for the autonomous decision makers at the government and user level. This leads to a hierarchical structure of decision making analogous to the leader-follower problem. At the upper-level the government or transport planner (leader) minimizes investment costs while accounting for the route choice behavior of network users. At the lower-level the users (followers) seek to optimize their mode/route choice by minimizing transportation costs in a user-equilibrium sense. In this thesis we posit a set of optimization models to answer the research question of efficiently selecting infrastructure investment projects on intermodal freight transportation networks in a developing economy. In answering such a question, we present a network design problem as a single level optimization model of an intermodal route assignment problem subject to an investment budget constraint. We then formulate another version of the network design problem as a bilevel mixed integer nonlinear programming problem where we include the investment costs as part of the decision variables in the objective function rather than through a budget constraint. We implement decomposition techniques (Generalized Benders Decomposition-GBD and Outer Approximation-OA) to solve the bilevel models. Using the single level and bilevel models, we evaluate a proposed inter- modal freight transportation corridor between the eastern region of Llanos and the western Pacific port of Buenaventura in Colombia. Our findings suggest that both the single level and the bilevel models have their advantages and limitations and neither is superior over the other as a tool for evaluating investments in intermodal freight routing. We find that these models are useful tools for assessing and evaluating transportation corridor projects at the strategic planning level. The planner could benefit from both model implementations when analyzing waterway investments in an intermodal freight transportation system as the case in Colombia.