A Multi-objective Evolutionary Methodology for an Interdependent Transportation Project Selection Problem

Most transportation planning evaluation practices involve the aggregation of impacts into a single utility function in order to be optimized. Techniques that translate different units of measure into monetary terms are highly controversial. The multi-objective transportation infrastructure project selection problem (MTPSP) can be mathematically modeled as a multi-objective 0-1 knapsack problem with some additional constraints. We solved a MTPSP for the Mexican environment using an evolutionary approach, while considering explicit modeling of interdependent projects effects (quadratic objectives). Such inter dependency relationships were estimated using a gravity model. Due to the distinctive "steps-shape" of the Pareto frontier, an evolutionary-guided search over knee points was used to avoid keeping non-attractive solutions in the final decision stage. Using context-dependent preferences and having explicit knowledge of the filtered Pareto solutions, the final choice was made using a systematic tradeoff procedure to find the one that best represented the decision maker's interests under the existing context