Selection and Scheduling of Interdependent Transportation Projects with Island Models

One major issue in solving the problem of selecting and scheduling interdependent transportation projects is the complex interdependence between projects that makes it difficult to evaluate the effects of individual projects or sets of projects. Previous researchers have developed mathematical models that capture some benefit and cost interactions among projects. Such models usually can handle some simple cases in which only pairwise interactions between projects exist, but they may not be realistic enough to capture the complex project interdependence for general transportation networks. Other issues arise in developing the algorithms to optimize the selection and scheduling of projects under resource constraints. For real applications, the developed algorithms should be effective and efficient in solving the problems for sizable transportation networks. To deal with these issues, this paper develops island models, which are variations of traditional genetic algorithms, for optimizing project selection and scheduling under resource constraints, while relying on a simulation model or an equilibrium traffic assignment model to evaluate the transportation project impacts. As a test case, a highway network is analyzed with equilibrium traffic assignment to evaluate the impacts of different combinations of transportation projects. The study investigates the island models with different domain partitioning and solution migration methods. The results show an island model generally can achieve better solutions than a traditional genetic algorithm.