Highway alignment optimization based on cost minimization requires comprehensive formulation of costs sensitive to alignment and development of efficient solution algorithms. The complexity of the applicable cost functions severely limits the search algorithms that can be employed. Recently, genetic algorithms, which can search very effectively through complex spaces with huge numbers of local optima, have been successfully developed for highway alignment optimization. However, in order to solve realworld problems the optimization algorithms should work directly with a Geographic Information System (GIS) which stores relevant geographic information, such as land boundaries, environmentally sensitive regions, and topographic data. This paper presents a model for highway alignment optimization that integrates a GIS with genetic algorithms, examines the effects of various costs on alignment selection, and explores optimization in constrained spaces that realistically reflect the limits on road improvement projects. The paper integrates several previously published developments and adds some new analysis approaches. A real-world problem using a GIS database for Maryland is solved using the proposed method. An example using an artificial map to investigate the effectiveness of the proposed model in mountainous terrain is also demonstrated. The results indicate that travel-time cost, which is often neglected by highway agencies in selecting alignments, significantly affects the alignment optimization. Computation time increases significantly with the number of properties affected by each alignment. The model can optimize alignments in mountainous terrain or regions with very complex geography. (C) 2004 Published by Elsevier Ltd.
