Multiobjective optimization for pavement maintenance programming

Abstract

Pavement maintenance planning and programming requires optimization analysis involving multiobjective considerations. Traditionally single-objective optimization techniques have been employed by pavement researchers and practitioners because of the complexity involved in multiobjective analysis. This paper develops a genetic-algorithm-based procedure for solving multiobjective network level pavement maintenance programming problems. The concepts of Pareto optimal solution set and rank-based fitness evaluation, and two methods of selecting an optimal solution, were adopted. It was found that the robust search characteristics and multiple-solution handling capability of genetic-algorithms were well suited for multiobjective optimization analysis. Formulation and development of the solution algorithm were described and demonstrated with a numerical example problem in which a hypothetical network level pavement maintenance programming analysis was performed for two- and three-objective optimization, respectively. A comparison between the two- and three-objective solutions was made to highlight some practical considerations in applying multiobjective optimization to pavement maintenance management.