Probabilistic Optimization of Pavement Preventive Maintenance Using Multi-Objective Genetic Algorithm

Abstract Highway agencies encounter the challenge of limited funding resources while striving to improve the condition of their pavement network. Therefore, optimization models have been developed to schedule maintenance and rehabilitation (M&R) activities for road networks. Previous probabilistic optimization models have primarily focused on the uncertainty of the budget constraint, neglecting other sources of uncertainty. Failure to account for the uncertainty of expected pavement condition and maintenance effectiveness in multi-year optimization models may lead to mistiming of maintenance applications and consequently result in suboptimal schedules. This paper develops stochastic optimization models that account for the uncertainty associated with the deterioration and improvement of pavement condition, as well as the budget constraint. The Multi-objective Genetic Algorithm (MOGA) is employed to create the optimization models due to its robust search capability, resulting in optimal or near-optimal global solutions. To reduce the computational expenses of the stochastic MOGA models, three approaches were implemented: (1) identifying and adopting the most commonly used maintenance treatments (2) clustering pavement sections based on their age, and (3) implementing a filtering constraint that enforces a rest period after treatment applications. The results demonstrate that the Pareto optimal solutions significantly change when considering the uncertainty of pavement condition deterioration and improvement. The developed stochastic MOGA models can provide highway agencies with probabilistic Pareto optimal solutions that account for the anticipated uncertainty in pavement condition data.