This paper focuses on the optimal maintenance and rehabilitation policiesof network-level pavement management problem associated with uncertainty in the deterioration process. Previous researchers have developed various optimization techniques for programming pavement maintenance and rehabilitation policies. However, pavement agencies usually deal with large and diverse networks consisting of thousands of pavement sections. As a result, the complexity of the pavement management problem increases exponentially as the size of the network increases. Optimization methods designed for small-sized problems thus suffer from the curse of dimensionality, significantly compromising the ability to solve problems in a reasonable time period. As a result, such problems should be approached from a different angle. For this reason, this paper proposes the Approximate Dynamic Programming method,which offers a practical solution to such pavement management problems that are difficult because of their size. The proposed method is a simulation-based forward recursion approach that avoids looping over every possible state of the value function. The proposed method also draws on past experience by incorporating the previous maintenance expenditures into theprocess of searching for future optimal policies. A case study is carriedout by applying the proposed method to both the budget planning and budget allocation problems of a typical pavement network.
Samenvatting