Efficient and Stable Road Maintenance Strategy Evaluation Algorithm Utilizing a Trigger Horizon

Abstract

The optimization of road maintenance on large road networks over long analysis periods is an extremely complex problem. Different approaches to the problem have been investigated and methods such as dynamic programming, integer programming, and genetic algorithms have shown promise in providing near-optimal solutions. One of the general approaches to the road maintenance optimization problem uses a sequential approach in which feasible candidate treatment strategies are first generated for each road segment. These strategies can then be evaluated using benefit cost analysis principles and the most promising strategies can be used in the network optimization phase. A key challenge in this approach is the tendency for the candidate treatment strategy set to increase exponentially as the modeling period increases. This leads at best to long modeling times, and at worst to computational instability. This paper presents a fast, approximate alternative to the conventional decision tree-based approach to strategy generation. The method involves the use of a specified “trigger horizon” to control the number of years a triggered treatment can be postponed. The proposed method is shown to have a more gradual increase in the number of strategies generated as the modeling period increases. The method is also shown to provide overall network costs that are within 2% for an unconstrained budget, and 11% for a constrained budget scenario, of the cost generated by the more comprehensive but slower, less stable, decision tree-based methods.