DYNAMIC DECISION MODEL FOR A PAVEMENT MANAGEMENT SYSTEM

Pavements represent gradually deteriorating structures for which observations of advance signs of impending failure are possible. Most agencies collect pavement condition data on a regular basis to identify such signs. However, neither the timing of occurrences of these signs nor the timing of actual failure following the signs can be predicted with certainty. Given this probabilistic behavior of pavements and the availability of periodic pavement condition data, a dynamic decision model is much more appropriate for such pavement management decisions as the selection of cost-effective pavement preservation actions and forecasting of future performance of a highway network. In this paper the basic structures of static and dynamic decision models, and a special class of dynamic decision models called a Markovian decision process, are described. Among the significant advantages of this model are reliable predictions of the future performance of a highway network and the identification of preservation actions that are generally less conservative (and less costly) than traditional choices of actions and yet maintain the network performance at prescribed standards. A successful application of the Markovian decision process to the pavement management system in Arizona is described.