Uncertainty, time, and risk in the optimization of maintenance and rehabilitation of pavements

A Markov decision model for the optimization of one section of highway or street pavement maintenance and rehabilitation incorporating utility theory is outlined. The model takes account of the uncertain pavement behaviour and of the interdependence of maintenance actions over time. An approach for estimating and updating Markov transition matrices for pavements is proposed.The objective function quantifies the pavement manager's attitudes toward the risk posed by the probabilities of pavement condition, magnitudes of consequences, and timing of decisions. Multiattribute utility theory is employed to aggregate multiple criteria, and to model the pavement manager's preferences in multiyear planning scenarios.The methodology is applied to the optimization of maintenance and rehabilitation of one highway pavement section. The preferences of five engineers are tested. These tests show that additive evaluation models are not appropriate for pavement management. Utility functions of one engineer are used in an illustrative example to demonstrate feasibility of the approach.The presented model deals with the project level of decisions. However, the Markov decision approach combined with multiattribute utility can also be useful when modified to deal with questions arising at the network level. At both levels, the approach provides a powerful research tool capable of answering a variety of pavement management questions. Key words: pavements, maintenance and rehabilitation, management aids, Markov decision model, multiattribute utility, probability updating.