There has been a concerted research effort to update the prediction and optimization capabilities of the micro paver pavement management system. An approach uniting the markov probability prediction curves and dynamic programming optimization has been proposed. This paper briefly outlines the background work done on both. It examines the sensitivity of the dynamic programming output to changes in input parameters. Data from three existing databases and a fourth, formulated database are used in the analysis. Output values from dynamic programming are also compared with results obtained from deterministic analysis using best-fit curves for prediction and cost versus condition data. It is concluded that the dynamic programming results are reasonable. A multizone dynamic programming approach was seen to give more consistent results than a one-zone approach. It was found that for life-cycle lengths greater than 15 years, there was little change in optimal decision or cost. Low interest rates favored more expensive, longer-lasting solutions. The minimum pavement condition index (pci) level specified affected the results appreciably, in general. These results were consistent through all four databases examined. It was concluded that the markov/dynamic programming approach was functioning satisfactorily, and was suitable and appropriate for use at the microcomputer level. This paper appears in transportation research record no. 1215, Pavement management and rehabilitation.
Abstract