On Markov Decision-making At The Network Level Pavement Management System

The application of Markov decision in PMS has been studied a lot in foreign countries, but in our country it is contrary. Based on internal and external experience of making Markov decision of PMS, this paper gives the correlative Markov decision theories. Then along these theories, this paper makes efforts in how to make the Markov decision model of Network Level PMS. This paper involves: classifying of pavement state and typical measures, transfer probability, cost model and Markov decision model. In the last part this paper provides some methods to solve the benefit model.Also, this paper puts forward to the classifying criteria of pavement state and summarizes the typical measures, gives the concept of "integration measure", and narrates how to harmonize and decompose the whole road network.In order to adapt the maintenance activities well, this paper makes some ameliorations, as follows:(1) It is considered not so pertinent in this paper to take road user cost into account of total cost, and an advice of using pavement serviceability performance for benefit is given.(2) When evaluating the average of the road network serviceability, this paper put forward to using weighed geometric mean to calculate PQI.(3) Bringing in a new idea of checking the reasonability of classifying pavement state and typical measure by transfer probability.(4) Besides linear programming, this paper also gives the other two methods to solve Markov decision model: policy iterative method, integer programming, as well as the method working out the optimal short-term maintenance policy.In chapter 5, an example is given to show how to use Markov decision.