An important function of a pavement management system (PMS) is to provide decision supports on pavement improvement decision-makings. In this regard, a decision model is used to establish an optimal improvement policy which prescribes an action pertaining to each pavement condition. A policy is optimal if the discounted value of all the expected costs within the planning horizon is minimum. With the successful application of Markov chain theory in pavement deterioration modeling, the theory of Markovian Decision Process (MDP) has become the current state-of-the-art method in decision models (Carnahan et al 1987, Butt et al). This approach treats the transition of pavement performance from the current state to another state under either an improvement action or no human intervention as a Markov chain. The Markov chain deterioration and decision models are well developed and are commonly used in existing PMSs (Golabi et al 1982). However, the underlying assumption of geometric (for discrete time) or exponential (for continuous time) holding times in the Markov-chain model is unneccessarily restrictive. Should we not let the available data determine the appropriate distribution? This paper discusses an alternative approach using the more general semi-Markov Theory.
展开▼
