Markov Decision Processes with Their Applications (Hardcover)

Markov decision processes (MDPs), also called stochastic dynamic programming, were first studied in the 1960s. MDPs can be used to model and solve dynamic decision-making problems that are multiperiod and in stochastic circumstances. There are three basic branches in MDPs: discrete-time MDPs, continuous-time MDPs and semi-Markov decision processes. Starting from these three branches, many generalized MDPs models have been applied to various practical problems. These models include partially observable MDPs, adaptive MDPs, MDPs in stochastic environments, and MDPs with multiple objectives, constraints or imprecise parameters. MDPs have been applied in many areas, such as communications, signal processing, artificial intelligence, stochastic scheduling and manufacturing systems, discrete event systems, management and economies. This book examines MDPs and their applications in the optimal control of discrete event systems (DESs), optimal replacement, and optimal allocations in sequential online auctions.The book presents three main topics: a new methodology for MDPs with discounted total reward criterion; transformation of continuous-time MDPs and semi-Markov decision processes into a discrete-time MDPs model, thereby simplifying the application of MDPs; application of MDPs in stochastic environments, which greatly extends the area where MDPs can be applied. Each topic is used to study optimal control problems or other types of problems.