Partially Observable Markov Decision Processes with Applications.
STANFORD UNIV CALIF DEPT OF OPERATIONS RESEARCH
Pagination or Media Count:
The study examines a class of partially observable sequential decision models motivated by the process of machine maintenance and corrective action or medical diagnosis and treatment. Emphasis is placed on the dynamics of the state, i.e., the possibility that the machine disease state changes during the decision process. This is incorporated in the form of a Markov chain. It is also assumed that the state is only indirectly observable via outputs probabilistically related to the state. The end result is a model which is a discrete time Markov decision process with a continuous state space, a finite action space, and a special transition structure. Modified author abstract
- Operations Research