MARKOVIAN DECISION PROCESSES WITH UNCERTAIN TRANSITION PROBABILITIES
Abstract
A dynamic programming formulation for the Markovian decision process when transition probabilities are unknown is proposed. This formulation is used to solve simple problems, but is shown to be too difficult to apply to more complex systems. Various approximate methods are then proposed and discussed. A simple approximating algorithm is finally presented.
Document Details
- Document Type
- Technical Report
- Publication Date
- Mar 01, 1965
- Accession Number
- AD0612601
Entities
People
- John M. Gozzolino
- Ralph L. Miller
- Romulo Gonzalez-zubieta
Organizations
- Massachusetts Institute of Technology