학술논문

New tests of optimality in Markov decision processes
Document Type
Conference
Source
Proceedings of 1994 33rd IEEE Conference on Decision and Control Decision and control Decision and Control, 1994., Proceedings of the 33rd IEEE Conference on. 1:180-183 vol.1 1994
Subject
Robotics and Control Systems
Computing and Processing
Testing
Cost function
State-space methods
Ear
Language
Abstract
Presents some new results on detection of optimal and non-optimal actions in Markov decision processes. In the discounted case the author provides two new tests to eliminate non-optimal actions and to keep optimal actions. These two conditions are direct extensions of similar conditions in the average cost case. The author then derives a new MacQueen type test for action elimination in the average cost case.ETX