In the paper, we consider killed Markov decision processes with uncountable sets of states and controls on a finite time interval. Definitions of killed Markov decision process and assessment of the way and optimal policy are given, as well as fundamental equation is proved in the case where the set of states and set of controls are measurable spaces. We also proposed a method to construct the optimal strategy and proved the existence of a uniformly optimal policy in case where the set of states and set of controls are separable metric spaces. Figs: 0. Refs: 5 titles.
Шпак Павел Романович, аспирант Львовского национального университета имени Ивана Франко,
e-mail: prshpak@gmail.com
Елейко Ярослав Иванович, доктор физ.-мат. наук, профессор Львовского национального университета имени Ивана Франко,
e-mail: yikts@yahoo.com