Título: Remark on Constrained Markov Decision Processes with Discounted Cost
Autor: Juan González Hernández


In the constrained Markov decision processes with discounted cost the objetive is to minimize an expected discounted cost I subjet to another expected discounted V to less than a constant K. In this paper we point out with an example some of the problems one way get when one works under constrains such as: Bellman's principle of optimality does not hold, optimal policies may depend upon initial state, optimal policies may be not deterministic, and finally may exist more than one optimal policy, even more in our example there are uncountables many optimal policies.

Key words: constrained, markov decision processes, discounted cost.

[prev.] [volumen] [cont.]
[foro] [red-mat] [galeria] [otros] [Visualización]