Occupation measures in average cost Markov decision processes
In the Borel Markov control model with average cost criterion the author shows existence of near-optimal, robust stationary (non-randomized) policies by occupation measure methods. The author is able to dispense with usual assumptions of continuity and universal measurability by restricting attention to the class of models where it suffices to consider the family of policies such that, for some state-action \((x^0,a^0)\), positive integer \(d\) and positive number \(\eta\), the probability of reaching the pair \((x^0,a^0)\) in any \(d\)-period time span (i.e. from \(nd\) to \((n+1)d\), \(n=0,1,2,\dots\)) is no less than \(\eta\). An assertion concerning the non-emptiness of this class of models is not included.
- scientific article; zbMATH DE number 825217
- Average cost optimal policies for Markov control processes with Borel state space and unbounded costs
- Markov Decision Processes with a Borel Measurable Cost Function—The Average Case
- The average cost optimality equation for Markov control processes on Borel spaces
- Weak conditions for average optimality in Markov control processes
- Adaptive Markov control processes
- General Irreducible Markov Chains and Non-Negative Operators
- scientific article; zbMATH DE number 49674 (Why is no real title available?)
- scientific article; zbMATH DE number 3338194 (Why is no real title available?)
- Markov Decision Processes with a Borel Measurable Cost Function—The Average Case
- Non-Existence of Everywhere Proper Conditional Distributions
- Probability Measures on Compact Sets
- Stochastic optimal control. The discrete time case
This page was built for publication: Occupation measures in average cost Markov decision processes
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4943243)