Controlled Markov Chains with AVaR Criteria for Unbounded Costs
Probability
2015-11-18 v4 Optimization and Control
Abstract
In this paper, we consider the control problem with the Average-Value-at-Risk (AVaR) criteria of the possibly unbounded -costs in infinite horizon on a Markov Decision Process (MDP). With a suitable state aggregation and by choosing a priori a global variable heuristically, we show that there exist optimal policies for the infinite horizon problem. To our knowledge, this is the first work of deriving dynamic programming equations with -unbounded costs via AVaR-operator.
Keywords
Cite
@article{arxiv.1501.02518,
title = {Controlled Markov Chains with AVaR Criteria for Unbounded Costs},
author = {Kerem Ugurlu},
journal= {arXiv preprint arXiv:1501.02518},
year = {2015}
}