In this paper, the authors study a set of stochastic dynamic problems from inventory and queuing theory where the transition probability distributions of the underlying Markov chains may be uncertain. To address the uncertainty on the probability distributions in such problems, they formulate a robust stochastic dynamic program with a maximin approach. This approach denes a game between the controller, i.e., the system manager, and Nature. Specically, they mainly focus on investigating the structural properties of optimal policies in a class of robust and semi-robust Markov decision processes (MDPs). They use event-based dynamic programming (EBDP) to formulate the robust MDPs and investigate the structure of their optimal policies. In this context, the authors describe a specic nominal EBDP framework, which provides the basis of their analysis, and introduce a robust counterpart of the nominal framework. They study the structural properties of nominal MDPs and extend these to robust MDPs within the EBDP framework. They also study structural properties of some queuing and inventory control problems. Leonardo Pasini
Recensione dell'articolo:(Turgay, Zeynep; Karaesmen, Fikri; Ormeci, Egemen Lerzan - "Structural properties of a class of robust inventory and queueing control problems." Naval Res. Logist. 65 (2018), no. 8, 699{716.) MR3905847 MathSciNet ISSN 2167-5163
Leonardo Pasini
2020-01-01
Abstract
In this paper, the authors study a set of stochastic dynamic problems from inventory and queuing theory where the transition probability distributions of the underlying Markov chains may be uncertain. To address the uncertainty on the probability distributions in such problems, they formulate a robust stochastic dynamic program with a maximin approach. This approach denes a game between the controller, i.e., the system manager, and Nature. Specically, they mainly focus on investigating the structural properties of optimal policies in a class of robust and semi-robust Markov decision processes (MDPs). They use event-based dynamic programming (EBDP) to formulate the robust MDPs and investigate the structure of their optimal policies. In this context, the authors describe a specic nominal EBDP framework, which provides the basis of their analysis, and introduce a robust counterpart of the nominal framework. They study the structural properties of nominal MDPs and extend these to robust MDPs within the EBDP framework. They also study structural properties of some queuing and inventory control problems. Leonardo PasiniI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
