Epsilon-Optimal Policies for Average-Cost Separable MDPs with Perturbations

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
1. Verfasser: Kantawala, Dhairya
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866918173153951744
author Kantawala, Dhairya
author_facet Kantawala, Dhairya
contents We study a class of infinite-horizon average-cost Markov Decision Processes (MDPs) whose reward and transition structures are nearly separable. For the totally separable baseline (that is, with no perturbation), we derive an explicit stationary decision rule that is exactly average-optimal. We then show that under an epsilon-perturbation of the separable structure, this policy remains epsilon-optimal, meaning that the loss in the average reward is of order O(epsilon).
format Preprint
id arxiv_https___arxiv_org_abs_2510_23335
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Epsilon-Optimal Policies for Average-Cost Separable MDPs with Perturbations
Kantawala, Dhairya
Optimization and Control
We study a class of infinite-horizon average-cost Markov Decision Processes (MDPs) whose reward and transition structures are nearly separable. For the totally separable baseline (that is, with no perturbation), we derive an explicit stationary decision rule that is exactly average-optimal. We then show that under an epsilon-perturbation of the separable structure, this policy remains epsilon-optimal, meaning that the loss in the average reward is of order O(epsilon).
title Epsilon-Optimal Policies for Average-Cost Separable MDPs with Perturbations
topic Optimization and Control
url https://arxiv.org/abs/2510.23335