Finite-Time Optimal Policy Identification for the Stochastic Shortest Path Problem
IEEE Control Systems Letters · 2026 · 10 · 409-414
DOI: 10.1109/lcsys.2026.3697256 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Control Systems Letters, 2026. Verfasst von Wangzhi Zhou und 2 weiteren Autor:innen. Die Arbeit nennt 27 Literaturangaben. Gefördert von: National Science and Technology Major Project of China, National Natural Science Foundation of China, Natural Science Foundation of Jiangsu Province.
Abstract (Originalsprache)
This letter investigates finite-time optimal policy identification for stochastic shortest path (SSP) problems. Unlike classical assumptions such as graph acyclicity or the prior requirement that all policies be proper, the proposed approach employs a Lyapunov-like function to guarantee that the optimal policy can be obtained within a finite number of iterations under both value iteration (VI) and policy iteration (PI). The proposed condition is further shown to be both necessary and sufficient for policy properness. Moreover, within this framework, the expected hitting time to the terminal state is guaranteed to be finite under any stationary policy. Simulation results are provided to validate the theoretical findings.
Förderung
- National Science and Technology Major Project of China
- National Natural Science Foundation of China
- Natural Science Foundation of Jiangsu Province
Zitieren
APA
Zhou, W., Mo, Y., & Dasgupta, S. (2026). Finite-Time Optimal Policy Identification for the Stochastic Shortest Path Problem. IEEE Control Systems Letters, 10, 409-414. https://doi.org/10.1109/lcsys.2026.3697256
Vancouver
Zhou W, Mo Y, Dasgupta S. Finite-Time Optimal Policy Identification for the Stochastic Shortest Path Problem. IEEE Control Systems Letters. 2026;10:409-414. doi:10.1109/lcsys.2026.3697256
BibTeX
@article{Zhou2026,
author = {Zhou, Wangzhi and Mo, Yuanqiu and Dasgupta, Soura},
title = {Finite-Time Optimal Policy Identification for the Stochastic Shortest Path Problem},
journal = {IEEE Control Systems Letters},
year = {2026},
volume = {10},
pages = {409--414},
doi = {10.1109/lcsys.2026.3697256}
}
EUVIMED