Temporal logic control of general Markov decision processes by approximate policy refinement

Sofie Haesaert, Sadegh Soudjani, Alessandro Abate

Research output: Contribution to journalArticleAcademicpeer-review

Abstract

The formal verification and controller synthesis for Markov decision processes that evolve over uncountable state spaces are computationally hard and thus generally rely on the use of approximations. In this work, we consider the correct-by-design control of general Markov decision processes (gMDPs) with respect to temporal logic properties by leveraging approximate probabilistic relations between the original model and its abstraction. We newly work with a robust satisfaction for the construction and verification of control strategies, which allows for both deviations in the outputs of the gMDPs and in the probabilistic transitions. The computation is done over the reduced or abstracted models, such that when a property is robustly satisfied on the abstract model, it is also satisfied on the original model with respect to a refined control strategy.
LanguageEnglish
Article number1712.07622v1
Number of pages23
JournalarXiv.org, e-Print Archive, Physics
Volume2017
StatePublished - 20 Dec 2017

Fingerprint

Temporal logic
Controllers

Bibliographical note

22 pages, 3 figures, submitted to the 24th International Conference on Tools and Algorithms for the Construction and Analysis of Systems (TACAS), 2018

Keywords

  • cs.SY
  • cs.LO
  • math.OC
  • 93E03, 93E20, 68W25

Cite this

@article{ed2776b9b3d246098d91c5a4e5fd0619,
title = "Temporal logic control of general Markov decision processes by approximate policy refinement",
abstract = "The formal verification and controller synthesis for Markov decision processes that evolve over uncountable state spaces are computationally hard and thus generally rely on the use of approximations. In this work, we consider the correct-by-design control of general Markov decision processes (gMDPs) with respect to temporal logic properties by leveraging approximate probabilistic relations between the original model and its abstraction. We newly work with a robust satisfaction for the construction and verification of control strategies, which allows for both deviations in the outputs of the gMDPs and in the probabilistic transitions. The computation is done over the reduced or abstracted models, such that when a property is robustly satisfied on the abstract model, it is also satisfied on the original model with respect to a refined control strategy.",
keywords = "cs.SY, cs.LO, math.OC, 93E03, 93E20, 68W25",
author = "Sofie Haesaert and Sadegh Soudjani and Alessandro Abate",
note = "22 pages, 3 figures, submitted to the 24th International Conference on Tools and Algorithms for the Construction and Analysis of Systems (TACAS), 2018",
year = "2017",
month = "12",
day = "20",
language = "English",
volume = "2017",
journal = "arXiv.org, e-Print Archive, Physics",

}

Temporal logic control of general Markov decision processes by approximate policy refinement. / Haesaert, Sofie; Soudjani, Sadegh; Abate, Alessandro.

In: arXiv.org, e-Print Archive, Physics, Vol. 2017, 1712.07622v1, 20.12.2017.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Temporal logic control of general Markov decision processes by approximate policy refinement

AU - Haesaert,Sofie

AU - Soudjani,Sadegh

AU - Abate,Alessandro

N1 - 22 pages, 3 figures, submitted to the 24th International Conference on Tools and Algorithms for the Construction and Analysis of Systems (TACAS), 2018

PY - 2017/12/20

Y1 - 2017/12/20

N2 - The formal verification and controller synthesis for Markov decision processes that evolve over uncountable state spaces are computationally hard and thus generally rely on the use of approximations. In this work, we consider the correct-by-design control of general Markov decision processes (gMDPs) with respect to temporal logic properties by leveraging approximate probabilistic relations between the original model and its abstraction. We newly work with a robust satisfaction for the construction and verification of control strategies, which allows for both deviations in the outputs of the gMDPs and in the probabilistic transitions. The computation is done over the reduced or abstracted models, such that when a property is robustly satisfied on the abstract model, it is also satisfied on the original model with respect to a refined control strategy.

AB - The formal verification and controller synthesis for Markov decision processes that evolve over uncountable state spaces are computationally hard and thus generally rely on the use of approximations. In this work, we consider the correct-by-design control of general Markov decision processes (gMDPs) with respect to temporal logic properties by leveraging approximate probabilistic relations between the original model and its abstraction. We newly work with a robust satisfaction for the construction and verification of control strategies, which allows for both deviations in the outputs of the gMDPs and in the probabilistic transitions. The computation is done over the reduced or abstracted models, such that when a property is robustly satisfied on the abstract model, it is also satisfied on the original model with respect to a refined control strategy.

KW - cs.SY

KW - cs.LO

KW - math.OC

KW - 93E03, 93E20, 68W25

M3 - Article

VL - 2017

JO - arXiv.org, e-Print Archive, Physics

T2 - arXiv.org, e-Print Archive, Physics

JF - arXiv.org, e-Print Archive, Physics

M1 - 1712.07622v1

ER -