Shield Synthesis for Reinforcement Learning

Bettina Könighofer, Roderick Bloem, Nils Jansen, Florian Lukas Lorber

Research output: Chapter in Book/Report/Conference proceedingConference paper

Abstract

Reinforcement learning algorithms discover policies that
maximize reward. However, these policies generally do not adhere to
safety, leaving safety in reinforcement learning (and in artificial intelligence in general) an open research problem. Shield synthesis is a formal
approach to synthesize a correct-by-construction reactive system called a
shield that enforces safety properties of a running system while interfering with its operation as little as possible. A shield attached to a learning
agent guarantees safety during learning and execution phases. In this
paper we summarize three types of shields that are synthesized from
different specification languages, and discuss their applicability to reinforcement learning. First, we discuss deterministic shields that enforce
specifications expressed as linear temporal logic specifications. Second,
we discuss the synthesis of probabilistic shields from specifications in
probabilistic temporal logic. Third, we discuss how to synthesize timed
shields from timed automata specifications. This paper summarizes the
application areas, advantages, disadvantages and synthesis approaches
for the three types of shields and gives an overview of experimental
results.
Original languageEnglish
Title of host publicationLeveraging Applications of Formal Methods, Verification and Validation
Subtitle of host publicationVerification Principles - 9th International Symposium on Leveraging Applications of Formal Methods, ISoLA 2020, Proceedings
EditorsTiziana Margaria, Bernhard Steffen
Pages290-306
Number of pages17
Volume1
DOIs
Publication statusE-pub ahead of print - 29 Oct 2020
Event2020 International Symposium on Leveraging Applications of Formal Methods - Virtuell, Greece
Duration: 26 Oct 202030 Oct 2020

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12476 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference2020 International Symposium on Leveraging Applications of Formal Methods
Abbreviated titleISoLA 2020
CountryGreece
CityVirtuell
Period26/10/2030/10/20

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Shield Synthesis for Reinforcement Learning'. Together they form a unique fingerprint.

Cite this