Shielding atari games with bounded prescience

M. Giacobbe, Mohammadhosein Hasanbeig, Daniel Kroening, Hjalmar Wijk

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

Abstract

We present the first explicit-state method for analysing and ensuring the safety of DRL agents for Atari games. Our method only requires access to the emulator. We give a suite of 42 properties that characterise "safe behaviour" for 31 games. We evaluate the safety of the best available DRL agents which, as our experiments show, violate most of our properties. We propose a countermeasure that implements shielding using bounded explicit-state exploration. Our method improved their overall safety, producing the safest DRL agents for Atari games currently available.
Original languageEnglish
Title of host publicationAAMAS '21
Subtitle of host publicationProceedings of the 20th International Conference on Autonomous Agents and MultiAgent Systems
PublisherInternational Foundation for Autonomous Agents and Multiagent Systems
Pages1507–1509
Number of pages3
ISBN (Electronic)9781450383073
Publication statusPublished - 3 May 2021
EventAAMAS '21: 20th International Conference on Autonomous Agents and Multiagent Systems - Virtual Event, United Kingdom
Duration: 3 May 20217 May 2021

Publication series

NameAAMAS Conference proceedings
Publisher International Foundation for Autonomous Agents and Multiagent Systems
ISSN (Print)2523-5699

Conference

ConferenceAAMAS '21
Country/TerritoryUnited Kingdom
Period3/05/217/05/21

Keywords

  • Safe AI
  • Deep Reinforcement Learning
  • Atari Games

Fingerprint

Dive into the research topics of 'Shielding atari games with bounded prescience'. Together they form a unique fingerprint.

Cite this