Online learning for patrolling robots against active adversarial attackers

Mahmuda Rahman, Jae C Oh

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

Abstract

We study the online route planning problem for patrolling robots, to assign them to optimal routes to patrol in a large crime-prone area. To model the actively engaging, intelligent, and adversarial opponents, we use the Stackelberg Security Game between the patrolling robots and the attackers. We leverage a graph-based bandit algorithm [16] with adaptive adjustment of the reward for the robots in this game to perplex the best response attackers and gradually succeed over them. Our graph bandits can outperform other stochastic bandit algorithms [10] when a simulated annealing-based scheduling mechanism is incorporated to adjust the balance between exploration and exploitation. Hence our method can successfully assign a small group of patrolling robots to cover a large number of routes.

Original languageEnglish (US)
Title of host publicationRecent Trends and Future Technology in Applied Intelligence - 31st International Conference on Industrial Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2018, Proceedings
PublisherSpringer Verlag
Pages477-488
Number of pages12
ISBN (Print)9783319920573
DOIs
StatePublished - Jan 1 2018
Event31st International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems IEA/AIE 2018 - Montreal, Canada
Duration: Jun 25 2018Jun 28 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume10868 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other31st International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems IEA/AIE 2018
CountryCanada
CityMontreal
Period6/25/186/28/18

Keywords

  • Mixed strategy
  • Stackelberg Game
  • UCB1 Bandits

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint Dive into the research topics of 'Online learning for patrolling robots against active adversarial attackers'. Together they form a unique fingerprint.

  • Cite this

    Rahman, M., & Oh, J. C. (2018). Online learning for patrolling robots against active adversarial attackers. In Recent Trends and Future Technology in Applied Intelligence - 31st International Conference on Industrial Engineering and Other Applications of Applied Intelligent Systems, IEA/AIE 2018, Proceedings (pp. 477-488). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10868 LNAI). Springer Verlag. https://doi.org/10.1007/978-3-319-92058-0_46