@inproceedings{915d6a1bd97b4fbeb1db9f1a3befdea4,
title = "Online learning for patrolling robots against active adversarial attackers",
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.",
keywords = "Mixed strategy, Stackelberg Game, UCB1 Bandits",
author = "Mahmuda Rahman and Oh, {Jae C.}",
note = "Publisher Copyright: {\textcopyright} 2018, Springer International Publishing AG, part of Springer Nature.; 31st International Conference on Industrial, Engineering and Other Applications of Applied Intelligent Systems IEA/AIE 2018 ; Conference date: 25-06-2018 Through 28-06-2018",
year = "2018",
doi = "10.1007/978-3-319-92058-0_46",
language = "English (US)",
isbn = "9783319920573",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "477--488",
editor = "{Ait Mohamed}, Otmane and Malek Mouhoub and Samira Sadaoui and Moonis Ali",
booktitle = "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",
}