Title
Optimal Patrol Planning for Green Security Games with Black-Box Attackers
Author(s)
Xu, Haifeng; Ford, Benjamin; Fang, Fei; Dilkina, Bistra; Plumptre, Andrew; Tambe, Milind; Driciru, Margaret; Wanyama, Fred; Rwetsiba, Aggrey; Nsubaga, Mustapha; Mabonga, Joshua
Published
2017
Abstract
Motivated by the problem of protecting endangered animals,there has been a surge of interests in optimizing patrol planning for conservationarea protection. Previous efforts in these domains have mostlyfocused on optimizing patrol routes against a specific boundedly rationalpoacher behavior model that describes poachers’ choices of areas to attack.However, these planning algorithms do not apply to other poachingprediction models, particularly, those complex machine learning modelswhich are recently shown to provide better prediction than traditionalbounded-rationality-based models. Moreover, previous patrol planningalgorithms do not handle the important concern whereby poachers inferthe patrol routes by partially monitoring the rangers’ movements. Inthis paper, we propose OPERA, a general patrol planning frameworkthat: (1) generates optimal implementable patrolling routes against ablack-box attacker which can represent a wide range of poaching predictionmodels; (2) incorporates entropy maximization to ensure that thegenerated routes are more unpredictable and robust to poachers’ partialmonitoring. Our experiments on a real-world dataset from Uganda’sQueen Elizabeth Protected Area (QEPA) show that OPERA results inbetter defender utility, more efficient coverage of the area and more unpredictabilitythan benchmark algorithms and the past routes used byrangers at QEPA.
Access Full Text
A full-text copy of this article may be available. Please email the WCS Library to request.
![]()
Back
PUB21234