Multiagent Patrol Generalized to Complex Environmental Conditions

• DOI: 10.1609/aaai.v25i1.7988
• Publication Date: 2022-09-13T04:37:48Z
• Authors: Noa Agmon; Daniel Urieli; Peter Stone
• Keywords: Autonomous agent; Point of interest

The problem of multiagent patrol has gained considerable attention during the past decade, with immediate applicability being one its main sources interest. In this paper we concentrate on frequency-based patrol, in which agents' goal is to optimize a frequency criterion, namely, minimizing time between visits set interest points. We consider environments complex environmental conditions that affect cost traveling from point another. For example, marine environments, travel ships depends parameters such as wind, water currents, and waves. demonstrate there need new strategy divides given area into parts more than agent active, for improving frequency. show general graphs intractable, therefore focus simplified (yet realistic) cyclic possible inner edges. Although remains generally intractable graphs, provide heuristic algorithm shown significantly improve point-visit compared other strategies. evaluation our work used custom developed ship simulator realistically models movement constraints engine force drag reaction changes.