Behavior Composition for Marine Pollution Source Localization Using a Mobile Sensor Network

Abstract

Marine pollution, which can cause damage to marine ecosystems, cut fishery production, and even harm human health, has aroused worldwide interest in recent years. Marine pollution reduction operations can stagnate in the case that the source of the pollution is unknown or hidden. In this paper, we present a novel method for marine pollution source localization using a network of mobile sensor nodes, such as autonomous underwater vehicles equipped with chemical sensors. Traditional reactive control methods can respond quickly to the shape dynamics of a chemical plume; however, they can hardly achieve intelligent cooperation unlike deliberative methods. In this study, we present a behavior composition method that attempts to combine the advantages of reactive and deliberative methods. An upwind-customized crossover operation based on the genetic algorithm was formulated as one of the elementary behaviors. The upwind sprint and movement away from the centroid of the sensor nodes were also modeled as another two elementary behaviors. Different sensor nodes are capable of different simultaneous elementary behaviors, enabling behavior composition in the mobile sensor network during plume source localization. The proposed method was evaluated using a widely used filamentous plume simulation platform, which has been used to facilitate field experiments in real marine environments. Simulation results indicate that the proposed method achieved high time-efficiency and localization accuracy during plume source localization in marine environments.