Application and Evaluation of Cognitive Search Algorithms for Pollutant Source Localization in Atmospheric Boundary Layer Flows: A Large Eddy Simulation-based Study

In the event of an accidental release of invisible harmful substances into the atmosphere, efficient tracing of the substance plume and quick identification of the source location are crucial for emergency response and hazard mitigation. Cognitive search algorithms are a category of search strategies that utilize information-theoretic rewards to optimally navigate a mobile sensor to the source location using sparse sensing cues along the way and sequential decision making under uncertainty. In this study, we demonstrate the applications of these cognitive search algorithms for pollutant source localization in atmospheric boundary layer turbulence using numerical simulations. In particular, a large-eddy simulation (LES) model is used to simulate the instantaneous flow velocity and pollutant concentration fields, using which the real-time progress of the source localization process of a mobile sensor (e.g., mounted on a drone) controlled by the cognitive search algorithms is simulated. This LES-based modeling framework allows for a comprehensive evaluation of various cognitive search algorithms under realistic environmental conditions.