An emulator-based rapid source localization approach in informative sensor planning

摘要

An emulator based Bayesian framework is proposed for sensor planning and interpretation of sensor measurements to infer relevant information for decision making. The framework uses Gaussian process based emulator to model dispersion of pollutant inside a building. Emulator is derived using integrated CFD (computational fluid dynamics) and multizone model CONTAM. On detection of contaminants by sensor system, Bayesian inference is used to infer location and characteristics of the source. Efficacy of the methodology is demonstrated for source localization in a five room building. The framework is investigated for different sensor networks and importance of sensor planning is demonstrated.