CBRN data fusion using puff-based model and bar-reading sensor data

This article provides a suboptimal approach to the measurement update of the state vector and the associated state error covariance in the data assimilation process of airborne material dispersion systems, in which the state vector consists of Gaussian puffs and the sensor measurements of the local material concentrations are bar readings. Based on the Bayes rule and numerical quadrature techniques, this approach approximates an interval in the concentration space associated with a sensor's bar reading by a set of discrete points and the integrals over the interval by sums of function evaluations at these points. An alternative approximation involving the Gaussian error function and the Hermite-Gaussian quadrature is also presented. The puff state is updated using a two-step procedure. First, the continuous-valued concentration forecast is updated with the bar-reading data. Second, the puff state is updated based on the correlation of it with the updated concentration estimate.