Batch CI-Based Kalman Smoother for PM2.5 Source Localization

This paper studies the source localization problem for particulate matter with aerodynamic diameter 2.5 μm (PM
2.5
). The PM
2.5
field is influenced by various meteorological factors and involved with wide geographic areas. However, only noisy concentration measurements are available from a limited number of sensors. Hence, a batch Covariance Intersection (CI)-based Kalman smoother is proposed to recover the PM
2.5
field, such that the position of maximum concentration, i.e., the source position, can be localized. The PM
2.5
field is first transformed to a linear large-scale system and partitioned into multiple subsystems with possibly overlapped state variables. Then, we design the local smoother for each low-dimensional subsystem with the batch CI algorithm, which fuses estimates of the overlapped state variables. Thus, each smoother is only responsible for the field over a small area, and computational cost is significantly reduced. Finally, simulations are included to validate the effectiveness of the proposed smoother.