Indoor contaminant source estimation using a multiple model unscented Kalman filter

The contaminant source estimation problem is getting increasing importance due to more and more occurrences of sick building syndrome and attacks from covert chemical warfare agents. To monitor a building contamination condition, a number of sensors are connected through a network, and the sensor measurements are sent to a fusion center to estimate contaminant source information. An estimation algorithm is required such that timely actions can be taken to mitigate the adverse effects. This paper proposes a multiple model unscented Kalman filter (MM-UKF) to estimate the contaminant source location, the source emission rate and the release time. A simulation test is conducted on a computer generated three-story building. The results show that the MM-UKF algorithm can achieve real-time estimation.