A probabilistic chemical sensor model for data fusion

For a data fusion system, taking data from chemical sensors and making inference about the source parameters of a chemical release in real time, a reversible probabilistic sensor model is proposed. Based on a simple ion mobility sensor with bar display, the proposed model represents a realistic sensor while yielding likelihood calculations fast enough to be used within a large inference engine. The sensor model and implementation strategy is described with initial results discussed in relation to a larger inference engine. The importance of approximating the integrals contained in the sensor models to ensure the model produces inference in the short time span required for real time inference is discussed and methods for improving the efficiency of the integral approximation proposed.