Sequential testing over multiple stages and performance analysis of data fusion

The JIEDDO Analytic Decision Engine (JADE) is a flexible software toolkit for studying the performance of sensor configurations for the detection of person-borne explosive compounds and other threat substances. JADE is designed to enable performance and tradeoff analyses between different, user-specified scenarios with given sensor placements and data fusion networks. JADE contains fundamental physics-based models of several sensor technologies of interest, such as nonlinear acoustic and radar-based detectors, along with a data fusion system that we focus on in this paper. The fusion system consists of a static component that combines the decisions of individual sensors at a fixed point in time, and a dynamic, time-dependent component that in turn fuses the outputs of the static structure at different times. The static component is based on a probabilistic graphical model, or Bayesian network, and accepts probability matrices from the physicsbased sensor models as inputs (the details of which are abstracted from the fusion system). Its outputs are fed into the dynamic fusion framework, which is based on sequential hypothesis testing and produces performance metrics for the entire, fused sensor configuration. The purpose of the system is to determine the performance of a given fusion structure, as opposed to doing fusion on actual measurements.