Optimal Multi-Sensor Fusion for BDA using Bayesian Networks

The application of HPEM weapons to disable enemies electronic capabilities is hampered by the lack of practical tools for electronic BDA. To provide real-time, high-resolution BDA in complex, noisy environments requires the use of multiple sensing modalities interpreting diverse signals related to different types of target electronic components. To aggregate and interpret the data from these varied sensors, we propose the development of novel statistical tools for modeling the interactions among and between target system components and BDA sensors.The proposed framework is built on a Bayesian network that encodes the statistical dependencies between the states of target components and the results produced by the various sensing modules. Bayesian inference methods allow us to infer a probability distribution over the state of each target component and capability post-HPEM. This approach can incorporate all available intelligence on the target system composition, environmental factors, and the BDA sensors themselves, optimally incorporating this information and propagating the associated uncertainty forward to provide robust and actionable BDA. This basic model will be augmented by leveraging queuing theory to infer probabilistic recovery times for each target component and capability, and by inverting the inference problem to infer the expected utility of any specified sensing module.