Factored Particle Filtering for Data Fusion and Situation Assessments in Urban Environments

We present and demonstrate a particle filtering approach to data fusion and situation assessment for military operations in urban environments. Our approach views such an environment as a physical system whose state vector is composed of a large number of both discrete and continuous variables representing properties of tracked entities. Inferencing on such vector-based models exploits both causal dependencies among variables in the state vector via its dynamic Bayesian belief network representation and vector decomposition into weakly interacting subcomponents. To effectively leverage the decomposition, instead of straightforward particle filtering based inferencing, the proposed algorithm maintains factored particles over clusters of state variables, thus resulting in smaller variance. The algorithm samples discrete modes and approximates the continuous variables by a multi-normal distribution updated at each time step by an unscented Kalman filter. The approach is demonstrated using a Marine Corps operational scenario involving a potential ambush on city streets.