Factorized covariance intersection for scalable partial state decentralized data fusion

We propose a new conditionally factorized covariance intersection (CI) algorithm for performing partial state decentralized data fusion (DDF). This is relevant for sensor networks where platforms must deal with mixed heterogeneous state estimation problems, e.g. due to coupling between uncertainties in shared subsets of externally monitored process states and private platform states. Our approach enables scalable robust conservative DDF in ad hoc networks through fusion only of external process state estimates that are jointly monitored by neighboring platforms, while still enabling updates to private platform state information via conditional inference. We also propose an additional enhancement to factorized CI which helps to minimize unnecessary local information losses due to conservative data fusion. A networked multi-platform target tracking simulation is provided to demonstrate the proposed approach, which can be extended to other Bayesian DDF applications involving networked fusion of heterogeneous state vectors.