ABSTRACT
This chapter discusses a general class of distributed estimation problems, including the state and measurement models, the centralized solution for both linear and nonlinear problems, and the distributed estimation architecture. It provides optimal fusion for nonlinear systems with process noise. The chapter presents optimal fusion for linear systems with process noise using augmented state. It also discusses alternative fusion approaches for linear systems with process noise, including maximum likelihood fusion with cross-covariance, minimum variance (MV) estimate, distributed Kalman filter (DKF), and accumulated state density (ASD) fusion. The chapter compares the performance of augmented state tracklet fusion with standard tracklet fusion, Bar-Shalom–Campo, and minimum variance fusion. The first type of fusion algorithms address dependence that can be characterized by cross-sensor covariances of the local estimation errors. The chapter presents the maximum likelihood estimate. It also discusses the MV estimate, which is equivalent to the best linear unbiased estimate.
