Dual Extended Kalman Filter Methods

This chapter presents a unified probabilistic and algorithmic framework for nonlinear dual estimation methods. Also included is a brief review of the EKF for both state and weight estimation, and an introduction of some of the complications in coupling the two. An example in noisy time-series prediction is provided. The development of a general probabilistic framework for dual estimation is detailed which allows the relating of various methods that have been presented in literature, and also provides a general algorithmic approach leading to a number of different dual EKF algorithms. Results on additional example data sets are also presented.