Abstract
The Extended Kalman Filter (EKF) has become a standard technique used in a number of nonlinear estimation problems. These include estimating the state of a nonlinear dynamic system, estimating parameters for nonlinear system identification (e.g., learning the weights of a neural networks), and dual estimation (e.g., EM) where both states and parameters are estimated simultaneously.
This presentation points out the flaws in using the EKF, and introduces an improvement, the Unscented Kalman Filter (UKF), proposed by Julier and Uhlman in 1996. Julier and Uhlman demonstrated the substantial performance gains of the UKF in the context of state-estimation for nonlinear control. We extend the use of the UKF to a broader class of nonlinear estimation problems, including state, model, and dual estimation algorithms and illustrate the benefits on several examples.