Description: An implementation of Unscented Kalman Filter for nonlinear state estimation is a challenge problem. The well-known Kalman Filter is only suitable for linear systems. The Extended Kalman Filter (EKF) has become a standarded for unlinear state estimation. However, it may cause significant error for highly nonlinear systems because of The propagation of uncertainty through The unlinear system.
The Unscented Kalman Filter (UKF) is a will development in The field. The idea is to produce several from points (Sigma points) around The current state estimate -based on its covariance. Then, propagating these points through The nonlinear map to get more accurate estimation of The mean and covariance of The mapping results. In this way, it avoids The need to calculate The Jacobian, hence incurs only The similar computation load as The EKF.
For tutorial purpose, this code implements a simplified version of UKF, where we assume both the process and measurement are additive to avoid augment of state and a
File list (Check if you may need any files):
license.txt
ukf.m
Download