Adaptive Unscented Kalman Filter and Its Applications in Nonlinear Control

Abstract

In this Chapter, two adaptive Unscented Kalman Filters (AUKFs), named MIT rule based AUKF and master-slave AUKF, are introduced respectively with the purpose of handling time-varying or uncertain noise distribution. According to the simulation results conducted on omni-directional mobile robot and model helicopter, we can conclude the followings: 1. With incorrect a priori statistic information, the AUKFs perform much better than the normal UKF does. 2. Although achieving a little higher estimation precision than the MS-AUKF does, the MITAUK suffers more complicated calculations. That means, the MS-AUKF is appropriate for the application where computation resources are critical. The MSAUKF, on the other hand, would meet the requirement of high precision estimation.