Font Size: a A A

The Research And Simulation Of Kalman Filter Algorithm Based On Unscented Transform

Posted on:2006-04-09Degree:MasterType:Thesis
Country:ChinaCandidate:Z XingFull Text:PDF
GTID:2178360182475291Subject:Signal and Information Processing
Abstract/Summary:PDF Full Text Request
Kalman filter is based on the data of output and input to estimate the signal. Thereare three kinds of the Kalman filters. The first is to manipulate the estimated valuebefore estimation time and the value at that time according to relatively precision;thesecond is to consider the dynamic characteristics completely;the third is therealization of the agonic estimation. The kalman filter is mainly divided into threeclasses:i. Optimal Kalman Filter (Linear time-varying system)ii. Sub-optimal Kalman Filter (Linear constant system)iii. Extended Kalman Filter (Nonlinear system)The extended Kalman filter is a widely used estimator for nonlinear systems. Inpractice, it has two drawbacks. Linearization can produce unstable filters and it isdifficult to implement the Jacobian matrices.In this paper, a new Kalman filter based on Unscented Transform substitute forEKF is studied. The principle of UT is analyzed and the algorithm of UKF isdiscussed in detail. In computer simulation, EKF and UKF methods are used toestimate the noisy chaotic time-series, and then the estimation error between twoalgorithms are compared. Finally the results are given to verify theoretical analysisof the UKF. EKF and UKF methods are based on the training algorithm of the NNsuch as Elman network which proves that the astringency, BER performance and soon based on UKF are better.
Keywords/Search Tags:Extended Kalman Filter (EKF), Nonlinear, Unscented Transform (UT), Chaotic time-series, Neural Network (NN), Bit Error Rate (BER)
PDF Full Text Request
Related items