Gaussian filtering traditionally suffers from two major drawbacks: 1) Gaussian approximation of the intrinsic non-Gaussian measurement noises and 2) ignoring delay in measurements. This article designs an advanced Gaussian filtering algorithm for addressing the two drawbacks and improving the accuracy. The proposed method is abbreviated as GFMCFD, indicating Gaussian filtering under the maximum correntropy (MC) criterion for fractionally delayed measurements. The MC-criterion-based design enables the proposed GFMCFD to handle the non-Gaussian noises. Moreover, to deal with the delay, the proposed GFMCFD stochastically identifies the delay and uses the current measurement to estimate the desired state at a past instant, depending on the preidentified delay. Thereafter, it later updates the estimated state till the current time instant using state dynamics to perform real-time estimation. Interestingly, the proposed GFMCFD considers the delay as a fractional multiple of the sampling interval. The improved accuracy of the proposed GFMCFD is validated for two nonlinear filtering problems.
Fractionally Delayed Bayesian Approximation Filtering Under Non-Gaussian Noisy Environment
IEEE Transactions on Aerospace and Electronic Systems ; 59 , 5 ; 5823-5834
2023-10-01
1680332 byte
Aufsatz (Zeitschrift)
Elektronische Ressource
Englisch
Driver-Gaze Zone Estimation Using Bayesian Filtering and Gaussian Processes
Online Contents | 2016
|GP-BayesFilters: Bayesian filtering using Gaussian process prediction and observation models
British Library Online Contents | 2009
|