Journals Proceedings

Abstract

Particle Filtering (PF) technique is functional approach for tracking application, navigation problem and in non-linear non-Gaussian system estimation problem. Technical diagnosis and prognosis is another promising area where PF is getting into foreground especially for Remaining Useful Life (RUL) estimation and potential fault prediction. The main reason is that PF could handle different types of noises and can handle measurement uncertainties, which are typically connected to problem of real degradation estimation. Even PF has different kind of filtering method and re-sampling algorithms there is still only limited number of real comparisons available in the state-of the art literature, especially from the point of root mean square error and computational demandingness. Main aim of this paper is to present different type of PF implementation and compare those on simple system simulation. At the same time an object oriented MATLAB Toolbox for Particle Filtering developed by Scientific Systems Company Inc. and University of North Carolina is introduced. This toolbox covers the gap in limited number of real, configurable and robust PF implementation available to broader pool of scientists and engineers.