Ensemble kalman filter-based particle filtering method

Abstract

The invention discloses an ensemble kalman filter-based particle filtering method, which comprises the following steps of: the initialization of sampled particles, the prediction of a background ensemble at a k moment, the calculation of a kalman gain, the fusion of the latest observation information, the updating of the background ensemble, the recalculation and analysis of the ensemble, the construction of a suggested Gaussian distribution function, the normalization of weights, and the like. The ensemble kalman filter-based particle filtering method provided by the invention avoids the calculation of a Jacobian matrix because a nonlinear system is not required to be linearized, and improves the calculation accuracy by adopting a sampling method to realize approximately linear distribution. In the particle filtering method, a sampling point number is heuristic and can be flexibly set, so the amount of calculation is not increased sharply with the increasing of dimensionality, and the real-time performance is effectively controlled.

Description

technical field [0001] The invention belongs to the technical field of nonlinear filtering, and in particular relates to a particle filtering method based on ensemble Kalman filtering. Background technique [0002] The basic task of nonlinear filtering is to recursively estimate the unobservable state value of the nonlinear system from the observations polluted by noise. Wide range of applications. For the following nonlinear discrete systems: [0003] x k =f(x k-1 , v k-1 ) [0004] z k =h(x k ,w k ) [0005] in is the state of the nonlinear discrete-time system at time k, is the observation vector of the nonlinear discrete system at time k; v k-1 ∈R n is the nonlinear discrete system process noise, w k ∈R m is the observation noise at time k; mapping and Both are bounded nonlinear functions, representing the state and observation model of the nonlinear discrete-time system, respectively. The purpose of filtering is to obtain the posterior distribution...

AI Technical Summary

Problems solved by technology

This invention mainly solves the problem that the current filtering method based on Monte Carlo sampling reduces the performance of nonlinear dynamic system state estimation when the number of samples is small.

Images