Distributed nonlinear Kalman filtering method based on alpha divergence

Abstract

The invention belongs to the field of signal processing, in particular to a distributed nonlinear Kalman filtering method based on alpha divergence, capable of effectively reducing the influence of noise of a system and external noise, obtaining high-precision estimation of a real value, and solving the filtering problem and the parameter estimation problem in signal processing, especially the target tracking problem. The distributed nonlinear Kalman filtering method comprises the following steps: firstly, obtaining intermediate state estimation by minimizing alpha divergence between an intermediate approximate posteriori distribution function and a real posteriori distribution function of each node; and then, calculating a final state estimation result of each node by minimizing a convexcombination of forward KL divergence between a final posteriori distribution approximation function of each node and an intermediate posteriori distribution approximation function of a neighbor node thereof. Different from the existing distributed particle filtering based on the minimum variance criterion and the distributed extended Kalman filtering based on the minimum mean square error criterion, the steady-state performance of the nonlinear distributed Kalman filtering is effectively improved.

Description

Technical field [0001] The present invention belongs to the field of signal processing, and relates to the target tracking problem in the signal processing field, in particular to the target tracking problem on a distributed wireless sensor network, and is specifically a distributed nonlinear Kalman filtering method based on alpha divergence. Background technique [0002] The Kalman filter algorithm can estimate not only the stationary one-dimensional random process, but also the non-stationary multi-dimensional random process, and the Kalman filter algorithm is recursive, with small storage capacity, fast convergence speed, and real-time processing speed Fast and other advantages, so Kalman filter has more extensive applications in complex systems, such as navigation, target tracking, positioning, etc. In addition, the Kalman filter algorithm is also used to predict dynamic systems, such as the trajectory of stars and the changing trend of commodity exchange prices. [0003] At p...

AI Technical Summary

Problems solved by technology

The extended Kalman filter algorithm makes a linear approximation for the nonlinear dynamic system, and then uses the method of the linear Kalman filter system for processing, which introduces linearization errors and has limitations such as slow convergence speed.

Images