Two-stage distributed Kalman filtering state estimation method based on multiplicative latent variables

Abstract

The invention discloses a two-stage distributed Kalman filtering method based on multiplicative latent variables. According to the method, the method comprises the steps of: firstly, defining a nonlinear multiplier in a model as a latent variable, so that the system model is rewritten into a pseudo-linear form based on combination of an original state variable and the corresponding latent variable; secondly, regarding the latent variable as a new system parameter variable, and establishing a dynamic correlation model between the latent variable and the original state variable; thirdly, writing a distributed observation model of to-be-estimated state variables and latent variables; and finally, designing a two-stage Kalman filter, and estimating a latent variable and a state variable in sequence. According to the method, a nonlinear multiplier in a model is defined as a latent variable, and then the states of the latent variable and a state variable are sequentially estimated through a two-stage Kalman filter; and therefore, truncation errors caused by Taylor expansion of traditional extended Kalman filtering are avoided, and the filtering effect is effectively improved.

Description

technical field [0001] The invention is applied to the field of state estimation, and in particular relates to a two-stage distributed Kalman filter state estimation method containing multiplicative latent variables. Background technique [0002] Estimation problems are increasingly encountered in areas such as automatic control, aerospace, communications, navigation, and industrial production. The so-called "estimation" is to extract information from observation data. For example, in order to facilitate the explanation of problems when doing experiments, the experimental results are often represented by curves, and some parameters in the equation of the curve need to be estimated according to the observation data. This process is called parameter estimation, and these estimated parameters are Random Variables. In aircraft navigation, it is necessary to estimate the position, velocity, acceleration and other motion state variables of the aircraft from observation data with...

AI Technical Summary

Problems solved by technology

However, for systems with uncertain system parameters, non-Gaussian, time-delay and nonlinear systems, the traditional Kalman filtering method will fail

Images