Chi-square adaptive factor-based robust filtering method

Abstract

The invention discloses a chi-square adaptive factor-based robust filtering method, which is used for improving the estimation error of a robust filter and the smoothness of a filtering curve under the condition of unpredictable sharp increase of observation noise so as to meet the filtering estimation requirement of higher precision. A standard robust filtering method requires a known observation noise covariance value, so that the method cannot be normally used under the condition of sharp increase of observation noise. According to the improved robust filtering method provided by the invention, the filtering effect of the robust filter can be effectively improved under the condition that the observation noise is sharply increased, and meanwhile, only less extra calculation complexity is consumed. According to the method, the limitation of a traditional robust filtering method is broken through, the statistics method is used for rapidly estimating the actual observation noise under the condition that the observation noise is sharply increased, the filtering precision and the filtering curve smoothness of the robust filter are effectively improved, and the estimation effect of the filter is optimized.

Description

technical field [0001] The invention belongs to the field of filtering estimation, and specifically relates to a robust filtering method based on a chi-square adaptive factor, which is mainly used to effectively improve the filtering accuracy and smoothing of the filtering curve of the robust filter under the condition of unpredictable surges in observation noise to optimize the estimation effect of the filter. Background technique [0002] The invention can be applied to photovoltaic systems. In these application fields, the state equation of the estimated object is often difficult to establish accurately, and its uncertainty generally exists in the motion model of the estimated object, process noise, and observation noise. The widely used Kalman filtering method has the optimal filtering characteristics for Gaussian white noise, but it requires an accurate state equation. When the parameters of the state equation are inaccurate, the filtering effect will be significantly...

AI Technical Summary

Problems solved by technology

However, in application fields such as photoelectric tracking systems, the observation noise may increase sharply due to factors such as atmospheric turbulence and tracking platform vibration, and traditional robust filters cannot effectively deal with the increase in observation noise.

In response to this problem, in 2019, a state observer based on the Bayesian method to identify observation noise was proposed and showed good performance, but although it can be applied to nonlinear systems such as the unscented Kalman algorithm, However, it cannot be applied to a robust state observer, so it cannot solve the difficulties mentioned above "Double-scale Adaptive Kalman Filter Based on Variational Bayesian Estimation Method" (Wu Junfeng, Xu Song

Images