Robust adaptive multi-model filtering method

Abstract

The invention discloses a robust adaptive multi-model filtering method, which specifically comprises the following steps of: 1, establishing a discrete state space model of an application object by adopting a multi-model filtering structure of three filters, and constructing a measurement noise variance matrix model set; 2, calculating a model mixing probability between filters, a mixing initial state of each filter and a mean square error matrix of each filter; 3, enabling the three filters to respectively carry out robust Kalman filtering at the same time; 4, updating the probability of the filter model by adopting a Bayesian hypothesis testing method; 5, performing probability weighted fusion on each filter estimation value and the corresponding filter model, and outputting joint state estimation and a mean square error matrix thereof; 6, adaptively updating the measurement noise variance matrix model set; and 7, repeating the steps 2-6 until the filtering is finished. According to the invention, the filtering robust capability can be effectively improved, the noise statistical characteristics can be adaptively and rapidly estimated and measured, and the filtering precision is improved.

Description

technical field [0001] The invention relates to a robust and self-adaptive multi-model filtering method, which belongs to multi-sensor information fusion technology and is especially suitable for the field of combined navigation. Background technique [0002] Kalman filtering is a linear, unbiased, and optimal estimation algorithm with the smallest error variance. It is widely used in multi-sensor information fusion. It has the characteristics of simple algorithm and easy engineering implementation, but the accuracy of the algorithm is highly dependent on the accuracy of the system model, and The statistical properties of system noise and measurement noise are required to be known accurately. In practical engineering applications, due to the complex and changeable sensor measurement environment, the statistical characteristics of measurement noise are difficult to obtain accurately and are time-varying. In addition, even for a very precise sensor, its measurement data will ...

AI Technical Summary

Problems solved by technology

Due to the complex and changeable underwater environment, the measurement data of the Doppler velocimeter will inevitably have gross errors. At the same time, the statistical characteristics of the measurement noise are often unknown and will cha

Images