Interactive multi-model estimation method based on covariance intersection

Abstract

Belonging to the field of maneuvering target tracking, the invention puts forward an interactive multi-model estimation method based on covariance intersection. The realization process of the method comprises: (1) model condition estimating and matching matrix interacting: used for calculating the initialization input of each model condition estimator; (2) model condition estimation filtering: used for updating the condition estimation of each model; (3) model matching matrix updating: using the covariance intersection method to calculate the matching matrix of each model; and (4) model condition estimation fusing: taking the matching matrix of each model as a weighted value, and calculating total estimation output by fusing the model condition estimation. The method of the invention overcomes the problem that a model probability does not match a model probability density function product existing in traditional interactive multi-model estimation, has the advantages of small calculation complexity and a good estimation effect, and is applicable to maneuvering target tracking for civil and military use.

Description

technical field [0001] The invention belongs to the field of maneuvering target tracking, and in particular relates to a method for realizing multi-model self-adaptive estimation by using covariance intersection. Background technique [0002] Since the Kalman filter technology was applied to the field of target tracking in the 1970s, the target tracking method based on state space description has attracted the attention of many research institutions at home and abroad, and has achieved fruitful theoretical results. All kinds of traffic control systems in civilian use have been widely used. With the development of electronic information technology and the application of new materials, the form of target movement tends to be more diverse and complex, which is also called the increasing mobility of the target. For example, the F-22 aircraft can still roll at a rate of 30° / s at an angle of attack of 60°, thereby rapidly changing the direction of the nose. After decades of deve...

AI Technical Summary

Problems solved by technology

However, in the traditional multi-model estimation method, due to the need to calculate the joint distribution of the continuous target state and the discrete switching variable, the product of the probability mass and the probability density function is inevitably introduced, and the different magnitudes of the two values ​​lead to different products. matching problem such that the population estimate obtained is only an approximation of the probability mass of the model

In recent years, many scholars have proposed various forms of improved methods from different perspectives, such as fuzzy reasoning, wavelet analysis, neural network and minimum variance fusion, etc., but these methods greatly increase the computational complexity while improving the estimation performance.

Images