Robust federated Kalman filtering method, device and system

Abstract

The invention discloses a robust federated Kalman filtering method, device and system, the system comprises an inertial navigation system and N other navigation systems arranged on a carrier, the inertial navigation system is corresponding to a main filter, and the N other navigation systems are in one-to-one correspondence with N local filters; wherein the N local filters are used for receiving input data of respective corresponding sensors and calculating an estimated value of a system state quantity; and the main filter is used for fusing the estimated values of the state quantities obtained by the N local filters to obtain global optimal estimation of the system state and a measurement noise variance matrix. According to the method, when the measurement information is increased or decreased, the federated filter can complete the switching of the fusion mode only by adjusting the number of local filters, and the method has a flexible data fusion structure, and has obvious advantages of robust capability, stability and robustness.

Description

technical field [0001] The present invention relates to the technical field of integrated navigation, in particular to a robust federal Kalman filtering method, device and system. Background technique [0002] With the development of navigation and positioning technology, there are more and more types of navigation systems, such as inertial navigation system, satellite navigation system, magnetic compass, odometer / Doppler speedometer / airspeed meter, barometric altimeter / radar altimeter, Landmark point / map matching etc. Each of these navigation systems has its own characteristics, with advantages and disadvantages. In many missions that have strict requirements on navigation performance, whether it is high precision or high reliability, a single navigation system may not be able to meet the requirements. Comprehensive processing of all measurement information (including detection, combination, correlation and estimation) results in more accurate and reliable navigation resu...

AI Technical Summary

Problems solved by technology

The conventional Kalman filtering algorithm is actually a kind of linear minimum variance estimation, which can only be applied to a specific mathematical model, and the system noise must also be Gaussian white noise whose statistical characteristics are known to have a mean value of zero, while When the mathematical model is uncertain or there are observed outliers, the reliability of the filter estimation will be disturbed

Because a ship sails in a complex ocean environment, it will be subject to various disturbances, such as the influence of weather and sea conditions. It may be very different from the actual speed of the hull. These values ​​​​are called observation outliers, which will have a great impact on the filtering accuracy, and may even cause the filtering to diverge.

Images