Multiscale Unscented Kalman Filter Estimation Method Based on Auv Tangential Velocity Model

Abstract

The underwater submersible vehicle field of the present invention discloses a multi-scale unscented Kalman filter estimation method based on an AUV tangential velocity model to obtain a state sequence X(N) and an observation value sequence Z(N); the state sequence X(N ) and the observation sequence Z(N) are decomposed to the scale i; obtain the state transition equations of the tangential velocity model of the two motion states under the discrete system; obtain the matrix transformation relationship between the state equation and the measurement equation between different scales, including System process noise, system measurement equation and system measurement noise; obtain the state one-step prediction equation at scale i, the update equation of the state vector, the update equation of the covariance and the gain K of MUKF k ; Noise reduction is performed by thresholding the detail information of the observation sequence; wavelet reconstruction is performed on the filtered approximate information and the noise-reduced observation detail information to obtain the optimal estimate on the original scale. The invention improves the accuracy of state prediction, reduces noise interference, and improves filtering estimation precision.

Description

technical field [0001] The invention belongs to the field of underwater vehicles, in particular to a multi-scale unscented Kalman filter estimation method based on an AUV tangential velocity model. Background technique [0002] The traditional nonlinear filtering method is mainly the extended Kalman filtering algorithm, but this algorithm has shortcomings such as low precision, poor stability, and slow response to target maneuvering. In recent years, a nonlinear filtering algorithm has been proposed in the literature--- -Unscented Kalman filter. It is an algorithm based on the combination of Unscented change and Kalman filter. This algorithm mainly uses the idea of ​​Kalman filtering, but when solving the predicted value and measured value of the target subsequent time, it needs to use sampling points to calculate. UKF approximates n-dimensional target sampling points by designing weighted points δ, calculates the propagation of these δ points through nonlinear functions, ...

AI Technical Summary

Problems solved by technology

The second-order filtering method considers the quadratic term of Taylor series expansion, thus reducing the estimation error caused by linearization, but greatly increasing the amount of calculation, so in practice, it is not widely used in the first-order EKF

However, due to the variety of underwater environments, the adaptability of this method is poor.

Images