捷联惯导系统初始对准的SVM自适应Kalman滤波算法(英文)

Support Vector Machine Adaptive Kalman Filtering for Initial Alignment of SINS

作为捷联惯导系统初始对准关键技术的Kalman滤波要求事先精确已知系统及量测噪声的统计特性,当在滤波过程中这些特性改变时,滤波器性能将会降低甚至发散,针对这一问题采用了一种支持向量机(SVM)自适应Kalman滤波(SVMAKF)算法,根据协方差匹配技术应用支持向量机来动态调谐量测噪声方差阵R,当量测噪声随时间改变时,SVMAKF可以实时的估计出准确的噪声方差阵,这就降低了系统对量测噪声先验统计特性的依赖性,能够改善kalman滤波器的状态估计效果。基于SVMAKF的捷联惯导系统初始对准计算机仿真结果表明在滤波精度和滤波器鲁棒性上,SVMAKF都有比传统Kalman滤波器好的表现。

The Kalman filtering is the most commonly used algorithm for the initial alignment of strapdown inertial navigation system （SINS）, but in Kalman filtering, the statistical characters of the process and the measurement noise must be prior known exactly, if these characters change, the performance of the filter will degrade or even make the filter unstable. To overcome this problem, a novel adaptive Kalman filtering （SVMAKF）, adjusted by Support Vector Machine, was introduced. The adaptation of the SVMAKF is in the sense of dynamically tuning the measurement noise covariance matrix R by employing the support vector machine （SVM） based on a eovarianee matching technique. If the measurement noise under it operates change or evolves with time, SVMAKF can give a new noise eovariance matrix. This relaxes the a priori measurement noise statistical assumptions and significantly benefits the Kalman filtering states estimate. The computer simulation about SVMAKF was done for the problem of initial alignment of SINS. The results illustrate that the SVMAKF gives better performance, in terms of accuracy and robustness, than traditional Kalman filtering.