一种抑制卡尔曼滤波发散的实时数据处理方法

A real-time data processing method for controlling Kalman filter instability

由于水声环境的复杂性和水声信道的时空变特性及水下航行载体的机动性,水声定位系统测量的弹道样点野值较多,平滑性差。介绍了一种野值的自动剔除和卡尔曼滤波递推处理方法,克服了滤波发散。文中选取距离D的倒数作为状态变量,使得1/D是近似线性变化的,此时量测方程的误差也近似是线性的,卡尔曼滤波器的表现是稳定的,并且是渐近无偏的。卡尔曼滤波的递推形式,滤波增益矩阵Kk的离线计算出,Qk和Rk值选取固定植,野值设定门限自动剔除,使滤波器收敛和稳定时间短,实现了对快速目标的跟踪和滤波输出,没有出现发散现象。该方法的特点是实时性好,对快速目标具有良好的跟踪能力,而且能达到工程上应用的精度要求。

Under complex ocean environment, the acoustic channel is a random channel in space-time domains, Because of this fact and the mobility of underwater carriers, many measuring samples that deviate from the true trajectory can be obtained in underwater positioning system, and lead to the worse smoothness of measurement. A real-time data processing and recursive algorithm, which gets rid of error trajectory samples automatically for Kalman filter, is proposed to solve the problem of filtering instability. The reciprocal of distance D is selected as a state variable, and 1/D is considered changing linearly. So the error of measuring equation is linear, the Kalman filter can continue operating stably with unbiased estimation. Using a recursive Kalman filtering method, calculating filtering gain matrix KK beforehand, fixing the value of Qk and Rk, and getting rid of error trajectory samples automatically by setting threshold, as a consequence, can make the filter operation convergent and stable in a short time with good performances in aspects of real-time processing and accurately tracking ability to high speed targets.