基于变权新息协方差的自适应卡尔曼滤波器

Adaptive Kalman Filtering Based on Variable Weight Innovation Covariance

针对传统卡尔曼滤波器鲁棒性差,无法实时精确跟踪系统突变状态的现实,设计了一款基于变权新息协方差的自适应卡尔曼滤波器。在传统卡尔曼滤波器的基础上,分析了突变状态无法跟踪的缘由;基于滤波发散判据,分析储备系数与均权新息协方差之间的关系,对状态突变程度进行分层;基于Sage-Husa估计原理与加权最小二乘准则,对于不同程度的突变状态,采用实时调整各历元新息协方差权重的策略,优化渐消因子,激活滤波增益,增权量测新息。实例研究表明,自适应卡尔曼滤波器鲁棒性强,能够精确跟踪系统突变状态,其状态收敛速度优于抗差卡尔曼滤波器,稳态精度提升了42.05%。

As for poor robustness of traditional Kalman filtering and bad behavior of accurate tracking breaking state of the system , variable weight innovation covariance was designed to regulate adaptive Kalman filtering. About the algorithm, this paper first analyzed the reason of bad behavior of accurate tracking in the breaking state based on traditional Kalman filtering. By using criterion of filtering divergence, degree of state mutation was layered on the basis of the relationship between reserve coefficient and innovation covariance.Based on Sage-Husa estimation principle and weighted least squares method, according to different degree of state mutation, the technology of dynamically adjusting the weight of innovation covariance in the filter estimation was introduced. Fading factor was optimized.Filtering gain was activated in real-time. The weight of measurement innovation was enhanced. The case study result shows that the adaptive Kalman filter has strong robustness. It can get accurate tracking breaking state of the system, and the convergence rate is superior to the other rate of robust Kalman filtering, and the steady precision can be improved to 42.05 percent.