摘要
针对卡尔曼滤波(KF)中噪声的统计特性与实际不符时滤波精度严重降低甚至引起滤波器发散的问题,提出一种基于支持向量机的自适应卡尔曼滤波算法(SVMAKF).根据新息理论方差与实际方差的比值,应用支持向量机产生自适应因子对卡尔曼滤波器的噪声方差阵进行在线修正,使噪声方差阵能够根据实际噪声的变化得到调整.通过对雷达目标跟踪系统的仿真表明,该算法对噪声有较强的自适应性,能够提高滤波精度和滤波器的鲁棒性.
As the accuracy will decrease or even divergence problems will occur while the theoretical statistical behavior of the Kalman filtering and its actual behavior do not agree, a new self-adaptive Kalman filtering, support vector machines adaptive Kalman filtering （SVMAKF）, is presented. In order to tune the noise covariance of the Kalman filtering on line, SVM is employed to generate the adaptive factor, according to the ratio of the theoretical covariance of the innovation sequence to its actual covariance. Simulation on target tracking shows that SVMAKF can increase the estimation accuracy and the robustness of the Kalman filtering remarkably, compared with the traditional Kalman filtering.
出处
《控制与决策》
EI
CSCD
北大核心
2008年第8期949-952,共4页
Control and Decision
基金
"十一五"国防预研项目(51309060401)
关键词
自适应卡尔曼滤波
新息序列
支持向量机
目标跟踪
Self-adaptive Kalman filtering
Innovation sequence
Support vector machine
Target tracking
