强跟踪求积分卡尔曼滤波算法  

Strong tracking quadrature Kalman filter

作　　者：马丽丽[1] 贺姗[1] 陈金广[1,2] 

机构地区：[1]西安工程大学计算机科学学院,陕西西安710048 [2]西安电子科技大学电子工程学院,陕西西安710071

出　　处：《计算机工程与设计》2014年第5期1802-1806,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61201118);中国博士后科学基金项目(2013M532020);陕西省教育厅科研计划基金项目(12JK0529)

摘　　要：在非线性系统滤波问题中,可能出现真实系统和滤波模型不匹配的现象,而标准形式的求积分卡尔曼滤波器对于这种具有模型不确定性系统的鲁棒性较差、滤波精度降低的问题。针对该问题,结合强跟踪滤波器的思想,提出了强跟踪求积分卡尔曼滤波算法。通过引入衰减因子对当前时刻的状态预测协方差矩阵进行修正,使得不同时刻的残差序列保持正交,减弱先前滤波结果对当前滤波过程的影响,增强量测值的作用,减弱模型的作用,克服模型的不确定性对滤波结果的影响。仿真结果表明,在具有模型不确定性情况的非线性滤波问题中,该算法与标准形式的求积分卡尔曼滤波算法相比,能够获得更高的滤波精度。In the nonlinear system filtering, maybe there is a phenomenon that the true system and the filtering model do not match very well. At this time, the robustness of the standard Quadrature Kalman filter is poor and the filtering precision is lower because of the uncertainty model. Based on the strong tracking filter idea, a strong tracking Quadrature,Kalman filtering algorithm is presented to handle this problem. In the process of filtering, a forgetting factor is introduced to modify the state predicted covariance matrix on the current time step so that the residual sequence at any time step can keep orthogonal. This also can weaken the effect on the current filtering process from the previous filtering results. Thereby, the role of the measurement value is enhanced, and the role of the uncertainty model is weakened. As a result, the uncertainty of model influencing on filtering result is overcome in some degree. The simulation shows that the new algorithm can obtain higher filtering accuracy than the standard Quadrature Kalman filter in the nonlinear filtering problem with uncertainty model.

关 键 词：非线性系统 强跟踪滤波 求积分卡尔曼滤波 状态估计 卡尔曼滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]