基于梯度自适应规则的自适应UKF算法及其应用  

An Adaptive UKF Algorithm Based on Gradient Adaptive Rule and Its Application

作　　者：郝顺义[1] 刘华伟[1] 黄国荣[1] 夏奇[1] 

机构地区：[1]空军工程大学航空航天工程学院,西安710038

出　　处：《计算机测量与控制》2014年第4期1205-1208,共4页Computer Measurement &Control

基　　金：航空科学基金项目(20100818017)

摘　　要：针对量测噪声方差统计值未知的非线性UKF(Unscented Kalman Filter)滤波问题,提出了一种基于梯度自适应规则的自适应UKF算法;在标准的非线性UKF算法基础上,根据残差方差阵的估计值与真实值之差构造代价指标函数,并将该函数相对于参数变化的负梯度方向作为参数更新的方向,构建自适应调节机制;将算法应用于GPS/DR(Dead-Reckoning)组合导航系统中,仿真结果显示状态估计误差具有良好的收敛性,估计精度较噪声观测器有明显改善,表明算法对量测噪声方差阵的动态变化具有较强的适应性。To consider the problem of non--linear UKF filtering under the circumstance of unknown covariance statistic of the measure- ment noise, an adaptive UKF algorithm was presented based on the gradient adaptive rule. On the foundation of standard UKF algorithm, an index function of cost was constructed according to the difference between estimated values o~ the innovation covariance and its real value, the negative gradient direction of parameter variance was taken as the parameter updating direction corresponding to the function, based on which an adaptive adjustment mechanism was founded. The algorithm was applied to the GPS/DR integrated navigation system, good convergence effects of estimating errors of the system states were achieved, the estimating precision was improved obviously compared to the noise observ- er method. Simulation results show that the algorithm is more adaptable to the dynamic changes of the measurement noise covariance.

关 键 词：组合导航 UKF算法 梯度自适应规则 代价函数 残差 

分 类 号：V249.32[航空宇航科学与技术—飞行器设计]