鲁棒化的变分贝叶斯自适应卡尔曼滤波算法  

Robust variational Bayesian adaptive Kalman filtering algorithm

作　　者：徐定杰[1] 沈忱[1] 沈锋[1] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《华中科技大学学报（自然科学版）》2013年第11期128-132,共5页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金资助项目(61102107)

摘　　要：提出了一种鲁棒化的基于变分贝叶斯的自适应卡尔曼滤波算法.该算法采用具有重尾特性的学生分布取代高斯分布来描述量测模型,减弱系统对于野值的敏感性;再利用变分贝叶斯方法对修正后的模型的时变参数进行逼近推断,在递推地估计状态的同时还能对变化的噪声方差进行跟踪,并更新引入的自由度参数,从而在自适应滤波的同时增强了鲁棒性.仿真实验证明了在野值存在且噪声变化的观测下该算法的自适应与鲁棒性.A robust adaptive Kalman ~iitering （RAKF） algorithm based on variational Bayes was pres- ented. This algorithm models the measurement model with a Studentrs distribution which could char- acterize the heavy-tailed phenomenon to replace Gaussian distribution and hence the sensitivity to out- liers decreases. Variational Bayes was also used in this algorithm to approximate time-variant parame- ters of the modified model. In this manner, the states were estimated recursively together with the time-variant noise covariance, and the introduced degree of freedom was updated. Therefore, the adaptive filtering was carried out with strong robustness. Simulation results demonstrate that the adaptability and robustness of the proposed filter are corrupted with outliers when the time-variant noisy is measured.

关 键 词：自适应滤波 卡尔曼滤波 鲁棒性 变分贝叶斯 野值 

分 类 号：TN965.8[电子电信—信号与信息处理]