首页|量测随机丢失下基于容积卡尔曼滤波的厚尾噪声处理方法

量测随机丢失下基于容积卡尔曼滤波的厚尾噪声处理方法

扫码查看
针对量测随机丢失和厚尾量测噪声条件下的非线性状态估计易发散问题,提出了一种新的非线性卡尔曼滤波方法.引入服从Gamma分布的辅助参数,将厚尾量测噪声建模为Student's t分布,以解决厚尾噪声导致的状态估计易发散问题,并采用服从Benroulli分布的随机变量来描述量测信号随机丢失的现象;在量测随机丢失下,基于目标状态和未知参数建立联合后验分布,并使用变分贝叶斯方法,联合估计系统状态、量测丢失概率和未知的厚尾噪声.非线性目标跟踪仿真实验表明,提出的算法可自适应估计未知的量测丢失概率,在野值概率为5%的条件下,算法目标跟踪的位置、速度和转动速率均方根误差分别为对比算法的37%、28%和60%;在野值概率为10%的条件下,其他算法均出现了发散现象,而提出的算法依然能够以较低的误差跟踪目标,体现了所提算法良好的鲁棒性和优越性.
Heavy-tailed noise processing method based on cubature Kalman filtering under random loss of measurement
A new nonlinear Kalman filtering method is proposed to address the problem of divergence in nonlinear state esti-mation under conditions of random measurement loss and heavy-tailed measurement noise.By introducing an auxiliary pa-rameter that follows a Gamma distribution,the heavy-tailed measurement noise is modeled as a Student's t distribution to solve the problem of state estimation divergence caused by heavy-tailed noise.A random variable that follows a Benroulli distribution is used to describe the phenomenon of random measurement loss.Under conditions of random measurement loss,a joint posterior distribution is established based on the target state and unknown parameters,and a variational Bayes-ian method is used to jointly estimate the system state,measurement loss probability,and unknown heavy-tailed noise.Non-linear target tracking simulation experiments show that the proposed algorithm can adaptively estimate the unknown meas-urement loss probability.Under conditions of a 5%outlier probability,the root mean square error of the position,velocity,and rotation rate of the algorithm target tracking are 37%,28%,and 60%respectively compared to the control algorithm.Under conditions of a 10%outlier probability,other algorithms have diverged,while the proposed algorithm can still track the target with low error,reflecting the good robustness and superiority of the proposed algorithm.

nonlinear state estimationmeasurement random lossheavy-tailed noiseStudent's t distributionvariational Bayes

李帅永、聂嘉炜、郭成春

展开 >

重庆邮电大学 工业物联网与网络化控制教育部重点实验室,重庆 400065

非线性状态估计 量测随机丢失 厚尾噪声 Student'st分布 变分贝叶斯

国家重点研发计划重庆市教委科学技术研究计划重庆市研究生科研创新项目

2017YFB1303700KJZD-M202300605CYS23465

2024

重庆邮电大学学报(自然科学版)
重庆邮电大学

重庆邮电大学学报(自然科学版)

CSTPCD北大核心
影响因子:0.66
ISSN:1673-825X
年,卷(期):2024.36(3)