电子学报 ›› 2019, Vol. 47 ›› Issue (9): 1830-1840.DOI: 10.3969/j.issn.0372-2112.2019.09.003

• 学术论文 • 上一篇    下一篇

量测随机延迟下带厚尾噪声的鲁棒Student's t随机容积卡尔曼滤波器

卢春光1, 张永顺1,2, 李志汇1, 葛启超1   

  1. 1. 空军工程大学防空反导学院, 陕西西安 710051;
    2. 信息感知技术协同创新中心, 陕西西安 710077
  • 收稿日期:2019-03-22 修回日期:2019-06-14 出版日期:2019-09-25
    • 作者简介:
    • 卢春光 男,1994年2月生于河南省太康县,现为空军工程大学防空反导学院研究生.主要研究方向为信号处理、目标跟踪等.E-mail:15891782912@163.com;张永顺 男,1961年生于陕西省,现为空军工程大学防空反导学院教授.主要研究方向为雷达阵列信号处理等.E-mail:yszhang@mail.xidian.edu.cn;李志汇 男,1991生于河南省西华县,现为空军工程大学防空反导学院研究生.主要研究方向为雷达阵列信号处理等.E-mail:lizhihui_16@163.com;葛启超 男,1993年生于安徽省亳州市,现为空军工程大学防空反导学院研究生.主要研究方向为雷达阵列信号处理等.E-mail:geqichao927@163.com
    • 基金资助:
    • 国家自然科学基金 (No.61571459)

Robust Student’s t-Based Stochastic Cubature Kalman Filter with Heavy-Tailed Noises and Randomly Delayed Measurements

LU Chun-guang1, ZHANG Yong-shun1,2, LI Zhi-hui1, GE Qi-chao1   

  1. 1. Air and Missile Defense College, Air Force Engineering University, Xi'an, Shaanxi 710051, China;
    2. Collaborative Innovation Center of Information Sensing and Understanding, Xi'an, Shaanxi 710077, China
  • Received:2019-03-22 Revised:2019-06-14 Online:2019-09-25 Published:2019-09-25
    • Supported by:
    • National Natural Science Foundation of China (No.61571459)

摘要: 针对量测随机延迟下带厚尾过程噪声和量测噪声的非线性状态估计问题,本文通过充分考虑量测一步随机延迟特性及过程噪声和量测噪声的"厚尾"特性,推导了一种新的鲁棒Student's t滤波器框架,并采用随机Student's t-球面相径容积规则近似计算Student's t权值积分,从而设计了一种新的鲁棒Student's t随机容积滤波器.首先,采用一组服从伯努利分布的随机序列来描述系统中可能存在的量测一步随机延迟现象,并采用Student's t分布刻画过程噪声和量测噪声中存在的"厚尾"特性;其次,从理论上证明了当自由度参数趋于无穷以及随机延迟概率为零时,该鲁棒Student's t滤波器就自动地降为标准的非线性高斯近似滤波器;最后,采用随机Student's t-球面相径容积规则给出了一种新的鲁棒Student's t随机容积滤波器,并通过协同转弯模型验证了该滤波器的有效性和优越性.

关键词: 随机延迟, 厚尾噪声, Student's t权值积分, 伯努利分布, 矩匹配方法, 非线性估计

Abstract: Aiming at the nonlinear state estimation problem with heavy-tailed process measurement noise and randomly delayed measurements, this paper deduces a new robust student's t filter framework by taking into one-step random delay characteristics and heavy-tailed characteristics of process noise and measurement noise account fully. Meanwhile, this paper proposes a new robust Student's t-based stochastic cubature filter(RSTCF)by approximately calculating the Student's t weight integral using stochastic Student's t-spherical radial cubature rule. Firstly, this paper uses a set of stochastic sequences obeying the Bernoulli distribution to describe the possible one-step random delay phenomenon in the system,what's more,this paper also uses the student's t distribution to characterize the heavy-tailed characteristics of the process noise and measurement noise. Secondly, it's theoretically proved that the robust Student's t filter automatically will be reduced to a standard nonlinear Gaussian approximation filter when the degrees of freedom in the posterior probability density function of the state and measurement noise are increasing continuously and randomly delay probability is equal to zero. Finally, this paper presentes a new robust Student's t-based stochastic cubature filter by use of stochastic Student's t-spherical radial cubature rule. The effectiveness and superiority of the filter are verified by the coordinated turn maneuvers model.

Key words: randomly delay, heavy-tailed noises, student's t weighted integrals, Bernoulli distribution, moment matching method, nonlinear estimation

中图分类号: