基于多元拉普拉斯分布的鲁棒非线性平滑器  

Robust Nonlinear Smoother Based on Multivariate Laplace Distribution

作　　者：何嘉诚 柏明明 王刚 彭倍 HE Jiacheng;BAI Mingming;WANG Gang;PENG Bei(School of Mechanical and Electrical Engineering,University of Electronic Science and Technology of China,Chengdu 611731;College of Control Science and Engineering,Zhejiang University,Hangzhou 310027;School of Information and Communication Engineering,University of Electronic Science and Technology of China,Chengdu 611731)

机构地区：[1]电子科技大学机械与电气工程学院,成都611731 [2]浙江大学控制科学与工程学院,杭州310027 [3]电子科技大学信息与通信工程学院,成都611731

出　　处：《导航与控制》2024年第4期100-107,共8页Navigation and Control

基　　金：国家自然科学基金(编号:62388101,62403422,51975107);中国博士后科学基金(编号:2023TQ0283,2023M743006,2024M750354);浙江省自然科学基金(编号:LQ24F030005);四川省重大科技专项(编号:2022ZDZX0039,2019ZDZX0020);四川省科技计划资助(编号:2022YFG0343)。

摘　　要：野值干扰具有不可预估性,将对导航与控制系统的平稳运行产生严重威胁。针对厚尾分布的测量噪声会诱导传统基于高斯假设的平滑器性能下降问题,提出了一种基于多元拉普拉斯分布的非线性平滑器。方法采用拉普拉斯分布建模含有随机野值干扰的量测噪声,并基于变分推断技术对多元拉普拉斯分布的模型参数进行在线辨识,从而提高平滑器在厚尾噪声下的鲁棒性和适应性。仿真结果表明,所提出的平滑器能够有效提升在厚尾噪声环境中对目标跟踪的精度,优于传统高斯假设下的平滑器。Outliers pose a significant threat to the stable operation of navigation and control systems with their unpredictability.To address the performance degradation of the traditional Gaussian-based smoothers caused by heavy-tailed measurement noise,a nonlinear smoother based on multivariate Laplace distribution is proposed.Laplace distribution is used to model the measurement noise containing random outlier disturbances,and the model parameters of the multivariate Laplace distribution are identified online based on the variational inference,so as to improve the robustness and adaptability of the smoother under heavy-tailed noise.Simulation results demonstrate that the proposed smoother effectively improves the target-tracking accuracy in heavy-tailed noise environments,which significantly outperforms the traditional Gaussian-based smoothers.

关 键 词：厚尾分布噪声 多元拉普拉斯分布 非线性 平滑器 

分 类 号：U666.1[交通运输工程—船舶及航道工程]