奇异值分解五阶容积卡尔曼滤波汽车状态估计  

Singular value decomposition fifth-order cubature Kalman filter for vehicle state estimation

作　　者：吴伟斌[1] 黄靖凯 曾锦彬 李浩欣 WU Weibin;HUANG Jingkai;ZENG Jinbin;LI Haoxin(School of Engineering,South China Agricultural University,Guangzhou 510642,China)

机构地区：[1]华南农业大学工程学院,广州510642

出　　处：《重庆理工大学学报（自然科学）》2024年第3期74-83,共10页Journal of Chongqing University of Technology：Natural Science

基　　金：广东省重点领域研发计划项目(2021B0101220003);广东省(深圳)数智农服产业园建设项目(FNXM012022020-1-03)。

摘　　要：针对三阶滤波对高维汽车非线性模型估计精度有限的问题,以电动汽车为研究对象,提出了一种基于奇异值分解的五阶容积卡尔曼滤波(SVD-FCKF)车辆状态估计器。首先基于Dugoff轮胎模型,构建高维非线性7自由度车辆动力学模型。然后根据三阶球面-径向容积规则将CKF拓展到五阶,使其具有五阶泰勒级数展开精度,同时利用奇异值分解代替传统Cholesky分解,提高估计器的鲁棒性。最后利用Carsim和Matlab/Simulink联合仿真平台对SVD-FCKF进行验证,结果表明:改进的SVD-FCKF估计器能够有效提高电动汽车纵向速度、侧向速度、质心侧偏角和四轮转速的估计精度和稳定性,多工况适应能力强,整体估计效果优于CKF估计器。研究结果为电动汽车主动安全研究提供了理论支撑,具有实际应用价值。To address the limited estimation accuracy of high-dimensional vehicle nonlinear model with third-order filtering,a fifth-order Cubature Kalman Filter vehicle state estimator based on singular value decomposition(SVD-FCKF)is proposed for electric vehicles.Firstly,based on the Dugoff tire model,a high-dimensional nonlinear seven-degree-of-freedom vehicle dynamics model is built.Secondly,CKF is extended to the fifth order according to the third-order sphere-radial volume rule,so that it has the fifth-order Taylor series expansion precision,and the singular value decomposition is employed to replace the traditional Cholesky decomposition to improve the robustness of the estimator.Finally,Carsim and Matlab/Simulink co-simulation platform are used to verify SVD-FCKF.Our results show the improved SVD-FCKF estimator effectively improves the estimation accuracy and stability of longitudinal speed,lateral speed,centroid sideslip angle and four-wheel speed of electric vehicles,and has strong adaptability to multiple working conditions.And the overall estimation is superior to that of CKF estimator.Our research may provide some theoretical support for the study of electric vehicles’active safety.

关 键 词：车辆动力学模型 状态估计 奇异值分解 五阶容积卡尔曼滤波 

分 类 号：U461.1[机械工程—车辆工程]