基于限定记忆随机加权扩展卡尔曼滤波的车辆状态估计  

Vehicle state estimation based on limited memory random weighted extended Kalman filter

作　　者：胡敬宇 汪䶮 严永俊 耿可可 柏硕 殷国栋[1] Hu Jingyu;Wang Yan;Yan Yongjun;Geng Keke;Bai Shuo;Yin Guodong(School of Mechanical Engineering,Southeast University,Nanjing 211189,China)

机构地区：[1]东南大学机械工程学院,南京211189

出　　处：《东南大学学报(自然科学版)》2022年第2期387-393,共7页Journal of Southeast University:Natural Science Edition

基　　金：国家自然科学基金资助项目(51975118,52025121);江苏省重点研发计划资助项目(BE2019004);江苏省科技成果转化专项资金资助项目(BA2018023,BA2020068)。

摘　　要：为了降低车辆状态估计过程中历史量测数据误差的影响,提出一种限定记忆随机加权扩展卡尔曼滤波(LMRWEKF)算法.首先,建立三自由度非线性车辆动力学模型;其次,将限定记忆滤波与扩展卡尔曼滤波融合,构成限定记忆扩展卡尔曼滤波;然后,依据随机加权理论,设计服从狄利克雷分布的加权系数来进一步提高滤波估计精度;最后,建立Carsim与Matlab/Simulink联合仿真平台,并进行了高附着系数和低附着系数2种不同工况下的仿真实验.结果表明:相比于标准扩展卡尔曼滤波(EKF)算法,高附着路面仿真工况下,基于LMRWEKF算法估计得到的横摆角速度、质心侧偏角和纵向速度的均方根误差分别降低了60.23%、19.63%、91.57%;低附着路面仿真工况下,基于LMRWEKF算法估计得到的横摆角速度、质心侧偏角和纵向速度的均方根误差分别降低了59.38%、13.92%、94.20%.所提出的LMRWEKF算法能有效抑制噪声波动,提高估计精度.To reduce the influences of historical data errors on the vehicle state estimation,a limited memory random weighted extended Kalman filter(LMRWEKF)algorithm was proposed.First,a three-degree-of-freedom nonlinear vehicle dynamics model was established.Secondly,the limited memory filter and the extended Kalman filter(EKF)were fused to form the limited memory EKF.Then,according to the random weighting theory,the weighted coefficient which obeys Dirichlet distribution was designed to further improve the estimation accuracy of the filter.Finally,a CarSim and Matlab/Simulink co-simulation platform was established.Simulation tests were carried out under the high adhesion coefficient and the low adhesion coefficient.The results show that compared with the estimation results of the standard EKF algorithm,under the high adhesion simulation condition,the root mean square error of yaw rate,sideslip angle and longitudinal velocity estimated based on LMRWEKF algorithm is reduced by 60.23%,19.63%and 91.57%,respectively.Under the low adhesion simulation condition,the root mean square error of yaw rate,sideslip angle and longitudinal velocity estimated based on LMRWEKF algorithm is reduced by 59.38%,13.92%and 94.20%,respectively.The LMRWEKF algorithm can effectively suppress the noise fluctuation and improve the estimation accuracy.

关 键 词：扩展卡尔曼滤波 限定记忆滤波 随机加权 车辆状态估计 

分 类 号：U461[机械工程—车辆工程]