基于交互式多模型和容积卡尔曼滤波的汽车状态估计  

Vehicle State Estimation Based on Interactive Multiple Model and Cubature Kalman Filter

作　　者：张家旭 李静[1] 

机构地区：[1]吉林大学,汽车仿真与控制国家重点实验室,长春130022 [2]中国第一汽车集团技术中心,长春130011

出　　处：《汽车工程》2017年第9期977-983,共7页Automotive Engineering

基　　金：国家自然科学基金(51275206)资助

摘　　要：基于UniTire轮胎模型建立了包含时变噪声统计特性的汽车动力学7自由度整车模型。针对系统状态噪声和观测噪声统计特性未知的问题,提出了一种基于交互式多模型和容积卡尔曼滤波(IMM-CKF)车辆状态估计算法。该算法采用包含不同系统状态噪声和观测噪声统计特性的汽车动力学模型作为交互式多模型算法的模型集,用容积卡尔曼滤波器对每个子模型的车辆状态进行估计,并使融合输出结果始终保持跟踪估计误差小的子模型输出。最后利用实车场地环境下多种驾驶工况的测试数据对IMM-CKF算法进行离线验证,并与容积卡尔曼滤波器的估计结果进行对比,结果表明其估计性能优于容积卡尔曼滤波器。A 7-DOF vehicle dynamics model with time-varying noise statistical characteristics is established based on Uni Tire model. For the unknown system statistical characteristics of state noise and observation noise,a vehicle state estimation algorithm based on interactive multiple model( IMM) and cubature Kalman filter( CKF) is proposed. The algorithm adopts the vehicle dynamics model with different system statistical characteristics of state noise and observation noise as model set of IMM algorithm and uses CKF to estimate the vehicle state of each submodel to make fusion output results constantly track the sub-model output with small estimation error. Finally,the measured data of several driving conditions under real vehicle test environment to conduct off-line verification on IMM-CKF algorithm with the results compared with that using CKF estimation. The outcomes show that the estimation performance of IMM-CKF algorithm is superior to that of CKF.

关 键 词：汽车动力学 容积卡尔曼滤波 交互式多模型 汽车状态估计 

分 类 号：U461.1[机械工程—车辆工程]