CAJ | 학술논문

汽车动力学%非追踪粒子滤波算法%汽车%多状态量%状态估计%定常统计特性噪声%Pacejka 轮胎模型%非追踪卡尔曼滤波算法%最小均方误差%粒子滤波算法汽車動力學%非追蹤粒子濾波算法%汽車%多狀態量%狀態估計%定常統計特性譟聲%Pacejka 輪胎模型%非追蹤卡爾曼濾波算法%最小均方誤差%粒子濾波算法
기차동역학%비추종입자려파산법%기차%다상태량%상태고계%정상통계특성조성%Pacejka 륜태모형%비추종잡이만려파산법%최소균방오차%입자려파산법
vehicle dynamics%unscented particle filter algorithm%vehicles%multi-states%state estimation%constant noise%Pacejka tire model%unscented Kalman filter algorithm%minimum mean-square error%particle filter algorithm

针对常用汽车状态估计算法计算复杂、精度低等问题，提出一种新的汽车多状态量估计方法。建立了包含定常统计特性噪声和 Pacejka 轮胎模型的七自由度非线性汽车动力学模型。针对一般粒子滤波(PF)算法存在的缺陷，使用非追踪卡尔曼滤波(UKF)算法产生粒子滤波的重要性概率密度。基于非追踪粒子滤波(UPF)算法实现对汽车多个关键状态量的最小均方误差估计。将基于 UPF 算法、UKF 算法与 PF 算法的估计器进行了比较，揭示了粒子数对汽车状态估计效果的影响。基于 ADAMS / Car 的虚拟实验和实车实验表明基于 UPF 算法的估计器精度高于基于 UKF 算法的估计器，估计值相对于实际值的平均绝对误差均控制在状态幅值的10%以内，且实时性优于基于 PF 算法的估计器。
침대상용기차상태고계산법계산복잡、정도저등문제，제출일충신적기차다상태량고계방법。건립료포함정상통계특성조성화 Pacejka 륜태모형적칠자유도비선성기차동역학모형。침대일반입자려파(PF)산법존재적결함，사용비추종잡이만려파(UKF)산법산생입자려파적중요성개솔밀도。기우비추종입자려파(UPF)산법실현대기차다개관건상태량적최소균방오차고계。장기우 UPF 산법、UKF 산법여 PF 산법적고계기진행료비교，게시료입자수대기차상태고계효과적영향。기우 ADAMS / Car 적허의실험화실차실험표명기우 UPF 산법적고계기정도고우기우 UKF 산법적고계기，고계치상대우실제치적평균절대오차균공제재상태폭치적10%이내，차실시성우우기우 PF 산법적고계기。
Aiming at the problems of complicated calculation and low precision for common vehicle state estimation algorithms,a novel vehicle multi-state estimation algorithm is proposed here. A 7 degrees of freedom ( 7-DOF) non-linear vehicle dynamic model containing constant noise and Pacejka tire model is established. Aiming at the defects of general particle filter(PF) algorithms,the unscented Kalman filter(UKF) algorithm is used to generate the importance density. The unscented particle filter(UPF) algorithm is used to realize the minimum mean-square error(MMSE) estimation of multiple key vehicle states. Estimators based on the UPF algorithm, UKF algorithm and PF algorithm are compared,and the results indicate the influences of numbers of particles on estimation accuracy. The results of a virtual experiment based on ADAMS / Car and a real vehicle experiment indicate that the accuracy of the estimator based on the UPF algorithm is higher than that of the estimator based on the UKF algorithm. The mean absolute errors of the estimate values of the estimator based on the UPF algorithm relative to the real values are lower than 10 percent of the modality amplitude. The real-time performance of the estimator based on the UPF algorithm is better than that of the estimator based on the PF algorithm.