基于双自适应无迹卡尔曼滤波的半挂车状态估计

Semi-trailer State Estimation Based on Double Adaptive Unscented Kalman Filter

针对半挂车辆状态估计过程中测量噪声不确定、累计误差影响严重、初值敏感等问题,提出一种适用于半挂车铰接角、车速等多个状态量估计的双自适应无迹卡尔曼滤波算法(FFUKF).基于搭建的半挂汽车12自由度非线性动力学模型和轮胎模型,通过测量的轮速与车辆加速度等信息,首先利用模糊控制自适应调整滑移率容差,综合判断每个车轮的稳定状态,通过轮速估算出一种车速;与此同时,模糊控制自适应调整测量噪声,利用无迹卡尔曼算法,依据动力学估计出铰接角和另一种车速;然后通过卡尔曼滤波算法融合两种方法估计的结果,实现车辆的纵向、侧向速度、横摆角速度和挂车与牵引车铰接角的实时估计.最后在Simulink/TruckSim联合仿真环境下进行多工况仿真试验,验证所提出的双自适应无迹卡尔曼估计算法(FFUKF)有较强的适应性、稳定性和鲁棒性,相比普通模糊自适应无迹卡尔曼(FUKF)有更高的估计精度,能有效克服累计误差,即便在估计初始值不准和有ABS控制输入的情况,仍可以较精确地对车速和铰接角进行实时估计.

Aiming at the problems of uncertain measurement noise,serious influence of accumulated error and sensitive initial value in the process of semi-trailer state estimation,a double adaptive unscented Kalman filter algorithm(FFUKF)is proposed,which is suitable for estimating several state variables of semi-trailer such as hinge angle and vehicle speed.Based on the established 12-degree-of-freedom nonlinear dynamic model and tire model of semi-trailer car,through the measured wheel speed and vehicle acceleration and other information,firstly,fuzzy control is used to adaptively adjust the slip rate tolerance,comprehensively judge the stable state of each wheel,and estimate a vehicle speed through the wheel speed;At the same time,fuzzy control adaptively adjusts the measurement noise,and estimates the hinge angle and another vehicle speed according to dynamics by using unscented Kalman algorithm;Then,Kalman filtering algorithm is used to fuse the estimation results of the two methods,so as to realize the real-time estimation of the longitudinal and lateral velocity,yaw rate and the articulation angle between trailer and tractor.At last,the multi-condition simulation experiment is carried out in Simulink/TruckSim cosimulation environment,which proves that the proposed double adaptive unscented Kalman estimation algorithm(FFUKF)has strong adaptability,stability and robustness,has higher estimation accuracy than the ordinary fuzzy adaptive unscented Kalman(FUKF),and can effectively overcome the cumulative error.Even if the initial estimation value is inaccurate and ABS control input is available,the vehicle speed and hinge angle can still be accurately estimated in real time.