基于卡尔曼滤波的车辆侧偏角软测量

Soft sensor of vehicle side slip angle based on kalman filter

针对车辆稳定性控制过程中难以在线测得车辆侧偏角的问题,该文基于参数软测量理论和离散信号滤波理论,并结合卡尔曼滤波和驾驶员—车辆闭环系统模型建立车辆侧偏角软测量模型。该方法通过易测变量横摆角速度和侧向加速度估算车辆侧偏角,以实现车辆侧偏角的状态估计。基于预瞄最优曲率控制理论和预测—跟随理论所建立的二自由度驾驶员—车辆闭环系统建立了软测量模型,并建立其状态方程和观测方程。为进行状态估计,对模型的状态方程和观测方程进行连续系统离散化,得到以横摆角速度和侧向加速度为观测量的系统离散观测方程。通过双移线试验与蛇形试验进行场地试验获取纵向速度、侧向速度、横摆角速度、侧向加速度及车辆侧偏角等试验数据。估计值和试验值比较显示,两者变化趋势一致,误差均值在试验值幅值的10%以内,试验表明,软测量算法能准确估算出车辆侧偏角是可行的。研究结果可为软测量技术在车辆稳定性控制系统上的应用提供参考。

Vehicle electronic control systems effectively improve vehicle stability performance by controlling the key parameters of the moving vehicle. The premise and necessary conditions to improve stability performance is to accurately measure of the stated vehicle parameters, which can reflect the vehicle changes in the control process. Yaw rate and vehicle sideslip angle are important parameters to the vehicle control system; the former can be directly measured through the corresponding sensors, while the latter is difficult or expensive to directly measure. Usually, vehicle sideslip angle is estimated by other vehicle parameters, such as vehicle longitudinal velocity and lateral velocity, etc. One is too direct integral method to calculate vehicle sideslip angle, but the noise signal is also involved in the integration process, and can cause greater errors. Another is to use GPS/INS and gyroscopes to measure, but the cost is too high to promote. In addition, some soft sensors based on state estimation have good applications in the field. To solve the problem that vehicle sideslip angles in the vehicle control process is too difficult to measure on-line, the soft sensor model is established with a Kalman filter and driver-vehicle closed-loop system, based on parameters of soft sensor theory and discrete signal filtering theory. This measurement can realize the on-line inference and estimation of sideslip angles by using yaw rate and lateral acceleration, which are easy to measure. Thus, the soft sensor of sideslip angles is achieved through state estimation. Based on the preview optimal control theory of curvature and the preview follow theory, the soft sensor model was developed by a two degree of freedom driver-automobile closed-loop system model. Its’ state equation and observation equation have been derived and discrete to draw the system discrete observation equation on the observation quantity of yaw rate and lateral acceleration and to do the simulation. To compare estimated value with test value, the researchers conducted road tests. One test was the double lane change （standard： ISO 3888-1999-1）, the other was a slalom （standard： GB/T 6323.1-1994）; these tests obtained the data of longitudinal velocity, lateral velocity, yaw rate, lateral acceleration, and vehicle sideslip angle. The values of yaw rate and lateral acceleration are collected real-time by a DMS-SGP01 gyroscope, and the values with noise are the input measurements of the state estimation by a kalman filter. Other values are collected by non-contact velocity meter systems that consist of LC-761, LC-1100, LC-2100, and LC-5200. The LC-1100, which is a longitudinal velocity sensor, and the LC-2100, which is a lateral velocity sensor, are spatial filter types used to collect the values of longitudinal velocity and lateral velocity. The values of sideslip angle are measured by using LC-5200 forward distance pulse in the system. The results of the values’ comparison show that there are the same changing trends and error within 10% of the test value amplitude. It is shown that the soft sensor can achieve the precise estimation results of the vehicle sideslip angle and is feasible. The realization of state estimation based on soft sensor technology provides a theoretical basis on soft sensors applied to vehicle stability control systems.