Estimation of Vehicle Centroid Side Angle Based on Neural Network

Abstract

Effective measurement and estimation of vehicle state parameters plays a crucial role in vehicle stability control. Among them, the measurement cost of vehicle sideslip angle is higher, and the value obtained by integral calculation in the actual test includes noise, so the accuracy is difficult to be guaranteed. In this paper, a neural network method is proposed to identify the system, and a system model of vehicle sideslip angle identification is established based on the vehicle body state quantity (yaw rate, longitudinal acceleration, lateral acceleration, etc.) which can be easily obtained. According to the experimental conditions, vehicle state parameters were obtained based on CarSim, and Matlab was used for network training. Finally, the network is verified, the applicability of the trained network under different working conditions is discussed, and some assumptions about the optimization of this method are put forward.

摘要

有效的测量和估计车辆状态参数对于车辆的稳定性控制起着至关重要的作用。其中车辆质心侧偏角的测量成本较高, 同时实际测试中通过积分计算得到的数值包含了噪声, 准确率难以得到保证。本文提出利用神经网络的方法对系统进行辨识, 通过车辆比较容易的得到的车身状态量 (横摆角速度, 纵向加速度, 侧向加速度等) 建立了车辆质心侧偏角辨识的系统模型。根据实验工况基于CarSim 得到汽车状态参数, 并利用Matlab 进行网络训练。最后对网络进行了验证, 讨论了不同工况下训练出的网络的适用性, 并对此方法的优化提出了一些设想。