The invention discloses a tire load state recognition method based on LSTM (Long Short Term Memory)-assisted extended Kalman filtering. The tire load state recognition method comprises the following steps: S1, establishing a 7-degree-of-freedom vertical dynamical model of a vehicle; s2, preparing observed quantity and state vector data as a training data set; s3, training the neural network based on the training set data; and S4, obtaining observed quantity through sensors mounted on the vehicle body and the axle bridge. S5, inputting observed quantity data, and predicting the vertical force of the wheel through an extended Kalman filter assisted by a neural network, wherein a Kalman gain coefficient is determined by LSTM (Long Short Term Memory); according to the method, the extended Kalman filtering and the LSTM network are fused, and the Kalman gain coefficient is predicted by using the LSTM network, so that the prediction precision of the extended Kalman filtering is improved while the demand on a precise dynamic model is reduced.
本发明公开了一种基于LSTM辅助扩展卡尔曼滤波的轮胎载荷状态识别方法,包括如下步骤:S1.建立车辆7自由度垂向动力学模型;S2.准备观测量和状态向量数据,作为训练数据集;S3.基于训练集数据对神经网络进行训练;S4.通过车体和轴桥安装的传感器获取观测量。S5.输入观测量数据通过神经网络辅助的扩展卡尔曼滤波器预测车轮垂向力,其中卡尔曼增益系数由LSTM确定;该方法融合扩展卡尔曼滤波和LSTM网络,使用LSTM网络对卡尔曼增益系数进行预测,从而在降低对精确动力学模型的需求的同时,提高扩展卡尔曼滤波的预测精度。
Tire load identification method based on LSTM-assisted extended Kalman filtering
一种LSTM辅助扩展卡尔曼滤波的轮胎载荷识别方法
2025-05-02
Patent
Electronic Resource
Chinese
AIAA | 2015
|AIAA | 2005
|AIAA | 2009
|