基于Elman神经网络的路面附着系数识别

准确、高效地识别路面附着系数为汽车主动安全系统提供了重要输入参数。笔者提出了基于Elman神经网络识别路面附着系数方法，采用Carsim/Simulink联合仿真，获取了某车辆的63个行驶工况，共20个重要动力学响应。构建了Elman神经网络的路面附着系数识别模型，对附着系数为0.2～0.9的路面进行了识别，识别平均绝对百分比误差为4.92%,准确率为91.22%。相对于传统的BP神经网络方法，该方法使路面附着系数的识别平均绝对百分比误差降低了2.24%,准确率提升了9.82%,并且在潮湿沥青路面以及干燥沥青路面进行了实车实验，验证了该方法的有效性、可行性。

Identification of road friction coefficient based on Elman neural network

Accurate and efficient identification of road adhesion coefficient provides important input parameters for active safety system. In this paper, an identification method of road friction coefficient based on Elman neural network was proposed. Through Carsim/Simulink co-simulation, 63 driving conditions and 20 important dynamics responses of a vehicle were obtained. The identification model of road friction coefficient based on Elman neural network was constructed. The road surface with friction coefficient from 0.2 to 0.9 was identified. The average absolute percentage error was 4.92% and the accuracy was 91.22%. Compared with traditional BP neural network method, this method reduced the average absolute percentage error of road friction coefficient by 2.24% and improved the accuracy by 9.82%. Vehicle experiments on wet and dry asphalt pavement verified the effectiveness and feasibility of the proposed method.