机器学习—动力学耦合车辆跟驰模型

摘要

目前,跟驰模型的建立主要基于动力学方法和机器学习算法,将两者耦合起来建立跟驰模型的研究还没有.以线性组合预测为基础,对最优加权法中的目标函数进行改进,将经典的Gipps模型和基于BP神经网络的跟驰模型(BP Car-following Model,BP)耦合起来,建立线性组合车辆跟驰模型(Linear Combination Car-following Model,LC-CF).结果表明:BP模型的预测结果更加贴近真实值,Gipps模型的预测结果更加贴近安全值;LC-CF模型可以通过调整参数,来控制BP模型和Gipps模型在LC-CF模型中的权重,进而达到控制预测速度的真实性和安全性的目的.%So far, the car-following model is mostly built by dynamic and machine learning algorithms, there remains no researches on the building of car-following model by coupling the two methods. On the basis of the linear combination forecast, this paper improves the objective function of the optimal weighting method and coupled the Gipps model and BP model based on back propagation neural network to establish the linear combination car-following model (LC-CF). The results show that the speed forecasted by BP model peforms better in accuracy and the speed forecasted by the Gipps model peforms better in security. LC-CF model can achieve the purpose of controlling the accuracy and security of the speed forecasting by adjusting the parameters and controlling the weight of BP model and Gipps model in the LC-CF model.