基于长短期记忆神经网络方法的车辆跟驰模型

Vehicle following model based on long short-term memory neural network

为模拟驾驶员的跟驰驾驶行为,并考虑驾驶员不确定性和记忆效应,基于实车跟驰实验数据,提出并训练了一种基于长短期记忆(LSTM)神经网络方法的车辆跟驰模型。基于该模型研究驾驶员的记忆效应影响时长并进行交通仿真。结果表明:与同体积隐藏层神经元的前馈神经网络比较,LSTM神经网络的跟驰模型预测结果更加贴近观测值且更加平滑,接近驾驶员的实际驾驶行为;驾驶员行为受当前环境及其前1.0~3.5 s内的记忆影响;该模型能够消散交通流中的扰动,模型具有较好的抗干扰能力和稳定性。

A vehicle following model based on Long Short Term Memory(LSTM)neural network was proposed in order to simulate car-following behavior considering drivers’ uncertainty and memory effect.The model was trained using the actual car following data. Based on the model,drivers’ memory effect was studied and a traffic simulation was carried out. The results show that compared with the feedforward neural network model with the same number of hidden layer neurons,the prediction results of the LSTM neural network model are more accurate,smoother and closer to drivers’ actual driving behavior. The car-following behavior is affected by the current environment and the memory within previous 1.0~3.5 s. The model can dissipate disturbances in traffic flow,and the model has good anti-interference ability and stability.