基于模糊推理的驾驶员车速决策行为建模及其标定

Driver's Speed Decision-making Behavior Modeling Based on Fuzzy Inference and Its Calibration

为模拟驾驶员在行驶过程中处理不确定信息的能力,并解决以往微观交通仿真中自由行驶与跟驰状态相互转换时的不平滑问题,对将2种行为综合起来的车速决策行为建立模糊推理(fuzzy inference system,FIS)模型。提出了一种隶属度函数参数化设计方法来解决FIS模型的标定问题。以车头时距、相对速度、期望速度差作为输入,加速度作为输出,结合专家经验构建模糊规则,建立了FIS模型。采用美国公路署NGSIM(next generation simulation)项目免费提供的车辆行驶数据,在对数据筛选、预处理的基础上,通过对驾驶员反应时间标定、期望车速设定,构造了模型的输入输出数据,并对模型参数利用遗传算法进行了标定。标定后的FIS模型对实际行驶数据有良好的拟合精度;通过想定实验,证明在模型较好地解决了状态转换的不平滑问题;最后对模型在跟驰状态下表现与GM模型进行了比较,结果表明标定后的FIS模型略优些,说明本文建立的参数化车速决策FIS模型具有良好的实用价值和应用前景。

To simulate the driver’s ability to deal with uncertainty and solve the unsmooth problem in the driving state transformation between free-traveling and car-following during microscopic traffic simulation, the fuzzy inference system （FIS ）was introduced to model the driver’s speed decision-making behavior which integrated the free-traveling and car-following behavior.A new way to calibrate the FIS model by pa-rameterizing the membership function was proposed.The inputs of the FIS mode are difference between ve-locity and desired speed,headway and relative velocity.The output is acceleration.The NGSIM （next generation simulation）data was used to calibrate and evaluate the model.With the analysis and pretreat-ment of NGSIMdata,drivers’reaction time was calibrated,and the desired speed of different vehicle type was approximated as the corresponding free speed.Using the driving data,the FIS model was calibrated and the model output was validated and compared with the real data.The results show that the calibrated＆amp;nbsp;FIS model simulates the drivers’behavior appropriately.A scenario experiment is designed to prove that the model can solve the unsmooth problem in the driving state transformation.In addition,the output of the calibrated FIS model under car-following state was compared with that of GM model.The comparison shows that the calibrated FIS performs are better in practical value and application prospect.