Abstract: To effectively solve the problem of driver’s uncertainty of surrounding perceived during his lane change, the fuzzy cluster analysis was initially applied to analyze driver’s behavior of lane changing. The fuzzy cluster analysis method was proposed to divide the fuzzy cluster after variable input to obtain corresponding Gaussian membership function. Takagi-Sugenoinference method was used in fuzzy reasoning and de-fuzzy treatment for behavior of lane changing.NGSIM data was used to calibrate the parameters of the fuzzy reasoning model established and analyze the reasoning results. The results show that: by the use of fuzzy clustering method to determine the membership function, the true data itself and driver’s psycho-physiological characteristics can be reflected and when compared with lane change decision made in reality , this inference method achieved correction rate up to 81%, which fully verified the feasibility of fuzzy inference in study of discrete issues and this method can be further used in development of automated driving, driver-aiding systems.

Key words: traffic and transportation engineering, lane changing, fuzzy clustering, gaussian membership function, Takagi Sugeno inference method

摘要： 为有效解决车辆换道行人研究中驾驶员对周围环境认知的不确定性，首次提出利用模糊推理系统对驾驶员换道行为进行分析。提出采用模糊聚类分析的方法进行输入变量的模糊集划分，求出对应的高斯隶属函数，首次引入Takagi-Sugeno推理方法进行车辆换道的模糊推理和去模糊化处理。利用NGSIM数据对建立的模糊推理进行参数标定，并进行推理结果分析，结果表明：利用模糊聚类确定隶属度函数的方法,能真实反映数据本身的特征和驾驶员的心理生理特性；而且推理结果与真实换道决策相比较时,其判断正确率高达81%，充分证明模糊推理在研究离散型推断问题中是可行的，而且此方法还可进一步应用到自动驾驶、驾驶员辅助系统的开发中。

关键词: 交通运输工程, 车辆换道, 模糊聚类, 高斯隶属度函数, Takagi-Sugeno推理方法