Abstract: In order to accurately grasp the travel mode selection rule in the autonomous driving environment, the panel data Mixed logit model was introduced to analyze the influence mechanism of the variables representing travelers’ individual socio-economic attributes such as age, income, education degree and the variables representing the travel mode characteristics such as travel cost, waiting time and in-transit time on travel mode choice behavior in an autonomous driving environment. Using 1050 valid sample data of 150 respondents in seven different scenarios from the Singapore Autonomous Driving Willingness Survey, a travel choice behavior model under the automatic driving environment was conducted. The results show that the p-value of the chi-square test of the model is less than 0.000, indicating the proposed model has good applicability. The parameter estimation results show that different individuals in the groups have strong heterogeneity in travel cost, waiting time and travel time. Marginal effect analysis shows that the probability of choosing autonomous driving, public transportation, and walking will rise as the age increases, but the probability of choosing online car hailing will go down. When the travel cost, travel time and waiting time are doubled, the probability of choosing to drive automatically will be reduced by nearly 4.0%.

Key words: traffic and transportation engineering, urban transportation, travel choice behaviour, panel Mixed logit, autonomous driving, marginal effect

摘要： 为准确把握自动驾驶环境下出行方式选择规律，引入面板数据Mixed logit模型分析在自动驾驶环境下表征出行者个体社会经济属性的年龄、收入、受教育程度等变量和表征出行方式特性的出行成本、等待时间、在途时间等变量对出行方式选择行为影响作用机理。利用新加坡自动驾驶出行意愿调研获得150位受访者在7个不同情景的1 050个选择意愿有效样本数据，构建自动驾驶环境下出行选择行为模型。结果显示：模型卡方检验p值小于0.000说明模型具有较好的适用性；参数估计结果显示人群中不同个体对出行成本、等待时间、出行时间存在较强异质性；边际效应分析显示随着年龄增大选择自动驾驶、公交和步行出行概率增大选择网约车概率降低，当出行成本、出行时间、等待时间增加一倍将使选择自动驾驶出行概率降低近4.0%。

关键词: 交通运输工程, 城市交通, 出行选择行为, 面板Mixed logit, 自动驾驶, 边际效应