Abstract: Different vehicles have different level of automation, which makes researchers confused how to do specific research. In order to embracing the new area of intelligent properly, According to the degree of automation, SAE set a standard to divide vehicles into six grades from 0-5.As we all know, during automatic drive of L3, when the vehicle incapable of dealing with difficulty, the AutoPilot system will send alarm signal to the driver, what we call this as Take-Over Request. In case of emergency, the driver can do nothing but pay attention to the traffic, and based on the information he obtained from traffic, should find an appropriate way to deal with the problems, we call this as Take-Over Process(Gold et al., 2015). Zeeb et al. (2015)&Petermeijer et al.(2016) who tend to combine Take-Over Request with time dimensions, and they devote themselves to define the specific TIME POINT of Take-Over Request, and find the phases of timeline correspond to certain psychological activities. Besides, they give a specific advice to improve the Take-Over Request’s proficiency. Based on their researches, the course of Take-Over Request is divided into three phases respectively. phase 1: evoking vigilant stat; Phase 2: Take-Over decision-making and Phase 3 Take-Over performing, which the process of three phases will take 0.8 seconds(Petermeijer et al., 2016)、4.2 seconds(Gold et al., 2013) and 2.4 seconds(鲁光泉等，2018)respectively. However, Other researchers, such as K?rber et al.(2016)&Naujoks et al.（2017）inclined to believe that the process of Take-Over Request is an integrated whole. In this word, in order to revise the Zeeb’s hypothesis and improve the theoretical models, combined many possible influential factors, including demographics, non-driving related tasks, driving fatigue and a sense of trust etc., we develop a theoretical model of three phases of Take-Over Request process. In this study, by means of questionaire and driving stimulator experiments, we investigate the relationship between the length of takeover process and other possible influential factors. The result shows that age difference is independent of the driver’s takeover performance. However, age has relations to driver’s operation style(compared to the younger driver, older driver intervene the AutoPilot system more frequent). There is significant positive relationship between a sense of AutoPilot and execute other non-driving related tasks. naturally, spending more in operating non-driving related tasks have negative relation to the proficiency of takeover process. Besides, compared to traditional driving style, it’s easier for drivers who as a monitor be passive fatigue state in the AutoPilot systems, which have adverse effect on the proficiency of takeover process. As a matter of fact, It’s easy to see that the process of Take-Over Request is totally complicate, including but not limited to the process of evoking vigilant state, making decision and operating. Unfortunately, in the intelligent transportation system field of study, researchers didn’t pay enough attention to this specific process of Take-Over Request. Beside, We should put the theory into practice by optimizing the design of human-computer interaction panel、training and guiding drivers how to use vehicle automation system properly. Enhancing driver’s abilities and their safety consciousness in monitoring can not overemphasized.

Key words: Autopilot, Takeover process, Psychological Model, Trust

摘要： 自动驾驶是当今智能交通的研究热点，目前正处于由L2等级向L3等级过渡的时期。接管过程作为L3等级自动驾驶车辆人机交互的核心概念，引起了大量研究人员的注意。文章在已有研究的基础上将接管过程划分为警觉唤醒、接管决策和接管执行三个阶段，结合人口学、非驾驶相关任务、驾驶疲劳、信任态度等影响因素加以完善，进一步提出接管过程的三阶段模型。并对接管过程影响因素的未来研究方向，心理机制的构建以及应用现状提出建议。

关键词: 自动驾驶, 接管过程, 心里模型, 信任