基于FSDT的驾驶员危险感知能力研究

Exploration of the drivers' risk-perception based on the fuzzy signal detection theory

驾驶员的危险感知与交通事故之间关系紧密。为了科学评价驾驶员的危险感知能力,采用情景试验的方法,利用UC-win/Road软件仿真驾驶场景,运用模糊信号检测理论(FSDT),结合模糊逻辑和传统信号检测理论(SDT)来衡量驾驶员的危险感知水平,对敏感性指标d和倾向性指标β进行了定量评价。最后,选取5名驾驶员对该方法进行验证,并对5名驾驶员危险感知水平进行对比,从中找出危险感知能力较低者,给予其针对性的培训。

This paper is inclined to make an exploration of the drivers＇ risk-awareness based on the fuzzy signal detection theory. It is said that a driver＇s ability to perceive traffic risks is closely related with the traffic accidents. In order to assess a driver＇s risk awareness, here we would like to adopt the method of scenario experiments, along with a UC-win/Road software, a useful tool to simulate the driving scenarios based on the research situation both at home and abroad. In proceeding with our research, we have first of all analyzed a series of theories about FSDT, including the risk perception, signal detection theory and fuzzy signal detection theory. As is known, fuzzy signal detection theory can be used to test the driver＇s ability to perceive the risks by combining the fuzzy logic and traditional signal detection theory （SDT）. According to FSDT, the risk perception research process involves four steps： the selection of mapping functions for signal and response, the application of implication functions, the computation of fuzzy hitting rates and false alarm rates, computation of fuzzy sensitivity and criterion. And, then, FSDT can be applied to evaluate the driver＇s risk perception. For example, five drivers can be chosen to validate the method, while eight scenarios can be selected from Graphic Driving Dangerous Projections 101 Cases to test their risk perception abilities, all of which belong to the grade of high-risks, though the difference among them is not very obvious. To make the test more persuasive, two following indicators are used for the quantitative evaluation of the aspects of their abilities： their sensitivity index d and response bias index β. The index d is used to measure the sensitivity of the drivers to safe and unsafe conditions, while the index β is used to measure their strategy tendency during the determination process. And, finally, a comparison has been done to test the levels of the five drivers＇ risk perception and awareness. While the lower perception levels were found among them, all of them were willing to take the risks, though the difference of their perception can be found apparent. In the above said examinations, the first driver is closest to the top grade： i.e., to β opt , whereas the other two, that is, No. 3 and No. 5, were found deviating from the optimal （ β〉〉β opt ）, which should be given targeted trainings. Thus, it can be seen that the paper has provided a scientific means in future to measure the drivers＇ ability to perceive risks, which are of great significance to reduce the traffic accidents caused by the driver errors.