【作者】 唐智慧；

【导师】 张殿业；

【作者基本信息】 西南交通大学 ， 交通工程， 2012， 博士

【摘要】 近年来,随着我国汽车保有量的不断增长,道路上新驾驶员越来越多,与新驾驶员息息相关的交通事故也与日俱增。众所周知,驾驶员是道路交通系统的主导,大多数的交通事故均直接或间接地由驾驶员所引发,所以加强新驾驶员的能力训练,尤其是新驾驶员的危险感知能力训练,使其在危险事件推演过程中能尽早感知危险的存在,避免交通冲突,降低事故的发生,这将是改善目前交通安全现状的一个比较重要的因素。然而,如何提高新驾驶员的危险感知能力,需从新老驾驶员在危险感知能力上的差异性及其引发差异的致因入手。论文首先研究了新老驾驶员在危险感知能力上的差异性。为了研究此问题,论文进行了新老驾驶员的对比实验,如下：(1)实验共选择79个非职业性的社会性车辆驾驶员参加,将其分成三组：一组为有经验驾驶员组,要求其驾驶时间超过10年且年均驾驶里程数超过8000km,另两组为新驾驶员,其驾驶时间少于4年。在新驾驶员中随机地分成两组,一组需要接受危险感知训练,而另外一组则无需训练。按此分组后,三组人员分别为：接受训练的新驾驶员组27人,无需训练新驾驶员组29人,无需训练经验丰富驾驶组23人。新驾驶员组在性别、年龄、平均年驾驶里程、总驾驶里程、过去三年的事故涉嫌数、简单反应时、驾驶经验训练等方面均无明显差异。(2)实验分两部分进行：第一部分为驾驶员的危险感知实验,通过测量三组驾驶员在各种设计场景中对危险的平均反应时距、组内反应灵敏度、对事件反应的原因口头阐述及其眼动方式,来统计分析新老驾驶员在危险感知能力上的差异性。第二部分为场景危险分级实验：在该实验中,共选择20个模拟交通场景作为危险分级任务的视频,让被试者根据交通突变所给的危险线索来判断该场景中隐藏的危险大小,并依照专家分级时所使用的7个分级点对其进行分级打分,最后将每个场景所给分值填写在实验报告上。(3)两个实验发现：在危险感知实验中,新驾驶员与经验丰富的驾驶员在危险感知能力方面确实存在差异性。经验丰富老驾驶员组和经过危险感知训练的新驾驶员组驾驶员对潜在危险(隐性事件)的感知均比未经危险感知训练的新驾驶员组驾驶员反应更加快。而在显性危险(显性事件)上,新老驾驶员的危险感知能力无显著性差异。且新老驾驶员对危险的感知随事件的时间推移也存在差异性：在计划性事件发生前,驾驶员对非计划性事件的反应率相对比较低且三组间比较接近,但在计划性事件发生后,经验丰富老驾驶员组和经过危险感知训练的新驾驶员组驾驶员对非计划性事件的反应率却是未经危险感知训练新驾驶员组的2倍以上在场景危险分级实验中,各组驾驶员之间并没有表现出显著性差异,说明他们都感觉到了场景中的危险存在,即：各组驾驶员对危险的敏感性无显著性差异。其次,论文研究了新老驾驶员差异性致因,剖析了新驾驶员对潜在危险反应慢的原因,是因为他们对反应证据的收集积累比经验丰富驾驶员慢?还是新驾驶员更加不愿意将某些情景标识为危险状态,从而使他们感受不到某些场景的危险性,抑或需要等待事态发展到更严重情形,新驾驶员才愿意将此场景标识为危险状态?换言之,论文进一步剖析了影响驾驶员危险感知能力差异性的致因到底是驾驶员对交通场景的信息采集能力差还是他们对危险判定所给的阈值不一样?为此,论文提出将模糊逻辑与信号检测理论相结合,通过模糊信号检测的方法,来分析各组驾驶员在危险感知测试和危险分级任务中的平均敏感度和反应标准值,统计发现：在危险感知测试中,经验丰富驾驶员并未比未经训练的新驾驶员表现出更敏感,t(50)=1.3317,P=0.1890,无显著性差异；经过训练的驾驶员与未经训练的驾驶员相比,也未表现出敏感度的差异,t(54)=1.4745,P=0.1462,无显著性差异；在危险分级任务中,也如此,经验丰富驾驶员并未比未经训练的新驾驶员表现出更敏感,t(50)=1.2192,P=0.2285；经过训练的驾驶员与未经训练的驾驶员相比,也未表现出敏感度的差异,t(54)=0.7316,P=0.4676。但在反应偏好测量中,实验却很好地反映了各组间的反应时距差异性：未经训练的新驾驶员组比经过训练的新驾驶员组明显要保守,t(54)=3.0969,P<0.01；未经训练的新驾驶员组与经验丰富驾驶组相比,也明显要保守,t(50)=3.7231,P<0.001。与此同时,反应偏好与反应时距之间存在明显关联关系,具有宽大偏好反应的驾驶员其反应时距小,反应也更快。以上实验结论说明：反应偏好模型能更好地解释新老驾驶员在危险感知中的差异性。不同反应偏好的驾驶员,由于其界定的危险阈值不一样,在其驾驶过程中对同一场景而言,他们对该场景的危险性采取的响应策略也将不一样。因此,在对新驾驶员的训练中,不能只注重驾驶员危险感知能力的训练,更重要的应该是通过训练使新驾驶员降低其心中的危险阈值判别标准,以便他能将更多的场景纳入危险场景,提前采取操作措施,防范事故的发生更多还原

【Abstract】 In recent years with the growing car ownership, more and more novice drivers come behind the driving wheel, which leads to a serious traffic safety issue:the accidents associated with the novice drivers are increasing. The driver behavior is the dominant factor in the road transport system, directly or indirectly causing the occurrence of accidents. Therefore, the exercise to strengthen the capacity of the novice drivers, especially the hazard perception, is an important factor to improve the traffic safety. In order to research this issue, the paper evaluates the differences of hazard perception between novice drivers and experienced drivers, and the contributing factors for the differences.Firstly, the paper researches on the differences of hazard perception between novice drivers and experienced drivers and proposes to conduct a comparative experiment of experienced and novice drivers as following:(1)participants consisted of79drivers who are further divided into3groups. One was the experienced driver group and in this group the drivers were those who had been driving for10years or more and driven more than8000km per year on average. The others were novice drivers who had been driving for4years or less and were randomly assigned to a trained and an untrained group. No participants had previously taken part in a hazard perception test. This assignment resulted in three independent groups:trained novices (n=27), untrained novices (n=29),and untrained experienced (n=23). The novice groups did not differ significantly in terms of gender, age, kilometers driven per year on average, total number of years of driving, accident involvement over the past3years, pure reaction time advanced driving qualification, and prior hazard perception training.(2)The experiment consisted of two tests.One is the hazard perception test, in which four dependent measures were available for each event, e.g., response time, response sensitivity (driver-group response distribution), the verbal description of the hazard instigator, and the eye movement pattern. Using these four dependent measures, the paper analyzes the difference in hazard perception between the experienced and novice drivers. The other is the hazard rating test which includes20traffic scenes. Participants were instructed to rate the potentials for a traffic conflict to occur in the next few moments after the occlusion on the same7-point rating scale used by the experts. Participants marked their ratings for each scene on a response sheet.(3)The hazard perception experiment revealed that the experienced drivers were significantly different from the novice drivers. The experienced drivers and trained drivers could perceive the potential hazard faster than the untrained drivers, but experienced drivers were not significantly different from any other groups in the obvious hazardous situations. At the same time, the hazard perception of experienced drivers and trained drivers varying with timing were marginally different from the novice drivers’as well. With regard to events that appeared "before-planned events",the response rate was relatively low and similar for all three driver groups, however, experienced and trained drivers responded nearly two times more than untrained drivers to events that appeared "after-planned events".However, in the hazard rating test there were not significant differences among three groups, suggesting that three groups of drivers perceived the hazard in the same magnitude, in other words, the three group drivers’sensitivity were not significantly different.Two comparisons revealed that novice drivers could identify the danger of scenes, which meaned there was no significant difference in sensitivity of hazard perception between the experienced drivers and novice drivers. So what caused the slow response of a novice driver to the less dangerous scene? It was because novices were slower to accumulate this evidence of potential hazard and allowed them to increase the response time. Alternatively, it was due to that novices could interpret the road situation just as accurately and quickly as experienced drivers did, but they were simply less willing to label situations as hazardous. Novices drivers needed to wait for developments to the more serious cases, they are willing to label situations as hazardous? In other words, how the mechanism affected the driver’s risk perception differences is subject to the drivers’information gathering capabilities. Or the threshold is not the same that drivers determine the risk?Secondly, the paper researches on the mechanism of difference between the experienced and novice drivers in term of risk perception, analyzes what caused the slow response of a novice driver to the less dangerous scene? It was because novices were slower to accumulate this evidence of potential hazard and allowed them to increase the response time. Alternatively, it was due to that novices could interpret the road situation just as accurately and quickly as experienced drivers did, but they were simply less willing to label situations as hazardous. Novices drivers needed to wait for developments to the more serious cases, they are willing to label situations as hazardous? In other words, how the mechanism affected the driver’s risk perception differences is subject to the drivers’information gathering capabilities. Or the threshold is not the same that drivers determine the risk? To address these issues, the paper presents to combine the fuzzy logic with signal detection theory, and analyzes the average value of sensitivity and response criteria in the hazard perception test and the hazard rating test through the fuzzy signal detection method. Statistical results found:In the hazard perception test, experienced drivers did not exhibit greater sensitivity than novices, t(50)=1.3317,P=0.1890.Similarly, the trained and untrained drivers did not differ in sensitivity, t(54)=1.4745,P=0.1462.The same outcomes happened in the hazard rating test, experienced drivers did not exhibit greater sensitivity than novices, t(50)=1.2192, P=0.2285.Similarly, the trained and untrained drivers did not differ in sensitivity, t(54)=0.7316,P=0.4676.However, the response bias measures reflected the latency differences between the groups better. The untrained novice group was significantly more conservative than trained novice groups, t(54)=3.0969,P<0.01.Similarly, untrained novice group was significantly more conservative than the experienced group, t(50)=3.7231,P<0.001.At the same time, response bias on the hazard perception test was significantly correlated with latency, such that more liberal responses were associated with faster latencies.Through experimental results described above,the response bias model can better explain the differences in the hazard perception test between the experienced drivers and the novice drivers. Because the drivers who have different response bias have the different response thresholds, They will take the different response strategies in the driving process of the same scene. Therefore, in the training for novice drivers the research cannot focus on the driver’s risk perception only. More importantly, it is recommended to reduce threshold of the novice driver through the training, so that the drivers will consider more scenes as hazard ones, and take operational measures in advance to prevent accidents.更多还原