基于驾驶行为个体特征的酒后驾驶状态识别方法研究

发布时间：2018-06-12 07:56

本文选题：驾驶行为 + 个体特征　；参考：《北京工业大学》2014年博士论文

【摘要】：驾驶员是道路交通系统中最重要的组成因素，在车辆运行中占据着主导作用。如果驾驶员自身出现异常，导致驾驶能力下降，势必会对其驾驶安全带来威胁。实际的道路交通事故统计也表明，大多数的交通事故是由驾驶员自身造成的，驾驶员不当的驾驶行为是导致交通事故的重要原因。因此，在交通安全领域研究中，对驾驶员的研究十分重要，特别是研究驾驶行为特征与交通安全的关系具有重要意义。在驾驶员危险驾驶状态中，酒后驾驶是典型的一种，尤其是容易造成恶性交通事故的一种危险驾驶状态。虽然世界各国均通过立法方式进行严厉禁止，但酒后驾驶现象仍然存在，因此需要探索研究新的识别方法，辅助酒后驾驶的检测。而驾驶行为受驾驶员状态的直接影响，同时驾驶行为是驾驶员持续表现出的参数，不接触驾驶员即可进行检测。那么，基于驾驶行为特征的酒后驾驶状态的检测为此问题提供了新的思路。然而，在驾驶行为研究中，驾驶员的个体差异性是降低驾驶行为研究准确性的重要因素，需要首先解决这项问题，以驾驶员个体特征为基础进行酒后驾驶状态的识别，以提高识别准确性。本研究中，，基于驾驶模拟综合实验平台，针对驾驶行为与酒后驾驶状态的关系展开研究。分析了饮酒影响驾驶员的机理，研究了酒后驾驶行为的一般特征，建立了驾驶员个体特征模型，并根据此模型提出了基于驾驶行为个体特征的酒后驾驶状态识别方法。主要内容如下：首先，基于驾驶模拟综合实验平台，设计综合实验进行了数据采集。针对驾驶员的一般研究内容和方法，介绍了面向人因因素研究的综合实验平台。该平台围绕驾驶模拟器，配合主观调查问卷、驾驶适应性特征检测平台以及驾驶员动态生理检测仪，组成了综合实验平台，使其能够实现对驾驶员的个性特征、驾驶适应性、动态驾驶特征等数据的采集。基于此平台，设计实现了驾驶员在正常状态和不同饮酒水平的酒后驾驶状态下的模拟驾驶实验，重点说明了实验中的虚拟场景及实验流程，详细阐述了实验中对于影响因素的控制方法，通过实验获得了支持本研究的基础数据。其次，分析了饮酒影响驾驶员的机理。从酒后驾驶时的主观特征和驾驶适应性特征两个方面进行了探讨。运用基本统计方法对主观特征进行了分布特征分析，说明驾驶员各方面的主观变化特征；运用重复测量方差分析和logistic回归分析对驾驶适应性指标特征进行了研究，说明了酒后驾驶影响驾驶员的基本机理。研究表明，酒后驾驶时驾驶员态度激进，倾向冒险，认知反应能力明显下降，更重要的是驾驶员的感知能力下降，道路环境信息的正确获取能力下降，尤其表现在视觉方面。基于此，从车辆的运行状态和驾驶员操作行为两方面的特征探讨说明了酒后驾驶行为特征指标的选取角度和方法。再次，分析了酒后驾驶状态下驾驶行为特征。主要运用统计分析方法对驾驶员在正常以及不同饮酒水平的酒后驾驶状态的驾驶行为数据进行了分析。首先对车辆运行状态，即速度、横向位置特征进行分析，分析中同时考虑道路条件的影响。通过研究明确了酒后驾驶时车速偏高，且车辆横、纵向运行稳定性下降的特征，并说明了酒后驾驶对驾驶行为的影响与道路条件相关。其次，对驾驶员操作行为，即加速踏板、制动踏板、转向盘、离合等数据进行了研究，证明了驾驶员在酒后驾驶时踩踏踏板更深更快，离合松弛更快、反应较慢，同时转向控制的细微操作减少、规律性变差等特征。通过影响分析研究，提取了受酒后驾驶状态显著影响的多种驾驶行为指标，并基于这些指标，探索了基于驾驶行为的酒后驾驶状态识别方法。第四，验证了驾驶员个体特征，建立了驾驶行为个体特征模型。运用相关性分析方法对驾驶员自身以及不同驾驶员之间的驾驶特征进行了分析，证明了个体特征和个体差异性。针对驾驶行为时序特征，运用隐马尔可夫模型（HMM）进行了建模探索，说明了驾驶行为个体特征来源于驾驶习惯，证明了基于时序特征研究驾驶行为的可行性。综合各项驾驶行为，运用高斯混合隐马尔可夫模型（GHMM）实现了驾驶员在典型道路条件下的个体特征建模，并验证了模型的有效性。最后，提出了基于个体特征模型的酒后驾驶识别方法。通过对正常与酒后状态下的个体特征模型参数的对比，确定了酒后驾驶时个体特征模型参数的变化倾向，提取出了6项识别指标，确立了基于这些指标的酒后驾驶状态基本识别方法，并证明了该方法在驾驶员状态识别方面的应用潜力。同时，对开发的基于驾驶行为的不良驾驶评估软件平台进行了简介，说明了其中酒后驾驶识别预警的基本算法。本次研究主要从驾驶行为特征研究出发，通过对每个驾驶员建立个体特征模型来解决驾驶行为个体差异性问题。通过对酒后驾驶影响机理分析以及对驾驶行为与酒后驾驶状态的关系研究，提出确定了一种新的酒后驾驶识别方法。整个研究结果不仅有利于酒后驾驶问题解决，而且对于驾驶行为相关其他研究具有推动作用。
[Abstract]:The driver is the most important component in the road traffic system and plays a leading role in the running of the vehicle . If the driver ' s own abnormality causes the driving ability to fall , the driver ' s driving safety is threatened . The actual road traffic accident statistics also show that the driver ' s improper driving behavior is the important cause of the traffic accident . Therefore , it is very important to study the driver ' s research in the field of traffic safety , especially to study the relationship between driving behavior characteristics and traffic safety .

In the driver ' s dangerous driving state , drunk driving is a typical one , especially one kind of dangerous driving state which can easily cause malignant traffic accidents . Although all the countries in the world are strictly prohibited by legislation mode , the driving behavior of drunk driving is still existed , so it is necessary to explore new identification methods to assist in the detection of drunk driving . However , the driver ' s individual difference is the important factor to reduce the accuracy of driving behavior research .

In this study , the relationship between the driving behavior and the drunk driving state is studied based on the comprehensive experimental platform of driving simulation . The mechanism of drunk driving behavior is analyzed . The general characteristics of drunk driving behavior are studied , and the driver ' s individual characteristic model is established . According to the model , a method for recognizing drunk driving state based on individual characteristics of driving behavior is proposed .

First , based on the comprehensive experiment platform of driving simulation , the data acquisition is carried out based on the comprehensive experiment platform . In view of the general research content and method of the driver , the comprehensive experimental platform for human factor research is introduced . Based on the platform , the simulation driving experiment of driver ' s personality characteristics , driving adaptability and dynamic driving characteristics is introduced . Based on the platform , the virtual scene and the experimental flow in the experiment are designed and realized , and the control method of influencing factors in the experiment is explained in detail , and the basic data for supporting the study are obtained through experiments .

Secondly , the mechanism of influencing the driver is analyzed . The subjective characteristics and the driving adaptability characteristics of drunk driving are discussed . The subjective characteristics are analyzed by using the basic statistical method , and the subjective change characteristics of the driver are explained .
This paper studies the characteristics of driving adaptability index by means of repeated measurement variance analysis and logistic regression analysis , illustrates the basic mechanism of drunk driving influence driver . The research shows that the driver ' s attitude is aggressive , the tendency to take risk and the ability of cognitive response decrease obviously when drunk driving , and it is more important that the driver ' s perceived ability decrease and the correct acquisition ability of road environment information falls , especially in the visual aspect .

This paper analyzes the driving behavior data of drunk driving under the condition of drunk driving .

Fourth , the characteristics of the driver ' s individual are verified , and the individual characteristic model of the driving behavior is established . The characteristics of the driver and the difference of the individual are analyzed by using the correlation analysis method . The characteristics of the driving behavior are derived from the driving habits , and the feasibility of the driving behavior based on the timing characteristics is proved .

Finally , a method for recognizing drunk driving based on individual characteristic model is put forward . By comparing the parameters of individual character models under normal and post - alcoholic conditions , six identification indexes are determined , six identification indexes are extracted , and the application potential of the method in driver ' s state identification is established . At the same time , a brief introduction is made on the development of the poor driving evaluation software platform based on the driving behavior , and the basic algorithm for recognizing early warning of drunk driving is described .

Based on the study of driving behavior characteristics , this paper proposes a new identification method for drunk driving by establishing an individual characteristic model for each driver . By analyzing the influence mechanism of drunk driving and the relationship between driving behavior and drunk driving state , a new identification method for drunk driving is put forward . The whole research result is not only beneficial to post - drunk driving problem solving , but also has a pushing effect on other research related to driving behavior .
【学位授予单位】：北京工业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：U491.254

【共引文献】