驾驶行为个性化操纵模式建模研究

发布时间：2018-11-03 19:52

【摘要】：在人—车—路三者构成的闭环交通系统中,人的因素与交通事故的发生概率相关性最高。基于此,为协调好人、车、路三者之间关系,以降低交通事故发生概率这一目标发展起来的驾驶安全辅助系统(Driving Safety Assistance Systems,DSAS)已成为智能车辆领域的重点研究内容之一。目前已有的DSAS系统基本上是在基准驾驶人的基础上进行辅助控制设计,导致DSAS系统对所有驾驶人均采取相同的辅助控制策略和控制参数,无法满足驾驶人的个性化需求。如果DSAS系统能够通过采集驾驶人的驾驶行为操纵数据和车辆运行状态信息自动完成对驾驶人驾驶的个体特性的学习并将学习结果用于DSAS系统的辅助控制策略,则可提高DSAS系统的个性化和智能化水平。实现这一目标的关键技术之一就是研究解决驾驶人的个性化驾驶行为操纵模式建模与驾驶行为识别问题。基于以上问题的出现,本文主要的研究内容如下:(1)通过采集驾驶行为操纵数据,对数据进行预处理后,用数据/机理混合的方法对驾驶人的个体特性进行分析,研究驾驶人操纵动作节点的提取方法以便进行驾驶行为操纵模式建模。(2)在对驾驶人操纵动作节点进行精确提取的基础上,基于有向图原理,研究个性化驾驶行为操纵模式建模方法。(3)利用ART2A-E自适应谐振神经网络具有稳定性、可塑性强、实时在线自学习的特点,研究基于ART2A-E算法与操纵模式反向映射的驾驶行为识别方法。研究结果表明:(1)在对驾驶人的个体特性进行分析的基础上,通过设置动态阈值uT来控制操纵动作节点生成的方法,能有效实现操纵动作节点的精确提取。(2)根据驾驶人在执行某一具体驾驶行为时操纵动作的变化规律,提出基于有向图的个性化驾驶行为操纵模式建模方法,测试表明该方法能有效地体现驾驶人的个体特性。(3)利用ART2A-E算法能够较好地实现基于个性化驾驶行为操纵模式反向映射的驾驶行为识别,为研究开发个性化和智能化的驾驶安全辅助系统提供了技术支撑。
[Abstract]:In the closed loop traffic system composed of man, vehicle and road, the correlation between the human factor and the probability of traffic accident is the highest. Based on this, in order to coordinate the relationship among people, vehicles and roads to reduce the probability of traffic accidents, the driving safety assistance system (Driving Safety Assistance Systems,DSAS) has become one of the most important research contents in the field of intelligent vehicles. At present, the existing DSAS system is basically designed on the basis of the reference driver, which results in the DSAS system adopting the same auxiliary control strategy and control parameters for all drivers, which can not meet the individual needs of the driver. If the DSAS system can automatically complete the learning of the driving characteristics of the driver by collecting the driver's driving behavior control data and vehicle running state information, and applies the learning results to the auxiliary control strategy of the DSAS system. It can improve the level of personalization and intelligence of DSAS system. One of the key technologies to achieve this goal is to study and solve the modeling and identification problems of driver's personalized driving behavior control mode. Based on the above problems, the main research contents of this paper are as follows: (1) by collecting driving behavior manipulation data, preprocessing the data, using the data / mechanism mixing method to analyze the individual characteristics of the driver. The extraction method of driver manipulating action node is studied in order to model the driving behavior control mode. (2) on the basis of accurate extraction of driver manipulating action node, based on directed graph principle, The modeling method of personalized driving behavior control mode is studied. (3) using ART2A-E adaptive resonant neural network has the characteristics of stability, plasticity and real-time on-line self-learning. A driving behavior recognition method based on reverse mapping of ART2A-E algorithm and control mode is studied. The results show that: (1) based on the analysis of the individual characteristics of the driver, the method of controlling the generation of manipulating action nodes by setting a dynamic threshold uT is proposed. It can effectively extract the manipulative action nodes. (2) according to the changing law of the driver's manipulating action while performing a specific driving behavior, this paper presents a modeling method of individualized driving behavior control mode based on directed graph. The test results show that the method can effectively reflect the individual characteristics of the driver. (3) using ART2A-E algorithm can better realize the driving behavior recognition based on the reverse mapping of personalized driving behavior control mode. It provides technical support for the research and development of individualized and intelligent driving safety assistant system.
【学位授予单位】：合肥工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U491.25

【相似文献】