【作者】 段立飞；

【导师】 高振海；

【作者基本信息】 吉林大学 ， 车辆工程， 2013， 博士

【摘要】 交通安全一直都是当今社会最为严峻的问题之一，统计数据表明我国已成为世界上交通事故死亡人数最多的国家之一。因此，需要对驾驶员、汽车和环境因素各角度全面深入地分析交通事故发生的原因并提出相应的解决措施，降低交通事故的发生频率和伤亡人数。汽车驾驶员负责环境信息的感知和汽车的操纵控制，是人车路闭环系统中最为关键的环节。驾驶员在驾驶过程中需要集中注意力，综合分析汽车的运动状态、交通状况等信息，决策出正确的驾驶策略并做出相应的驾驶动作。因此，汽车驾驶不仅需要驾驶员付出一定的体力还需要承担一定的精神负荷，而当驾驶员承受的负荷过高或过低时，就可能引起决策或操作的失误，从而导致交通事故的发生。随着交通密度、道路复杂性的增加以及车内信息设备的使用，驾驶成为了更加复杂的交互活动，驾驶员经常处于双任务或多任务的工况下，其注意力可能从环境感知、汽车操纵等与驾驶任务直接相关的任务中脱离。然而，大脑处理信息的能力存在一定的局限性，当处在复杂的多任务环境中时，与驾驶任务不直接相关的其他任务在脑资源方面势必会与驾驶任务形成竞争关系，从而使驾驶员承受的精神负荷增加。当处在较高的精神负荷水平时，及容易导致驾驶员出现紧张感和压力感，并在感知和决策方面出现分心或失误，引发交通事故的发生。针对上述问题，本文在分析总结国内外驾驶员行为研究现状基础上，借鉴心理生理学和实验心理学研究方法，分析了多任务工况下驾驶员生理参数的动态变化规律，建立了基于生理参数的精神负荷评价方法，并将关于驾驶员心理生理特性的研究方法和规律应用于辅助驾驶系统关键问题之中，为辅助驾驶系统开发提供一定的理论基础和技术支撑。本文的具体研究内容如下：1.反映驾驶员精神负荷的生理信号变化规律分析。这一部分的研究目的是通过多任务模拟驾驶试验探索和验证典型生理参数随驾驶员主观精神负荷的动态变化，初步建立典型生理参数与精神负荷之间的定量关系。借鉴人因工程研究的双任务试验范例，在驾驶模拟器中设定了驾驶任务与语音交互任务的双任务工况，记录整个试验过程中的心电、皮电和呼吸在内的三种生理信号以及转向盘角度的变化情况。在数据分析阶段，对心率变异性、皮电水平和呼吸频率三种指标的动态变化情况进行统计分析，并运用信息论中的熵分析方法对单一任务与双任务工况下的转向盘角度变化规律进行了对比分析，在此基础上运用多元回归分析方法建立了精神负荷的统一评价模型。2.基于多导生理参数和支持向量机的驾驶员精神负荷评价。针对实际驾驶过程中具体的多任务工况如接打电话、发送信息和播放音乐等，通过驾驶模拟器试验采集真实驾驶员驾驶过程中的脑电和皮电两种类型的生理信号，运用独立分量方法对原始数据进行滤波，通过小波包分解技术提取多维时频域参数，构建反映驾驶员精神状态的特征空间。之后，利用核主成分分析方法对脑电信号小波包参数和皮电信号统计学指标进行降维处理。在对比了Logistic回归模型、BP神经网络和支持向量机模型的多项性能指标后，选取支持向量机作为不同精神负荷水平的识别模型，实现对驾驶员精神负荷水平的识别。3.驾驶员精神负荷评价方法在换道意图识别中的应用。这一部分的研究目的是将前文中进行的驾驶员心理生理特性研究方法和规律应用于辅助驾驶系统的关键问题之中，为辅助驾驶系统的设计提供理论依据。目前在车道偏离预警系统产品开发过程中，主要通过转向盘角度的变化、汽车在车道中的位置以及转向灯是否开启等操纵信息甚至是在车道变换之前的视线分布信息识别驾驶员的换道意图，在检测到汽车在驾驶员无意识情况下偏离原车道时及时发出预警信息。本文将驾驶员心理生理特性分析方法引入驾驶意图识别之中，分析了换道过程中生理参数的动态变化规律，将反映驾驶员精神状态变化的多生理参数指标应用到意图识别模型的构建之中，得到了具有较高识别精度的换道意图识别模型。本文借鉴心理生理学和实验心理学的研究方法，建立了多任务工况下驾驶员精神负荷的评价方法，并将驾驶员心理生理特性应用于辅助驾驶系统关键问题的解决之中。研究结果服务于驾驶状态和车内预警系统等汽车主动安全和辅助驾驶系统的理论研究和工程化应用，提高驾驶员的驾驶安全性和舒适性。更多还原

【Abstract】 Traffic safety has always been the most serious problem in current society, statisticsshow that China has the highest number of traffic fatalities in the world. Therefore, depthanalysis of the driver, vehicle and traffic condition is needed in order to find appropriatesolutions to reduce the frequency of traffic accidents and casualties.Driver is responsible for the perception of traffic information and vehicle status, playingthe most important role in the closed-loop of driver-road-vehicle system. A driver needs toconcentrate during driving and analyze the state of vehicle dynamics and traffic conditions,then making a proper driving strategies and maneuver. Therefore, a driver does not only needto afford some physical workload but also mental workload. When the load is too high or toolow, it may lead the driver making wrong decisions or driving maneuver, resulting in trafficaccidents.Changes of the driving task makes the problem above particularly prominent. Driversoften undertake dual-task or multi-task operating conditions because of the growing trafficdensity, complexity of road condition and usage of vehicle information system, theirattention may be detached from the current driving condition that directly related to thedriving task. However, the brain’s ability to process information has certain limitations.When in a complex condition, unrelated tasks are bound to compete for the brain resourcethat should be used in driving task, so that driver’s mental workload increased in thiscondition. When the mental workload reaches a certain level, it makes the drivers feelpressure and tension, leading to errors in the perception and decision-making.In order to solve the problems above, this paper analyzed and summarized the researcharticles in the field of driver behavior home and abroad, and based on the knowledge andresearch methods of experimental psychology and psychophysiology, evaluation method ofdriver mental workload under dual-task condition using psychological and physiologicalparameters is established. The methods and results in the research of driver mentalworkload was then used in driver assistance systems to solve the key issues, providing atheoretical basis and technical support for the development of driver assistance system.Specific contents of this paper are as follows:1. Research on dynamic changes of driver physiological signal under multi-taskconditions. The purpose of this part of research is to explore and verify the relationshipbetween driver mental workload and physiological information, establishing an evaluation method of mental workload based on typical physiological indexes.Dual-task paradigm in experimental psychology is introduced into our experiment. Adual-task condition is carried out in driving simulator containing driving task and voiceinteraction task. ECG, skin conductance, respiration and the steering wheel angle is recordedacross the entire experimental trial. In the data analysis stage, three kinds of physiologicalindexes including heart rate variability, skin conductance level and respiratory rate areanalyzed. A combined measure of driver mental workload is then created using multipleregression analysis and physiological indexes above. Finally, steering entropy method isused in the analysis of steering wheel angle under dual-task condition.2. Driver mental workload evaluation based on physiological parameters and supportvector machine. In this part, a driver mental workload recognition model is established basedon kernel principal component analysis and support vector machine.A field study is carriedout aiming at the events in actual driving conditions such as receiving calls, sendingmessages and playing music. Two types of physiological signals including EEG and skinconductance are recorded. Independent component method is used as the filter of EEG, andmulti-dimensional wavelet packet decomposition technique is used to extract thetime-frequency domain parameters. Kernel principal component analysis method is used toreduce the dimension of input data. At last, support vector machine is selected as the mentalworkload evaluation model after the comparison of model performance among Logisticregression, BP neural network and support vector machine.3. Application of driver psychological and physiological characteristics in the driverassistance system. The research purpose of this part is the application of driver psychologicaland physiological characteristics in design of driver assistance system. In the developmentprocess of lane departure warning system, driving intent is recognized mainly through thechanges of steering wheel angle, position of the vehicle, turning signal and gaze direction.When unintentionally lane departure is recognized through these parameters, a warningsignal is released to the driver. Lane change experiment is designed and physiological andvehicle status information are collected. Typical physiological indexes is analyzed and usedin the recognition model of lane change intention. Finally, a recognition model with higheraccuracy is derived compared with the model using vehicle status information only.Research method of psychophysiology and experimental psychology is used in thispaper, a mental workload evaluation method under dual-task condition based onphysiological indexes is created. Research method of driver’s psychological andphysiological characteristics is then used in the key issues in driver assistance systems.更多还原