Fundamental Study on Human Model Synthesis Based on Extraction ofAccident Risk Variable and Its State Transition Probability

本研究提出了一种通过人-车系统模型参数分析和驾驶操作行为分析的理论方法，特别是在困倦驾驶状态和注意力不集中状态下。这项研究可能有助于进一步减少追尾事故，追尾事故占日本所有事故的30%。（1）基于驾驶模拟器和试验车的试验，研究了与车速控制相关的驾驶行为参数的提取方法。(2)使用多名受试驾驶员收集乔夫市和河口湖町之间的公共高速公路驾驶数据。(3)通过将驾驶数据与驾驶员模型进行比较来进行困倦驾驶行为分析。(4)通过与驾驶员面部摄像头图像主观评价结果的对比，验证了该方法的有效性。本研究中采用的速度控制驾驶员模型基于风险势理论表示为弹簧-质量-阻尼器系统，并通过引入侧向控制驾驶员模型，提高了困倦驾驶检测的准确性。本文的研究成果有望应用于主动安全系统，以防止事故的发生.

This research proposes a theoretical method of by driver-vehicle system model parameter analysis and driving operational behavior analysis especially in drowsiness driving state and inattention states. This research potentially contributes to the further reduction of rear-end collision accidents which accounts for 30% of all accidents in Japan. The following research issues were conducted during the research term:(1) The method to extract driving behavior parameters which relate to the velocity control based on experiments using Driving Simulator and a test car.(2) Public expressway driving data collection between Chofu city and Kawaguchiko town using a number of subject drivers.(3) Drowsiness driving behavior analysis by comparing the driving data with the driver model.(4) Verification of the effectiveness of the drowsiness detection method by comparing with the subjective evaluation results using driver face camera image.The velocity-control driver model used in this study is expressed as a spring-mass-damper system based on risk potential theory.Furthermore, by including the lateral-control driver model, the higher accuracy of drowsiness driving detection can be obtained. The achievement from this researchis expected to be applied to active safety systems to prevent accidents.