以既有路段为研究对象,分别基于理想道路中线平面坐标数据和实测道路中线平面坐标数据对比分析三次样条插值法、最小二乘拟合法、纬地平面智能布线法的适用性和稳定性,提出等距分组最小二乘拟合圆曲线法.结果表明,对于理想道路中线平面坐标数据,3 种方法恢复的圆曲线半径均可靠,其中最小二乘拟合法恢复缓和曲线长度的结果不稳定;对于实测道路中线平面坐标数据,三次样条插值法的适用性差且增大测量点密度不能提升恢复结果的可靠性,最小二乘拟合法和纬地平面智能布线法恢复圆曲线半径较可靠且通过增大测量点密度可在一定程度上提升缓和曲线长度恢复的可靠性.

为探寻城市道路交叉口交通事故预防重点,更好地预防城市道路交叉口交通事故,提取交通事故双方交通方式、行驶方向、碰撞方式三要素,将城市道路交叉口因交通参与者之间发生交通冲突导致的交通事故细化为 3 个大类、18 个小类,并制定命名规则分别对 18 种事故类型进行命名;分别以事故实际发生频率和事故危险性作为风险矩阵中风险概率和风险影响的定量衡量指标,提出一种将风险矩阵方法与有序样品聚类方法相结合的事故风险分级方法,分别对无信号控制交叉口和信号控制交叉口 18 种事故类型的风险进行分级,梳理出无信号控制交叉口 6 种中等风险以上交通事故类型、信号控制交叉口 3 种中等风险以上交通事故类型,并分别提出相应预防对策.

为探索有条件自动驾驶对非驾驶相关任务的允准边界,基于实车驾驶模拟器,设计自动驾驶接管试验典型场景,招募30名被试者开展驾驶模拟试验;要求驾驶人执行3种分心形式的驾驶次任务,系统发出接管请求提示后,驾驶人接管车辆控制权以避免险情发生,并分析驾驶人接管反应时间、驾驶负荷以及驾驶绩效等相关数据.结果表明:驾驶次任务涉及的分心形式越复杂,接管过程安全性越差,视觉分心任务与操作分心任务对接管行为影响显著;驾驶人更倾向于选择制动操作接管车辆,次任务分心程度越高,制动接管比例越大;与乘客聊天对接管行为影响不显著,看视频和玩手机游戏均会显著延长接管反应时间,增加工作负荷与车辆纵向减速度,玩手机游戏还会显著提升车辆横向加速度.

In view of the mass of accidents currently caused by distracted driving, research on distracted driving recognition methods was conducted. Driving simulation tests were carried out in which drivers were driving normally, were using a hand-held answering phone, and were using a hand-free answering phone during the lane change process. A radial basis function neural network model based on orthogonal minimum square method for distraction detection was built. Results show that drivers needed more time to complete a lane-change process and made more operation mistakes when he is answering a phone. Six driving performance parameters as vehicle's longitudinal velocity, lateral velocity, lateral acceleration, steering angle, steering wheel velocity, and opening degree of accelerator, changed significantly and were chosen as input variables for distraction recognition model. The average recognition accuracy of the model reached 88.7%, which could accurately identify if a driver is distracted in real time.