换道决策中的车辆速度估计模型
Speed estimation model during lane-changing decision

为研究换道决策阶段驾驶人对后方车辆的速度感知特征,以小型乘用车为平台,利用毫米波雷达、车载总线数据仪、音视频监测系统等搭建了试验车。招募了15名驾驶人,在某高速公路完成后方车辆速度估计试验,试验车速度分别设置为60、70、80、90 km？h-1,最后获取了1 625组数据。采用显著性分析方法,分析了相对速度、后方车辆速度与相对距离对驾驶人速度估计行为的影响特性。利用多元线性回归理论建立了驾驶人速度估计模型,并对模型进行了检验。分析结果表明:约60%的速度估计误差绝对值不大于10 km？h-1,且驾驶人的速度估计误差满足正态分布; 驾驶人速度估计误差随两车相对速度和后方车辆速度的增大而减小,相对速度和后方车辆速度较低时,易高估后方车辆速度,相对速度和后方车辆速度较高时,易低估后方车辆速度; 随两车相对距离的增大,驾驶人速度估计误差变化趋势不显著,但两车相对距离较小时,驾驶人易高估后方车辆速度; 速度估计模型的平均误差为-0.56 km？h-1,因此,估计结果可靠。
In order to research the driver’s estimation characteristic for rear-vehicle speed during lane-changing decision, small passenger car was used as test platform and established by using microwave radar, vehicle CAN-bus data logger, audio-video monitoring system. 15 drivers were recruited, and the estimation test of rear-vehicle speed was carried out in a normal highway. The speeds of test vehicle were set as 60, 70, 80, 90 km？h-1 separately, and 1 625 sets of data were obtained finally. The impact characteristics of relative speed, rear-vehicle speed, and relative distance on driver’s speed estimation behavior were analyzed by using significance analysis method. A driver’s speed estimation model was established by using multiple linear regression theory, and the model was examined. Analysis result shows that 60% of absolute values of speed estimate errors are no more than 10 km？h-1, and the driver’s speed estimation error follows a normal distribution. Driver’s speed estimation error decreases with the increase of relative speed and rear-vehicle speed, when the relative speed or rear-vehicle speed is lower, rear-vehicle speed is overestimated, and when the relative speed or rear-vehicle’s speed is higher, rear-vehicle speed is underestimated. When the relative distance of vehicles increase, the driver’s speed estimation error change very small. However, when the relative distance is smaller, rear-vehicle speed is overestimated by driver. The average estimation errors of speed estimation model is -0.56 km？h-1, so the model is feasible. 3 tabs, 12 figs, 23 refs