视觉类次任务驾驶安全性预测模型

Safety Prediction Model of Visual Secondary Task Driving

为了建立一种视觉类次任务驾驶安全性预测模型,实现对驾驶人执行视觉类次任务时的行车安全进行预测,设计并进行了城际高速公路场景中三种视觉类次任务驾驶试验;以眼球运动状态参数集和车身行驶状态参数集为安全评价指标,使用模糊层次分析法(F-AHP)计算得到测试驾驶人的安全评价数字等级;以此为基础,利用BP神经网络建立视觉类次任务驾驶安全性预测模型,以30名测试驾驶人的评价指标数据作为神经网络的输入训练样本、安全评价数字等级作为神经网络的输出训练样本,调试网络参数直至满足本研究的精度需要;最后,使用其余10名测试驾驶人的数据对模型进行适用性测试。研究结果表明,在操作车载收音机、发信息以及操作触摸屏设备这三种次任务情境下,本预测模型的适用性较好,可以比较准确的实现对驾驶人安全等级的预测。

In order to establish a vision-based sub-task driving safety prediction model, and to predict driving safety when the driver performs visual sub-tasks, three visual sub-task driving tests in the intercity highway scenario are designed and developed;The eye movement state parameter set and the body driving state parameter set are safety evaluation indexes. The fuzzy analytic hierarchy process (F-AHP) is used to calculate the number of the test driver's safety evaluation. Based on this, the BP nerves are used to establish the visual class times. The characteristic evaluation index data of 30 test drivers as the input training samples of the neural network is wsed in the mission driving safety prediction model, and the digital level of safety evaluation as the output training sample of the neural network, and debugs the network parameters until it meets the accuracy requirements of the research. Finally, the model was tested for suitability using data from the remaining 10 test drivers. The research results show that under the three sub-task situations of operating the car radio, sending information, and operating the touch screen device, the prediction model has good applicability and can accurately predict the driver's safety level.