基于BP神经网络的驾驶员昼夜动态空间距离判识规律

Law of BP Neural Network-based Space Distance Cognition of Driver in Dynamic Environment at Day and Night

为研究昼夜动态环境中驾驶员对空间距离判识的规律,进行了实际道路试验。随机选取32名驾驶员分别在昼、夜环境中不同深度距离和速度下,判识红、绿色障碍物的空间距离,统计并检验驾驶员对红、绿色障碍物判识距离的差异,获得判识特征值;运用BP神经网络拟合距离判识结果,分析距离判识变化规律。结果表明:BP神经网络可以很好地拟合距离判识变化规律,精度优于现有模型;绝对距离和相对距离判识结果均随速度增加而减小,随深度距离增加而增大;夜间判识距离大于白天,驾驶员对相对距离判识准确性高。

In order to study the law of distance cognition of driver in dynamic environment at day and night, authors carried out the actual road test. 32 drivers were randomly selected to separately recognize space distances of red and green obstacles under different depth distances and velocities at day and night. Authors analyzed cognition difference between red obstacles and green obstacles and got character values by statistical method. Distance cognition results were simulated by BP neural network and variation laws of space distance cognition were analyzed. Results show that BP neural network can well simulate variation laws of space distance cognition and precision of the model is higher than that of other models. Cognitive absolute distances and relative ones decrease with increase of speed and augment with increase of depth distance. Cognition value at night is larger than that at day and driver＇s cognition accuracy of relative distance is higher than that of absolute ones.