基于高德导航数据与FOA-GRNN模型的驾驶倾向性辨识方法

A Method for Identifying Temperament Propensity of Drivers Based on AutoNavi Navigation Data and a FOA-GRNN Model

为提升汽车主动安全功能,研究了1种基于高德导航数据的低成本、高精度驾驶倾向性辨识方法。基于高德软件开发工具构建动态驾驶数据采集应用程序,并融入个人智能终端以实现对行车数据的实时采集、处理与网络化存储。通过驾驶员生理、心理测试和实车实验获取不同驾驶倾向性驾驶员在导航行驶过程中由时间、速度和加速度推演的驾驶行为信息,采用主成分分析法(PCA)提取驾驶倾向性主要因子,并将驾驶倾向分为激进型、普通型和保守型这3类。构建基于果蝇优化算法(FOA)和广义回归神经网络(GRNN)的高精度驾驶倾向性辨识模型,利用特征变量集对模型进行训练和验证。验证结果表明:该模型总体准确率可达94.17%,对激进型、普通型和保守型的驾驶倾向性的辨识精确度分别为95.06%,92.5%,94.93%;进一步对比发现,该模型比单一的GRNN模型总体准确率提高5%~10%,与现有基于惯性传感器数据和离散小波变换结合自适应神经模糊推理系统的方法相比,该方法更具实用性且模型总体辨识准确率提升了2.17%。

In order to improve the capacity of automobiles in active safety,a method for identifying driving propensity with a low-cost and high accuracy based on AutoNavi navigation data is proposed. An application to collect driving data is developed based on Amap software development tool,which is further integrated into an intelligent terminal for data collection,procession,and storage in real time. Driver behavior data inferred from the time,speed,and acceleration of vehicles controlled by drivers with different temperament propensity are obtained through physiological,psychological and driving experiments. The principal component analysis(PCA)technique is used to extract the important factors for studying the temperament propensity of drivers,and the drivers are grouped into three driving propensities: radical,common and the conservative. A Fruit-fly optimization algorithm(FOA)and a generalized regression neural network(GRNN)are integrated to establish a high-precision model for driving propensity identification,which is further trained and verified using collected data. The verification results show that: the overall accuracy of the identification model is 94.17%,and the identification precision of the radical,common and the conservative types are 95.06%,92.5% and 94.93%,respectively;compared to the simple GRNN model,the overall precision of the proposed model is improved by 5%~10%;and compared to the previous method based on inertial sensor data and the integrated method of discrete wavelet transformation and adaptive neuro fuzzy inference system,the FOA-GRNN model is more practical,and its overall precision is improved by 2.17%.