基于SOM神经网络的多工况驾驶风格识别

不同的驾驶员对车辆的各项性能可能有个性化地要求,因此有必要对驾驶风格的分类与识别问题进行研究。首先在驾驶模拟器上采集不同驾驶员在多工况下的数据,利用主成分分析法选取驾驶员在各个工况下的特征参数,SOM神经网络分别对起步、加速及制动工况下的驾驶数据进行了聚类分析,然后以驾驶风格聚类分析结果为基础,建立了基于SOM神经网络的驾驶风格识别系统,该系统可根据驾驶员驾驶历史数据来判断其驾驶风格,最后以某一温和型驾驶风格识别结果为例验证了系统的合理性。

Driving style recognition based on SOM neural network under multiple driving conditions

The performance of different drivers is usually personalized,so it is necessary to study the classification and recognition of driving style.Firstly,the driving data of some drivers in different conditions were collected in the driving simulator.Then the driver’s clustering characteristic parameters in such conditions were selected by principal component analysis.The collected driving data were clustered by SOM neural network clustering algorithm.Based on the cluster results,the diver’s characteristic identification model that was based on SOM neural network was built.It can identify the diver’s characteristic through the driving data.Finally,taking a conservative driving style recognition result as an example,the rationality of the recognition system is verified.