车载信息融合下的混合驾驶工况识别

Recognition and Classification of Vehicle Driving Cycles based on Information Fusion

整车驾驶性表现隐含在各驾驶工况下汽车客观性能参数的数据特征中,基于车载数据自动快速地识别与分类驾驶工况,对于提高各大主机厂整车驾驶性开发效率与结果准确性的意义重大。根据整车道路试验数据,结合专家知识与汽车试验理论,对所需识别的各驾驶工况进行工程定义,并采用数据层信息融合设计各工况下的特征参数。通过特征提取与组合构建驾驶工况识别样本集,并进行归一化处理。分别基于C4.5决策树及朴素贝叶斯方法构建了混合驾驶工况识别与分类模型,并采用测试数据集进行了验证。根据上海某自主品牌车型车载多传感器数据进行模型测试,结果表明,模型对多维数据中隐含工况的分类准确率均达到85%以上,在大幅提高分析效率的同时保证了较高的精度,基于C4.5决策树构建的工况分类器性能优于朴素贝叶斯,精度可达95.5%。

The vehicle drivability performance is manifested in the data characteristics of the objective parameters of the vehicle for different driving cycles, and it is of great significance to automatically identify and classify the driving cycles on the basis of on-board multi-sensor data for vehicle drivability development and evaluation. According to the vehicle road test data, combined with expert knowledge and vehicle test theory, the driving cycles to be identified were defined, and the features for each cycle were designed based on information fusion. Then, the sample set for driving cycle recognition was constructed by feature extraction and combination, and the feature parameters were normalized. Finally, the driving cycle classification model was established based on the C4.5 decision tree and Naive Bayesian method respectively, and each model was tested and verified by using the test data set. The results were obtained on the basis of on-board multi-sensor data from an independent brand vehicle in Shanghai. The accuracy of the driving cycle identification and classification of each model reaches more than 85%, maintaining a high accuracy while improving the efficiency of drivability analysis. The classification accuracy of the C4.5 decision tree is 95.5%, which is better than that of Naive Bayesian.