基于改进DBSCAN算法的驾驶风格识别方法研究

Research on a driving style recognition method based on improved DBSCAN algorithm

新能源汽车监测平台由于需要记录的数据量大项多,为降低成本,通常采用10 s或30 s的数据采集周期,导致在使用平台数据识别驾驶风格时易出现信息缺失、识别不准的问题。为此,文章提出了一种基于改进DBSCAN算法的驾驶风格识别方法,从参数和权重2个方面改进算法;对741台电动汽车的自然行驶数据聚类处理,得到驾驶风格评分,根据评分将驾驶风格分为5类,并讨论了采集周期对识别结果的影响。研究结果表明,在1 s和10 s 2种采集周期下,驾驶风格识别结果基本相同,而30 s的数据则会出现较大误差;对5类风格车辆的平均车速、平均加速度、加速度标准差和车辆行驶能耗等指标的计算结果表明,驾驶风格越激进,指标值越大,从而验证了该文方法的合理性。

Due to the large amount of data to be recorded,in order to cut the cost,many new energy vehicle monitoring platforms usually use a data acquisition cycle of 10 seconds or 30 seconds,which leads to a problem that is the lack of information and identification when using the data from platform to identify the driving style.To solve this problem,this paper proposed a driving style recognition method based on the improved DBSCAN algorithm,improving the algorithm from the aspects of parameters and weights,and clustered the natural driving data of 741 electric vehicles to obtain the driving style score,then divided the driving style according to the score into five categories and discussed the impact of the acquisition cycle on the recognition results.The result shows that under the two acquisition cycles of 1 second and 10 seconds,the driving style recognition results are basically the same,but the 30-second data will have a big error.The calculation of the average speed,average acceleration,standard deviation of acceleration and vehicle energy consumption of the five types of vehicles indicates that a more aggressive style tends to have a higher value among the above indices,which verifies the rationality of the recognition method.