Research On Driving Behavior Based On Natural Driving Data

Nowadays,with the rapid increase in the number of cars in China,problems such as road congestion,reduced driving safety,serious environmental pollution,and reduced driving experience have become more and more serious.Many relevant studies have confirmed that improving the driving behavior of drivers is an important means to solve the above problems.At present,the natural driving data obtained by using the Internet of Vehicles technology provides a new way to analyze and understand the driving behavior of the driver.In order to analyze the driving behavior accurately,we propose a scenario recognition method firstly.By using the method,various scenarios experienced by the driver in the driving process can be identified and the driving characteristics in typical scenarios can also be analyzed.Secondly,we take driving characteristics as an important basis for determining driving behavior evaluation indicators and propose a comprehensive evaluation method of driving behavior based on multi-scenario characteristics.The main research contents of this article are as follows:(1)Firstly,the sources and characteristics of natural driving behavior data are described in detail.Then,a scenario recognition method which can identify driving scenarios from natural driving data and extract corresponding driving scenario data is proposed.In addition,methods such as mean filtering are used to preprocess data items such as acceleration in natural driving data,which provides convenience for subsequent data analysis.(2)Secondly,based on the study of driving scenarios,parameters related to driving behavior are selected from natural driving data to analyze the characteristics of these parameters in specific scenarios.Besides,the effect of scenario time nodes on driving characteristics is studied by means of variance analysis.(3)Thirdly,a comprehensive evaluation method of driving behavior based on multiscenario driving features is proposed.In this paper,the driving behavior evaluation indicators and evaluation system are determined based on the characteristics of the scene.The weights of each indicator are determined by the combination of extension analytic hierarchy process,entropy weight method and combination weighting method.Then,the gray whitening weight function clustering method and fuzzy comprehensive evaluation method are used to evaluate the driver’s driving behavior.(4)Finally,based on natural driving data and driving behavior analysis,this paper develops a natural driving data analysis system.The system uses Hadoop-based offline computing technology and Spark Streaming real-time computing technology to realize the vehicle condition monitoring,driver behavior analysis and evaluation and so on.Finally,the system can be used to provide suggestions for drivers to improve their driving behavior and provide feasible solutions of operation and maintenance services and iterative development of products for auto makers.