Risky driving is a major reason of traffic accidents, thus detecting risky driving behavior is an important part of advanced driving assistance systems (ADAS) in reducing traffic accidents caused by human operation errors. However, the fixed warning strategy of these system is used for different types of drivers. In this paper, we propose a personalized online driving risk evaluation method. Firstly, velocity, longitudinal and lateral accelerations as well as time headway data of real vehicle are collected to analyze the lateral and longitudinal risky driving behaviors under different driving events include free driving, car following and lane changing. And then, fuzzy models are established for each driving event to get the quantified driving risk degree within [0,1]. More important, to ensure the effectiveness of warning information, the driving risk degree in medium range which are difficult to define the risk values are optimized by driving styles. Fuzzy c-means (FCM) algorithm is utilized to cluster the driving features of multiple vehicles. An optimal coefficient is obtained by centers and membership degrees of aggressive and safe driving styles clusters. The final driving risk degree is obtained after the output of fuzzy models has been optimized by the optimal coefficient. At last, the real driving data in NGSIM dataset are used to verify the validity of this method. The result shows that this method can evaluate driving risk online and optimize the degree according to different driving characteristics. This method can not only quantitatively evaluate the driving risk under different driving events, but also achieve a personalized warning to different drivers.
A Personalized Driving Risk Evaluation Method Considering Driving Style
2019
Aufsatz (Konferenz)
Englisch
Europäisches Patentamt | 2024
|