Driving risk assessment under the connected vehicle environment:a CNN-LSTM modeling approach

Connected vehicle(CV)is regarded as a typical feature of the future road transportation system.One core benefit of promoting CV is to improve traffic safety,and to achieve that,accurate driving risk assessment under Vehicle-to-Vehicle(V2V)communications is critical.There are two main differences concluded by comparing driving risk assessment under the CV environment with traditional ones:(1)the CV environment provides high-resolution and multi-dimensional data,e.g.,vehicle trajectory data,(2)Rare existing studies can comprehensively address the heterogeneity of the vehicle operating environment,e.g.,the multiple interacting objects and the time-series variability.Hence,this study proposes a driving risk assessment framework under the CV environment.Specifically,first,a set of time-series top views was proposed to describe the CV environment data,expressing the detailed information on the vehicles surrounding the subject vehicle.Then,a hybrid CNN-LSTM model was established with the CNN component extracting the spatial interaction with multiple interacting vehicles and the LSTM component solving the time-series variability of the driving environment.It is proved that this model can reach an AUC of 0.997,outperforming the existing machine learning algorithms.This study contributes to the improvement of driving risk assessment under the CV environment.

Literature review of driving risk identification research based on bibliometric analysis

In order to understand the current research status and development direction of driving risk identification at home and abroad,relevant literatures in the field of driving risk identification from the China National Knowledge Infra-structure(CNKI)and Web of Science(WOS)in recent 12 years(2011-2022)were selected as research samples,and literature metrology tools VOSviewer and Citespace were used for visual analysis.The situation was analyzed from the aspects of chronological distribution,national cooperation network,distribution of domestic institutions,journal performance and keywords overview,literature coupling clustering and research hotspots.The results show that the number of published papers fluctuates year by year,and China,the United States and Germany have the largest number of published papers.The United States is at the center of international cooperation.The CNKI shows that universities in China such as Chang’an University and Chongqing Jiaotong University have published a large number of documents.According to the statistics of WOS,Accident Analysis&Prevention is the most widely published journal in the world.The average level of the journal is high and the quality of articles is better.Combining the research contents of CNKI and WOS,the main research directions can be clustered into five cluster themes by using the coupling function in VOSviewer,including driving risk assessment considering driver factors,the influence of driving environment on driving risk,driving risk assessment considering multi-source characteristic data,multiaspect research on driving risk and risk identification of non-traditional vehicles in specific scenarios.Human-machine co-driving,artificial intelligence,intelligent driving,risk identification and natural driving are the current research hotspots and the future research trends.

Towards Human‑Vehicle Interaction: Driving Risk Analysis Under Different Driver Vigilance States and Driving Risk Detection Method

The driver's behavior plays a crucial role in transportation safety.It is widely acknowledged that driver vigilance is a major contributor to traffic accidents.However,the quantitative impact of driver vigilance on driving risk has yet to be fully explored.This study aims to investigate the relationship between driver vigilance and driving risk,using data recorded from 28 drivers who maintain a speed of 80 km/h on a monotonous highway for 2 hours.The k-means and linear fitting methods are used to analyze the driving risk distribution under different driver vigilance states.Additionally,this study proposes a research framework for analyzing driving risk and develops three classification models(KNN,SVM,and DNN)to recognize the driving risk status.The results show that the frequency of low-risk incidents is negatively correlated with the driver's vigilance level,whereas the frequency of moderate-risk and high-risk incidents is positively correlated with the driver's vigilance level.The DNN model performs the best,achieving an accuracy of 0.972,recall of 0.972,precision of 0.973,and f1-score of 0.972,compared to KNN and SVM.This research could serve as a valuable reference for the design of warning systems and intelligent vehicles.