Recognition And Early Warning Of Cognitive Distraction Driving State Based On Human Factors Data And Vehicle Operation Information

Security has always been the focus of people’s livelihood that cannot be ignored.However,distracted driving has a great impact on the driver’s information processing ability and poses a serious threat to traffic safety.This topic mainly studied distracted driving behavior in order to provide some help for the development of automobile safety warning system.The main research contents are divided into the following five parts.First of all,the research background and significance of distracted driving are described.Through the study of relevant literature,the deficiencies of the current research are summarized from three aspects of distracted driving intervention technology,distracted characteristics monitoring technology and distracted driving identification technology,and the research content and research route of this paper were also clarified.Secondly,a questionnaire survey of 300 drivers was conducted by multi-stage and multi-level sampling method,and four types of typical driving behaviors with high risk causing drivers’ cognitive distraction were obtained,as well as the ranking of acceptability of different warning methods.It provided the basis for the design of the driving task and the setting of the early warning mechanism.Thirdly,according to the results of the questionnaire survey,four kinds of driving tasks were designed to induce cognitive distraction,so as to achieve the purpose of inducing cognitive distraction of drivers in the experiment.In the simulated urban Road driving environment built by UC-win/road simulation software,driving simulator and Tobii glasses,the simulation driving experiments of driving without distraction and performing four tasks were carried out.The corresponding vehicle running information and human factor information were collected,and the cognitive distraction database was established.At the same time,the division of the following segments,the data preprocessing and the calculation of the characteristic indexes of the original data were completed.Through descriptive analysis,three characteristic indexes that affect the risk of distracted driving are obtained,which are the standard deviation of longitudinal speed,the standard deviation of steering wheel angle and the driver’s braking reaction time.On this basis,an early warning grading model of distracted driving was established by using analytic hierarchy process and fuzzy comprehensive evaluation.According to the early warning level,five groups of data were divided into three levels: no early warning,secondary early warning and primary early warning.And the grading effect was verified through the human factor data index.Finally,based on genetic algorithm,35 characteristic indexes of vehicle running information were extracted,and 8 representative characteristic indexes were obtained.Using time series data composed of selected features,a cognitive distraction recognition and early warning model based on LSTM was established,which effectively avoided the negative effect of neural network in feature extraction and realizes real-time recognition and early warning of cognitive distractions with different risk levels.In addition,the SVM recognition model was established.By comparing and analyzing the accuracy,F1-Score and recall rate of the two models,it was proved that the LSTM model was obviously superior to the established SVM model.And early warning mechanism was introduced,and distraction early warning systems with different early warning levels were set up.