Research On Distracted Driving State Of Truck Drivers Considering Driving Fingerprint

In the road transportation system,the transport safety of dangerous goods vehicles(dangerous trucks)has always been a matter of concern.At the same time,distracted driving is an important factor leading to traffic accidents.Previous studies on driving behavior or distracted driving at home and abroad often regard all drivers as a whole.when comparing the differences in driving performance of drivers under different driving conditions and establishing a distracted driving identification model,the driver’s driving fingerprint characteristics are often ignored,which may lead to poor accuracy of the model.This study takes the drivers of dangerous trucks as the research object.First of all,the driving fingerprint characterization index system is analyzed and established.Secondly,the simulated driving experiment is designed and carried out.Then,the individual differences of driving performance data of different drivers under different driving conditions are analyzed,and the driving fingerprint file of truck drivers is established.Finally,a personalized truck driver distracted driving state identification model and distracted driving prediction model are established.The main contents of this paper are as follows:(1)driving fingerprint characterization method and simulated driving experiment.First of all,the driving fingerprint index system is established by reading the literature at home and abroad.Secondly,the simulation driving experiment for dangerous truck drivers is designed and carried out.Nine dangerous truck drivers were recruited and their driving performance data were collected.(2)the establishment and validity verification of dangerous truck driver’s driving fingerprint file.First of all,statistical methods are used to explore the individual differences of driving performance data among individual drivers.Secondly,select the indicators with significant differences among individual drivers to establish the driving fingerprint file.Finally,a driver identification model is established to verify the validity of the driver pattern file.(3)the establishment of personalized distracted driving state identification model.First of all,take the individual driver as the research object to explore the impact of each driver’s distracted driving on driving performance.Secondly,through the feature selection algorithm,the feature index set is formed to establish the personalized distracted driving state identification model.Then,by comparing the three algorithms of Gradient Boosting Decision Tree(GBDT),e Xtreme Gradient Boosting(XGBoost)and Light Gradient Boosting Machine(LGBM),the LGBM algorithm with the best performance is selected and the super parameters of the model are adjusted by Bayesian optimization algorithm.Finally,a personalized distracted driving state identification model is established and compared with the general model.By comparison,the performance of the personalized distracted driving state identification model considering driving fingerprint characteristics is better than that of the general model.(4)the prediction model of personalized distracted driving behavior is established.Using the index set formed by the above feature selection,a personalized distracted driving behavior prediction model based on LGBM is established,and the Bayesian optimization algorithm is used to adjust the super parameters of the model.It is proved that the personalized distracted driving prediction model considering driving fingerprint characteristics can accurately identify the premonitory performance before the occurrence of distracted driving behavior.