Research On Driving Characteristics Of Lane-changing Vehicle Based On Driver Behavior

As a typical driving behavior pattern,lane change has a significant impact on driving safety.The accidents caused by inadequate information perception and incorrect decision-making during the lane change has continued to rise.If the driving intention is predicted relying on multi-dimensional signals before lane change occurs,the safety of driving environment can be assessed effectively in advance,and then making accurate judgments so the traffic accidents can be avoided.As a result,this thesis focused on driver's visual characteristics and vehicle operation characteristics to deeply study the driving characteristics of lane-changing vehicles.First of all,based on the relevant researches at home and abroad,lane-changing behavior,eye movement behavior and data processing were deeply analyzed.The lanechanging process was divided into two phase: intention phase and implementation phase,and five fixation areas were also obtained.According to the research requirements,a driving simulation experiment scheme was designed,and video footage and relevant data of driving process were collected from a simulated driving experiment of 22 subjects,which provided data for follow-up analysis.A questionnaire was used to quantify and classify the driving style,the subjects finally were divided into three driving styles: impulsive,common,and cautious.Secondly,time window of three style drivers which could characterize driver's lane changing intent was respectively determined as 2s,3s and 5s based on rearview mirror visual characteristics and steering wheel angle,then a sample set containing 397 different types of samples was obtained.Besides,statistical methods such as mean analysis and standard deviation analysis were used to analyze the change law of driver's visual parameters and vehicle operating parameters.Based on the visual analysis,the independent-samples T test(T-test)and analysis of variance(ANOVA)were used to quantify the effects of driving behavior and driving style on each parameter.The results show that most parameters had significant differences.Finally,based on differential analysis and principle of indicators selection,the indicator system that characterizes different driving behaviors was determined,which made up of five parameters,namely Speed,Vehicle offset angle,Lateral acceleration,Distance to vehicle in front and Time to collision.Then,the behavioral prediction model based on multi-valued Logistic regression was established by integrating the five indicators.In addition,the test samples were utilized to realize the prediction of lanechanging intention,the overall prediction accuracy rate reached 89.5%,which indicated that the model has a good prediction effect.By the research of this thesis,based on change law of driver's visual characteristics and vehicle operation characteristics in highway scene as well as the difference in driving behavior and driving style of different parameters,enterprises can further improve driving assistance system and active safety warning system.At the same time,the lane-changing intention prediction model which integrated multi-parameter will enable driver of rear vehicle to evaluate the safety of driving environment in advance,which is be of great significance for improving vehicle safety as well as provides a powerful technology for future active safety of smart vehicles.