Driving Behavior Pattern And Establishment Of Behavior Spectrum Under Typical Operating Conditions

An important development trend of current road traffic safety is to shift from accident prevention and control to risk prevention and control,and risk prevention and control needs to focus on the probability prediction of possible future accidents.As the core element of road safety system,the prediction of traffic behavior is of great importance for accident risk prevention and control.The key to accurate traffic behavior prediction is how to grasp the driving behavior pattern through driving behavior analysis and find out the individual characteristics of drivers with differentiation,which is also the scientific problem to be focused on in the thesis.Based on the pre-study,this paper takes the driver’s focus on dynamic risk sources as a perspective and classifies the traffic scenarios into four typical operating conditions:lane changing on multi-lane highway,overtaking on two-lane highway,crossing crosswalks in the middle section of urban roads,and overtaking cyclists on single-lane highways.The common driving behaviors are classified into two categories: basic(free driving,stable following)and transitional(lane changing,overtaking,avoiding,etc.).Transitional behaviors for each operating condition are divided into several driving behavior patterns from the driver’s level of risk perception and differentiated response.Based on the similarity of driving behavior time series data and musical scores,a driving behavior spectrum definition and a normalized data structure are proposed.To study how to go from data acquisition to behavior spectrum establishment,an traffic video processing software was developed for drone video acquisition method,which realizes fast extraction of speed and trajectory of vehicles,non-motorized vehicles and pedestrians.For the driving simulation experimental data,a basic paradigm of data export is proposed,based on which a driving behavior spectrum establishment method based on a linear reference system is proposed.A large number of drone video and driving simulation experimental data were analyzed to establish a driving behavior spectrum,and the critical conditions that can reflect the differentiated driving behavior patterns under typical operating conditions were studied.The statistical analysis showed that for the lane changing behavior on multi-lane highway,when there is a slow moving vehicle blocking in current lane and the rear vehicle in the target lane is within 50 m of the lane changing vehicle,the driving behavior pattern is differentiated into two types: direct changing lane and changing lane after temporarily following.For overtaking behavior on two-lane highway,when the vehicle in the opposite direction is 100-300 m away from the overtaking vehicle,the driving behavior pattern will be differentiated into two types: direct overtaking and overtaking after temporarily following.For crossing crosswalks in the middle section of urban roads,when Tadv is between-4s and 2s,the driving behavior pattern is differentiated into two types: pedestrian game failure and motor vehicle game failure.For overtaking cyclists on single-lane highway,when the lateral distance between the outer edge of the motor vehicle and the pedestrian is greater than 0.5m,the driving behavior pattern will be differentiated into two types: steering avoidance and no steering avoidance.Based on the results of statistical analysis of behavioral spectrum data,a risk quantification calculation method based on probabilistic generalization and collision detection is proposed.Based on the similarity between driver risk perception and human ear perception of sound,a risk metric in d B is defined.The risk quantification model was used to quantitatively analyze the driving behavior risk of 29 drivers under four typical operating conditions.Based on the stability of the intrinsic value orientation of individual drivers,the behavioral spectrum characteristic parameter RTF,which can reflect the personality characteristics of drivers,was proposed.It is verified that the RTF has the stability of the same-scenario and crossscenario dimensions,and the prediction method of driving behavior based on RTF has high accuracy.