Research On Path Following Control Method Based On Driver Deceleration Behavior Analysis

In the closed-loop driving system consisting of human-vehicle-road,the driver is considered to be the most uncertain and weakest part in this system due to the different driving skills,physiological and psychological state,personality and physical strength.For this reason,smart vehicles are considered to be a great solution to traffic safety problems,which partially or completely replace the role of human.Then,as one of the key technologies in the field of intelligent vehicles,path following control has become a research hotspot.However,most of the current research methods on path following control generally pay attention to improve the accuracy of following and ensure safety.In contrast,it is not enough to consider whether the smart car control strategy is in line with the driver’s general driving habits.In this paper,the driver’s performance in the natural driving environment is analyzed by statistical theory,and the analysis results are applied to the algorithm design of the vehicle path following control.Finally,the effectiveness of the control algorithm are verified in the simulation environment.The main research is as follows:1.The driver’s deceleration behavior analysis has become the focus of research,as the basis of the design of intelligent vehicle humanized brake control system.Generally,there are two modes for driver deceleration:(1)press brake pedal,applying brake system to reduce speed;(2)release accelerator pedal,releasing the gas pedal without pressing brake pedal,thus decelerating by naturalistic driving resistance.In the previous research,a large number of scholars have studied the "press brake pedal" mode,and the "release the accelerator pedal" mode has not received sufficient attention.This paper relies on the naturalistic driving data collected by the Automated Vehicle Testing and Evaluation Technology Project of China Automotive Engineering Research Institute(CAERI).By setting the boundary conditions,the sample cases consisting of digital data and video data are intercepted from the tens of thousands of kilometers of vehicle driving data.Then,the potential factors that may affect the driver’s braking behavior are divided into continuous variables and discrete variables.Then,by watching the video,the discrete variables of each sample are uniformly assigned.At the next,the significant variables affecting the driver’s braking behavior are extracted by Student’s t test.Finally,this paper build a logistic regression model with the independent variables,and the dependent variable.2.As one of the key technologies in the field of intelligent car,path following control has become a research hotspot.And,based on the previous analysis of the driver’s braking behavior and the logistic regression model,a path tracking longitudinal controller based on driver behavior is designed.The controller can enable the vehicle to travel at a desired speed by adjusting the vehicle throttle opening and the brake system wheel cylinder pressure to achieve the vehicle’s longitudinal speed(vehicle travel direction)tracking.the optimal preview theory is widely used in the design of path tracking controllers because it is more in line with the driver’s handling habits.Further more,in order to improve the computational efficiency and adaptability of path control algorithm based on preview theory,in this paper,an arc-length preview method is proposed to obtain the lateral displacement of preview point based on actual traveled distance of the car,which originate from the preview optimal curvature model.Under this method,a relationship is deduced that between the lateral displacement of preview point and the rotation angle of front wheel.Subsequently,the lateral control model of path tracking is established.3.In this paper,the path-following controller is established in Simulink according to the previous research.And,this controller is used to establish the “human-vehicle-road” closed-loop simulation system in the Carsim/Simulink co-simulation environment.Finally,the effectiveness of the controller are analyzed by establishing several typical simulation conditions.The research results show:(1)There are 7 factors,which significantly impact the driver’s deceleration behavior,including intersection type,traffic flow,echo-vehicle motion state,target motion state,ego-vehicle braking initial speed,ego-vehicle braking time THW(time head way),the expected acceleration;(2)The logistic regression model can quantitatively describe the relationship between the seven significant factors and the driver’s deceleration behavior,and the probability that the driver will adopt the "release accelerator pedal" mode;(3)The steering wheel stability of the human-vehicle-road closed-loop system under the arc length pre-sight method is simultaneously impacted by the response delay time of driver(or vehicle),the preview distance and the continuity of the path point direction.When the desired path point direction has a high continuity,and the system preview distance is greater than the critical front line of sight,the steering wheel tends to be stable;(4)Based on the preview optimal curvature model,this research proposed the arc length preview method,which has a high operation speed in the human-vehicle-road closed-loop simulation system(the radius of 80 m Path)with the average calculation time is 3.75s~4.56s;(5)This research designed the longitudinal tracking three-phase closed-loop control system with the probability of the model as the logic control valve,which significantly reduced the switching times of the throttle opening,the total duration of the throttle opening,the brake wheel cylinder pressure switching times and the total duration.