Research On Driver Steering Model Considering Driving Style Classification

In general,driving styles among drivers differ within a wide range.Even with the same driving task,drivers with different styles have different ways to control the vehicle,which requires that the control mode of the intelligent vehicle should be flexible.Currently,it is one of the research hotspots to maintain the tracking accuracy and ensure the driving comfort by considering the driver's driving style at the same time.Meanwhile,this trend is also the inevitable requirement for future development of intelligent and personalized vehicle active control technology.This thesis aims to use steering control as an entry point to establish a driver model that takes driving style into consideration,i.e.,steering control model considering driving style,to study the characteristics of vehicles controlled by drivers with certain driving style and their effects to tracing performance and stability of vehicles.It can provide new ideas for improving the adaptability of vehicle control systems for different driver groups.The influences of driver's driving style on the vehicle manipulation are represented by the vehicle stability and the randomness of a driver who may belong to cautious,moderate,or aggressive driving style.Research contents and conclusions of this thesis are as follows:First,based on the linear two-degree-of-freedom vehicle dynamics model,lateral kinematics parameters are selected as state variables to establish a predictive control model.To accurately track the target trajectory,the limit constraints of the vehicle dynamics are considered to determine the constraints of the state variables,controlled variables and controlled increment through model predictive control theory.Meanwhile,the solution of the steering controller is converted into a quadratic programming problem.Secondly,on the basis of real vehicle data,the average speed,number of overspeeds,maximum lateral acceleration,standard deviation of lateral acceleration,maximum value of the product of steering wheel angle and speed,proportion of time when the driving speed is over 80% of the limited speed,standard deviation of longitudinal acceleration and minimum time headway are extracted as indicators.Principal component analysis is used to reduce dimensions of these indicators and they are extracted as longitudinal driving characteristics,lateral driving characteristics and coupled driving characteristics.Based on K-means clustering method,30 drivers are classified into cautious drivers,moderate drivers and aggressive drivers.An analysis of 30 subjects shows that 9 drivers are cautious drivers,17 drivers are moderate drivers and 4 drivers are aggressive drivers.Finally,maximum lateral acceleration is extracted as an indicator of random characteristics of driving characteristics by studying the driving characteristics of different types of drivers.Based on stochastic programming model,random properties of driving characteristics are quantified as constraints.At the same time,based on the assumption of normal distribution,the distribution characteristics of the maximum lateral acceleration of three types of drivers are quantitatively described using point estimation and interval estimation.Driver steering model considering driving style is then established.By classifying driver's driving styles and analyzing the maximum lateral acceleration and other vehicle kinematic parameters,driving characteristics of drivers with different driving styles can be comprehensively understood.Based on this,a vehicle model with different driving styles is constructed.It can provide theoretical support for personalized control design of self-driving vehicles.