Research On Curve Free-driving Longitudinal Personalized Control Of Autonomous Vehicle

In recent years,with the continuous development of autonomous driving technology,the planning and control technology has been more mature but far away from fully automated self-driving,the human-machine co-driving will continue for some time.In the humanmachine co-driving phase,the role of the driver is faced with the change of from the operator to the supervisor.Therefore,the feelings of human driver cannot be ignored.Most of the traditional longitudinal planning and control technology uses control theory method such as PID and its derivative algorithm,MPC,LQR and so on,which achieving good performance in the vehicle straight drive situation.But in the vehicle curve driving situation,the traditional control method only limit the maximum speed according to the curvature of the road.The timing of acceleration and deceleration of algorithm is different from that of human driver.In order to make the driver have a better driving experience,taking the driving data as the propeller and integrating driving behavior characteristics into algorithm design.This paper focus on longitudinal personalized control in the curve free-driving situation.By analyzing the driver's curve free-driving data and considering the road construction,the driving phase can be divided into three section,close-curve driving,on-curve driving and leave-curve driving.When in on-curve driving phase,the driver tends to keep a particular comfortable speed,and when in close-curve driving and leave-curve driving phase,the driver will take a certain deceleration and acceleration.Based on the three-phase driving characteristics,a curve-free driving model is established,which includes an acceleration / deceleration model based on the GMM/GMR and a speed tracking model.In order to achieve the consistency of the driving trajectory under the premise of precise path following and considering the robustness and computational efficiency of the algorithm,the improved pure pursuit algorithm is used to track the desired path.In order to verify the curve free-driving model proposed,a model verification experiment was designed based on the CMSP simulation platform.First of all,using simulation platform to collect driving data.Then,in order to realize the precise control of the model on line,the split-mode speed control algorithm and the improved pure tracking algorithm are respectively carried out to verify the validity of the algorithm.After verifying the validity of these two algorithm,an online verification experiment of the curve free-driving model are conducted.The experimental results show that the algorithm can effectively imitate the driver's longitudinal manipulation and realize the consistency with the driver' s drive path.