Personalized Driving Decision-Making And Motion Control Method For Automated Vehicle

Autonomous vehicle technology is to provide services for humans,and its system should consider personalized characteristics of human drivers.In real traffic environment,different drivers demonstrate different driving behavior characteristics,and a single driving mode of autonomous vehicle is not able to meet different driving requirements.For high-level autonomous vehicle,if the autonomous vehicle does not match human driver’s expectations,the acceptance of the autonomous vehicle system will be reduced,and sometimes may even lead to accidents.However,in most current researches,the autonomous vehicle ignores the personalized characteristics of different drivers.This thesis focuses on the personalized decision-making,planning and control method of autonomous vehicle.Based on decision-making frameworks of hybrid automation,with the results of naturalistic driving experiments,personalized methods of autonomous vehicle considering human driving behaviors and surrounding vehicle characteristics are proposed.The specific research work included in this paper is as follows:(1)This thesis uses freeway naturalistic driving experiment data to study personalized driving behaviors.With naturalistic driving experiment data,based on the C-mean fuzzy clustering algorithm,lane-changing and car-following data are divided into different stages.And then we study the personalized driving behavior variables(headway,lane offset,velocity,acceleration,yaw rate,and yaw acceleration)in lanechanging and car-following scenes.The results present that the headway,acceleration and yaw acceleration can reflect the personalized driving behaviors for different drivers.(2)Aimed at personalized characteristics of human driving behaviors and surrounding vehicles,this thesis establishes decision-making frameworks of hybrid automata algorithms.Based on the developed APF(Artificial Potential Field)algorithm,traffic environment is rebuilt,and with human driver’s personalized driving behaviors we calibrate the critical parameters of the APF model,with which the personalized planned trajectories for autonomous vehicles come out.(3)This thesis designs a personalized tracking controller in a variety of typical traffic scenes based on the MPC(Model Predictive Control)algorithm.The PID feedback is introduced into the conventional MPC algorithm to eliminate the steadystate errors caused by the internal model distortion.The explicit MPC is also used to reduce the computing burden.Based on the hybrid automaton algorithm,we realize control mode switching for typical traffic scenes.In addition,chance constrained programming is used to include personalized constraints into MPC.(4)Based on the distributed MPC algorithm,this thesis studies the decisionmaking,planning and control method considering surrounding vehicles’ personalized characteristics.Based on the proposed driving compatibility indicator,different collaborative strategies are adopted between autonomous vehicles and surrounding vehicles.And then we design the distributed MPC to improve the control performance and meet drivers’ personalized needs,which can avoid "misjudgments" without considering surrounding vehicles’ personalized characteristics.In this thesis,the decision-making,planning and control methods are verified in the software-in-the-loop,hardware-in-the-loop and field vehicle experiment platform.The experimental results show that the method proposed in this thesis can control the vehicle smoothly in different typical driving scenes,such as car-following and lanechanging,and reflect some personalized driving characteristics.The personalized decision-making,planning and control algorithm proposed in this study can not only improve the level of road traffic safety,but also improve the driver/occupant’s acceptance of autonomous vehicles,which make a significant difference in the development of the autonomous vehicle in our country.