Research On Motion Planning Of Automated Vehicle Based On Driver's Hazard Perception

In recent years,countries around the world have taken autonomous driving technology as an important direction for future traffic development,and actively promoted the research and application of autonomous driving technology.China attaches great importance to the development of autonomous driving and intelligent connected cars,and includes intelligent connected cars in the important areas of national intelligent manufacturing development in the next decade.Motion planning is one of the core technologies of autonomous driving.The main task of motion planning is to plan the execution strategy of autonomous vehicles in a dynamic environment.In the increasingly complex traffic environment,the driving safety of vehicles is affected by many factors,which can be divided into drivers,vehicles,roads and other factors.In the study of traditional motion planning methods,safety is mainly based on vehicles(kinematic and dynamic constraints of autonomous vehicles,other traffic participants)and road rules,and little consideration is given to driver.This paper mainly studies the motion planning methods of autonomous vehicles driving on structured roads.Based on the guarantee of safety,this paper deeply studies the influence of driver's driving style on motion planning The main research contents of this paper are as follows(1)This paper constructs a driving hazard field for obstacles,which maps potential dangers caused by obstacles to the traffic environment.During the driving process,the driver recognizes the potential risks in the traffic,evaluates the risks according to the corresponding influencing factors,and then selects the appropriate driving behavior based on the risk assessment results.This paper constructs the driving risk field of obstacles to quickly evaluate the driving risk in the traffic environment.This paper divides the construction of driving hazard field into two parts,namely static hazard field and dynamic hazard field.The driving hazard field is mainly affected by the external dimensions of the obstacle and the relative movement of the obstacle and the autonomous vehicle.The static risk field is constructed based on the high-dimensional Gaussian function,and its field strength and influence range are affected by the shape and size of the obstacle;while the dynamic hazard field is constructed based on the asymmetric two-dimensional Gaussian function,and its field strength and influence range are affected by the relative motion direction and relative speed of the obstacle and the autonomous vehicle(2)Based on the human driving data set,this paper analyzes the driving hazard levels that drivers with different driving styles can accept during driving.The driving behavior of vehicles on structured roads is divided into lane keeping and lane changing.Because lane changing behavior,especially the lane changing behavior to avoid collisions,is more difficult and dangerous,this paper mainly studies the lane changing process to avoid collision.Trajectory samples of lane changing to avoid collision are extracted from the human driving data set according to time-to-collision and time headway in the vehicle lane changing process.In this paper,the K-means algorithm is used to cluster the samples into three types of conservative,general,and aggressive driving styles based on the minimum time-to-collision and time headway.Based on the constructed driving hazard field,the driving hazard levels of different driving styles are analyzed,and the safety threshold of each driving style is defined.For a specific driving style,the driving hazard of autonomous vehicles in the traffic environment cannot exceed the safety threshold of the style(3)Based on different driving styles,construct collision-free trajectories that meet vehicle kinematics and dynamic constraints.In this paper,the driving behavior of autonomous vehicles is determined based on the driving hazard of the vehicle.Based on the safety threshold,a planning threshold for each driving style is defined.The planning threshold for each driving style is determined according to the driving risk at the beginning of the lane change trajectory.When the driving hazard of the autonomous vehicle in the traffic environment exceeds a planned threshold,the autonomous vehicle plans a lane change trajectory or a lane keeping trajectory to reduce the vehicle's driving hazard.This paper uses polynomial spirals to construct multiple candidate paths that meet vehicle kinematic constraints,then generates candidate trajectories based on fixed speed profiles,and finally selects the optimal trajectory based on the travel time and lateral acceleration(4)The joint simulation of MATLAB and CarSim is used to analyze the safety and feasibility of motion planning algorithms based on different driving styles in lane changing scenarios.The simulation results show that the following accuracy of planned trajectory is high,which is safe and feasible for the vehicle;different driving styles affect the behavior of autonomous vehicles in the traffic environment,such as the timing of lane changes,the acceleration and deceleration of the trajectory,and the lateral acceleration of the lane change trajectory,etc.