Research On Local Path Planning And Tracking Of Autonomous Vehicle Based On APF And MPC

Autonomous driving is an important trend of vehicle development,which can replace the human driver to accomplish some driving tasks.Recently,autonomous driving attracts much attention since it can effectively alleviate the problems of traffic safety.The key technologies of autonomous driving include environmental perception,decision-making,path planning,trajectory tracking,etc.The local path planning is the same as the trajectory planning,it aims to plan a suitable trajectory based on the environmental perception,which is the prerequisite for autonomous driving.And the trajectory tracking is the key to autonomous driving,which ensures that the vehicle follows the planned trajectory.In this thesis,the methods of the local path planning and trajectory tracking are researched.The main research contents are as follows.(1)In order to fully describe the interaction between various factors(such as obstacles,road structures,environmental vehicles,and target point)and vehicles in local path planning,the artificial potential field(APF)method is chosen to describe the driving environment.The current driving environment potential field models do not take all environmental factors into consideration and the same kind of factors are treated uniformly,without considering the difference in intra-class.To solve these problems,obstacles,road structures,environmental vehicles and target point are considered as the key factors in the thesis.Moreover,the obstacles are divided into crossable obstacles and noncrossable obstacles,then according to the difference of the two types of obstacles,the potential field functions of Gaussian-like and power functions are used to describe the obstacles respectively.In the same way,the road factors are divided into lane line and boundary line,and the straight road and the curve road are considered.And in the function of the environmental vehicle potential field,the vehicle is classified by large vehicles and small vehicles,and the longitudinal potential field function is adjusted according to the difference in vehicle types.Finally,a complete driving environment potential field model is proposed in this thesis,which is the basis of the subsequent trajectory planning.(2)Combining the proposed driving environment potential field with model predictive control(MPC)algorithm,a unified autonomous vehicle trajectory planning and tracking method is proposed,which can plan and track the trajectory simultaneously under various traffic scenes.Considering the requirement of vehicle characteristics in trajectory planning,the traffic environment potential field is added to the objective function of the MPC algorithm,then the problem of trajectory planning and tracking is transformed into a unified constrained optimization problem.In addition,the current unified trajectory planning and tracking methods do not consider the impact of road surface.In order to adapt various conditions of the road surface,such as wet and slippery road surface,and improve the universality of the method,a more accurate vehicle dynamics model is used as the MPC prediction model.Moreover,the slip angle is added into constraints,which improves the stability of driving.(3)The current trajectory planning methods seldom consider driving style,so a trajectory planning and tracking method considering driving style in car-following mode and lane change(overtaking)mode is proposed.Driving style aims to improve the user experience,so the user can select appropriate driving style mode.In the carfollowing mode,the time headway potential field is added to reflect the difference of car-following distance between different drivers,and different longitudinal acceleration constraints are used to reflect the difference of speed control between different drivers.In the lane change(overtaking)mode,the lane change distance is used as the judgment of changing from the free driving mode to the lane change mode.In the course of lane change,the lane change duration potential field is used to represent the difference of lane change behavior between different drivers,and the comfort in the lane changing process is ensured by lateral acceleration constraint.The proposed methods are validated via CarSim and MATLAB/Simulink.A variety of driving scenarios(car-following,lane change,obstacle avoidance,straight experiment,curve experiment,etc.)are built,and the results demonstrate that the proposed unified trajectory planning and tracking method can deal with various traffic scenes.Furthermore,the trajectory planning and tracking method considering driving style is verified through car-following scene,lane change scene,the results reflect the characteristics of personalized trajectory planning.