Research Of Trajectory Planning And Control For Autonomous Driving Vehicles Under Multi-scenarios

Autonomous driving is an effective methods to decrease serious road traffic accidents,promote traffic efficiency and satisfy the diversified travel demand of passengers,and has been proved to be a significant direction for the evolution of the automobile industry.Autonomous driving is also a complex and huge system.Vehicle trajectory planning and trajectory tracking control are significant components of this system.This paper mainly focuses on the intersection and non intersection scenarios in urban roads.Based on the decoupling theory of path and speed,this paper decrease the dimension of three-dimensional programming problems including time,lateral and longitudinal into twodimensional programming problems.By searching and optimizing,a trajectory planning framework for autonomous driving vehicles under multi-scenarios is realized,and a trajectory tracking controller based on model predictive control is designed.The main contribution of this paper are as follows:（1）In the part of trajectory planning,this paper devises a trajectory planning algorithm based on path and speed decoupling.Firstly,continuous and smooth discrete reference lines are generated by optimization.The SL graph is constructed based on the reference line,use A* algorithm to search the original feasible path,and then the method based on quadratic programming is used to plan the sophisticated path.Finally,the sophisticated path is smoothed on the premise of ensuring security.The ST graph is constructed based on the optimized path and predicted trajectory of obstacles,the rough speed profile is searched by dynamic programming.In order to improve the computational efficiency,an accelerated jump out mechanism is added to the dynamic programming algorithm.Based on the rough profile,the feasible region of speed planning is selected and the simplified relationship between time and curvature is constructed.Therefore,the lateral acceleration constraint of the vehicle is added to the quadratic programming,and the final speed profile is obtained.In addition,for common traffic factors in non intersection scenarios,such as reverse direction of opposite vehicles and signal lights in intersection scenarios,measures such as obstacle screening,introduction of virtual obstacles and obstacle expansion are adopted in the planning algorithm to limit the feasible region of the planning problem.These behaviors enhance the generalization ability of the planning method,and ensure the planning results more practical.（2）In the part of trajectory tracking control,a trajectory tracking controller based on the model predictive control method and coupled lateral and longitudinal is devised in this paper.Based on the 3-dof dynamic vehicle model considering longitudinal acceleration,the model predictive control algorithm is devises and deduced theoretically.In order to ensure the smoothness of the control,the incremental model predictive control is adopted,and the model is linearized to ensure the solution speed.A simple bottom controller is designed.The driving model of the vehicle is switched by using the taxiing curve of the vehicle,and the final control signal is obtained through the longitudinal dynamic model of the vehicle.（3）In Algorithm verification part,a simulation architecture based on Prescan,Matlab /Simulink and ROS is constructed,firstly.Prescan is used for scenarios construction.ROS is responsible for the communication between modules,and Simulink is used as the interface of simulation signal.In order to prove the planning algorithm devised in the paper,different simulation conditions are designed for intersection and non intersection scenarios,such as lane change,uncontrolled intersection,cut in and so on.The co-simulation results show that the trajectory planning algorithm devised in this paper can plan the feasible trajectory satisfying various constraints in both urban intersection and non intersection scenarios,and the model predictive controller based on the 3-dof dynamic model can keep the tracking error within a small range.In addition,the running frequency of both the planning algorithm and the tracking control algorithm can meet the harsh demand of the autonomous driving system.