Research On Algorithms Of Longitudinal And Lateral Control For High-speed Autonomous Vehicle

Autonomous vehicles are an important part of intelligent transportation systems.They are mobile robots that realize autonomous driving through technologies such as environmental perception,path planning,decision-making,and control.Use both hands to ease traffic jams and improve road utilization.Lateral and longitudinal control is one of the key technologies of autonomous vehicles.The research content of this thesis focuses on the design of the control system for autonomous vehicle in highway scenarios,including three parts: state observer,lateral control system and longitudinal control system,which can achieve speed tracking and safe distance maintenance,as well as lane keeping.First of all,in order to implement feedback control,a reliable vehicle motion state observer is needed.In this thesis,a single-track kinematics model and a particle kinematics model are sequentially simplified from the Ackerman steering model of the vehicle.The particle kinematics model is used to derive the linear state transition equation,apply the extended Kalman filter formula to obtain the observer of the vehicle movement state.Simultaneously,a method for estimating the noise variance is designed to obtain a more accurate observation noise covariance matrix.Secondly,the lateral and longitudinal control related algorithms are designed respectively.For longitudinal speed planning,the MPC-based distance-preserving algorithm and smooth speed planning algorithm are designed,as well as the switching logic of the two algorithms;For longitudinal speed control,a feedforward compensated PI control algorithm is designed for vehicle speed tracking;For lateral control,a linear lateral dynamics error model about the road is derived,based on the model,an LQT optimal control algorithm is designed to track the target path.Third,the LQT optimal control algorithm requires vehicle model parameters,in order to accurately estimate the model parameters,the kinematic parameter estimation and dynamic parameter estimation methods are designed respectively.Through measurement,trajectory fitting,and system identification,the wheelbase,steering ratio,fixed steering angle error,vehicle mass,moment of inertia,centroid position and lateral stiffness of front and rear wheel are obtained.These parameters have an important affect on the accuracy of LQT lateral control algorithm.Finally,the method co-simulation of Car Sim and Simulink was used to test the function implementation and control performance of the lateral and longitudinal control correlation algorithms respectively.For the longitudinal simulation experiments,the smooth speed planning and distance keeping,as well as the vehicle speed tracking performance were tested respectively;For the lateral simulation experiments,the circular path tracking and the curvature-varying path tracking were tested respectively.The results show that the lateral and longitudinal control related algorithms designed in this thesis can well implement the function and have good control performance.