Research On Lateral Control Algorithm Of Self-driving Vehicle

Since this century,China's economic and social development has continued to develop rapidly.By the end of 2017,the number of motor vehicles in China has exceeded 300 million.The traffic accident casualties,atmospheric haze pollution and energy crisis and other social problems caused by the rapid growth of vehicle ownership in our country have become more and more serious at the same time.The research and development of self-driving vehicle and its related technology has received extensive attention from all fields of society.The selfdriving vehicle has great significance for the future urban intelligent transportation system.Vehicle motion control is one of the key technologies of the self-driving vehicle.Vehicle motion control can be divided into longitudinal speed control and lateral trajectory following control,this paper mainly studies the lateral control algorithm of the self-driving vehicle.Lateral motion control of self-driving vehicle is the basis of vehicle safety and stability and has been the hotspot and difficult point in the research field of self-driving vehicle in recent years.Lateral control system of self-driving vehicle requires perception system to provide environment information to judge the position and posture of ego vehicle and the location of desired trajectory so that the steering system can be controlled to ensure the accurate tracking of the desired trajectory and avoid the occurrence of the deviation accident.ASCL of Jilin University has been doing research on lateral control of self-driving vehicle related fields since last century and accumulated valuable experience and achievements.According to the human drivers' real control process,a series of self-driving vehicle lateral control algorithms based on driver's preview following theory have been put forward,and the robustness and feasibility of these algorithms have been verified by extensive mount of simulation and test.In this paper,the following aspects are addressed:1.The vehicle dynamic model is derived.Aiming at the stable and accurate tracking of desired trajectories,a vehicle location prediction method and vehicle dynamics model,which can characterize vehicle kinematics characteristics,are established.Based on the vehicle kinematics model,the steady state trajectory of the vehicle is deduced,and the method of vehicle location prediction is established.Then combined with the seven degree of freedom vehicle dynamic model and tire model,vehicle's longitudinal dynamics,lateral dynamics and yaw dynamics were analyzed to derive the vehicle dynamic model with longitudinal,lateral,yaw motion and tires' rotation.2.The design of vehicle feedforward-feedback lateral controller.Based on the preview following theory,vehicle feedforward-feedback lateral controller which contains a feedforward controller of road curvature and a feedback controller of vehicle lateral deviation is completed.Compared with the traditional preview following lateral controller,this controller takes expected trajectory's curvature information and the current vehicle position deviation in to account to make sure the self-driving vehicle have better dynamic response performance and lateral control volatility and less time delay.The simulation results show that the controller has good stability and speed adaptability.3.Design of lateral controller based on model predictive control theory.In order to ensure real-time control,a linear time-varying model predictive controller is applied to design the lateral controller of the self-driving vehicle.The derivation of the linear time-varying model predictive control algorithm is completed.Then,combined with tire dynamic characteristics,vehicle dynamic characteristics and vehicle steering system constraints,the state and output of the controller in lateral control are constrained such vehicle and feasibility in lateral control are ensured.