Research On Path Tracking Control Of Autonomous Vehicle

Safety is an eternal topic in the field of automobile.In the "driver vehicle road"closed-loop system,drivers are considered to be the weakest link in the system due to their driving ability,inborn physiological differences,subjective emotions and other aspects,and accidents caused by driver factors account for an important part of road traffic accidents.Automatic driving vehicle is considered as the main direction of future automobile development and technology innovation because it can partially or completely replace the driver's work,reduce the influence of driver's operation factors and probability of error,and improve driving safety,comfort and efficiency.The motion control of the automatic driving vehicle is the core technology of driving safety and comfort.The lateral motion control mainly realizes the tracking of the vehicle's lateral path,that is,controlling the steering wheel angle,making the vehicle drive according to the planned path,and ensuring the safety.stability and comfort of the vehicle.In this paper,the lateral path tracking control of autonomous vehicles is studied,mainly from four aspects:vehicle dynamics modeling,driver steering behavior modeling,driver tracking characteristics analysis,path following control.In terms of vehicle dynamics modeling,the modular and parametric modeling idea is used to benchmark a certain type of vehicle in hanteng,and the 14 DOF vehicle dynamics model of the target vehicle is established.Then,according to the different research purposes,the 14 DOF vehicle model is simplified and designed:1.According to the demand of real-time response of the control algorithm,the 2-DOF vehicle model is established as the prediction model in Chapter 4 of this paper In order to initially reflect the dynamic characteristics of all aspects of the vehicle and realize the rapid verification of the "driver vehicle road" closed-loop system,a 7-DOF dynamic model is established.Finally,through the comparison of real vehicle data and simulation data,the correctness and validity of 14,7,2-dof vehicle model are verified,which lays the foundation for the following algorithm verification;In the aspect of driver's steering behavior modeling,firstly,based on the"preview follow" theory,the coordinate transformation relationship of the desired path,the geometric relationship between the self position and the desired path,the dynamic relationship between acceleration and steering wheel angle are derived,and the "predictor" and "follower" of driver's lateral control model are established;secondly,based on Logitech G29 set up a hardware in the loop simulation platform driver simulation system to collect real driver's data.Finally,combined with the 7-DOF vehicle model established in this paper,the driver's lateral control model and real driver's "driver vehicle road" closed-loop verification are carried out;In the aspect of driver tracking characteristic analysis,the typical double lane moving obstacle avoidance working condition is designed.The path tracking characteristic analysis is carried out through the established driver lateral control model,and the influence of preview time and speed on the path tracking process is clarified.Then,the semi physical simulation platform is used to collect multiple driver steering operation data to simulate skilled driving Driver's path tracking behavior in driver vehicle road closed-loop system.In the aspect of path tracking control,the model predictive control(MPC)algorithm is used to build the path tracking controller.In order to prevent the noise interference of state variables and the mismatch of model parameters in nonlinear area,Kalman filter state observer is designed.Secondly,based on the 2-DOF vehicle model and driver tracking characteristics,a predictive model with steering wheel angle as input and lateral displacement and yaw angle as output is established.After the predictive transformation,the MPC control is finally designed The optimization solution of each step of the controller is transformed into a constrained quadratic programming problem.The steering wheel angle required to track the reference path is solved by using the characteristics of model predictive control rolling optimization,which acts on the 14 DOF vehicle model.Compared with the skilled driver model,the MPC controller is better than the skilled driver,and can adapt to a variety of complex environments.Finally,in the aspect of real vehicle verification,three working conditions are designed,which are straight road keeping experiment,curve tracking experiment and double lane shifting obstacle avoidance.Through the comparative analysis with the data of skilled drivers,we can see that the path tracking controller based on MPC designed in this paper has better performance.