Lane Detection And Trajectory Following Control Of Driverless Vehicle Based On Model Prediction Algorithm

With the development of computer technology and sensor technology,as well as the increasing demand for intelligentialize in nowadays,people's demand for intelligent cars going rapidly.However,an accurate and stable trajectory tracking control system is necessary for the unmanned vehicle.The research object of this thesis is unmanned vehicles equipped with camera,and the theme is how to detect the lane by image processing and follow the desired trajectory precisely through front steering in different scenarios.In this research,it manly involves the following aspects:(1)Lane detection,based on varies sensors around host vehicle to detect lane marker,extract information and process information,then the prediction lane model is built.(2)Decision making and path planning: replacing driver to make the safest decision and planning corresponding path after receiving surrounding information.(3)Trajectory tracking: Propose 3 DOF vehicle dynamic model,Magic formula tire model is applied to describe the linear constraint region.Designing and applying a linear time-varying model prediction controller to meet the demand of trajectory tracking through controlling front wheels.Firstly,Udacity driverless road data were first introduced in this paper,and MATLAB image processing tools are applied on the lane line identification,including RGB to gray processing,image enhancement processing,dynamic region of interest extraction,inverse perspective map transformation,hough straight line detection and other image processing technologies,and then relative relevant lane shape information is collected.Secondly,a vehicle dynamic model in three dimensions of freedom and its magic formula tire model are proposed,and deducing state equations by combine them.The nonlinear system is converted to the linear time-varying system considering the real-time requirement of the trajectory tracking control algorithm.Then the LTV MPC formula is derived,as well as related prediction model constraint conditions.The LTV MPC optimal problem is converted to the standard quadratic programming(QP)problem that is easy to calculate in computer.Finally,built Simulink/Carsim co-simulation platform,then trajectory tracking control algorithm was verified by tracking the lane that is produced in the image process detected.The simulation results show good tracking accuracy,while the performance of the controller could not be measured in this condition.Therefore,the double lane change trajectory,a more complicated working condition,is designed and the trajectory tracking simulation is carried out.These simulation results show that vehicle trajectory tracking accuracy and vehicle stability are influenced by the speed and road friction coefficient.For the sake of improving control precision and stability of system,a fuzzy controller was carried out,which acts on the front wheel angle for compensation.The result dedicate 43.21% less position error and smaller 82.5% root-mean-square error than the result before optimization.Due to the obstacles exist in the coming lane or lane changing demand,this article in view of the working condition of target lane design and related path planning strategy,then combine with model predictive controller for trajectory tracking simulation.Simulation results show this controller has good robustness and adaptability.