Vision System And Lateral Control System For Autonomous Vehicle Considering Nonlinearity And Uncertainty

Increasing traffic accidents,traffic jam and environmental pollution motivate the development of intelligent transportation system.Autonomous vehicles(AVs),with the improved security and better road utilization,have become an emerging research focus.Basically AVs control can be divided into two types:lateral control and longitudinal control.One of the most rudimentary issues for AVs is lateral control,whose control objective is to converge the path tracking errors(i.e.,the lateral offset and heading error)to zero.However,it is worth attention that AVs' researches are only based on the traditional front-wheel steering(FWS)vehicles and put in moderate driving conditions,i.e.,in low-speed driving or around low-curvature road.In these condition,the vehicle tires are in the linear region.So,AVs control system should consider the nonlinearity of tire in the extreme condition.Sideslip angle can affect the tracking.Also,the line detection arithmetic has a good robustness and accuracy.In this research,Vision system and lateral control system are studied based on the SEUAV-1 which is taken as the platform for experimental research and developed by the School of Mechanical of Southeast University.The research work of this paper consists of the following major parts:(1)Establish the SEUAV-1 experience platform and discuss the system composition.Introduce the environmental awareness system,information fusion system,planning and decision system and control system of SEUAV-1.Set the design targets of the AVs lateral control system according the characteristic of AVs.(2)Ensure the way of line demarcate of vision system.Considering robustness and accuracy of line detection,we design the algorithm of single line detection and multi-line detection using image gray processing,image filtering,image edge enhancement,image binaryzation,Canny operator,Hough transform and kalman filter.The algorithm can lay the foundation of control system.(3)Kinematic model,three DOFs vehicle dynamic model and magic tire model,two DOFs AVs lateral dynamic model and path tracking system model are established for the autonomous vehicle.The co-simulation between Simulink model and CarSim model is conducted and shows the similar characteristics of dynamics.(4)An analysis of steady state error point that 4WS AVs can eliminate steady state error in ideal condition but traditional Front-Wheel steering(FWS)AVs cannot achieve,because the sideslip angel which can affect the steady state error is existed in extreme condition,inevitably.By introducing sideslip angle,two sliding mode controllers are utilized to converge lateral offset and heading error to zero and keep vehicle stability in extreme condition.(5)Considering sensor and actuator nonlinearity,and the parameter-varying property of tire cornering stiffness in extreme handling situations,a Takagi-Sugeno(T-S)fuzzy model is established to represent the nonlinear characteristics of tires and control input nonlinearity.A robust H? adaptive sliding mode controller is designed to achieve the path tracking and vehicle lateral control simultaneously.The condition of H ? control constraint can be converted to the convex optimization problem.Two simulation cases are presented with a high-fidelity and full-car model based on the CarSim-Simulink joint platform,and the results confirm the effectiveness of the proposed control approach.