| With the development of society and the progress of science and technology,people hope that vehicles are not only a basic tool for walking,but also to develop in a more intelligent way to meet people’s needs for comfort and ease of operation.This promotes the development of advanced driving assistance system and driverless vehicle to a certain extent.The main research contents of driverless vehicle include environment perception,positioning and navigation,path planning,motion control,integrated design and so on.The research content of this paper involves two aspects of lane detection and lane keeping control,and the purpose of the research is to make the vehicle drive more safely and accurately along the lane centerline.In the aspect of lane detection,several classical semantic segmentation networks are compared and analyzed.From the specific needs of lane detection,the lightweight network e Net is selected as the basic network of lane segmentation.In order to improve the ability of e Net network to extract spatial context information from images and improve its robustness in complex road environment,this paper improves the e Net network combined with spatial CNN network structure.After the segmentation result of lane line is obtained,it is converted to aerial view by using inverse perspective transformation,and then the lane line is fitted by using least square method.Finally,the improved e Net convolutional neural network model is compared with the traditional threshold segmentation method in tusimple data set.The test results show that the accuracy of lane line segmentation of the improved model is significantly improved.At the same time,the improved e Net convolutional neural network model is compared with the traditional threshold segmentation method.The results show that the convolutional neural network lane line detection algorithm can adapt to the detection task in complex environment,Better robustness.In the aspect of lane keeping control,firstly,based on the vehicle tire model,a nonlinear 3-DOF vehicle dynamic model is established as the controlled object of the controller;Based on the model predictive control theory,a lane keeping controller is developed.In order to improve the real-time performance of the controller,the vehicle model is linearized and discretized,and the corresponding constraints are added.After constructing an appropriate objective function,the objective function is transformed into a quasi quadratic programming problem to complete the controller design.Finally,a co simulation environment of Car Sim and Simulink is built to verify the path following performance of the controller.At the same time,the MPC controller is compared with the preview follow optimal controller in different working conditions.The comparison results show that the controller has better control effect and better path tracking performance.In order to improve the robustness of the controller,the original MPC controller is improved.The parameters of the common speed range are matched.The parameter selection module of the controller is integrated into the lane keeping controller,and the parameter adaptive MPC controller is designed.The simulation results show that the parameter adaptive lane keeping controller has better tracking accuracy and driving stability when the vehicle speed changes. |
