Research On Road Perception And Trajectory Planning Algorithms For FSAC

Nowadays,autonomous driving technology has shown immense potential and value in fields such as transportation,logistics,and commercial transport.To ensure the safe and efficient operation of autonomous vehicles,reliable environment perception information and driving paths are crucial.This thesis focuses on the issues of track perception and trajectory planning in the unmanned formula racing car competition,using "Formula Student Autonomous China FSAC" as a platform.The main research contents of this thesis are as follows:(1)Track cone identification based on YOLOv7.This thesis uses the YOLOv7 deep learning framework to train the cone detection network and achieve perception of the cone track in the FSAC.To train a high-performance detection network,a large number of cone labeling datasets were created to train the network model.The trained model is used to complete cone detection,positioning,and sorting tasks,providing a perception basis for track mapping and path planning.(2)Track positioning and mapping based on binocular vision.Firstly,the binocular camera is calibrated using MATLAB to obtain the camera’s intrinsic and extrinsic parameters.Then,the SGM stereo matching algorithm is used to calculate depth information from the left and right camera images.Next,the inertial navigation system is calibrated to obtain the vehicle’s positioning coordinates.Finally,the cone world coordinates and the vehicle’s positioning coordinates are fused to establish a track map.(3)Global reference line generation and trajectory planning for high-speed track following.Firstly,based on the track map,the Delaunay triangulation algorithm is used to solve the center points of the track,and then the cubic spline interpolation is used to fit the centerline of the track.The global reference line is obtained by smoothing the driving path of the vehicle using the cubic spline interpolation,and the Frenet coordinate system is used for high-speed track following trajectory planning,including steps such as initial state calculation,lateral and longitudinal state sampling,and optimal trajectory generation.Finally,the feasibility and effectiveness of the algorithm are verified through simulation.(4)Real car verification of track mapping and trajectory planning.Firstly,an unmanned formula racing car experimental platform is established,describing the overall hardware architecture and the entire vehicle control system.Then,the track mapping and trajectory planning algorithms are designed using ROS.Finally,the feasibility of the unmanned formula racing car’s track perception and trajectory planning system is verified through real-car experiments.