http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
K-city 자율주행 경진대회 참가를 위한 자율주행 플랫폼 개발
김태호(Teaho Kim),김동진(Dongjin Kim),민동규(Dongkyu Min),서현지(Hyeonji Seo),윤호진(Hojin yun),정은빈(Eunbin Jung),이진강(Jingang Lee),권혁재(Hyeokjae Kwon),김진석(Jinseok Kim),김대국(Daekuk Kim),문일주(Iljoo Moon),유정흠(Jeongheum You 대한기계학회 2021 대한기계학회 춘추학술대회 Vol.2021 No.11
In this paper, a ROS-based autonomous driving framework designed by team ACCA from the School of Mechanical Engineering, Soongsil University, for the 2021 international college student creative car challenge held in K-city. The autonomous vehicle’s chassis used in this challenge is equipped with Velodyne 3D lidar, a Sick down-looking 2D lidar, Xsens MTi-30 AHRS, CCD camera, webcam, and PC-based controller. First, before the challenge in K-city, we evaluated the ROS-package-based SLAM such as LIO-SAM in a ring-shaped road environment on the campus of Soongsil University. After the successful SLAN and mapping process, the hdl_localization, which is a 3D lidar-based real-time 3D localization package, is used to estimate the global pose with respect to the global frame using NDT scan matching. For lane detection, traffic sign, and traffic signal recognition, the two well-known DNN models are utilized. Based on experimental results from both simulation and an actual autonomous vehicle platform, the Point Instance Network (PINet) for lane detection shows 88% of test accuracy, and the YOLO V4 for the traffic light and sign recognition offers 95% test accuracy.