http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
A Research on Fail-Safe System by Watch Dog for Multi Sensor Fused Autonomous Vehicle
Sunyoung Park(박선영),Hyunjun Gil(길현준),Sangjun Kim(김상준),Juwon Seo(서주원),Yunjung Choi(최윤중),Hyunwoo Kim(김현우),Jungha Kim(김정하) 한국자동차공학회 2022 한국자동차공학회 학술대회 및 전시회 Vol.2022 No.11
In this study, the Watch Dog monitoring system was designed by installing ROS Melodic based on Ubuntu 18.04. Watch Dog is a system that detects programatic errors in autonomous driving controllers and manages sensors such as LiDAR, Camera, GPS, and Imu in real time. First, Watch Dog checks the sensor for power and lan connections by default. It also analyzes the reliability of the data obtained from NDT and GPS and the speed difference between GPS and encoder to determine whether localization and mapping are going well. Visual monitoring is performed to recognize the part of the problem in the event of an error, and the control command is passed on to the control PC after determining whether it is possible to drive.
Graph SLAM을 통한 산악형 도심지역의 고정밀지도 작성 및 위치 인식 시스템에 대한 연구
김현우(Hyunwoo Kim),최윤중(YunJung Choi),강동완(DongWan Kang),김상준(SangJun Kim),길현준(HyunJun Gil),서주원(JuWon Seo),박선영(SunYoung Park),김재일(JaeIl Kim),임효진(HyoJin Lim),김정하(JungHa Kim) 한국자동차공학회 2022 한국자동차공학회 학술대회 및 전시회 Vol.2022 No.11
This paper suggests a localization system architecture of unmanned vehicle for autonomous driving. In this system, Graph SLAM algorithm is used for correction of accumulated errors acquired from scan matching algorithm. On this paper, the experiments are proceeded in environment of mountainous city which have continuous elevations. The acceleration and magnetic data from IMU, RTK corrected GNSS are used for graph optimization and loop closure on mapping. NDT scan matching algorithm is used for localization.
라이다-카메라 캘리브레이션을 통한 동적 장애물 회피를 위한 자율주행 인지시스템에 대한 연구
김현우(Hyunwoo Kim),최윤중(YunJung Choi),김상준(SangJun Kim),길현준(HyunJun Gil),서주원(JuWon Seo),박선영(SunYoung Park),김재일(Jaell Kim),임효진(HyoJin Lim),김정하(JungHa Kim) 한국자동차공학회 2022 한국자동차공학회 학술대회 및 전시회 Vol.2022 No.11
Sensor fusion area are known to great solution for the edge cases of perception system of automated car. This system uses camera and LiDAR on front side of the vehicle and integrated on a coordinate system for the projection. Each point data on image pixel integrated point cloud gets its class by semantic segmentation. This paper suggests a perception system of autonomous vehicle for moving objects by deep learning based image semantic segmentation and lidar points projection to camera pixel image.
자율주행 시스템의 Pure-pursuit 알고리즘을 이용한 Path Management 시스템 개발
최윤중(YunJung Choi),김현우(Hyunwoo Kim),신희석(HeeSeok Shin),김명준(MyeongJun Kim),김상준(SangJun Kim),길현준(HyunJun Gil),서주원(JuWon Seo),박선영(SunYoung Park),김재일(Jaell Kim),임효진(HyoJin Lim),김정하(JungHa Kim) 한국자동차공학회 2022 한국자동차공학회 학술대회 및 전시회 Vol.2022 No.11
Research on autonomous vehicles consists of perception, judgement, and control. This paper is about the Ld method of the Pure-pursuit algorithm among the control algorithms based on the kinematic method in the control field. In the process of finding the correct L<SUB>d</SUB> for actual control, an error occurred between the detected L<SUB>d</SUB> value and the correct L<SUB>d</SUB>, and only a part of the path was detected L<SUB>d</SUB>. In order to store only the coordinates to be detected by LD in the path reference, the interval between paths was set very small, and the average interval and error of the detected LD values were calculated. The average interval and error of the detected LD value was proportional to the distance the sensor moved during the period of receiving a new signal, and accordingly, the path interval and path error were set appropriately. In the existing algorithm, the optimal interval and error are determined using the experimental data of the vehicle. The algorithm in this study may be less accurate than the existing algorithm. However, using only the target speed, the period of the sensor, and the minimum turning radius of the vehicle, it is possible to set the distance and the error of the path with high accuracy. Finally, the verification was carried out by directly controlling the autonomous vehicle.