범퍼 장착형 Dual-LiDAR 시스템에서 저상 및 제한된 FOV를 고려한 SLAM 및 정밀 위치추정 = SLAM and Localization considering Low-Mounted and Limited-FoV Constraints in Bumper-Mounted Dual-LiDAR System

다국어 초록 (Multilingual Abstract)

This study proposes a bumper-mounted LiDAR sensor system (BMLSS) to achieve 3D environmental perception and precise localization for autonomous vehicles operating under low-height sensing conditions. Conventional roof-mounted LiDAR configurations can suffer from near-range observation gaps, and their installation is often constrained by vehicle structures, design requirements, and aerodynamic considerations. To address these issues, the proposed system leverages geometric information from low-height, limited-FOV point clouds acquired at the vehicle bumper, aiming to realize SLAM and robust real-time localization on a real vehicle platform. The proposed framework combines SLAM-based local estimation with map-matching-based global correction to validate the applicability of the bumper-mounted configuration. To improve data quality and mitigate ground-induced interference, it performs inter-sensor spatial calibration and time synchronization, and applies preprocessing steps including ROI filtering, statistical outlier removal (SOR), and radius outlier removal (ROR), together with a preprocessing-driven targetless calibration procedure. In addition, an SLAM algorithm tailored to the bumper LiDAR setup is introduced, and an IMU–EKF-based sensor fusion module is implemented to enhance robustness beyond LiDAR-only estimation, resulting in a real-time localization system based on a 3D point-cloud map. Experimental results on a real autonomous vehicle platform show that the proposed system maintains map alignment errors within 0.05 m in the SLAM stage and achieves stable mapping and real-time localization in complex urban and curved-road environments. Moreover, map-based real-time localization achieves positioning accuracy within 1 m, demonstrating the feasibility of a cost-effective bumper-mounted LiDAR perception and localization architecture for low-speed urban driving, automated parking, and operation in narrow spaces.

목차 (Table of Contents)

• Ⅰ. 서론 1
• 1.1. 연구 배경 1
• 1.2 연구 목표 및 범위 7
• Ⅱ. 자율주행 LiDAR 인식 시스템 10
• 2.1 자율주행 인식·위치추정 시스템 개요 10
• 2.2 LiDAR 기반 SLAM 및 정밀 위치추정 15
• 2.3 저상·제약 시야(FOV) 등 제약 점군 환경에서의 인식 19
• Ⅲ. 저상·제약 시야 조건을 고려한 범퍼 LiDAR 통합 점군 기반 SLAM·정밀 위치추정 방법 24
• 3.1 점군 매칭 구조를 이용한 타깃리스 LiDAR 교정 및 통합 점군 생성 24
• 3.2 통합 점군 입력을 통한 저상·제약 FOV에서 특징 기반 SLAM 35
• 3.3 LiDAR odometry-Map 2단계 점군 매칭 기반 실시간 정밀 위치추정 47
• Ⅳ. 실차 기반 성능 검증 및 실험 60
• 4.1 실차 실험 플랫폼 및 데이터 구성 60
• 4.2 범퍼 Dual-LiDAR 통합 점군 생성 성능 검증 및 분석 70
• 4.3 저상·제약 FOV 점군 특징 추출 기반 SLAM 성능 및 지도생성 평가 79
• 4.4 2단계 매칭을 통한 저상·제약 FOV 점군에서 위치추정 성능 평가 87
• Ⅴ. 결론 98
• 참고문헌 101