Enhancing Road Anomaly Detection in Autonomous Vehicles through Cooperative Perception = 협력적 인식을 통한 자율주행차의 도로 이상 감지 향상

다국어 초록 (Multilingual Abstract)

The proposed system introduces a dual-mode architecture that integrates direct Vehicle- to-Vehicle (V2V) communication with Server-assisted Vehicle-to-Infrastructure (V2I) align- ment. Unlike traditional approaches that rely on noise-prone GPS localization, the V2I module utilizes a pose-free visual alignment technique based on deep feature matching to accurately register anomalies across different viewpoints. Simultaneously, the V2V mod- ule facilitates rapid hazard propagation by sharing lightweight metadata, ensuring early warning capabilities even when visual confirmation is unavailable to the following vehi- cle. Experimental validation was conducted in a custom CARLA simulation environment populated with realistic road anomalies. The results confirm that the proposed framework effectively overcomes the limitations of ego-only perception. The system demonstrates a robust trade-off between the high precision provided by server-side alignment and the low- latency efficiency of direct vehicle communication. Consequently, this study establishes that cooperative perception significantly improves detection robustness and safety margins, of- fering a scalable solution for autonomous navigation in occlusion-prone environments.

목차 (Table of Contents)

• Table of Contents iv
• List of Figures vi
• List of Tables viii
• List of Abbreviations ix
• Abstract x
• 1 Introduction 1
• 1.1 Overview 1
• 1.2 Statement of the problem 4
• 1.3 Research contribution 5
• 1.4 Thesis organization 5
• 2 Related Works 7
• 2.1 Computer Vision Approaches to Road Anomaly Detection 7
• 2.2 Cooperative Perception Frameworks (V2V, V2I, V2X) 9
• 2.3 Multi-View Alignment and Feature Matching 12
• 2.4 Simulation Platforms for Autonomous Vehicle Research 14
• 2.5 Cooperative Perception Under Communication Constraints 16
• 2.6 Research Gap Analysis 18
• 3 Proposed System 21
• 3.1 System description 21
• 3.2 Cooperative Perception Module 25
• 3.3 Performance Evaluation Criteria 47
• 4 Development of a Simulator for Cooperative Perception 50
• 4.1 CARLA Modification 50
• 4.2 Custom Map Development 51
• 5 Experiments and Results 57
• 5.1 Experimental Setup 57
• 5.2 Experiments 61
• 5.3 Results 64
• 5.4 Numerical Results 77
• 6 Conclusion 84
• References 87
• Abstract in Korean 96