Autonomous Environmental Service Vehicles (AESVs) require safety verification that is distinct from passenger AVs because of their dual-task nature. This study proposes a dual-perspective ODD risk assessment framework using three months of field data (Oct–Dec 2025) from an AESV deployed in Jeju City, South Korea. The PCA-weighted Enhanced CII quantifies the internal mechanical interference, whereas entropy-weighted risk scores combined with k-means clustering identify external hazard zones. The CII declined from 1.25 to 1.12, confirming progressive adaptation of compensatory control, whereas spatial mapping revealed persistent hotspots at intersections, sharp curves, and commercial districts, indicating that ODD boundaries are governed by road geometry and land use.

ABSTRACT
1. 서론
2. 이론적 배경 및 선행 연구 고찰
    2.1. 운행설계영역 프레임워크
    2.2. 차량 이탈 분석 및 현장 데이터 기반 안전성 평가
    2.3. 도시환경관리 자율주행 차량
3. 연구 방법론
    3.1. 데이터 수집 및 운행 맥락
    3.2. 내부 안정성 평가: Enhanced CII
    3.3. 외부 위험 매핑: ODD Risk Score
    3.4. 통계 검정 방법
4. 분석 결과 및 고찰
    4.1. 내부 안정성 평가 결과
    4.2. 공간 위험 매핑 결과
    4.3. 거시적 운행 추이
5. 결론 및 향후 연구과제
REFERENCES

• 이은호(아주대학교 일반대학원 D.N.A.플러스융합학과 석사과정) | Lee Eunho
• 홍지호(아주대학교 일반대학원 D.N.A.플러스융합학과 박사과정) | Hong Jiho
• 김주영(아주대학교 일반대학원 D.N.A.플러스융합학과 석 · 박사통합과정) | Kim Juyeong
• 고준혁(아주대학교 공과대학 교통시스템공학과 학사과정) | Koh Junhyeok
• 소재현(아주대학교 공과대학 교통시스템공학과 부교수) | So Jaehyun (Jason) Corresponding author