Comparative Study on Driving Environment Complexity Based on Open Autonomous Driving Datasets

Miyoung Bhin; Seulki Son

doi:10.7470/jkst.2026.44.5.820

All Issue

2026 Vol.44, Issue 5 Preview Page Next Page

Article

Comparative Study on Driving Environment Complexity Based on Open Autonomous Driving Datasets 공개 자율주행 데이터셋 기반 주행환경 복잡도 비교 연구

31 October 2026. pp. 820-831

PDF XML

Abstract

Quantitative assessment of driving environment complexity is essential for the safe expansion of the Operational Design Domain (ODD) of Automated Driving Systems (ADS). This study develops a Driving Environment Complexity (DEC) indicator system consisting of Objects Per Frame (OPF), Traffic Composition Entropy (TCE), Composite Complexity Index (CCI), and Vulnerable Road User ratio (VRU%), and compares six open autonomous driving datasets by directly parsing their bounding-box annotations: Waymo (US), ONCE (CN), nuScenes (SG), AI Hub (KR), GGData (KR), and 42dot (KR). Dataset heterogeneity was explicitly considered, and direct comparisons were restricted to datasets collected under identical conditions (10 Hz, 3D LiDAR). The OPF ranking was Waymo (US) (48.26) > ONCE (CN) (29.16) > GGData (KR) (12.77) > AI Hub (KR) (3.34), with statistical significance confirmed by bootstrap confidence intervals (n = 10,000) that were non-overlapping except for a slight (2.4%) overlap between Waymo (US) and ONCE (CN). Applying the ISO 34503:2023 ODD taxonomy revealed that VRU exposure contributes more strongly to driving environment complexity than road type, and the empirical results confirmed that a higher OPF does not necessarily indicate a higher Time-To-Collision (TTC) critical rate. The lower complexity observed in domestic datasets reflects limited real-world driving experience in complex mixed-traffic environments rather than differences in sensing technology. These findings provide empirical evidence supporting the phased expansion of autonomous driving ODD into VRU-mixed urban environments in South Korea.

Keywords

autonomous driving dataset

driving environment complexity

operational design domain

surrogate safety measure

자율주행 시스템(Automated Driving System, ADS)의 안전한 운행설계영역(Operational Design Domain, ODD) 확장을 위해서는 주행환경 복잡도의 정량적 평가가 필수적이다. 본 연구는 Waymo(US), ONCE(CN), nuScenes(SG), AI Hub(KR), GGData(KR), 42dot(KR)의 6개 공개 자율주행 데이터셋의 바운딩박스 어노테이션을 직접 파싱하여 주행환경 복잡도(Driving Environment Complexity, DEC) 지표 체계인 프레임당 객체 수(Objects Per Frame, OPF), 교통 참여자 구성 다양성(Traffic Composition Entropy, TCE), 복합복잡도지수(Composite Complexity Index, CCI), 취약 도로 이용자 비율(Vulnerable Road User ratio, VRU%)을 설계하고 비교 분석하였다. 데이터셋 간 이질성을 명시적으로 고려하여 동일 조건(10Hz, 3D LiDAR)에서 수집된 데이터셋 간 직접 비교를 원칙으로 하였다. OPF 순위는 Waymo(US)(48.26) > ONCE(CN)(29.16) > GGData(KR)(12.77) > AI Hub(KR)(3.34) 순으로 Waymo(US)와 ONCE(CN) 간 신뢰구간의 소폭 중첩을 제외하면 통계적으로 유의하게 확인되었으며, ISO 34503:2023 ODD 분류체계 적용 결과 도로 유형보다 취약 도로 이용자(Vulnerable Road User, VRU) 노출이 주행환경 복잡도를 더욱 강하게 결정하는 것으로 나타났다. Waymo(US)와 nuScenes(SG)의 실측 결과, OPF가 높다고 해서 충돌까지의 시간(Time-To-Collision, TTC) 위험률이 반드시 높아지는 것은 아님을 확인하였다. 국내 데이터셋의 낮은 복잡도는 센싱 기술의 차이가 아니라 복잡한 혼합교통 환경에서의 실주행 경험 부족을 의미하며, 본 연구 결과는 취약 도로 이용자(VRU) 혼합 구간으로의 운행설계영역(ODD) 단계적 확장을 위한 실증적 근거로 활용될 수 있다.

키워드

자율주행 데이터셋

주행환경 복잡도

운행설계영역

대리안전지표

References

Arun A., Haque M. M., Bhaskar A., Washington S., Sayed T. (2021), A Systematic Mapping Review of Surrogate Safety Assessment Using Traffic Conflict Techniques, Accid. Anal. Prev., 153, Elsevier, 106016.

10.1016/j.aap.2021.106016

Caesar H., Bankiti V., Lang A. H., Vora S., Liong V. E., Xu Q. et al. (2020), nuScenes: A Multimodal Dataset for Autonomous Driving, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, 11618-11628.

10.1109/CVPR42600.2020.01164

Das T., Samandar M. S., Autry M. K., Rouphail N. M. (2023), Surrogate Safety Measures: Review and Assessment in Real-World Mixed Traditional and Autonomous Vehicle Platoons, IEEE Access, 11, IEEE, 32682-32696.

10.1109/ACCESS.2023.3248628

Efron B., Tibshirani R. J. (1993), An Introduction to the Bootstrap, Chapman & Hall (New York, USA), 436.

10.1007/978-1-4899-4541-9

EPC (2019), Regulation (EU) 2019/2144 on Type-Approval Requirements for Motor Vehicles (General Safety Regulation), Article 3(21).

42dot Inc. (2022), MCMOT: Multi-Camera Multi-Object Tracking Dataset [Data set], 42dot AI Kit, https://42dot.ai/akit/dataset/, 2026.03.15.

Gyeonggi Data Dream (2025), Zero-Shuttle Vehicle Sensor Data [Data set], https://data.gg.go.kr/portal/data/service/selectServicePage.do?infId=CN5XBMCX5O7V31WD30UT31907399&infSeq=1, 2026.03.15.

Huang P., Ding H., Chen H. (2024), An Entropy-Based Model for Quantifying Multi-Dimensional Traffic Scenario Complexity, IET Intell. Transp. Syst., 18(7), IET, 1289-1305.

10.1049/itr2.12510

International Organization for Standardization (ISO) (2023), ISO 34503:2023 Road Vehicles – Taxonomy for Operational Design Domain for Automated Driving Systems.

Mao J., Niu M., Jiang C., Liang H., Chen J., Liang X. et al. (2021), One Million Scenes for Autonomous Driving: ONCE Dataset, Advances in Neural Information Processing Systems 34 (NeurIPS 2021), Datasets and Benchmarks Track.

Minderhoud M. M., Bovy P. H. L. (2001), Extended Time-to-Collision Measures for Road Traffic Safety Assessment, Accid. Anal. Prev., 33(1), Elsevier, 89-97.

10.1016/S0001-4575(00)00019-1

National Information Society Agency (NIA) (2022), Passenger Autonomous Vehicle Daytime Urban Road Data v2 (Dataset No. 71621) [Data set], AI Hub, https://aihub.or.kr/aihubdata/data/view.do?dataSetSn=71621, 2026.03.15.

Shannon C. E. (1948), A Mathematical Theory of Communication, Bell Syst. Tech. J., 27(3), American Telephone and Telegraph Co., 379-423.

10.1002/j.1538-7305.1948.tb01338.x

Sun P., Kretzschmar H., Dotiwalla X., Chouard A., Patnaik V., Tsui P. et al. (2020), Scalability in Perception for Autonomous Driving: Waymo Open Dataset, Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, 2446-2454.

10.1109/CVPR42600.2020.00252

Yu R., Zheng Y., Qu X. (2021), Dynamic Driving Environment Complexity Quantification Method and Its Verification, Transp. Res. Part C Emerg. Technol., 127, Elsevier, 103051.

10.1016/j.trc.2021.103051

Information

Publisher :Korean Society of Transportation
Publisher(Ko) :대한교통학회
Journal Title :Journal of Korean Society of Transportation
Journal Title(Ko) :대한교통학회지
Volume : 44
No :5
Pages :820-831
Received Date : 2026-04-15
Revised Date : 2026-05-12
Accepted Date : 2026-07-22
DOI :https://doi.org/10.7470/jkst.2026.44.5.820

Journal of Korean Society of Transportation ISSN:1229-1366(Print) 2234-4217(Online) 대한교통학회지

All Issue