ETRI-Knowledge Sharing Plaform



논문 검색
구분 SCI
연도 ~ 키워드


학술지 A Formal and Quantifiable Log Analysis Framework for Test Driving of Autonomous Vehicles
Cited 3 time in scopus Download 125 time Share share facebook twitter linkedin kakaostory
성경복, 민경욱, 최정단, 김병철
Sensors, v.20 no.5, pp.1-29
20HS2300, 고정밀 맵 음영 환경의 완전자율주행 네비게이션 인공지능 기술개발, 민경욱
We propose a log analysis framework for test driving of autonomous vehicles. The log of a vehicle is a fundamental source to detect and analyze events during driving. A set of dumped logs are, however, usually mixed and fragmented since they are generated concurrently by a number of modules such as sensors, actuators and programs. This makes it hard to analyze them to discover latent errors that could occur due to complex chain reactions among those modules. Our framework provides a logging architecture based on formal specifications, which hierarchically organizes them to find out a priori relationships between them. Then, algorithmic or implementation errors can be detected by examining a posteriori relationships. However, a test in a situation of certain parameters, so called an oracle test, does not necessarily trigger latent violations of the relationships. In our framework, this is remedied by adopting metamorphic testing to quantitatively verify the formal specification. As a working proof, we define three metamorphic relations critical for testing autonomous vehicles and verify them in a quantitative manner based on our logging system.
KSP 제안 키워드
Autonomous vehicle, Detect and analyze, Formal specifications, Metamorphic relations, Metamorphic testing, analysis framework, chain reaction, log analysis, logging architecture
본 저작물은 크리에이티브 커먼즈 저작자 표시 (CC BY) 조건에 따라 이용할 수 있습니다.
저작자 표시 (CC BY)