http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
On-Chip Debug를 이용한 Windows 환경의 Time-critical S/W Fault Injection Test
김기철(Kicheol Kim),박민기(Minki Park),성기태(Gitae Seong),이진환(Jinhwan Lee) 한국자동차공학회 2020 한국자동차공학회 부문종합 학술대회 Vol.2020 No.7
Automobile’s convenient services and functions are increased recently. That cause increase of Software(S/W) size and S/W complexity. To keep safety and reliability from S/W fault, Fault Injection(FI) is used in S/W testing. Fault Injection Test(FIT) is performed with defect-caused conditions. However, in Automotive systems, Fault injection test is difficult to keep injection continuously and set fault conditions on intended timing. In this paper, Firstly, we address fault injection test method environment for AUTomotive Open System Architecture(AUTOSAR) platform. To keep fault injection, we proposed On-Chip Debug(OCD) in updating or overwriting fault values. Then, we address time-critical test methods and compare timing difference and accuracy between C/C++ time-measurement functions in Windows environment.
머신 러닝 기반 엔지니어링 메트릭을 고려한 차량용 임베디드 소프트웨어 결함 예측에 관한 실험적 연구
김대성(Daesung Kim),이성훈(Sunghoon Lee),성기태(Gitae Seong),김대성(Daesung Kim),이진환(Jinhwan Lee),배홍용(Hongyong Bhae) 한국자동차공학회 2020 한국자동차공학회 학술대회 및 전시회 Vol.2020 No.11
The complexity of automotive software has rapidly been increased with vehicular intelligence. Software fault prediction (SFP) can be helpful for improving the software quality and efficiently managing limited testing resources with early identification of faulty module. Many papers have been proposed to predict the potential software fault. Most of them are only considering source code attributes, as a result, it has been remained a challenge for industrial area to adopt which method is fit, robust, and provide most accurate model. In this paper, the framework is presented for SFP model design of the automotive steering system software obtained by mining historical repositories. For best performance model, we have additionally chosen engineering features like difficult level of requirements, competence of engineer, complexity of ports, test coverage with code metrics. Our model is evaluated with AUROC which is a performance measurement for classification problem. The Random Forest Classifier shows the best performance with 92.31%.