2021 Fiscal Year Research-status Report
Surrogate Model-based Integration Testing of CPS with Complex Black Box Components
| Project/Area Number |
20K23334
|
|---|
| Research Institution | National Institute of Informatics |
|---|
Principal Investigator |
KLIKOVITS Stefan 国立情報学研究所, アーキテクチャ科学研究系, 特任研究員 (10875983)
|
|---|
| Project Period (FY) |
2020-09-11 – 2023-03-31
|
| Keywords | Automated driving / Search-based testing / Genetic algorithms / noisy systems / scenario generation |
|---|
| Outline of Annual Research Achievements |
My goals for FY2021 were to use my understanding of ADSs and select a component for surrogate-based testing. Specifically, I selected the Rule Engine component, which is the core of the behaviour planning of the ADS. I am currently in the process of creating a surrogate that I can later use for scenario generation.
The results are promising and I have regular fruitful discussions with the developer team in Waterloo, Canada to receive feedback on my plans and valuable suggestions and insights.
Inspired by my ADS work, I established a line of research on kNN-Averaging for Multi-Objective Optimization. This work was published at QUATIC 2021 and PRDC 2021 and is currently being prepared for journal publication. Furthermore, I presented the concept of multi-level scenario modelling at the MPM4CPS 2021 workshop.
Finally, with my colleagues at NII, I participated in the CommonRoad competition, where I learned about motion planning for ADS and the SBST2022 competition, where our tool again was among the most successful ones. These participations also allowed me to gain valuable insight and better knowledge on ADSs and the problems, hopefully leading to much deeper research in the ADS domain.
|
| Current Status of Research Progress |
Current Status of Research Progress
3: Progress in research has been slightly delayed.
Reason
Due to the continued COVID-situation, I could again not travel to Canada. Nonetheless, our regular collaboration meetings helped me gain a good understanding of the systems, even though the timezone difference slows down collaboration.
Despite these difficulties, I selected the Rule Engine component for the research project and started development of a method for surrogate model extraction. The basic structure is set up and I am busy working on the methodology. The regular collaboration meetings with the University of Waterloo help me refine my ideas and gather feedback from knowledgeable insiders.
|
| Strategy for Future Research Activity |
As my project is slightly delayed, the plan for the future remains similar. My goal is to develop an automated technique for surrogate model extraction and the scenario generation based on this surrogate model. My goal is to use the extension period in FY2022 to first wrap up the research and compile a technical report on the findings. Furthermore, I would like to submit a paper to a respectable conference or journal for publication.
|
| Causes of Carryover |
As I am becoming more proficient with the research, I believe to make more progress on the development side. Thus, after a little more development, I will use most of the money for computing infrastructure (e.g. AWS computing power).
Furthermore, I plan to visit conferences to learn more about techniques to efficently perform the computation that I am researching.
|
