Utilizing Modeling Approach Patterns in the Embedded Software Industry

Abstract

To cope with the growing complexity of softwareintensive embedded system development, modeling is a widely used approach. The modeling approaches in the embedded software industry vary depending on many modeling characteristics (e.g., purpose, modeling rigor, medium type used, modeling stakeholder profile, etc.). In the light of our previous studies, we have identified modeling approach patterns of embedded software development projects and constructed a characterization model. This model not only identifies and defines modeling approach patterns of the stakeholder in embedded software development projects, but also gives recommendations for commonsense modeling practices. In this article, one of the application results of this characterization model, which was performed in Defense & Aerospace sector is presented. The model was successfully applied to the case, in which a series of both structured and semi-structured interviews with 17 embedded software professionals were conducted. The results show that identification of individual patterns provide insight for improvement both for individuals as well as larger units of operations such as projects and organizations.

Towards Modeling Patterns for Embedded Software Industry: Feedback from the Field

“…53 interviews in 14 companies within different target sectors had been carried out. Interviewees represented about different SE roles and academic background [10]. They have, cumulatively, 756 years of software development experience.…”

“…During this process, grouping 11 main characteristics on survey data, which includes ~80 attributes would generate lots of combinations [9]. In order to eliminate unnecessary combinations, we reduced the number of attributes by generating derived attributes after determining the most critical ones [10]. After visualizing these groups (e.g., with scatter and bars stacked charts), we derived the necessary characteristics, which have critical importance on the categorization: "purpose", "medium type", "archivability", "modeling language, if any" and "SDLC phase" [10].…”

“…The semi-structured interviews were conducted mostly in face-to-face, but if it is inconvenient, on Skype as in the case of intercontinental interviews with different industrial sectors, roles, experiences and practices in embedded software industry [10]. All interviewees were promised that only anonymous data would be presented and the interviewer would take notes on what he spontaneously found relevant for analysis.…”

“…All interviewees were promised that only anonymous data would be presented and the interviewer would take notes on what he spontaneously found relevant for analysis. During the interview, there was a clear and complete list of general topics (i.e., interview instrument), which cover about modeling usage patterns besides their success & failure stories, the attitudes to the adoption of modeling and so on [10]. The interviewer asked the set of questions, listened to the answers and followed up answers with additional questions when necessary [16].…”

“…), the interviewer tried to understand their modeling practices. Then, by giving the necessary terminology on MDE, MBE and sketching, the second key/critical question about model-driven usage was asked [10]. Since we previously found out that modeling purpose directly affects the modeling rigor (hence modeling approach) [6], the interviewer also asked "Is there any listed purpose while you are modeling?"…”