Model-based Systems Engineering (MBSE) is a promising approach to handle the increasing complexity of embedded systems through the use of models that can be analyzed for early design errors detection. A critical phase of systems engineering is Requirements Engineering (RE), which occurs at the very beginning of the development cycle and aims at stating precisely the problem that a system is intended to solve. Requirements errors are numerous and persistent, and are often the most dangerous and expensive errors. It is well known that a clear statement of the problem is an essential step towards a correct solution.
Coupled with the requirements is a system design specification, which states a solution to the problem formulated by the requirements. Many Architecture Description Languages (ADLs) have been proposed for modeling system designs, providing support for predicting system Non-Functional Properties (NFP) such as resource consumption, schedulability, availability, etc. However, some of these ADLs such as the SAE Architecture Analysis and Design Language (AADL) currently have no means to model the problem domain. In addition the analysis of NFPs is performed with external tools and integrating the analysis with the design remains a problem.
To solve these problems, this thesis proposes two new modeling languages that can be used with any ADL to model the problem and analysis domains. The Requirements Definition and Analysis Language (RDAL) has been developed to support the modeling and analysis of requirements, and the automated evaluation of requirements verification by the design. RDAL is currently under the process of being adopted as a standard annex of the AADL.
The Quantitative Analysis Modeling Language (QAML) allows for modeling quantitative analysis models of many kinds and supports their seamless integration with design models. Automated evaluation of system NFPs by interpretation of QAML models ensures that estimates are always consistent with the evolving design. QAML is also well suited to for the representation of complex component data sheets to ease their integration in model-based designs.
The validation of both languages is presented with a set of example models showing how the proposed languages can help in building correct by construction systems through early error discovery.
- 09 May 2014
| Title || Modeling Languages for Requirements Engineering and Quantitative Analysis of Embedded Systems |
| Authors || Dominique Blouin |
| Type || Thesis |
| Conference/Journal Title || Doctorat |
| Volume/Number || |
| Editors || |
| Publisher || Université de Bretagne-Sud, France |
| Month || December |
| Year || 2013 |
| Pages || 163 |
| DOI || |
| Keywords || Model-based System Engineering, Requirements Engineering, Non-functional Properties, Architecture Description Languages, AADL, Use Case Maps, jUCMNav |