Publication

Method and Tool Support for Automotive Software Engineering

Modelling is an essential activity in all engineering disciplines, and automotive software engineering is no exception. Model-driven software engineering – which has been shown with extensive evidence to improve efficiency and effectiveness of software development – acknowledges the centrality of models and advocates the use of models as primary development artefacts; it is in the models that work should be done, and other development artefacts – such as requirement specifications, documentation and source code – should be generated from the models. Introducing a model-driven development approach in a previously non model-driven process, however, presents specific challenges to the development organization; there is, furthermore, no consensus in the current state of research how to best make such a transition. In the automotive domain, where software engineering is only one of several engineering disciplines involved in development of the car, best practices for transitioning to a model-driven approach are yet more unclear. In addition, the overarching development paradigm in the automotive domain is often document-centric – i.e. it is in textual documents that development information is officially disseminated in the development organization. Nevertheless, software models have been found to play a central role within projects in the organization. In this thesis, the main research question is: How can model-driven development methods improve the development of automotive software? The approach taken is to acknowledge that the overarching development paradigm will remain unchanged, and to examine how – in the current development practice – software models are used within individual projects. Moreover, the thesis examines how these models relate to the documents in which development information is disseminated in the organization-wide development process. Using triangulation of empirical and theoretical research methods, the results in the thesis show that there can be fundamental differences between an organization-wide process and how that process is implemented within individual project; the results also identify issues related to this fact. The results, furthermore, provide insights into opportunities of how modelling methods may improve the development of automotive software.

Author(s)
Niklas Mellegård
Research area
Systems for Accident Prevention and AD
Publication type
Licenciate thesis
Project
ASIS - Algorithms and Software for Improved Safety
Year of publication
2010