Thesis: Exploration Mode Requirements Engineering

Background


With the rise of AI and Machine Learning (ML), researchers have recently begun to examine the effects of ML inclusion as part of software development. ML is often part of a larger complex system, and most parts of such systems can follow standard development processes, e.g., scaled agile (SAFe), scrum or the V-model. But when it comes to the ML-specific development, these methods typically fail to account for the explorative nature and non-determinism of ML. Particularly, we are interested in requirements engineering as part of software engineering for ML. How does the representations and processes associated with requirements need to change in the face of ML? Initial, exploratory findings have shown that ML-enabled systems follow standard development up to a point, but then work in a much more bottom-up or middle-out experimental mode for ML-related requirements and development. Data and data requirements play a key role.
At Volvo Service Market Logistics we are in the process of fully digitalizing the supply chain planning processes, making use of intelligent microservices to implement business processes. We face challenges in short cutting the requirements-development-testing cycle, different development/deployment/release cycles for ML-enabled products, and different data needs for ML testing.


Thesis questions and expected outcome
The proposal for this thesis is to examine RE methods, representations, processes, and challenges for ML-enabled exploration mode development in Volvo Service Market Logistics. In exploration mode development you don’t know what you will develop when starting out, you explore the data and find opportunities, apply business knowledge, and continue model exploration. However, even in such a context, as part of a large organization, it is desirable to have requirements processes which are somewhat structured and systematic, and which discover and make use of best practices. We envision that this thesis will involve one or more case studies focusing on exploration developments working towards digital products. Such developments often “fail” and do not make it to a digital product state, but it is still interesting to examine how this process works from a requirements perspective, collecting best practices and working towards process and representation recommendations. Could RE methods and processes facilitate explorations reaching production state without reducing the explorative freedom needed to find opportunities?


Student profile and application
One or two Master students in Computer Science, Software Engineering, or other relevant fields.
Application deadline: TBD, we will continuously review the applications so don’t wait with submitting.


For further information, please contact:




Recruiting contact:
Mattias Jonhede at Advanced Analytics
E-mail: mattias.jonhede@volvo.com
Phone: +46739020425