Test automation is key not only in CD/CI environments. And yet, it is a necessary not a sufficient condition for high quality code. In this talk, I will tackle the question of what makes a test a good test (the audience is invited to think about that question before we meet!), how a useful definition naturally implies that high code coverage also is necessary but not sufficient, how such a definition gives rise to testing based on defect hypotheses, and how to transfer that idea to testing automated and autonomous driving systems. I will show that testing based on pre-recorded drives in general is very problematic, the consequences of this insight for regression testing, and how to resolve this issue by deriving system-specific test cases. Time permitting, we can discuss the intricacies of testing AI-based components in these systems.
Recommended pre-reads: Defect-Based Testing Re-Using Concrete Test Scenarios Generally is a Bad Idea Fitness Functions for Testing Automated and Autonomous Driving Systems
(looking up and documenting facts / links / infos during lectures and discussions):
(putting researched material, emerging arguments, and ideas in context of the session's topic):
(documenting the essence and some highlights of the session to share them with the whole group in a 2 minute pitch during the wrap-up):