Trial lecture
AI-Based Software Testing Approaches: From Classical Automation to Autonomous AI Test Agents”
Time and place: June 26 , 2026 11:15 AM. Kristen Nygaards sal (5370), Ole-Johan Dahls hus.
Main research findings
Software is getting smarter. It can now sense its surroundings, adapt, and make decisions on its own. We already rely on these intelligent software systems to drive cars, run industrial robots, and support medical decisions. But these systems operate in open, dynamic, and unpredictable environments, posing significant challenges for quality assurance. Failures, particularly in safety-critical contexts, can lead to severe consequences, even fatal accidents. Ensuring their safety and trustworthiness is therefore essential.
This thesis develops practical solutions to advance quality assurance for intelligent software systems. Specifically, it spans four perspectives: it advances system validation, assesses the fidelity gap in simulation-based validation, quantifies uncertainty in a system's decisions, and evaluates the quality of our tests under that uncertainty. To do so, it draws on AI-driven validation methods, digital twin technologies, uncertainty quantification methods, and foundation models such as GPT and vision-language-action models.
Demonstrated across open-source and industrial cases including self-driving cars, industrial robots, and healthcare systems, this thesis lays foundations for building safer, more trustworthy intelligent software, a vital step toward a future where we can safely depend on the systems we increasingly rely on.
Adjudication committee
- Professor Dietmar Pfahl, Institute of Computer Science, University of Tartu, Estonia
- Researcher Irune Agirre Troncoso, IKERLAN, Spain
- Professor Ozgu Alay, University of Oslo, Norway
Supervisors
Professor Are Magnus Bruaset, Simula, Norway
Professor Shaukat Ali, Simula, Norway
Chair of defence
Associate Professor Ellen Munthe-Kaas, University of Oslo
Contact information at Department: Mozhdeh Sheibani Harat