AI Foundation Models for Quality Assurance of Robotic Software

AI foundation models trained on extensive multimodal data, such as text, images, audio, and video, are transforming how we design, develop, test, and debug robotic software. To this end, this keynote will explore the emerging opportunities and challenges of applying AI foundation models to robotic software quality assurance. Drawing on real-world and industrial case studies, the keynote will demonstrate how these models can be integrated into quality assurance workflows to support test generation and uncertainty identification. Moreover, the keynote will discuss how uncertainty quantification and testing of vision-action-language models embedded in robotics can help improve robotic software quality. The keynote will conclude with key future research directions from a research roadmap on how foundation models can facilitate the software engineering of self-adaptive robotics.