Speaker: Ali Aminzadeh, postdoctoral fellow at Department of Physics, Digital Physics Group, Umeå University
Title: “Towards Trustworthy Physical AI: From Formal Safety Verification to Physics-Informed Learning-Based Adaptive Control”
Abstract:
Physical AI is becoming increasingly relevant to modern life, extending artificial intelligence beyond digital environments to systems that interact with and operate in the physical world. As these systems take on greater responsibilities in uncertain and dynamically changing environments, a fundamental challenge is to equip them with learning and adaptation capabilities while establishing rigorous guarantees on their behavior. This seminar presents my research on the control-theoretic foundations of trustworthy Physical AI, spanning formal safety synthesis and verification and physics-guided data-efficient control. The seminar concludes with an outlook on my ongoing research into learning-based adaptive control, exploring the challenges of incorporating physical knowledge into learning and enabling adaptation under uncertain and changing system dynamics.
