Physics-Guided Motion Generation and Control: From Autonomous Racing towards General Robotic

Abstract: Emerging embodied AI applications in autonomous driving and robotics hold great promise for transforming society, yet they face key challenges in generalization to unseen scenarios, interpretability, and data efficiency. This talk will discuss how physics-guided learning-based methods can address these issues by embedding prior physical knowledge into data-driven models. It will highlight recent advances in physics-guided motion generation and control for autonomous driving and racing in real-world scenarios, along with preliminary extensions to general robotics.

Bio: Mattia Piccinini is a Humboldt Post-doctoral Fellow at the Professorship of Autonomous Vehicle Systems (AVS) at the Technical University of Munich, Germany. He received an M.Sc. (cum laude) in mechatronics engineering and a Ph.D. (cum laude) in autonomous systems from the University of Trento, Italy, in 2019 and 2024 respectively. His Ph.D. thesis obtained the 2025 Best Dissertation Award from the IEEE Intelligent Transportation Systems Society. He is the recipient of a TUM Global Post-doctoral Fellowship (2024) and a Humboldt Post-doctoral Fellowship (2025). His research lies at the intersection of machine learning, physics-based modeling, motion generation and control for mobile robotics and autonomous vehicles.