Workshop on Physics Meets Data:
Learning, Optimization, and Control

In recent years, research at the interface of machine learning and classical engineering disciplines has grown rapidly. Data-driven methods are increasingly used for modeling and control of dynamical systems, complementing and, in some cases, replacing traditional physics-based models and numerical solvers. At the same time, concepts from engineering and scientific computing are reshaping machine learning by promoting physics-aware, structured approaches that move beyond black-box models and explicitly encode prior knowledge in model architectures and loss functions.

Physics Meets Data captures a paradigm in which data-driven learning and physical modeling are tightly integrated rather than treated as competing approaches. Physics-informed machine learning embeds physical laws, domain knowledge, and structural priors directly into learning architectures and objectives, allowing control systems to benefit from the flexibility of data-driven methods while retaining a principled understanding of the underlying dynamics. This synergy between physics and data leads to more efficient, trustworthy, and interpretable learning, with improved performance and robustness in control applications.

The impact of this paradigm spans a wide range of control and optimization problems. In robotics, physics-meets-data approaches enhance the control of complex manipulators and autonomous systems by explicitly respecting mechanical constraints, kinematics, and dynamics. In power systems and industrial processes, they enable the design of data-driven control strategies that remain grounded in physical phenomena such as heat transfer, fluid flow, and thermodynamics.

This workshop aims to present recent advances in physics-informed machine learning for modeling, control, and optimization while tightly integrating conceptual discussions with hands-on experience. The program interleaves lecture-style talks on state-of-the-art physics-informed learning methods with interactive coding sessions, allowing participants to directly implement and experiment with the presented ideas. Through this combined format, the audience will gain both theoretical insight and practical skills to apply physics-meets-data approaches to real-world control problems.

The workshop targets an audience from graduate level to experienced theoretical and practically oriented control engineers who aim to improve their knowledge in physics-informed machine learning for control and optimization.