This course provides an introduction to Gaussian Processes (GPs) and their role in modern learning-based control. The main goal is to present Gaussian Process regression as a flexible probabilistic framework for modeling unknown dynamics, quantifying uncertainty, and incorporating data-driven corrections into control design. After introducing the foundations of Gaussian Process regression, kernels, hyperparameter learning, and prediction with uncertainty estimates, the course will focus on applications motivated by geometric mechanics and control.
In particular, we will discuss how Gaussian Processes can be used to learn and approximate dynamics in mechanical systems of Lagrangian type, where structure, physical interpretability, and control relevance play a central role. We will then illustrate these ideas in applications to quadrotor systems, emphasizing residual dynamics learning, model improvement, and uncertainty-aware control. Finally, the course will address the learning of nonholonomic systems, highlighting both the opportunities and the challenges that arise when data-driven methods are combined with kinematic and dynamic constraints.
The course is intended to build a bridge between machine learning and control theory, showing how Gaussian Processes can serve as a rigorous and practical tool for modeling, prediction, and control of complex dynamical systems.
Schedule
© 2026 | Università degli studi di Verona
******** CSS e script comuni siti DOL - frase 9957 ********
