Dynamic Identification

Dynamic identification focuses on extracting governing equations or models from observed data of complex systems, aiming to understand and predict their behavior. Current research emphasizes developing efficient and robust algorithms, including neural networks (like autoencoders and recurrent networks), sparse identification methods, and Bayesian approaches, to handle high-dimensional, noisy, and non-stationary data, often incorporating physics-informed constraints. These advancements are crucial for diverse applications, such as improving robotic control (especially in human-robot interaction), understanding complex biological systems (e.g., neuroscience and ecology), and creating accurate reduced-order models for computationally expensive simulations. The ultimate goal is to achieve accurate, interpretable, and computationally efficient models for a wide range of dynamic systems.