Nonlinear System Identification

Nonlinear system identification aims to build accurate mathematical models of complex systems whose behavior isn't easily described by linear equations, primarily using data-driven approaches. Current research emphasizes developing robust methods that handle noisy data, sparse data, and adversarial attacks, often employing neural networks (including recurrent, convolutional, and neural ODEs), sparse identification of nonlinear dynamics (SINDy), and subspace methods, along with advanced optimization techniques like L-BFGS-B and ADMM. These advancements are crucial for improving control, prediction, and understanding in diverse fields, from robotics and industrial processes to aerospace and audio signal processing.