Control-oriented system identification: Classical, learning, and physics-informed approaches

S. Sivaranjani

doi:10.1016/j.arcontrol.2026.101067

Control-oriented system identification: Classical, learning, and physics-informed approaches

2026-07-25

0

PRE

AI

DOI：10.1016/j.arcontrol.2026.101067

Original

Original request for help

Save

Abstract

En 中文

This article surveys classical, machine learning, and data-driven system identification approaches to learn control-relevant and physics-informed models of dynamical systems. In recent years, machine learning approaches have enabled system identification from noisy, high-dimensional, and complex data. However, their utility in control applications is limited by their ability to provide provable guarantees on control-relevant properties. Meanwhile, traditional control theory has identified several properties of physical systems that are useful in analysis and control synthesis, such as dissipativity, monotonicity, energy conservation, and symmetry-preserving structures. In this paper, we postulate that merging system identification algorithms with such control-relevant or physics-informed properties can provide useful inductive bias, enhance explainability, enable control synthesis with provable guarantees, and improve sample complexity. We formulate system identification as an optimization problem where control-relevant properties can be enforced in three ways, namely, direct parameterization (constraining the model structure to satisfy a desired property by construction), soft constraints (encouraging control-relevant properties through regularization or penalty terms), and hard constraints (imposing control-relevant properties as constraints in the optimization problem). Through this lens, we survey methods to learn physics-informed and control-relevant models spanning classical linear and nonlinear system identification techniques, machine learning-based approaches, as well as direct identification through data-driven and behavioral representations. Taken together, these perspectives suggest that control-oriented identification should be viewed not only as a problem of minimizing prediction error, but also as one of selecting model structures and learning methods that preserve the properties needed for downstream analysis and control synthesis. Throughout the paper, we provide several expository examples that are accompanied by code and brief tutorials on a public Github repository. We also describe several challenging directions for future research in this area, including identification in networked, switched, and time-varying systems, experiment design, and bridging the gaps between data-driven, learning-based, and control-oriented system identification.

Control-oriented system identification: Classical, learning, and physics-informed approaches

Abstract

Journal

Organization