Core Concepts
Mathematical framework for modeling nonlinear control systems using Koopman Control Family.
Abstract
The paper introduces the Koopman Control Family (KCF) as a mathematical framework for modeling general discrete-time nonlinear control systems. It aims to provide a solid theoretical foundation for the use of Koopman-based methods in systems with inputs. The concept of KCF captures the behavior of nonlinear control systems on a function space, establishing a universal form for finite-dimensional models. The paper discusses how the proposed framework naturally lends itself to data-driven modeling of control systems by approximating models in general form and characterizing model accuracy using invariance proximity.
The Koopman operator approach to dynamical systems has gained attention due to its linear formulation, providing beneficial algebraic properties for analyzing complex dynamical systems. The paper extends this approach to control systems, addressing challenges related to input signals fundamentally altering system behavior. By defining the KCF and exploring common invariant subspaces, the paper provides a comprehensive mathematical framework for Koopman operator-based modeling of control systems.
Literature review highlights applications of the Koopman operator in various fields like fluid dynamics, stability analysis, reachability analysis, safety-critical control, and robotics. Data-driven methods like Dynamic Mode Decomposition (DMD) are discussed along with error bounds for accuracy assessment. The work emphasizes the importance of subspace invariance and proposes an algorithm to approximate Koopman-invariant subspaces with tunable accuracy.
Overall, the paper presents a systematic approach to modeling nonlinear control systems using the Koopman Control Family, offering insights into common invariant subspaces and accurate model approximations.
Stats
This work was supported by ONR Award N00014-23-1-2353 and NSF Award IIS-2007141.