Finite Dimensional Koopman Form of Polynomial Nonlinear Systems

Lucian-Cristian Iacob, Maarten Schoukens, Roland TóthIFAC-PapersOnLine 2023 · DOI · OpenReview

TL;DR

This paper introduces a systematic method to convert specific polynomial nonlinear systems into an exact, finite-dimensional linear representation using the Koopman operator framework. This breakthrough allows for the application of powerful linear control and analysis tools to a broader class of nonlinear systems without approximation errors.

What Problem Does This Paper Address?

The Koopman operator framework offers a powerful way to linearize nonlinear systems by lifting them into an infinite-dimensional linear space. However, for practical applications, especially in control, a finite-dimensional linear representation is highly desirable. The challenge has been that there's no constructive theory for selecting the "observable functions" needed to achieve this finite embedding, often leading to heuristic, data-driven approaches with unknown approximation errors. This paper tackles the problem of finding an exact finite-dimensional Koopman embedding for a specific class of polynomial nonlinear systems.

Methods

The core of the paper lies in a constructive procedure to derive the observable functions that guarantee an exact finite-dimensional Koopman embedding. The authors focus on polynomial nonlinear systems with a "lower triangular" structure in their state equations. They prove, through induction, that for such systems, an exact finite-dimensional lifting exists. The method involves systematically constructing a set of lifting functions that includes the original states and specific polynomial terms derived from the system's dynamics. This construction ensures that the dynamics of the lifted system are linear and perfectly capture the original nonlinear dynamics. The paper also extends this method to control-affine systems by showing how to incorporate input terms into the lifted representation, resulting in a linear parameter-varying (LPV) model.

Datasets

Not discussed in the paper. The paper focuses on a theoretical framework and demonstrates its application with a simulated example.

Key Insights and Ideas

  • Exact Finite Embedding is Possible: The paper demonstrates that for a specific class of polynomial nonlinear systems, an exact finite-dimensional linear representation via the Koopman operator is not just a theoretical possibility but can be systematically computed. This is a significant step beyond approximate methods.
  • Constructive Procedure for Observables: The key contribution is a concrete, step-by-step method to identify the necessary observable functions. This moves away from heuristic choices and provides a deterministic approach.
  • Lower Triangular Polynomial Systems are Key: The "lower triangular" structure of the polynomial dynamics is identified as a crucial property that enables this exact embedding.
  • Extension to Input-Driven Systems: The framework seamlessly extends to control-affine nonlinear systems, yielding an LPV representation, which is highly valuable for control design.

Key Results

The paper provides a theoretical proof for the existence of an exact finite-dimensional Koopman embedding for autonomous polynomial nonlinear systems with a lower triangular structure. The method is demonstrated through an example of a 4-dimensional system. Simulation results show that the state trajectories of the original nonlinear system and its Koopman embedding are virtually identical, with errors on the order of $10^{-15}$, attributed to numerical artifacts. This indicates that the derived finite-dimensional linear model perfectly captures the nonlinear dynamics. The paper also shows that for input-driven systems, the lifted model exhibits the same exactness.

Limitations and Future Work

The primary limitation acknowledged is that the proposed method is specific to polynomial nonlinear systems with a lower triangular structure. The authors suggest that future work should focus on extending this approach to a more general class of nonlinear systems. This would involve exploring how to handle systems with different polynomial structures or even non-polynomial nonlinearities.

Summary updated

Abstract

The Koopman framework is a popular approach to transform a finite dimensional nonlinear system into an infinite dimensional, but linear model through a lifting process, using so-called observable functions. While there is an extensive theory on infinite dimensional representations in the operator sense, there are few constructive results on how to select the observables to realize them. When it comes to the possibility of finite Koopman representations, which are highly important form a practical point of view, there is no constructive theory. Hence, in practice, often a data-based method and ad-hoc choice of the observable functions is used. When truncating to a finite number of basis, there is also no clear indication of the introduced approximation error. In this paper, we propose a systematic method to compute the finite dimensional Koopman embedding of a specific class of polynomial nonlinear systems in continuous-time such that, the embedding, without approximation, can fully represent the dynamics of the nonlinear system.
BibTeX
@inproceedings{iacob2023finite,
  author = {Lucian-Cristian Iacob and Maarten Schoukens and Roland Tóth},
  title = {{Finite Dimensional Koopman Form of Polynomial Nonlinear Systems}},
  booktitle = {IFAC-PapersOnLine},
  year = {2023},
  doi = {10.1016/j.ifacol.2023.10.849},
}
Figure