Computationally Efficient Data-Driven Discovery and Linear Representation of Nonlinear Systems For Control

TL;DR

This paper introduces a deep learning-based data-driven approach using Koopman operator theory for efficient system identification and linearization of nonlinear systems, demonstrated on a pendulum example with noisy data.

Contribution

It presents a novel recursive deep learning framework for Koopman-based system linearization that improves training efficiency and accuracy over existing autoencoder methods.

Findings

More efficient training process

Higher accuracy in system identification

Effective on noisy data simulations

Abstract

This work focuses on developing a data-driven framework using Koopman operator theory for system identification and linearization of nonlinear systems for control. Our proposed method presents a deep learning framework with recursive learning. The resulting linear system is controlled using a linear quadratic control. An illustrative example using a pendulum system is presented with simulations on noisy data. We show that our proposed method is trained more efficiently and is more accurate than an autoencoder baseline.