Deep unfolding-based output feedback control design for linear systems with input saturation

TL;DR

This paper introduces a deep unfolding-based, data-driven control design framework using Neural Ordinary Differential Equations to effectively handle input saturation in linear systems, improving stability and performance.

Contribution

It presents a novel deep learning approach for output feedback control with input saturation, combining neural ODEs with theoretical validation to enhance control performance.

Findings

Significantly outperforms traditional LMI-based methods

Enlarges the stability region of the closed-loop system

Provides a computationally efficient control design approach

Abstract

In this paper, we propose a deep unfolding-based framework for the output feedback control of systems with input saturation. Although saturation commonly arises in several practical control systems, there is still a scarce of effective design methodologies that can directly deal with the severe non-linearity of the saturation operator. In this paper, we aim to design an anti-windup controller for enlarging the region of stability of the closed-loop system by learning from the numerical simulations of the closed-loop system. The data-driven framework we propose in this paper is based on a deep-learning technique called Neural Ordinary Differential Equations. Within our framework, we first obtain a candidate controller by using the deep-learning technique, which is then tested by the existing theoretical results already established in the literature, thereby avoiding the computational challenge in the conventional design methodologies as well as theoretically guaranteeing the performance of the system. Our numerical simulation shows that the proposed framework can significantly outperform a conventional design methodology based on linear matrix inequalities.