One controller to rule them all

TL;DR

This paper proposes a novel in-context learning-based approach to design a single, universal controller for an entire class of dynamical systems, potentially revolutionizing control design by moving away from individual controllers.

Contribution

It introduces the first in-context learning-based method for creating a unified controller applicable to all systems within a class, rather than individual systems.

Findings

Numerical results demonstrate the effectiveness of the proposed approach.

The paradigm shift from one-system-one-controller to one-class-one-controller is promising.

Potential for broad applicability across different dynamical systems.

Abstract

Imagine having a system to control and only know that it belongs to a certain class of dynamical systems. Would it not be amazing to simply plug in a controller and have it work as intended? With the rise of in-context learning and powerful architectures like Transformers, this might be possible, and we want to show it. In this work, within the model reference framework, we hence propose the first in-context learning-based approach to design a unique contextual controller for an entire class of dynamical systems rather than focusing on just a single instance. Our promising numerical results show the possible advantages of the proposed paradigm, paving the way for a shift from the "one-system-one-controller" control design paradigm to a new "one-class-one-controller" logic.