Control of chaos in the Frenkel-Kontorova model using reinforcement learning

TL;DR

This paper introduces a reinforcement learning approach to control chaos in the Frenkel-Kontorova model without requiring explicit system knowledge, successfully achieving synchronization in coupled nonlinear pendulums.

Contribution

A model-free reinforcement learning method is proposed for chaos control in the FK model, enabling synchronization without explicit system or target state information.

Findings

Reinforcement learning effectively controls chaos in FK model.

The method achieves perfect synchronization among oscillators.

Perturbations lead to more regular, synchronized patterns.

Abstract

The spatiotemporal chaos in the Frenkel-Kontorova (FK) model is studied. A model-free reinforcement learning algorithm is proposed to the design of a controller. There is no need for explicit knowledge on system, target states and unstable periodic orbits. In numerical experiments, the proposed method is used in the coupled array of nonlinear pendulum. It is shown that the perturbations (prescribed) are applied on the part or all of oscillators parameters will lead to differently regular and high-level synchronal patterns. Specially, the system achieves perfect synchronization acting the all of oscillators.