Generalized synchronization between two distinct chaotic systems through deep reinforcement learning

Abstract

This study presents a novel model-free deep reinforcement learning (DRL) approach to control generalized chaotic synchronization between two distinct chaotic systems, referred to as the slave and master systems. Our method enables an agent to progressively optimize its strategy for applying external compensation to the slave system through continuous interaction with both systems, achieving generalized synchronization without requiring prior knowledge of the underlying chaotic dynamics. We validate our approach on several classical chaotic systems, including the Lorenz system, the Rössler system, and the Chua circuit, demonstrating rapid synchronization and exceptional performance. A key contribution of this work is the ability of a single learned agent to control generalized chaotic synchronization across various slave–master system configurations. The DRL approach shows remarkable robustness, maintaining effective synchronization control even with limited information about the state variables of both systems. Notably, our method successfully controls high-dimensional coupled chaotic systems while requiring only minimal state variable information. The results underscore the potential of DRL for applications in secure communication and other fields where chaos control is crucial.