Constrained Reinforcement Learning-Based Chaos Synchronization and Its Applications in Secure Communications

This paper applies a recently developed stabilityconstrained reinforcement learning scheme to address a chaos synchronization problem. The scheme employs a tensor product model transformation method to convert the synchronization error system into weighted sum linear systems, which are then used to establish the stability constraints. The deep deterministic policy gradient algorithm serves as the foundation for the learning process, and the Lyapunov stability theory guarantees that the slave and master systems are asymptotically synchronized. The effectiveness of the scheme is demonstrated through simulation studies of Lorenz chaotic systems with secure communication applications. The results show that the synchronization errors rapidly converge to zero, and the messages are then successfully recovered.