Bounded Antisynchronization of Multiple Neural Networks via Multilevel Hybrid Control

Area of Science:

• Control Theory
• Artificial Intelligence
• Network Science

Background:

• Investigates the bounded antisynchronization (AS) problem in multiple discrete-time neural networks (NNs) with fuzzy models.
• Addresses challenges posed by varying group sizes, communication links, dynamic variabilities, and environmental factors affecting network topology.

Purpose of the Study:

• To propose an energy-efficient cluster pinning communication mechanism for multiple NNs.
• To design an impulsive observer for accurate state estimation of target NNs.
• To develop a multilevel hybrid controller integrating AS and bounded synchronization (BS) controllers.

Main Methods:

• A cluster pinning communication mechanism is introduced to minimize communication energy consumption.
• An impulsive observer is designed for state estimation of the target NN.
• Stability analysis of the BS augmented error (BSAE) and AS augmented error (ASAE) using a fuzzy-based Lyapunov functional (FBLF) establishes sufficient conditions for bounded AS.

Main Results:

• Sufficient conditions for achieving bounded antisynchronization in the considered fuzzy neural network systems are derived.
• The proposed cluster pinning mechanism and impulsive observer effectively manage communication and state estimation.
• The multilevel hybrid controller demonstrates capability in achieving bounded AS.

Conclusions:

• The study successfully establishes conditions for bounded antisynchronization in complex fuzzy neural network systems.
• The developed methods offer a viable approach for energy-efficient control and state estimation in distributed NN systems.
• Numerical and application examples validate the effectiveness of the proposed theoretical framework.