Equivalent-input-disturbance estimator-based event-triggered control design for master-slave neural networks

Area of Science:

• Control Systems Engineering
• Computational Neuroscience
• Networked Systems

Background:

• Master-slave neural networks (MSNNs) face challenges with network-induced delays, uncertainties, and external disturbances.
• Existing synchronization methods may not adequately address these combined issues, impacting system performance and reliability.
• Efficient bandwidth utilization and reduced communication load are critical for practical MSNN applications.

Purpose of the Study:

• To investigate and solve the robust synchronization problem for MSNNs under complex operating conditions.
• To develop an event-triggered control protocol that enhances communication efficiency and avoids Zeno behavior.
• To propose a method for compensating unknown uncertainties and exogenous disturbances.

Main Methods:

• Application of an equivalent-input-disturbance (EID) estimation technique to counteract uncertainties and disturbances.
• Development of an event-triggered control protocol with periodic verification to ensure synchronization and avoid Zeno behavior.
• Formulation of an augmented system by combining synchronization error, observer, and filter states.
• Utilization of Lyapunov stability theory and the reciprocally convex approach to establish delay-dependent stability conditions.
• Co-design of event-triggering parameters, controller, and observer gains based on stability conditions.

Main Results:

• The proposed EID estimation technique effectively compensates for unknown uncertainties and disturbances.
• The event-triggered control protocol successfully achieves MSNN synchronization while optimizing communication bandwidth.
• Delay-dependent stability conditions for the augmented system were rigorously derived.
• Feasible solutions were obtained for co-designing control and observer parameters.
• Simulations on two examples validated the theoretical findings.

Conclusions:

• The developed framework provides a robust solution for MSNN synchronization in the presence of delays and disturbances.
• The event-triggered approach offers a practical and efficient method for networked neural systems.
• The study contributes theoretical insights and practical design methodologies for advanced neural network control.