A two-level neurodynamic approach for heterogeneous networked game under event-triggered quantized mechanism

Area of Science:

• Control Theory
• Artificial Intelligence
• Game Theory

Background:

• Deep learning applications in complex, multi-player systems face challenges in communication costs and convergence rates.
• Cyber-physical systems necessitate agents with identical intrinsic dynamics and efficient communication protocols.
• Networked games are crucial for modeling interactions in multi-agent systems.

Purpose of the Study:

• To develop a game theory-based deep learning framework for networked games with reduced communication.
• To ensure prescribed-time convergence and handle heterogeneous dynamics in multi-agent systems.
• To address communication burdens using event-triggered and quantized communication strategies.

Main Methods:

• Utilizing a one-to-one gradient-based event trigger and logarithmic quantizer to minimize communication.
• Employing passivity-based strategies to manage incomplete information.
• Designing control inputs for heterogeneous dynamics to track Nash equilibrium (NE).
• Applying Lyapunov methods to prove convergence within adjustable time and exclude Zeno behavior.

Main Results:

• The proposed two-level neurodynamic system demonstrates convergence within adjustable time.
• Communication burden and frequency are significantly reduced through event-triggered and quantized communication.
• The approach effectively handles heterogeneous dynamics and ensures convergence to Nash equilibrium.
• Zeno behavior is successfully excluded, ensuring practical applicability.

Conclusions:

• The developed neurodynamic approach offers an effective solution for networked games in cyber-physical systems.
• The method enhances convergence rates and reduces communication costs, outperforming existing strategies.
• The framework is validated through a connectivity control problem for autonomous mobile robots.