Distributed Model-Free Adaptive Control for Learning Nonlinear MASs Under DoS Attacks.

Area of Science:

• Control Theory
• Artificial Intelligence
• Cybersecurity

Background:

• Multiagent systems (MASs) are increasingly complex and vulnerable to cyberattacks.
• Denial-of-Service (DoS) attacks disrupt communication and control in MASs.
• Model-free adaptive control (MFAC) offers a promising approach for systems with unknown dynamics.

Purpose of the Study:

• To develop a robust distributed model-free adaptive control (DMFAC) algorithm for nonlinear MASs facing DoS attacks.
• To create a unified framework addressing leaderless consensus, leader-following consensus, and containment control.
• To enhance the learning and adaptive capabilities of MASs under adversarial conditions.

Main Methods:

• An improved dynamic linearization technique to derive an equivalent linear data model.
• A novel attack compensation mechanism to mitigate DoS attack impacts.
• A learning-based DMFAC algorithm integrating the linearization and compensation strategies.

Main Results:

• The proposed DMFAC algorithm effectively learns and controls nonlinear MASs despite DoS attacks.
• The dynamic linearization method successfully transforms complex systems into manageable linear models.
• The attack compensation mechanism significantly reduces the influence of DoS attacks on system performance.

Conclusions:

• The developed DMFAC algorithm provides a robust and unified solution for controlling nonlinear MASs under DoS attacks.
• The approach demonstrates effectiveness in leaderless consensus, leader-following consensus, and containment control scenarios.
• Simulation results validate the algorithm's capability to maintain system stability and achieve control objectives.