Prescribed-Time Human-in-the-Loop Optimal Synchronization Control for Multiagent Systems Under DoS Attacks via...

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Network Security

Background:

• Multiagent systems (MASs) are vulnerable to denial-of-service (DoS) attacks disrupting communication links.
• Human-in-the-loop (HiTL) control offers enhanced system governance but requires robust synchronization strategies.
• Prescribed-time (PT) control aims for convergence within a user-defined finite time, crucial for time-sensitive applications.

Purpose of the Study:

• To address the prescribed-time (PT) optimal synchronization control for MASs under link-based DoS attacks.
• To develop a distributed observer capable of estimating leader output within PT under switching topologies.
• To implement a model-free Q-learning algorithm for optimal policy learning with reduced computational load.

Main Methods:

• A fully distributed observer with a prescribed finite-time function is proposed for follower agents.
• An augmented system combines follower dynamics with the observer for stability analysis.
• A single-critic neural network (NN) Q-learning algorithm, trained via least-squares, is employed for policy optimization.

Main Results:

• The proposed observer guarantees global practical PT convergence with bounded gain, independent of global topology.
• The Q-learning algorithm demonstrates convergence of Q-functions, enabling optimal synchronization policy learning.
• Simulation results validate the efficacy of the developed control scheme against DoS attacks.

Conclusions:

• The study successfully develops a PT HiTL optimal synchronization control for MASs under DoS attacks.
• The proposed observer and Q-learning approach provide a robust and computationally efficient solution.
• The findings contribute to enhancing the resilience and performance of MASs in adversarial environments.