Data-Based Output Synchronization of Multi-Agent Systems With Actuator Faults.

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Robotics

Background:

• Multi-agent systems (MAS) are susceptible to actuator faults, compromising their operational integrity.
• Existing fault-tolerant control (FTC) methods often require system models and struggle with heterogeneous MAS.

Purpose of the Study:

• To develop a model-free fault detection mechanism for MAS with actuator faults.
• To propose a novel FTC strategy that accommodates heterogeneous MAS and avoids complex parameter tuning.
• To enhance the robustness and reliability of MAS in the presence of actuator failures.

Main Methods:

• A backward input-driven fault detection mechanism (BIFDM) was developed for fault identification without a system model.
• A fault-tolerant controller (FTC) was designed using reinforcement learning (RL) and backward information (BI).
• The proposed FTC integrates BI to eliminate the need for additional fault-specific parameters.

Main Results:

• The BIFDM successfully detected actuator faults in MAS without requiring a system model.
• The RL-based FTC demonstrated effective fault tolerance, even in heterogeneous multi-agent systems.
• The proposed methods were validated through two simulation examples, confirming their efficacy.

Conclusions:

• The developed BIFDM and FTC provide a robust and adaptable solution for managing actuator faults in MAS.
• The model-free nature and applicability to heterogeneous systems represent significant advancements in fault-tolerant control.
• The findings contribute to the reliable deployment of MAS in complex and dynamic environments.