Distributed Fault-Tolerant Containment Control Protocols for the Discrete-Time Multiagent Systems via Reinforcement...

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Robotics

Background:

• Multi-agent systems (MASs) are susceptible to actuator faults, compromising their operational integrity.
• Achieving containment control in MASs with unknown dynamics and faults presents a significant challenge.

Purpose of the Study:

• To develop a model-free fault-tolerant containment control strategy for MASs with time-varying actuator faults.
• To ensure uniform boundedness of containment errors despite system uncertainties and faults.

Main Methods:

• A distributed containment control method based on reinforcement learning (RL) is employed, utilizing relative state information.
• The optimal control problem is reformulated as an optimal regulation problem for a derived containment error system.
• An RL-based policy iteration method is used to derive a nominal controller, followed by a fault-tolerant controller design.

Main Results:

• The proposed RL-based method achieves containment control objectives without prior knowledge of system dynamics.
• The developed fault-tolerant controller effectively compensates for time-varying actuator faults.
• Numerical simulations validate the uniform boundedness of containment errors and the method's effectiveness.

Conclusions:

• The model-free, RL-based approach offers a robust solution for fault-tolerant containment control in MASs.
• This method demonstrates significant advantages in handling actuator faults and unknown system dynamics.
• The guaranteed uniform boundedness of containment errors highlights the reliability of the proposed control scheme.