Distributed NN-Based Formation Control of Multi-Agent Systems: A Reduced-Order Appointed-Time Observer Approach

Area of Science:

• Robotics and Control Systems
• Distributed Systems
• Artificial Intelligence

Background:

• Formation control in multi-agent systems is challenging, particularly with distributed output-feedback controllers.
• Existing methods often require input information exchange, limiting applicability.
• Designing controllers without inter-agent input communication remains an open problem.

Purpose of the Study:

• To develop a novel distributed reduced-order observer for formation error estimation.
• To design a neural-network-based formation controller for multi-agent systems.
• To address distributed output-feedback formation control without input information exchange.

Main Methods:

• Utilized relative output information for distributed observer design.
• Developed a reduced-order observer for real-time formation error estimation.
• Employed a neural network for the formation controller design in connected graph topologies.

Main Results:

• Successfully designed a distributed observer estimating formation error at a predefined time.
• Developed and validated a neural-network-based formation controller.
• Demonstrated the effectiveness of the proposed methods through theoretical analysis and simulations.

Conclusions:

• The proposed distributed observer and neural-network controller effectively solve the distributed output-feedback formation control problem.
• The approach enables formation control without requiring agents to exchange input information.
• The findings are robust and verified by simulation examples.