Centralized and distributed model predictive control for consensus of non-linear multi-agent systems with...

Area of Science:

• Robotics
• Control Theory
• Artificial Intelligence

Background:

• Multi-agent systems require robust control for coordinated behavior.
• Model Predictive Control (MPC) is effective but often complex to design, especially for nonlinear systems with constraints.
• Achieving consensus in multi-agent systems with obstacles presents significant control challenges.

Purpose of the Study:

• To develop and evaluate centralized and distributed Model Predictive Control (MPC) methods for nonlinear multi-agent systems.
• To address the consensus problem in the presence of fixed and time-varying obstacles.
• To simplify MPC design by eliminating the need for terminal ingredients, enhancing applicability to higher-order systems.

Main Methods:

• Centralized and distributed Model Predictive Control (MPC) strategies were implemented.
• The distributed MPC was executed using both serial and parallel approaches.
• Contractive constraints were applied to guarantee system stability and agent convergence.
• Obstacles were integrated as constraints within the optimization control problem.

Main Results:

• The proposed MPC methods successfully achieved consensus in nonlinear multi-agent systems.
• Stability and convergence of agents to a consensus point were guaranteed.
• The necessity of designing complex terminal ingredients for MPC was eliminated.
• Simulations demonstrated the effectiveness of both centralized and distributed methods for wheeled mobile robots.

Conclusions:

• The developed centralized and distributed MPC methods offer a simplified yet effective approach to solving the consensus problem for nonlinear multi-agent systems.
• These methods are robust to fixed and time-varying obstacles and are suitable for higher-order systems.
• The elimination of terminal ingredients reduces design complexity, making MPC more accessible for advanced applications.