A Novel Multiagent Neurodynamic Approach to Constrained Distributed Convex Optimization.

Area of Science:

• Control Systems Engineering
• Optimization Theory
• Computational Neuroscience

Background:

• Distributed convex optimization problems are crucial in multiagent systems.
• Existing methods often struggle with general constraints and communication bandwidth.
• Nonsmooth convex functions and local constraints present significant challenges.

Purpose of the Study:

• To propose a novel continuous-time multiagent neurodynamic approach for distributed convex optimization with general constraints.
• To ensure asymptotic consensus and convergence to optimal solutions for all agents.
• To develop a method with low structure complexity suitable for constrained problems.

Main Methods:

• A continuous-time multiagent neurodynamic system is formulated.
• Each agent individually solves its objective function.
• Analysis guarantees the existence, boundedness, and convergence of state solutions.

Main Results:

• The proposed neurodynamic system guarantees the existence and boundedness of global state solutions.
• State solutions asymptotically reach the feasible region of the equivalent optimization problem.
• Agent outputs converge to the optimal solution set of the primal distributed problem.

Conclusions:

• The novel neurodynamic approach is effective for general constrained distributed optimization problems.
• The method offers low structure complexity, reducing communication bandwidth requirements.
• Demonstrated efficiency in numerical examples and power system optimal load-sharing problems.