Robust pole assignment for synthesizing feedback control systems using recurrent neural networks.

Area of Science:

• Control Systems Engineering
• Computational Neuroscience
• Optimization Theory

Background:

• Robust pole assignment is crucial for designing stable linear control systems.
• Existing methods often face challenges with model complexity and convergence guarantees.
• State and output feedback are key control strategies.

Purpose of the Study:

• To develop a neurodynamic optimization approach for robust pole assignment.
• To formulate the problem as a pseudoconvex optimization task.
• To ensure guaranteed convergence to global optimal solutions.

Main Methods:

• Formulating robust pole assignment as a pseudoconvex optimization problem.
• Utilizing the spectral condition number as a robustness measure.
• Employing two coupled recurrent neural networks for real-time optimization.

Main Results:

• The proposed neurodynamic approach demonstrates exponential convergence to global optima.
• It achieves this with lower model complexity compared to traditional methods.
• Simulations on 11 benchmark problems confirm its superior performance.

Conclusions:

• The neurodynamic optimization method offers a robust and efficient solution for pole assignment.
• It provides guaranteed convergence and improved performance.
• This approach advances the synthesis of linear control systems.