Neural-genetic synthesis for state-space controllers based on linear quadratic regulator design for eigenstructure...

Area of Science:

• Control Systems Engineering
• Computational Intelligence
• Applied Mathematics

Background:

• State-space controllers are crucial for multivariable dynamic systems.
• Eigenstructure assignment is a key challenge in control design.
• Existing methods may face limitations in complex system synthesis.

Purpose of the Study:

• To present a novel neural-genetic model for state-space controller synthesis.
• To apply the model to eigenstructure assignment in multivariable dynamic systems using linear quadratic regulator design.
• To evaluate the performance and convergence of the proposed computational intelligence approach.

Main Methods:

• A hybrid model fusing a genetic algorithm (GA) and a recurrent neural network (RNN).
• GA is used for weighting matrix selection; RNN solves the algebraic Riccati equation.
• A fourth-order electric circuit model serves as the testbed for evaluation.

Main Results:

• The genetic search convergence was assessed via fitness function statistics.
• Recurrent neural network convergence was analyzed using energy and norm landscapes.
• Control design performance was validated through time and frequency domain analyses (impulse response, singular values, modal analysis).

Conclusions:

• The neural-genetic model demonstrates effective eigenstructure assignment for state-space controllers.
• The computational intelligence paradigms show reliable convergence and robust control performance.
• This fusion approach offers a promising direction for advanced control system synthesis.