Tuning of a neuro-fuzzy controller by genetic algorithm.

Area of Science:

• Control Systems Engineering
• Artificial Intelligence
• Computational Intelligence

Background:

• Genetic algorithms (GAs) offer powerful optimization for adaptive control systems.
• Fuzzy logic controllers (FLCs) are widely used but often require complex manual tuning.
• Neuro-fuzzy logic controllers (NFLCs) integrate neural networks and fuzzy logic for enhanced capabilities.

Purpose of the Study:

• To present a novel neuro-fuzzy logic controller (NFLC) where all parameters are simultaneously tuned using genetic algorithms (GAs).
• To demonstrate that GA-tuned NFLCs can automate and simplify the design process of fuzzy logic control systems.
• To evaluate the performance of the proposed GA-tuned NFLC against conventional fuzzy controllers and GA-tuned PID controllers.

Main Methods:

• The proposed NFLC utilizes a radial basis function neural network (RBF) architecture with Gaussian membership functions.
• Genetic algorithms (GAs) with dynamic crossover and mutation rates were implemented for parameter optimization.
• A flexible position coding strategy was employed within the GA to achieve near-optimal solutions.

Main Results:

• The GA-tuned NFLC significantly reduces the need for manual tuning of membership functions and fuzzy rules.
• Simulation results indicate that the proposed NFLC demonstrates superior performance compared to a conventional fuzzy controller.
• The NFLC also outperformed a Proportional-Integral-Derivative (PID) controller that was tuned using genetic algorithms.

Conclusions:

• The GA-tuned NFLC presents a robust and efficient approach for developing adaptive control systems.
• This method streamlines the design and tuning of fuzzy logic controllers, making them more accessible.
• The proposed controller shows significant advantages and improved performance in control applications.