Decomposed fuzzy systems and their application in direct adaptive fuzzy control

Area of Science:

• Control Engineering
• Artificial Intelligence
• Computational Intelligence

Background:

• Traditional fuzzy systems face limitations in adaptive control due to complex rule adaptation and computational burden.
• Existing fuzzy approximators may struggle with scalability and efficiency when dealing with complex control tasks.

Purpose of the Study:

• To introduce a novel fuzzy structure, the Decomposed Fuzzy System (DFS), as an improved fuzzy approximator for adaptive fuzzy control.
• To enhance function approximation capabilities and learning efficiency in adaptive fuzzy control systems.
• To develop a simplified DFS for real-time applications with reduced computational demands.

Main Methods:

• Decomposing fuzzy variables into layers, where each layer represents a traditional fuzzy set.
• Forming component fuzzy systems from layers of different variables, enabling independent learning.
• Developing a simplified DFS structure to address real-time computational constraints.

Main Results:

• The DFS demonstrates superior function approximation capability and learning efficiency compared to traditional fuzzy systems in adaptive control.
• DFS exhibits nearly constant learning time even with an increased number of fuzzy rules.
• The simplified DFS achieves satisfactory performance with a more concise structure, suitable for real-time applications.

Conclusions:

• The Decomposed Fuzzy System (DFS) offers a significant advancement in adaptive fuzzy control by improving approximation and learning efficiency.
• DFS provides a scalable and computationally efficient approach to fuzzy control, particularly with its simplified variant.
• The proposed DFS structure facilitates adaptation without introducing substantial learning burdens, making it practical for various applications.