Related Experiment Videos
Adaptive fuzzy control of unstable nonlinear systems
1Department of Control Engineering, National Chiao-Tung University, Hsinchu, Taiwan, R.O.C.
International Journal of Neural Systems
|September 1, 1995
Summary
This study introduces the Fuzzy Adaptive Learning Control Network (FALCON), an adaptive AI system that learns fuzzy logic control rules dynamically. It offers improved structure and learning capabilities for complex control tasks.
Area of Science:
- Artificial Intelligence
- Control Systems Engineering
- Computational Intelligence
Background:
- Traditional fuzzy logic controllers often lack adaptability and can struggle with complex systems.
- Existing neural network approaches may require extensive pre-configuration or lack dynamic learning capabilities.
Purpose of the Study:
- To propose a novel feedforward multilayer connectionist network, Fuzzy Adaptive Learning Control Network (FALCON), for fuzzy logic control.
- To introduce an online structure/parameter learning algorithm, FALCON-ART, for dynamic network construction.
- To enable adaptive control without prior knowledge of system dynamics or initial parameters.
Main Methods:
- Developed FALCON, a connectionist network integrating fuzzy logic principles.
- Implemented FALCON-ART, combining backpropagation for parameter tuning and Fuzzy ART for structure learning.
- Utilized dynamic input/output space partitioning based on training data distribution.
Main Results:
- FALCON-ART dynamically partitions input/output spaces and tunes membership functions.
- The system learns fuzzy logic rules online without a priori information.
- Successfully controlled two unstable nonlinear systems: the seesaw and inverted wedge systems.
Conclusions:
- FALCON offers a flexible and adaptive approach to fuzzy logic control.
- FALCON-ART overcomes limitations of fixed-grid fuzzy systems and enables autonomous learning.
- The proposed method demonstrates significant potential for controlling complex nonlinear systems.