Cyber-Physical System for Synthesizing Adaptive Fuzzy Controllers Based on Modified ANFIS Neuro-Fuzzy Network

A system for synthesizing adaptive fuzzy controllers is proposed using a modified ANFIS neuro-fuzzy network. The controller is adjusted by combining gradient descent with the least squares' method. Studies show that this adaptive fuzzy controller greatly improves transient process quality over PID controllers and non-optimized fuzzy controllers, reducing overshoot by four times compared to an untuned controller and by half compared to a PID controller, with transient time cut by three times. Analysis confirms its adaptive properties and reliable operation with other control loops. The main limitation is needing membership functions that allow differentiation, and fitting onboard aviation systems.