Self-Constructing Neural Network Controllers with Adaptive Structural Learning: Application to Stepper Motor Positioning

The performance of neural network controllers is highly dependent on the number of neurons in the hidden layers. An insufficient neuron count leads to slow response, while an excessive number may cause instability. This paper presents a self-constructing neural network controller that automatically adjusts the number of neurons to ensure both efficiency and stability. The proposed method is validated on stepper motor position control and compared with conventional neural network controllers.