Embedding the Radial Basis Function Neural Network based PID Controller in a Microcontroller for Controlling the cart's Position in Real-time

Proportional integral derivative (PID) is a classic conventional controller and widely utilized in industry. However, its coefficients are normally chosen by using empirical methods. If the parameters of the plant are changed in time or affected by uncertain noises, the conventional controller is not stable because of its fixed coefficients. Therefore, this paper proposes a method to embed the radial basis function neural network based PID controller in the STM32F4VE microcontroller to control the cart's position. Its coefficients are estimated in real-time using the Gradient Descent approach and the radial basis function neural network. Two controllers utilized to control the cart added 2.5 kg payload and 20cm movement. The conventional PID controller made the overshoot of 29.5 % while the proposed method is 2.2 %. The experimental results show that the proposed method can perfectly control the cart's position with the movement distance and the cart's payload changed as well.