Hardware Implementation Neural Network Controller on FPGA for Stability Ball on the Platform

Abstract

This study is about a development and investigation of the neural network controller of the stabilization system of a moving object on a plane with its hardware implementation on the FPGA. It consists of a designing balloon balancing model, hardware and software for this layout. The platform ball balancing system consists of a black plastic plate with a white table tennis ball, a drive mechanism for tilting a plate around two axes, a digital video camera, tracking the position of the ball, hardware and software that processes the information and manages the system in real time mode. The physical modeling of the system is carried out and the equation of motion of the ball in the plane is deduced. The equations of motion are nonlinear and unsuitable for the synthesis of control systems based on linear control theory. A generalized block diagram of a neurocontroller based on the FPGA, which implements the basic components of neural network control systems, is developed. To control the position of the ball on the platform a neural network control system with feedback was developed. The scheme is a classical scheme of specialized inverse training.