Automated Control System of a Modular Robot Manipulator for Sorting Objects Based on Neural Networks

This article is devoted to the development of an automated robot manipulator control system to sort objects based on the use of neural networks. This article discusses the processes associated with the creation of an automated control system and the development of its components. Specifically, a system is described that is designed to detect, classify, and determine the position of objects, as well as a system that calculates the rotation angles of the robot manipulator modules. The principles of operation of the control system and its components, as well as the architecture of neural networks used in their development are given. The results of the experiments demonstrate the effectiveness of using neural networks to solve the tasks of automating the control of robot manipulators. The accuracy of object recognition on the scene was achieved equal to 82.7 percent for a test sample of data. In addition, the selected neural network architecture ensures the positioning accuracy of the system equal to 99.7 percent with an acceptable error of 1 degree.