Robotic System Position Control Algorithm Based on Target Object Recognition

Abstract

Creation of robotic systems capable of manipulating objects of the real world is an actual problem allowing both raising labor efficiency and reducing traumatism risk for a person. In the paper, the task of large-node replacement of the outdoor lighting luminaires with the use of robotic system is considered. For this, the accompanying tasks connected with the object (luminaire) identification and positioning of robot gripper concerning object have been solved. The resulting algorithms allow us to solve tasks in conditions of varying visibility, different backgrounds, overlapping objects, if necessary, positioning on certain parts of the target object. They can be used to identify of any elongated shape objects after appropriate training of the neural network. These algorithms allow us to position the robot arm in such a way that it can take the necessary object with the help of the gripper. The practical significance of the solved problem is connected with the possibility of robotics systems practical application in the human environment and the creation of anthropomorphic robots.

Keywords

• pattern recognition
• image recognition
• object detection
• neural network
• robotic system
• identification
• self-positioning