Sistema autónomo de organización usando visión artificial RGB-D y robot colaborativo en banda transportadora

Abstract

This project addresses the design and implementation of an autonomous manipulation system that integrates RGB-D computer vision and collaborative robotics. The objective is to develop a solution capable of detecting, segmenting, and classifying moving parts on a conveyor belt, justified by the industrial need to optimize operational efficiency with minimal human intervention. For the implementation, a hardware architecture was developed combining processing on a NVIDIA Jetson Nano, an Intel RealSense depth camera, and a robotic arm, communicating via MQTT protocol and ROS2 topics. The system core integrated the YOLOv5-seg model with the inverse kinematics algorithms to synchronize detection with the robot actuation in real-time. Experimental results validated that the Deep Learning model successfully identified objects and predicted their position using the camera image, demonstrating a high capacity for generalization across different categories. It is concluded that the proposed control architecture exhibits stability against position and speed variations, being able to guarantee robust autonomous operation. For future work, implementing dynamic calibration modes and speed estimation systems is recommended to refine individual trajectory calculations. Keywords: Computer vision, collaborative robotics, industry 4.0, deep learning, automation.