분리수거 로봇 파지 정확성 향상을 위한 지능형 제어 시스템 설계 및 구현

Abstract

This paper presents an intelligent control system that combines deep-learning-based perception with adaptive robot control to enhance the grasping performance of a recycling robot. Conventional grasping methods lack adaptability to diverse waste shapes and materials, and their abrupt movements reduce dynamic stability. To address these issues, the proposed system is implemented using a Robot Operating System (ROS) and the You Only Look Once (YOLO) detection model, integrating three core techniques. First, a deep-learning-based contour detection method computes rotated bounding boxes to determine the optimal grasping angle. Second, adaptive grasping control is applied according to the object class identified by YOLO, enabling material-aware and stable grasping. Third, an S-curve velocity profile smooths acceleration and deceleration, enhancing motion stability during high-speed operations. Experimental evaluations show that the proposed system overcomes the limitations of traditional approaches and significantly improves grasp accuracy and overall waste-sorting efficiency, demonstrating its effectiveness for intelligent recycling robots.