IMPLEMENTASI ALGORITMA YOU ONLY LOOK ONCE (YOLO) UNTUK SISTEM DETEKSI OBJEK PADA MOBILE ROBOT PENGANGKUT BARANG FORKLIFT

Abstract

The advancement of robotics and computer vision technology plays a vital role in supporting modern industrial
automation systems. One of the main challenges in developing autonomous goods carrier robots, such as self-driving forklifts,
is achieving accurate and real-time object detection in complex environments. This study implements the latest version of the
You Only Look Once algorithm, YOLOv11, as an object detection system for a mobile goods carrier robot. The dataset
consists of eight object classes—wooden pallets, milk boxes, Le Minerale cartons, Aqua cartons, mineral water gallons, Aqua
gallons, safety helmets, and Indomilk cartons—comprising 4949 images and 1250 manually annotated objects. The model
was trained for 100 epochs on an NVIDIA A100 GPU using 640×640-pixel images. The experimental results achieved a
Precision of 0.897, Recall 0.882, mAP50 0.928, and mAP50–95 0.870. The highest accuracy was obtained for the mineral
gallon class (mAP50 = 97.3%), while the lowest was for Aqua cartons (mAP50 = 84.6%). These results demonstrate that
YOLOv11 provides excellent, stable, and efficient detection performance, making it suitable for real-time object recognition
in industrial mobile robots.