Image Processing Supported Robot Mapping System

VØBEST is an image-processing-supported robot mapping system developed for ATP-Tech's autonomous service robot PuduBot. The system fuses LiDAR, IMU and odometry data in a ROS 2 framework, integrating a YOLOv8 object detector optimized via Teacher-Student knowledge distillation and an OCR layer with dictionary-based fuzzy matching. The result is a semantically enriched real-time map in .pgm/.yaml format, achieving 20–40 FPS inference and 91.5% mAP@50 on a Jetson Orin Nano—enabling intelligent autonomous navigation in dynamic indoor environments such as shopping malls.