AUTONOMOUS ROBOT NAVIGATION BETWEEN ROWS OF CROPS

Abstract

The paper describes autonomousUGV robot navigation among rows ofcrops. The objective of the study was to explore the most relevant approaches and robots used for this task. Two methods were compared for row tracking: traditional image processing and machine learning methods. A specific task included recognizing the end of row crops and navigating the robot from one row to another. An RTK-GPS sensor was used for detecting the end of the row crops, while a NAV 2 Waypoint Follower based on the PurePursuit algorithm was embedded for generating and tracking paths which are guiding the robot to the next row crop. The entire system was integrated into the ROS 2 environment. The approaches were tested in both real and simulation environments. NVIDIA Isaac Sim simulation was used for the simulation environment, from which semantic datasets were collected for training theYOLOv8 row crop tracking model. Finally, an evaluation and comparison of the methods were conducted.