April 2025 | Computers and Electronics in Agriculture | Volume 231
Introduction: Agricultural robots could help automate labor-intensive tasks such as harvesting, pruning, and crop monitoring, but deploying them in small-scale greenhouse farms remains challenging. Fujinaga focused on high-bed cultivation environments, where uneven surfaces, narrow passages, changing objects, and repetitive cultivation structures can make conventional self-localization unreliable. In particular, LiDAR data collected between parallel cultivation beds contain few distinctive features, increasing the risk of localization errors. To address this problem without requiring additional infrastructure such as landmarks, floor lines, or magnetic tape, the study developed a hybrid autonomous navigation method for a compact crawler-type agricultural robot. The system switches between waypoint navigation for moving across the farm and real-time, LiDAR-based cultivation-bed navigation to maintain precise movement along crop rows. The method was first evaluated in a Gazebo virtual environment and then tested in a real high-bed strawberry greenhouse.
