Challenges in Orchard Weed Management: Perspectives on the Use of 3D Cameras and LiDAR to Develop a Low-Cost Small-Scale Robotic Weeder

In orchard weed management, accurate information about the orchard environment, such as tree positions and diameters, weed locations, and obstacles, is essential for robotic weeders to enable autonomous navigation. Three-dimensional (3D) cameras and light detection and ranging (LiDAR) sensors have the potential to provide good perception for a weeding robot, as GNSS cannot provide consistent point cloud data due to canopy coverage in dense-canopy areas. However, the success of robotic weed control comes not only from navigation but also from the proper use of weeding tools. This chapter describes the potential of 3D cameras and LiDAR to support the development of low-cost small-scale robotic weeders for orchard operations. First, we discuss the complexity of the environment in an orchard, where a robot may have difficulty navigating under the canopy of fruit trees while avoiding surrounding obstacles. After that, the advantages and disadvantages of various perception sensors are detailed based on the literature. Finally, we introduce a weed control strategy for orchard weeding operations in terms of the navigation strategy and weed cutting mechanisms of small robots. The robotic weeder can be used for interrow and intrarow weeding, utilizing 3D cameras and LiDAR as the main sensors with the support of perception sensors other than those using satellite-based navigation.