3.2 Prior Map Construction Test

To equip our robots with comprehensive environmental information, we initially establish prior maps across the four scenarios in Fig. 3. This vital stage paves the way for accurate navigation and path planning. The resulting laser maps and point cloud maps are shown in Fig. 4.
The mapping results within crop rows are showcased in Fig. 4(a) and (e). Both parallel crop rows and on-road obstructions are distinctly depicted. Furthermore, even the detailed features of plant foliage on both sides of the path are effectively represented. Fig. 4(b) and (f) display the mapping results within a landscape greenhouse. The paths are initially and clearly depicted. Following this, the precise details of the plants bordering the paths are effectively mapped. These accurate maps are crucial for navigation, as the determination of whether a path is feasible relies on precise measurements of plant height. Fig. 4(e) and (g) capture the mapping outcomes in complex greenhouse environments, characterized by overlapping plants, irregular surfaces, and small streams. The pathways are distinctly segmented, with even the uneven road sections, indicated by the red circles in the figure, being well represented. Despite a few obstructions, detailed features of the plants alongside the pathways are also clearly depicted in the map. Lastly, Fig. 4(d) and (h) reveal the mapping results of low-lying plants. The brick-paved road and the plants on either side are accurately reconstructed. As demonstrated in the figure, low-lying vegetation is effectively represented. The muddy path between the plants, though segmented, is less intuitively discernible.
Fig. 4 showcases the distinctive mapping capability of the improved LeGO-LOAM algorithm in diverse greenhouse environments. The algorithm’s robustness and high accuracy in large-scale complex greenhouses provide a crucial basis for precision-oriented subsequent robotic navigation.