UGV–UAV Bounding Box Mapping for Occluded and Dense Orchard Environments

Autonomous 3D mapping in agricultural environments is challenging due to dense vegetation, canopy occlusion, and degraded GPS signals, which hinder consistent spatial representation. While dense point clouds offer high fidelity, they are computationally intensive for real-time deployment. This paper investigates a collaborative unmanned ground vehicle (UGV) and unmanned aerial vehicle (UAV) mapping approach for orchard environments using a compact Bounding Box (BB)-based representation to preserve geometry for navigation. Three simulated scenarios with increasing complexity assessed robustness: aligned rows, misaligned rows, and canopy occlusion. Results showed the BB representation reduces map size by approximately 43%, maintains planar coverage above 95%, and preserves vertical canopy structure. Collaboration enhances spatial completeness, with the UGV contributing 64–79% of total coverage, complemented by UAV sensing. The BB abstractions also support simplified 2D projections and corridor analyses. Overall, the proposed BB representation is an efficient alternative, improving coverage in occluded regions for autonomous agricultural use.