ArborSplat improves online 3D mapping for orchard robots
ArborSplat: Online Semantic Gaussian Splatting SLAM for Orchards
Robotics
Summary
Orchard robots need detailed 3D maps that accurately show small but important features like tree trunks and trellises. Existing mapping methods often miss these thin structures because they rely mainly on image appearance. The authors created ArborSplat, which combines 3D Gaussian splatting with semantic information and height constraints to better identify and preserve these features. This system runs fast, uses LiDAR for position tracking, and improves accuracy in various orchard conditions such as dormancy and flowering.
What this means in practice
- •For agricultural robotics teams: Build real-time 3D semantic maps of orchards that accurately preserve thin important structures for better robot navigation and task execution.
- •For autonomous vehicle developers: Enhance online semantic mapping in environments with thin or small structures by integrating class-specific geometric constraints and multi-view fusion.
Authors
Alessandro Masini, Matteo Frosi, Mirko Usuelli, Matteo Matteucci
Abstract
Orchard robots need maps that preserve small but semantically important structures such as trunks, trellises, and fruit. 3D Gaussian Splatting (3DGS) SLAM achieves high photometric fidelity. However, its optimization remains appearance-driven, and transferring image semantics to 3D points is unreliable for thin structures, whose pixels may receive depth from background surfaces. We present ArborSplat, an online semantic 3DGS SLAM system that tracks with LiDAR odometry and optimizes semantics directly on the Gaussian map, constrained by class-specific height bands above a ground plane fitted to each keyframe's stereo point cloud, and fuses multi-view evidence into a semantic point cloud online while rejecting labels inconsistent with the local ground surface or with monocular depth. Class-constrained refinement reserves Gaussian capacity for underrepresented structures and, under reduced budgets, increases training-view accuracy on tree classes. We evaluate the approach on apple and pear orchards during dormancy, flowering, and harvesting. On full routes, it keeps ATE below 0.5 m on all 12 traversals. On shared 301-frame segments, it exceeds SGS-SLAM and GS3LAM by 0.23 to 0.50 training-view and 0.15 to 0.36 held-out mIoU while running 1.7 to 7.5 times faster, whereas SemGauss-SLAM runs out of GPU memory on all six.