FOSCo-Net: Ego-centric Voxel Completion for Forest LiDAR Occlusions
FOSCo-Net: Ego-centric Voxel Completion for Forest LiDAR Occlusions
This paper introduces FOSCo-Net, a self-supervised voxel completion framework designed to address systematic LiDAR occlusions in temperate forests caused by tree trunks. The method reconstructs occluded geometry from a single scan using temporally accumulated observations as automatic supervision. The authors evaluate the framework using both cylindrical and Cartesian ego-centric grids with multiple decoder designs, finding that a gated skip decoder achieves the most reliable performance by significantly reducing false obstacle predictions compared to baseline methods.
The authors propose FOSCo-Net, a self-supervised voxel completion framework for ground robots in temperate forests that reconstructs occluded geometry from single LiDAR scans using temporally accumulated observations to mitigate blind regions caused by tree trunks.
Source: arXiv Robotics — research abstracts · Read original article ↗
Loading article text…
What the source reports
Publisher-reported claims, with original evidence. These results have not been independently verified by RoboSignal.
- paper: not_reported
Source excerpts and review record
Automatically extracted; no manual editorial approval recorded.
arXiv:2611.06588v1 Announce Type: new Abstract: Ground robots operating in temperate forests suffer from systematic LiDAR occlusions caused by tree trunks, creating persistent blind regions that degrade traversability estimation. In this paper, we propose FOSCo-Net, a self-supervised voxel completion framework that learns to reconstruct occluded geometry from a single scan using temporally accumulated observations as
Open source S2
automatic supervision, and we evaluate cylindrical and Cartesian ego-centric grids with multiple decoder designs. FOSCo-Net consistently improves reconstruction quality and traversal safety, with a gated skip decoder achieving the most reliable performance and significantly reducing false obstacle predictions compared to baselines.
Open source S3
Source：arXiv Robotics — research abstracts · arxiv.org