Underground-mine navigation paper validates GPU-free autonomous stack on Spot
Evidence notes
- The paper presents a fully autonomous underground-mine navigation stack for Boston Dynamics Spot running on a low-power Intel NUC without GPU or network dependency.
- The system combines LiDAR-inertial odometry, scan-matching localization, terrain segmentation, visibility-graph planning and local path following.
- Field trials covered more than 700 m of autonomous traversal across 20 runs in an experimental underground mine.
Company context
Boston Dynamics builds high-performance mobile robots including Spot for industrial and field inspection, Stretch for warehouse trailer unloading, and Atlas for advanced humanoid mobility.