Spot terrain-classification paper uses proprioceptive sensing for traversability

Evidence notes
- The paper develops a terrain classifier for Boston Dynamics Spot using proprioceptive signals from the robot instead of external perception.
- Field tests classify terrain types from leg and body signals such as joint angles, forces, and foot interaction measurements.
- The work gives Spot a benchmark for traversability-aware terrain classification using onboard robot state.
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.