Quadrupedal wheel-legged robots can use wheel kinematics to estimate motion, especially when LiDAR geometry offers few constraints. But the abstract notes that slip and contact loss make rolling predictions unreliable, and that contact state alone is insufficient.
The paper, titled "Learning Wheel Motion Uncertainty for Degeneracy-Aware LiDAR-Inertial Odometry on Quadrupedal Wheel-Legged Robots," proposes learning a model of wheel motion uncertainty. This is intended to let the odometry system weigh wheel predictions appropriately in degenerate conditions.
The significance is practical: robots operating in corridors, open spaces, or other geometrically weak environments may maintain better localisation by accounting for how much to trust the wheels.