The paper introduces SHOES, a method for legged robots that borrows an idea from human locomotion: changing footwear to match the terrain. Just as hikers switch to boots or skaters to skates, the approach adapts the robot's ground-contact interface to enable traversal of otherwise difficult surfaces.

SHOES stands for Sim-to-real Hardware-adaptation via On-policy Expansion of Skills. The key move is to treat the contact interface as a modifiable component and to expand the robot's skill set through on-policy learning, which helps bridge the gap between simulation and real-world deployment.

The abstract is brief, so details on the specific hardware and experiments are not yet available. Still, the framing suggests a shift from fixed robot morphology toward adaptable, terrain-specific contact designs—a practical step for legged robots operating in varied environments.