Researchers have built a robotic foot with motor-driven retractable cleats that automatically adjust their depth based on terrain penetrability, enabling bipedal robots to walk on steep granular slopes like sand or gravel. This work demonstrates that active terrain manipulation via cleats—commonly seen in animal feet—can be transferred to humanoid robots, opening the door for reliable locomotion on loose, flowable surfaces.
What the Researchers Built
The team designed and tested a novel robotic foot that can extend or retract cleats—small, blade-like protrusions—depending on the ground's hardness. The foot monitors motor current as a proxy for terrain penetrability: when the cleat meets resistance (rigid ground), negative current signals retraction; when it sinks into granular media (loose sand), positive current deploys the cleat to its full depth.
Two robots were used. BLUEY is a small bipedal platform (15‑cm feet, 13‑slit sole) used for systematic experiments with varying cleat spacing (1, 4, and 12 cm) and depth (1–3 cm). HECTOR is a full-size autonomous biped (85 cm tall, 15 kg) with 5‑DOF legs and 17‑cm × 10‑cm feet. HECTOR’s cleats are 3D‑printed as integral parts of the foot. Both robots were tested on slopes of poppy seeds (a safe granular analog) at inclines up to 25°.
Key Results
Without cleats, BLUEY stalled on a 20° granular slope—friction alone could not overcome the downslope gravitational pull. Adding dense, deep cleats (1 cm spacing, 3 cm depth) allowed BLUEY to climb the same slope reliably. In contrast, sparse cleats (4 cm spacing, 2 cm depth) actually worsened performance: the trailing cleat dragged through the material, fluidizing the soil and causing backward pitch.
HECTOR validated the principles at scale on a 15° slope. With 1 cm‑spaced cleats, the robot walked autonomously without slipping. When cleats were too sparse (16 cm spacing), the robot struggled to maintain forward progress.
Quantitative force measurements from a dual‑plate intrusion apparatus showed that closely spaced plates (1 cm) generated up to 1.8× higher granular resistive forces than widely spaced plates (5 cm) on a 30° incline. This matches the intuition that dense cleats engage more granular material, creating a larger reaction force.
How It Works
The physics of cleated locomotion on granular media hinges on how the foot interacts with the flowing soil. As the foot sinks and pushes backward, cleats act like small plates that shear the granular material. Closely spaced cleats (≈1 cm) create an almost continuous blade, mobilizing a large volume of grains and generating high resistive forces. Widely spaced cleats allow material to flow around each individual blade, reducing effective friction.
The researchers used Particle Image Velocimetry (PIV) to visualize soil flow around the cleats during walking. They observed that a trailing cleat—especially when shallow—acts as a dragged plate that fluidizes the adjacent material, creating a low‑density zone that cannot support the robot’s weight. Deeper cleats (≥3 cm) cut below this fluidized layer and engage denser, stronger material.
The active retraction mechanism uses a simple control law: monitor the motor current driving the cleat extension. On hard surfaces, the motor encounters resistance and current becomes negative; the controller retracts the cleat to avoid stalling or damaging the foot. On soft granular slopes, positive current indicates the cleat is sinking in, and it is fully deployed. This allows the robot to transition seamlessly between rigid and loose terrain.
A summary of BLUEY’s experimental outcomes on a 20° slope:
| Cleat Spacing | Cleat Depth | Result |
|---|---|---|
| No cleats | – | Stagnation (failure) |
| 4 cm | 2 cm | Backward pitch (failure) |
| 1 cm | 3 cm | Successful climb |
Why This Matters for Robotics
Humanoid robots are currently confined to structured, hard floors. This research provides a low‑cost, mechanically simple method to extend bipedal locomotion to outdoor terrains like beaches, construction sites, and disaster rubble—where granular soils are the norm. The cleat approach mimics biological solutions (e.g., bird feet, desert lizards) and could be integrated into any legged robot.
For operators of humanoid robots used in search‑and‑rescue or outdoor logistics, this technology means robots can now traverse sloped sand, loose gravel, and icy snow without custom tracks or constant gait tuning. The active retraction also prevents damage when walking indoors, making the cleat foot a “set‑and‑forget” solution.
Manufacturers of warehouse robots could adapt the concept for wheeled or tracked platforms operating on uneven granular floors, though the paper focuses on bipeds.
Limitations and Open Questions
The experiments used a single granular material (poppy seeds) at fixed slopes and walking speeds. Real‑world soils vary in grain size, moisture, and compaction—each may shift the optimal cleat spacing and depth. The active retraction controller is also heuristic (based on motor current) and may fail on mixed terrain where penetrability changes rapidly.
Scaling to heavier, faster robots remains untested. The paper’s force measurements suggest that cleat geometry must be carefully tuned for each robot’s weight and foot size. Finally, the long‑term durability of retractable cleats under repeated impact on hard surfaces is unknown.
Frequently Asked Questions
Why do cleats help on granular slopes but not on rigid ground? Cleats penetrate loose grains and mobilize friction, but on hard surfaces they cannot dig in, so they become obstacles that reduce contact area and stability. The active retraction mechanism solves this by withdrawing cleats automatically.
How does the robot know whether to extend or retract the cleat? It monitors the motor current driving the cleat actuator. On rigid ground, the motor encounters resistance and draws negative current; on soft ground, positive current signals the cleat is sinking in.
What is the best cleat spacing for granular walking? The paper found that 1 cm spacing (dense) works well for both BLUEY and HECTOR. Wider spacings (4 cm and above) caused the trailing cleat to fluidize the soil and reduce traction.
Can this approach work on wet sand or mud? The experiments used dry poppy seeds as a granular analog. Wet sand has cohesive forces that may improve or impede cleat penetration—further study is needed, but the principles likely apply with re‑tuned cleat geometry.
Conclusion
This work demonstrates that retractable cleats can give bipedal robots the ability to walk on steep granular slopes without sacrificing performance on flat hard ground. By combining careful cleat spacing and depth with a simple terrain‑sensing mechanism, the researchers have made a practical stride toward all‑terrain humanoid locomotion.