← All work

CMU, 2024

Soft robot legs that bend the right way

Robots that crawl inside pipes can inspect and repair them without digging them up. We studied a soft crawler design driven by a single air muscle, where each leg bends and buckles to grip the pipe wall more in one direction than the other, and asked how the leg's angle and small notches change that behavior.

With Ben Spin, Aviva Young and Ella Zhao, as a team of four. Soft robotics course project (24-673), Carnegie Mellon, 2024.

Final report · 8 pages Download PDF

The problem

The design we started from, by Lin et al., uses one pneumatic McKibben actuator (an air muscle that shortens when inflated). Its legs press against the pipe wall through rubber pads, and the robot moves because the legs slip forward more easily than backward.

That direction preference depends on details of the leg: its angle, its cross-section and where it buckles. Notches can steer buckling, but there is little theory saying a simple notch will reliably send it the right way. We wanted to characterize those details.

Diagram and photos of the Lin et al. soft pipe crawler: a McKibben actuator between angled legs that press against the pipe wall.
The design we built on (Lin et al.). One air muscle compresses the legs, which buckle against the pipe wall.

What we did

In Ansys, we simulated twelve leg variants: angles of 30, 40 and 54 degrees with a notch on the left, middle or right, plus the unmodified designs. We ran a static analysis, then an eigenvalue buckling analysis, then a second static run seeded with the buckled shape, which gives realistic large deformations. A second model squeezed each leg between two rigid beams standing in for the pipe walls.

For the physical side, our first PLA leg cracked under repeated bending, so we switched to Tough 1500 resin on a Form 4 printer (roughly 10-hour prints). We printed legs at 54, 40 and 30 degrees, plus two 30-degree legs with different notch positions, and cast silicone rubber feet in a 3D-printed mold.

The test rig was a hand-crank platform with a digital force gauge. It pulled each leg between two acrylic sheets that acted as the pipe walls, and we recorded the forward and backward force needed to make the feet slip.

Hand-crank force gauge rig with acrylic walls on brackets, next to several 3D-printed resin legs.
Our friction test rig and printed legs. The gauge pulls a leg between two acrylic walls.

Results

Steeper legs pushed harder on the walls, and simulated shear stress nearly doubled going from 30 or 40 degrees to 54. But the number that matters for crawling, the ratio of backward to forward force, favored the smaller angles. That was the opposite of what we expected.

On buckling, a double-sided notch consistently produced one dominant buckling mode, which made behavior predictable, though not always in the direction we wanted. Most original designs buckled inward. Only the 30-degree leg with a left notch buckled outward, and only under extreme deformation.

We redesigned that leg with an extra notch to weaken its stiff ribbed section, and it then reliably buckled outward. A printed prototype matched the Ansys shape, and a combined squeeze-then-compress simulation showed it still buckled outward inside the pipe.

Line plot of the backward-to-forward force ratio over time for 54, 40 and 30 degree legs, with error bars.
Backward-to-forward force ratio. The smaller angles (30 and 40 degrees) held a higher ratio, which is what crawling needs.
Ansys post-buckling result for the altered 30-degree leg beside a photo of the printed leg bending the same way.
The redesigned 30-degree leg. Left: Ansys post-buckling. Right: the printed prototype bending to match.
Ansys total-deformation plot of the altered leg squeezed between two beams and buckling outward.
Squeeze-then-compress simulation. The altered leg still buckles outward and touches the wall.

What I took away

Notches promote buckling, but a simple notch does not decide its direction on its own: the analysis found more than ten possible modes. Double-sided notches are what made the direction predictable.

Simulating first paid off, because each flexible print took about half a day. And our intuition about leg angle was wrong, which is exactly why the physical tests were worth doing.

Tools ANSYS, 3D printing (Tough 1500 resin), Silicone casting

Report

Loading…