A Wormlike Robot for Ferromagnetic Surface Inspection: A Novel System Inspired by Caterpillars
Parisa Parhami, Mohammad Hossein Salehpour, Hamed Hamzeh, Rezvan Nasiri, Hadi Moradi
IEEE Robotics & Automation Magazine (Early Access) · 2026
Overview
The Wormlike Inspection Robot is a bio-inspired system for inspecting ferromagnetic structures such as steel pipes, vessels, and walls. Inspired by the segmented body of a silkworm, it links three identical modules with spring joints. This flexible, modular body can follow curved surfaces and cross corners that are difficult for a rigid inspection robot to reach.
Each module carries a passive magnetic adhesion system, so the robot holds on without spending energy to generate adhesion — including during a power loss. The paper evaluates the design's maneuverability, stability, robustness, and energy efficiency through analytical models and physical experiments.
Demonstrations
Robot design
The design targets four goals at once.
- ManeuverabilityReach confined spaces and traverse internal and external corners, cylinders, and pipes with a three-module body.
- StabilityStay attached during complex maneuvers with a passive magnetic adhesion system.
- Energy efficiencyExploit the natural dynamics of cyclic crawling through a compliant, spring-linked body.
- RobustnessKeep maneuverability and stability in uncertain conditions and survive partial falls.
WEach module weighs about 400 g and measures roughly 8.5 × 12.5 × 6.5 cm. It has two identical driving sections joined by a passive central joint, allowing the module to adapt to uneven surfaces and edges. Each section includes a DC motor, timing pulleys, and a Hall-effect sensor for speed feedback. Lightweight PLA and Plexiglas parts help keep the center of mass close to the surface, reducing the chance of sliding.
We embedded neodymium magnets near the outer surface of 3D-printed flexible belts. This keeps the magnets close to the metal while letting the belts conform to irregularities; the grip needs no power to remain attached. Pairs of replaceable springs connect adjacent modules, helping the body bend around corners and supporting cyclic, worm-like locomotion over longer distances.
Results and analysis
The analytical model asks whether each module has enough magnetic grip to resist sliding as the direction of gravity changes. Corner transitions are especially demanding because only one magnet may be in contact while the body rotates across an edge. The model also identifies a design trade-off for turning: stiffer spring joints increase the minimum turning radius, while stronger adhesion permits a tighter turn. For a partial fall, the required adhesion is based on the case where one module must support the others.
The physical tests covered five internal and five external 90° corner transitions at different orientations to gravity. The robot completed a wall U-turn at its 25 cm minimum turning radius and climbed and circled a 15 cm pipe. In a demanding fall test, one attached module held equal-mass substitutes for the detached modules, plus an additional module. During some corner transitions, modules that briefly detached were brought back into contact by the rest of the body.
To see whether the compliant body could make travel more efficient, we compared simple locomotion with a cyclic pattern that oscillates the middle module relative to the outer two. We ran ten trials at each of nine frequencies on a 4 m flat, non-ferromagnetic path. At the estimated natural frequency of 1.5 Hz, the cyclic pattern increased forward speed by 7.4% and reduced cost of transportation by 5.6%. The strongest measured result came at 0.4 Hz: 10.2% higher speed and 6.7% lower cost of transportation. That lower-frequency result suggests the robot's nonlinear dynamics matter alongside its estimated natural frequency.
| Metric | Value |
|---|---|
| Min. adhesion force for falling robustness | ≈ 33 N |
| Min. turning radius | 25 cm |
| Pipe diameters (inside / outside) | ≥ 20 cm / ≥ 15 cm |
| Forward-velocity gain at 0.4 Hz | +10.2% |
| Cost-of-transport reduction at 0.4 Hz | −6.7% |
Limitations
- Simplified dynamics: The analytical models ignore the wheels' rolling dynamics, which is valid only at constant speed or when acceleration is small.
- Fall testing: Falling robustness was verified with equal-mass substitutes rather than full failure modes.
- Test surface: The energy-efficiency gains were measured on a flat, non-ferromagnetic path, not during on-surface inspection.
Future work includes tuning the joint stiffness, adding image-based scanning and path planning, and tracking planned paths under stability constraints.