Reliable communication is a bottleneck for many real-world deployments of unmanned systems—especially in environments where conventional radio struggles. Underground tunnels, caves, mines, and seawater are all “lossy” media for high-frequency (HF) electromagnetic waves: signals can attenuate so quickly that links become unreliable or disappear entirely. In these settings, teams often need a fallback channel to pass critical information—coordinates, status updates, simple commands—between agents operating in different media (e.g., underground to surface, underwater to air).
Low-frequency (LF) and extremely low-frequency (ELF) signals can propagate more effectively through rock and seawater. The challenge is practical hardware: traditional LF transmitters rely on electrically resonant antennas whose size is tied to wavelength, making them extremely large and inefficient for mobile platforms. This is where mechanical antennas become attractive.
Why mechanical antennas
A mechanical antenna generates a time-varying field through motion rather than a long resonant conductor. In simple terms, instead of driving a large electrical antenna, we physically oscillate a magnetized element to create a changing magnetic field at very low frequencies. This approach relaxes the “antenna must be huge at low frequency” constraint and opens a path to LF transmitters that are small enough to mount on robots.
However, many existing mechanical antennas still face a practical trade-off: designs that produce strong fields often require bulky drives (motors, heavy structures), while miniaturized designs tend to emit weaker fields that limit range. Our goal in this work was to push mechanical antennas toward field-deployable cross-medium links: compact, lightweight, and compatible with real unmanned platforms—without sacrificing tens-of-metres communication capability.
What we built
We developed a Flexible, Magnet-Based Miniaturized Low-Frequency Mechanical Antenna (FMLMA). The device combines a flexible permanent-magnet film (a silicone-rubber matrix loaded with NdFeB magnetic powder) and a thin piezoelectric actuator (macro-fibre composite, MFC) that provides controlled vibration.
A key design requirement was conformability. In practical deployments, antennas often need to mount on curved shells, protective housings, or airframes. Our antenna can bend beyond 180°, enabling attachment to curved surfaces. At the same time, it remains small and light: < 6.8 cm³ in volume and < 50 g in mass, driven by a compact 12 V electronics module (about 26.6 g). These engineering details matter because they determine whether a device can be carried by a UAV or a legged robot without compromising mobility.
A key engineering trade-off: magnet strength vs flexibility
Achieving both strong emission and flexibility required careful material choices. Increasing magnetic powder loading raises remanent magnetization, which can improve emitted field strength. But it also increases stiffness, reducing flexibility and potentially degrading vibration performance.
We evaluated NdFeB loadings between 60–85 wt% and found a practical “sweet spot” around 80 wt%: magnetic performance continues to improve with loading, but the stiffness increase becomes increasingly costly above this point. Selecting 80 wt% allowed us to keep the antenna flexible while maintaining strong magnetic output—an important step toward a conformal, miniaturized LF transmitter.
Manufacturing for shape and repeatability
To support integration on real platforms, we fabricated the magnet film directly on the actuator using five-axis 3D printing/dispensing, followed by curing and pulsed magnetization (~3.0 T) to align magnetic domains. This workflow is compatible with customized shapes and geometries, and it helps reduce assembly complexity—important for future iterations where arrays or platform-specific form factors may be required.
Performance: range scaling, arraying, and durability
For low-frequency magnetic near-field transmission, field strength typically decays rapidly with distance. We measured the radiated magnetic field versus distance and observed a strong inverse-cubic (1/r³) relationship, consistent with near-field magnetic dipole behaviour. Using this measured trend, the single-antenna signal remains detectable down to a 1 pT threshold at an estimated distance of about 60 m.
We also explored a straightforward route to stronger fields: arraying. A two-antenna array achieved a measured 1.74× improvement over a single antenna. The gain is below the ideal factor of two primarily due to real-world mismatches (e.g., small differences in resonant frequency from manufacturing tolerances). This points to a clear path forward: tighter fabrication control and closed-loop synchronization can push array performance closer to the theoretical limit.
Durability is another practical requirement. In many deployments, a conformal antenna will experience repeated bending during installation, transport, or operation. After 1,000 bending–unfolding cycles beyond 90°, the antenna’s amplitude variation remained within roughly 1%, with negligible change in measured field strength. This stability supports the idea that conformal mounting can be practical without frequent recalibration.
Cross-medium validation: underground, rock, and seawater
To evaluate real cross-medium usefulness, we tested the antenna in scenarios that mimic operational constraints.
Underground to surface (through-the-earth):
We transmitted a “Hello” voice signal (encoded message) from approximately 20 m underground to the surface. In this environment, a mobile phone and GPS signals were fully blocked underground, while our LF magnetic link remained measurable at the surface with signal strength consistently above 20 pT.
Through rock (between adjacent tunnels in a cave):
At the same separation, the measured field in air was 37.2 nT, while transmission through rock was 36.7 nT—only ~1.34% attenuation through rock. For comparison, under these conditions HF alternatives were far more limited (e.g., mobile phone loss of signal, and walkie-talkie performance constrained through rock).
Through seawater:
In a seawater test, the LF signal strength was about 12% lower than in air at the same distance. Extrapolating from measured trends suggests an approximate detection distance of ~52.4 m at a 1 pT threshold. Again, the key point is not high throughput, but the feasibility of a robust control-link class channel in seawater where many HF systems become ineffective over very short ranges.
Application focus: enabling unmanned collaboration in a cave
Beyond link-level metrics, we wanted to demonstrate a capability that matters operationally: collaboration between unmanned systems when conventional communication fails.
We built a field demonstration in a natural cave environment. A UAV (DJI Matrice 30) carried the antenna and drive electronics. The UAV identified target markers inside the cave and transmitted encoded coordinate information to a quadruped robot (Unitree Go2) positioned outside. On the receiver side, the system extracted the carrier component and used short-time Fourier transform (STFT) analysis and a simple threshold decision to decode binary symbols. The quadruped then navigated using the decoded coordinates and reunited with the UAV within about one minute in the demonstrated case.
This demonstration emphasizes the intended value of the system: in environments where RF links, GPS, and vision-based communication can all degrade, a compact LF mechanical antenna can still deliver mission-critical information for coordination.
What this link can carry today
LF links are inherently bandwidth-limited. Our current system is best viewed as a reliable low-rate channel for commands, coordinates, and status rather than for high-data-rate payloads.
In our experiments with a 4.0 Hz antenna, we achieved error-free decoding at about 2 bit/s using the threshold method. At higher rates, bit errors appear, but we also observed high waveform consistency for identical symbols/letters; leveraging this, we demonstrated decoding using a correlation-coefficient recognition approach at 4 bit/s. Future work can extend these results with improved modulation, coding, and synchronization strategies—especially important for arrays.
Outlook: toward practical cross-medium LF infrastructure
Because the antenna is miniaturized, lightweight, and conformal, it is suited for integration into space-constrained platforms. More broadly, we see this as a building block for a 3D cross-medium communication concept spanning subsurface, underwater, and open-air domains (as illustrated in Fig. 1). One promising direction is hybrid relaying: using LF links to cross difficult media boundaries (ground–air, water–air), then translating to higher-frequency channels for longer-range backhaul.
Overall, our results suggest that flexible, magnet-based mechanical antennas can move LF transmission closer to field deployment—enabling robust coordination where conventional RF methods struggle.