Underwater robots often lose the clearest clues about nearby objects once cameras face murky, low-light water. Researchers built flexible, whisker-like probes using Fiber Bragg Grating (FBG) optical sensors and showed how two channels can capture relative flow, foil-generated wakes, and signals useful for choosing a robot route. The approach brings tactile-style hydrodynamic sensing to platforms that cannot rely on vision alone.
What Did the Researchers Build?
The research team developed a bioinspired hydrodynamic sensor that gives an underwater robot a sense of local water movement. The device uses flexible whisker-like structures instrumented with Fiber Bragg Grating sensors. An FBG is an optical sensing element embedded in a fiber; when the fiber bends or stretches, the wavelength of reflected light shifts. That shift provides a measurement of mechanical strain without requiring conventional electrical wiring at the sensing point.
The study examines two sensing channels, identified as FBG 1 and FBG 2. Rather than attempting to reconstruct a complete underwater image, the system measures changes in the surrounding flow. Those changes include steady relative motion caused by towing, oscillatory disturbances generated by a foil, and the downstream wake left by a moving source.
The experiments were designed around a practical robotic question: can a robot use hydrodynamic signals to decide between two possible routes or branches? The classifier therefore used only the two FBG channels. Robot pose, velocity, yaw, and visual tracking were deliberately excluded, isolating the information available from the whiskers themselves.
This design targets situations where cameras struggle, including turbid water, darkness, bubbles, and environments with limited visibility. It also offers a lightweight sensing concept that can be attached to a vehicle without relying on a large sonar or a complex underwater vision system.
What Were the Key Results?
The experiments established three basic operating behaviors under controlled towing: how the whisker handles vortex-induced vibration, how its signal changes with relative flow speed, and how its response depends on angle of attack. Angle of attack describes the orientation of the whisker relative to incoming water, so this test indicates how mounting direction affects the signal.
The wake experiments then separated two physically different cases. One used a pitching foil held in place in otherwise still water; the other used a pitching foil that translated through the water. That distinction matters because a moving source leaves a downstream-convected wake, which is closer to the trail a robot would encounter while following another moving object.
For the robot decision task, the system used a binary classifier trained only on the two optical channels. The reported data split contained 463 training windows, 114 validation windows, and 186 held-out test windows. The split occurred at the trajectory level before extracting windows, preventing samples from the same trajectory from leaking into both training and testing.
| Evaluation detail | Reported setting |
|---|---|
| Input channels | FBG 1 and FBG 2 |
| Training windows | 463 |
| Validation windows | 114 |
| Held-out test windows | 186 |
| Classifier window length | 2.0 seconds |
| Resampled classifier rate | 500 Hz |
The supplied paper text does not include the final accuracy, precision, recall, or comparison score, so those metrics cannot be stated reliably. The strongest documented result is the creation of a controlled sensing and classification pipeline that tests whether wake information alone can support branch selection.
How Does the Fiber-Optic Whisker Work?
The sensing process begins when water pushes against the flexible whisker. Bending and vibration create strain in the embedded optical fiber. The FBG converts that strain into a wavelength shift, producing a time-varying signal that reflects local hydrodynamic conditions.
The towing experiments provide a clean baseline. By moving the sensor through water at controlled speeds, the researchers can separate the effects of relative flow speed from disturbances created by a wake. The tests also examine vortex-induced vibration, or VIV, in which alternating vortices shed from a structure and drive it into oscillation. A useful whisker should provide a measurable flow response without allowing uncontrolled structural vibration to dominate the signal.
The foil tests add a more realistic disturbance. A pitching foil periodically changes its orientation and generates unsteady flow. When the foil also translates, its disturbance is carried downstream. The study uses the Strouhal number to describe the relationship between oscillation frequency, source size, and translation speed. This nondimensional quantity helps compare wake conditions across different operating regimes.
The classifier receives a trailing 2.0-second window from each FBG channel. Each signal is resampled to 500 Hz, detrended, mean centered, and low-pass filtered at 25 Hz with a sixth-order Butterworth filter. Zero-phase filtering avoids shifting features forward or backward in time. When a window reaches the beginning of a record, the earliest available value is repeated to pad the missing portion.
The towing geometry comparison follows a separate processing path: FBG 2 traces are interpolated to 1 kHz, baseline corrected, and smoothed with a third-order Savitzky–Golay filter using a 31-sample window. A common fixed-duration segment is used across geometries, avoiding geometry-specific peak searches that could make comparisons unfair.
Why Does This Matter for Robotics?
Underwater robots need reliable perception when cameras cannot see far enough or when sonar is too expensive, heavy, or slow for a particular task. A flexible hydrodynamic whisker offers a different sensing mode: it detects what the water is doing rather than trying to directly image an object.
That capability could support wake following, obstacle awareness, relative positioning, and route selection around submerged structures. A robot might detect the signature of a nearby vehicle, recognize that it has entered a disturbed flow region, or compare two channels or passages based on their water-motion patterns.
The approach also has potential value for compact inspection robots and bioinspired vehicles. Optical fibers are lightweight and immune to electromagnetic interference, while the flexible mechanical element can extend sensing beyond the robot body. The results are not yet a product-ready perception stack, but they identify a path toward sensor-rich underwater platforms with fewer assumptions about visibility.
For comparison with other autonomous mobile systems, buyers can browse warehouse robots on BotMarket, where route selection and environmental sensing are also central concerns. The broader integration challenge resembles sensor retrofits used with used industrial robots: the value depends not only on the sensor, but also on mounting, calibration, data processing, and task-specific software.
What Are the Limitations and Open Questions?
The supplied results do not report final classifier accuracy or show how the method compares numerically with sonar, cameras, conventional strain sensors, or earlier artificial whiskers. That makes it difficult to judge readiness for deployment.
The classifier also uses only FBG 1 and FBG 2. Excluding robot pose, velocity, yaw, and visual tracking is useful for isolating whisker information, but real systems will normally have access to those signals. More sensors could improve robustness, while also increasing calibration and integration demands.
Controlled towing and foil-generated wakes cannot represent every underwater environment. Future tests need to cover changing current direction, turbulence, vehicle speed, sensor fouling, salinity, depth, and irregular wakes. Long-duration durability and performance during full robot maneuvers remain important practical questions.
Frequently Asked Questions
What is a Fiber Bragg Grating whisker?
It is a flexible whisker-like structure containing an optical fiber sensor that measures bending and strain through changes in reflected light wavelength.
What can the sensor detect?
The experiments examine relative flow, vibration behavior, angle-of-attack effects, and wakes generated by pitching foils, including moving foil sources.
Can the whisker replace underwater cameras or sonar?
The study does not establish replacement performance; it demonstrates a complementary sensing channel that works from local water motion rather than visual information.
How was the robot classification test structured?
A binary classifier used only two FBG channels and 2.0-second signal windows, with 463 training, 114 validation, and 186 held-out test windows.
Conclusion
Fiber Bragg Grating whiskers give underwater robots a compact way to sense flow and wake structure when vision is unreliable. The research establishes controlled flow tests and a leakage-resistant two-channel classification setup for hydrodynamic route selection. Further field testing and published task-performance metrics will determine how far the concept can move toward operational underwater autonomy.