A team from Southampton, Edinburgh, Tokyo, and Delft just published a paper in npj Robotics demonstrating a soft underwater wing that senses its own deformation and uses that signal to reject flow disturbances autonomously.
The wing uses a liquid-metal capacitive e-skin. Six EGaIn electrodes in silicone produce nine capacitance signals to estimate its camber in real time. When a sudden gust hits, the wing’s flexibility causes a characteristic shape oscillation. The controller detects that signature and hydraulically morphs the wing to compensate.
The result is an 87% reduction in unwanted lift impulse compared to a rigid wing, using nothing more than a proportional controller and threshold-based detection.
What matters here is the design philosophy. The softness is the sensing mechanism. Deformation under fluid load becomes the primary signal for disturbance detection, the same way fish fin rays and bird feather mechanoreceptors work. The passive compliance of the material handles baseline gust mitigation on its own (three times better than rigid), and active camber control mops up the residual bias. They call it hybrid passive-active disturbance rejection, and it performs roughly twice as well as a barn owl’s gust-rejection maneuver, with the caveat that cross-domain comparisons are imprecise.
The implications for personal submarines and long-range UAVs are immediate. Underwater vehicles burn enormous energy on thruster-based station-keeping in currents — compliant control surfaces that passively absorb disturbances while actively trimming residual error could extend operational range significantly.
The e-skin is body-shape agnostic, meaning it can wrap around different fin geometries without redesigning the sensor architecture. The current limitation is actuator speed (1.7-second rise time on hydraulic), which the team’s follow-on ICRA 2026 work addresses with a formal disturbance observer.
The real story here is when you let the structure do the sensing instead of bolting instruments onto rigid frames, you get embodied intelligence that scales naturally with the problem it’s trying to solve.