🛸 This drone navigates in the dark with whiskers

A small drone can now move through darkness by gently touching its surroundings with two long artificial whiskers.

The idea comes directly from rats and moles. In the dark, their whiskers allow them to sense nearby objects and follow a wall. Researchers adapted this principle to a drone weighing just 44.1 g. Two flexible rods positioned at the front touch obstacles to prevent the body of the device from colliding with them.

The 44.1 g Crazyflie drone equipped with two artificial whiskers, showing the tactile sensor and its biological inspiration.

The 44.1 g Crazyflie drone equipped with two artificial whiskers, showing the tactile sensor and its biological inspiration.
Credit: Ye et al., Nature Communications (2026), CC BY 4.0.

Each whisker is 20 cm long and only 0.4 mm in diameter. When it bends against an object, tiny sensors at its base measure the pressure variations caused by the contact. The drone can then estimate how far away the obstacle is. The entire detection system weighs 3.2 g.

Contact therefore does more than simply signal that a wall is there. The drone obtains an estimate of the obstacle’s position with an accuracy of a few millimeters. It then adjusts its trajectory in real time. The researchers also carefully selected the whiskers’ orientation, because friction against a surface can destabilize such a lightweight device.

The tests show what this approach makes possible in practice. In complete darkness, the drone can move forward until it encounters a wall, then follow it. It also navigated a course containing several transparent panels. These surfaces can be difficult to detect with some systems that rely on light.

Tests of different artificial-whisker orientations while the drone followed a wall.

Tests of different artificial-whisker orientations while the drone followed a wall.
Credit: Ye et al., Nature Communications (2026), CC BY 4.0.

More surprisingly, the device can gradually build a representation of an enclosed space by exploring its walls through contact. It does not need to send its measurements to an external computer. Information processing, decision-making, and control are carried out directly on board, using a microcontroller with just 192 KB of RAM.

This method does not necessarily replace cameras or distance sensors. It mainly provides another way to perceive the environment when vision becomes unreliable. A small drone could therefore continue detecting obstacles in darkness, smoke, or certain confined spaces. Its low mass also allows it to reach places where a heavier device would be less suitable.

The technique nevertheless has limitations. Tests on flexible surfaces caused oscillations, particularly because of folds and friction.

The researchers envision this tactile perception for search and rescue, inspection, and the exploration of difficult-to-access locations. It remains to be determined how it will perform in real-world environments.

FI
Filou

The material of the mustache is not specified: fiberglass, kevlar, silk, nylon?
I think it will be adapted to the projected circumstances.
And it could be funny if it tickles!

NO
Nox42

192 KB of RAM to estimate obstacles, correct the trajectory, and build an on-board map is probably the most interesting detail. The system remains usable on a very limited microcontroller.

LU
LuluCosmos

A drone that navigates by touching walls like an animal is awesome 😄 Could we combine these whiskers with a camera so it can choose the best sensor on its own depending on the location?

SI
sicep

In a laboratory with clean walls and a few panels, the principle works. In rubble, cables, or fabrics that catch the whiskers, I would wait to see if the drone remains just as stable.

TS
Tsetse

I honestly love the 44-gram drone that feels its way around with its whiskers and maps itself out. However, in heavy debris, it might get some severe shaking.

ME
mecaNico

On a 44 g machine, 20 cm whiskers have to provide serious leverage at the slightest snag. The real test will be in cables, dust, and debris, not against a clean wall.

AR
ArianeB

On a drone aussi léger, 3.2 g of detection already represents more than 7% of the total mass. Achieving a localization within a few millimeters without offloaded calculation is an impressive integration performance.