A 3.2-gram whisker array guides a palm-sized drone through the dark by touch alone

Each flexible whisker sits over a MEMS pressure sensor at its base, an unusual thing to build a touch sensor out of, and contact bends the shaft into a differential pressure signal that the onboard code converts into tactile depth, meaning how far the obstacle has pushed into the whisker ^[14][6]. A depth reading puts the array in a different class from a bumper: earlier contact-based methods reported binary contact or relied on passive bumpers, while this system reports continuous feedback with explicit localization of where the contact happened ^[9].

A barometer's zero wanders, and that is the hard part here, because sustained contact is exactly when drift matters: a drone following the contour of a wall keeps a whisker loaded for seconds at a time. The pipeline the authors describe compensates for noise and drift and estimates depth to millimetre-scale accuracy ^[3]. Contact also disturbs the aircraft, and the paper frames the base-mounted barometers as a way to estimate depth in flight while minimizing contact-induced destabilization ^[2].

The paper, published by Nature ^[13], puts the ceiling for these platforms at 100 grams, where sensing, computation and power are all severely limited ^[7]. At 3.2 grams, the array takes 3.2 percent of that ceiling ^[1], and the whole system fits in a microcontroller with 192 KB of memory ^[4]. The abstract and introduction report the mass and the memory; power draw, trial counts and a comparison against a camera-based baseline on the same airframe are missing ^[17]. In place of that, the paper argues sufficiency: with touch as the primary modality, the drone avoided obstacles and contoured surfaces, and explored confined spaces, in both rigid and soft environments ^[5][15].

Touch has one hard limit: the array reports an obstacle only once a whisker reaches it. Two design choices make that workable: the low interaction force keeps contact safe with delicate or lightweight obstacles, and sliding contact along the shaft enriches the data as the whisker slips across a surface ^[10].

The alternatives in the dark are poor. Cameras, LiDAR and other rangefinders fail in smoke-filled buildings, unlit caves and around transparent or reflective obstacles ^[8], and aerodynamic proximity sensing from wall and ground effects is imprecise and sensitive to flight dynamics ^[12]. On that evidence I would expect whiskers to be added to drones that still carry an optical sensor for the cases where light works, because an array this light and this small in memory is affordable next to one ^[1][4].