Arcani's HARK portable acoustic drone detection sensor approaching parabolic sound mirrors in Kent, United Kingdom. The sound mirrors were built in World War 2 to 'listen' for incoming Luftwaffe aircraft before radar existed. The HARK sensor is the modern version of the passive detection method used in the parabolic mirrors.

Harry Howe at Arcani Systems Ltd.

Researchers at Worcester Polytechnic Institute (WPI) have developed palm-sized drones that use bat-inspired ultrasound and AI to navigate through fog, smoke, and other challenging conditions. The device mimics how bats use simple echolocation to navigate dark, cluttered environments with minimal neural processing.

The Ghost Murmur mission report illustrates how real-world operations depend on detecting and interpreting faint, ambiguous signals, aligning with the broader challenge of separating meaningful signals from noisy environments.

But sound has historically been difficult to use in operational environments. Acoustic data is inherently chaotic in real conditions, and isolation is the limiting factor. If sound can be altered, then isolating and interpreting it becomes more complex.

“The hard part is hearing it in the real world, where you have wind, vehicles, generators, surf, birds, echoes from buildings or trees, and often several sound sources at once,” said Harry Howe, CEO and founder of Arcani. “A small drone is rarely the loudest thing in the scene, and its signature changes with speed, aspect angle, throttle, altitude, terrain and weather."

Howe says that means the challenge is less about how a microphone can hear a sound and more about whether the system separates the right sound from everything else early enough to make it useful operationally.

For Arcani, its HARK (Hostile Acoustic Reconnaissance Kit) drone detection sensor can detect, classify and localize drone and ground-based threats in real time. HARK has been exercised in two military Operational Exercises: NATO’s Steadfast Dart 2026 in northern Germany, which tested rapid Allied deployment and Project Vanaheim 2025, a UK-US counter-drone experimentation program in Germany.

“If the acoustic layer is reliable enough, operators do not have to treat every anomaly equally,” said Howe. “They can rank alerts by confidence, cross-check across nodes, and decide whether to escalate, observe or ignore.”

Howe says the shift is not just in detection but in how operators make decisions in real time, which changes the workflow from “investigate everything” to “prioritize what matters.”

HARK listens, deciphers, interprets, and then notifies of an incoming drone threat at the edge within the sensor itself. “We then structure that into a network-level picture: what each node is hearing, how confident it is, whether different nodes are correlating to the same event, and whether the event is stationary, moving, approaching or departing,” said Howe.

In practice, says Howe, that means they aren’t just collecting audio; they are building a spatial and temporal representation of the environment across the sensor network.

Howe added that this reduces cognitive load and improves time-to-decision, while also enabling power- and bandwidth-intensive systems to be used more efficiently by cueing them with stronger evidence rather than scanning continuously.

“Reliable sound changes acoustics from background context into a practical part of the command-and-control loop,” said Howe.