Watchpoint is a mesh of acoustic sensors that detects, classifies, and locates drones by sound. That includes fiber-optic FPVs, which carry no radio link and give radar nothing to find. The kit itself emits no RF and runs without infrastructure.
A modern FPV, especially a fiber-optic variant, carries no electronic signature, cannot be jammed, and flies below radar coverage. Most of the equipment fielded against it was built for an older threat.
A fiber-optic FPV has no radio transmitter. It is small, flies low, and gives conventional radar nothing to lock onto. Any detection system built around electromagnetic emissions will miss it.
A jammer defeats radio-linked drones, but it also broadcasts your location to every direction-finding receiver in range. Transmitting in order to defend yourself makes you a target.
Many detection products need fixed power, an internet connection, GPS, or a central server. A forward position has none of that, and when the infrastructure drops out, the sensor stops working.
Every drone makes sound. Watchpoint nodes capture that signature, classify it on the device with a trained model, and triangulate the position across the mesh. The operator sees bearing and location on a map as it happens.
Each node runs a trained classifier on a custom microcontroller board. Detection happens on the device, with no cloud dependency and no single point of failure. Nodes are weatherproof, battery powered, and fit in a backpack.
Nodes share detections over a self-forming mesh and compute time difference of arrival across the array. The network needs no internet or GPS and reroutes around lost nodes on its own.
A web-based operator interface receives bearing and position data streamed from the sink node. The display shows the classification, the bearing from each node, and the estimated position on the map.
Built to be carried into a forward position and set up without a technician.
Place four sensor nodes around the area you want to cover. They organize themselves into a mesh without any configuration.
Each node samples audio continuously. The onboard classifier picks drone signatures out of the noise floor in real time, including in loud environments.
Detection timestamps are synchronized across the mesh. The array computes a bearing from each node and estimates the target position from the time differences.
The sink node streams position and classification data to the tactical map. The operator watches the bearing, the estimated location, and the threat class update live.
Every item below can be verified.
Won the European Defense Tech Hackathon three times, judged by defense operators, procurement officers, and investors.
The concept and system architecture were validated by the Cyber Innovation Hub, the innovation unit of the German armed forces.
Three Ukrainian battalions are testing Phase 1 hardware in operational conditions. Their feedback goes straight into the next build.
When the v2 mesh localizes a target, it cues an interceptor FPV, computes an intercept vector, and guides the interceptor onto the target. Nobody has to aim it.
The operator approves the intercept and the system flies it, with no requirement for line of sight or manual piloting.
The mesh computes the intercept vector from acoustic position data and uplinks it to the interceptor FPV without aiming input from the operator.
The interceptor receives position updates as the target moves and corrects its flight path until terminal engagement.
Target acquisition is entirely acoustic, so the target never needs to emit a signal and the interceptor needs no radar lock or RF seeker.
Authorization to intercept stays with the operator. Everything below that decision runs automatically.
We build and validate in parallel, and every subsystem goes from breadboard to field within weeks. The list on the right reflects the current state of the hardware and software.
A model trained on drone audio signatures runs on the microcontroller board with classification latency under 100 ms.
A microcontroller sensor board with an acoustic front end, on-device inference, and a mesh radio interface. Prototype builds are in active testing.
Time synchronization between nodes and TDOA-based bearing triangulation across the sensor array.
A web-based operator display with live bearing lines, the position estimate, and the classification from the sink node.
Mesh-to-interceptor cueing, continuous position uplink, and guided FPV intercept without operator aiming.
If you operate in contested environments, buy counter-drone equipment, or invest in early-stage defense hardware, write to us.
We can walk you through the technology, the traction, and the roadmap in a short call.
Contact for briefingWe will demonstrate detection, classification, and mesh triangulation live. Field-test slots are open.
Request demoDefense units and integrators with operational context, investors with a record in deep-tech or dual-use hardware, and engineers working in embedded systems, acoustics, or counter-drone technology.