GROUNDTRUTH — Digital Mines & Sensors: Deny Ground in 15 Minutes, Watch It for Weeks

In 20 seconds

The concept in motion

Illustration. Sound: pings rise and quicken as a contact closes in.

01 The problem

Terrain you cannot watch is terrain you do not hold

Ground surveillance in contested terrain rests on a category of device that has barely changed in twenty years. Four deficits compound one another, and one of them is political.

01
The eastern flank is re-mining. Germany cannot.

Estonia, Latvia and Lithuania left the Anti-Personnel Mine Ban Convention on 27 December 2025, Finland on 10 January 2026, Poland on 20 February 2026. Germany did not — it chairs the convention and is the world's second largest bilateral donor to humanitarian demining. Finland now plans to build a million landmines; Poland, the Czech Republic, the Netherlands, Italy and France are tendering anti-tank mines. The operational problem at that border is identical. The lawful answer is not.
02
Ground sensors are assets, not consumables

Nodes in the multi-sensor class run from around EUR 2,500 to over EUR 15,000 apiece. They ship with rechargeable and inductively charged batteries. That is a design decision with a consequence: a field you must go back and collect is a field you cannot scatter across ground you do not control.
03
What is cheap is blind in every sense but one

Genuinely expendable sensors exist and have done since 2010 — seismic only, one event per node, no classification across modalities and no track. They tell you that something crossed. They cannot tell you what it was, where it went, or where it will be in four minutes. Ukraine's Sky Fortress has shown that cheap sensors scale — over ten thousand acoustic nodes at roughly USD 400 to 1,000 each — but they listen to the sky from fixed masts, not to the ground.
04
The whole category hangs on satellite navigation

The market leader states that its sensors carry a "GPS receiver for self-localization"; the long-standing US supplier lists a "GPS Receiver for Location". Against an opponent whose demonstrated core competence is jamming and spoofing that exact signal, the entire category shares one point of failure — and without position there is no field, only litter.

Sources: Arms Control Association, January 2026; Federal Foreign Office; Bertin Exensor, Flexnet UGS; McQ RANGER; IndexBox, Europe UGS; The War Zone, 21 October 2026; TED notices 406637-2026, 142573-2026, 278766-2026, 715137-2025, 766164-2025; Militarnyi on Sky Fortress.

02 Why now

Ground sensing is being bought — in thousands, and for a decade

Market estimates in this segment contradict each other. Procurement does not. These are public awards and requirements from the last twelve months.

Recent procurement evidence
Date	Buyer	Supplier	Scope	Volume
16 Oct 2026	Polish armed forces		Acoustic detection to complement radars; system to manage up to 5,000 sensors simultaneously
20 Nov 2025	COMMIT, Netherlands	Bertin Exensor	Frame contract, unmanned unattended ground sensors, nine years	EUR 45 m
Sep 2025	Sweden, Finland, Denmark, Norway	Bertin Exensor	Joint ten-year frame contract for unattended ground sensors
2028	European Union		Eastern Flank Watch to reach full operational capability

Sources: Defence24; Bertin Exensor; Calibre Defence; European Security & Defence.

03 Two capabilities, one field

Close the ground fast. Never stop listening.

A minefield denies ground because nobody crossing it knows what is there. GROUNDTRUTH keeps that uncertainty and removes the explosive.

A · Rapid denial

Seeded in 12 to 18 minutes

Drones scatter the field across an approach route, a gap or a flank. Some nodes can deliberately announce themselves. The opponent sees that the ground has been seeded and knows it is being watched — but cannot tell from the outside whether it is only watching. Every metre has to be reconnoitred before it is crossed. That cost in time and certainty is the denial.

No effector in the node; any effect stays with own forces and a human decision
The same announce signal finds the field again — for lanes, inventory and end of life

B · Persistent surveillance

Twelve to fifteen weeks of tracks, not alarms

Once seeded, the field detects, classifies and hands every mover from node to node. Days of movement logs become corridors, timings and forecasts. Own forces carry an identification token and pass through lanes that can be opened and closed digitally; everything else is tracked and reported.

Friend-or-foe discrimination inside the field
Lanes for own forces, opened and closed without going back in

Denial with friendly lanesMovement entering the field is tracked and continued as a forecast. Own units carrying a token pass through an open lane without an alert; the lane can be closed again from outside.

An unmarked utility vehicle drives along a dirt track at dusk; the field on both sides is seeded with small sensor discs, the track itself stays clear as an open lane. — IllustrationAn open laneOwn forces keep moving. The field around them keeps listening.

04 The solution

A field that reports a trajectory, not an alarm

Two parts, and only one of them is hardware. The node is the entry ticket. The field — the ability of many cheap nodes to agree on where they are and to pass a target between them — is the product.

A · The node

Coin-sized, twelve to fifteen weeks, then inert

Seismic, acoustic, sound-level, pressure, magnetic, thermal and gas sensing in an attritable package on a primary cell, with specialist nodes for further modalities. Classification runs on the node (edge processing) against a signature library; the raw signal is reduced to a class and a confidence value and then discarded. Nothing that could be a voice ever leaves the device.

Sovereign root of trust, open RISC-V, fabricated in Germany
Cost per node set by the requirement never to recover one
Reports its own position, its own health and its own end of life

B · The field

Ranging in the mesh replaces the satellite

Nodes measure distance to their neighbours and solve their own relative geometry, anchored to the carrier's release track — no GNSS at any point. Findings leave the field through a few cigarette-pack-sized transmitter nodes, which pass them rearward along a chain over 3 to 5 km. Transmission hops along that chain at random, so there is no fixed emitter to locate or engage — a low-probability-of-intercept and -detection (LPI/LPD) design with emission control built in. If a transmitter fails, the chain routes around it. The same measurements carry the handover: the track is passed on rather than restarted.

Position without GNSS, and therefore under jamming
Transmitter chain 3–5 km to the rear, hopping at random against direction finding
Self-healing mesh: sensor nodes and transmitters take over for failed ones — no single point of failure
Designed to serve as one bearer in a non-linear PACE plan (primary, alternate, contingency, emergency)
Detections become tracks, tracks become movement patterns
Patterns become forecast — route corridors, timing, intent

A flat sage-olive sensor disc beside a euro coin on weathered concrete. — Design mock-up · illustrationCoin-sizedThe node is the entry ticket. The field is the product.

Node architectureA node optimised for size, weight and power (SWaP): 1 cover with acoustic, pressure and gas ports · 2 mesh and ranging antenna · 3 multi-sensor board · 4 processor with secure element · 5 primary cell · 6 base plate for ground coupling.

Sensing modalities
Modality	What it delivers	Role in the field
Seismic	Footsteps, wheeled and tracked vehicles, gait frequency	Core sensor for ground targets
Acoustic	Engines and movement, classified on the node, never recorded	Classification, bearing in the mesh
Sound level	Level trend per node: approaching, receding	Low-power wake-up threshold
Pressure · infrasound	Heavy vehicles, pressure events, weather reference	Far field, calibration
Magnetic	Metal mass	Vehicle versus person
Thermal (passive infrared)	Warm bodies at close range	Confirmation person or animal
CO₂	Breath and exhaust at very close range, in low wind	Confirmation, not search
Exhaust gas	Diesel exhaust signature	Vehicle confirmation (specialist node)
Passive RF	Drone control links and radios nearby	Detects emitters without emitting (specialist node)
Attitude · motion	Orientation after landing, movement of the node	Ground-coupling check, anti-tamper

Sensor mesh and transmitter chainCoin-sized sensor nodes form the mesh and hold their geometry without GNSS. A few cigarette-pack-sized transmitters collect the findings and pass them 3 to 5 km to the rear as a chain. Which link transmits changes at random; a failed transmitter is bypassed.

Dispersal and drop mapping Scattering is the easy half. Knowing where three hundred nodes actually came to rest, in the right order and to the right accuracy, is what turns a scatter into a field. The release track is the anchor; the mesh does the rest.

Handover and forecast Existing expendable sensors emit an event per node and leave the assembly to a human at the other end of a radio link. Here the track is assembled in the field, node to node, and continues past the last detection as an explicit forecast with a stated cone of uncertainty. A minefield reports that something happened. This reports where it is going.

05 How it works

From payload container to forecast in five stages

The carrier is deliberately irrelevant. Any platform with a payload container and a recorded release track can seed a field.

01

Dispense

Drone-agnostic container, field live in 12–18 minutes. The release track is logged as the geometric anchor.

02

Self-locate without GNSS

Mesh ranging between neighbours resolves relative geometry under jamming.

03

Sense and classify on board

Many modalities in, one class and one confidence out. The raw signal is discarded.

04

Hand the track over

The mover is passed node to node. The track persists across the field, not per device.

05

Fuse and forecast

Movement logs across days produce corridors, timings and a stated forecast envelope.

A small quadcopter at dusk releases a trail of flat sensor discs over a misty field bordered by pine forest. — IllustrationStage 01 · DispenseAny platform with a payload container and a recorded release track can seed a field.

Classification at the node Reducing the signal on the device is not only a bandwidth decision. A node that never transmits raw audio cannot record a conversation — which is what makes the same hardware defensible on a pipeline or an airport perimeter, not just on a front.

The signature library is the part that cannot be copied. Silicon can be reproduced; a curated corpus of vehicle and weapon signatures with traceable provenance, grown over years and held in Europe, cannot.

06 Requirements & hard problems

Twenty-four problems a ground sensor field has to solve

What a requirement owner will ask, what the market answers today, and where GROUNDTRUTH stands. The status column is deliberately conservative: nothing is marked tested until there is a date and a measurement behind it.

Addressed by design

Solved by the architecture itself; no technology needed.

Building block proven

Core building block demonstrably exists at third parties.

Approach defined

Our solution path is defined, not yet built.

Open — talent wanted

Unsolved — this is where we are hiring and partnering.

Requirements, market state and status
Requirement	Why it is hard	Market today	GROUNDTRUTH approach	Status
R1GNSS-independent self-localisation	Jamming and spoofing are the opponent's core competence	Every unattended ground sensor (UGS) with stated positioning uses GPS (Exensor, McQ, Textron)	Inter-node ranging, geometry solved in the mesh, anchored to the logged release track	Approach defined
R2Field-wide time synchronisation	Ranging, time-difference-of-arrival and track fusion need a common clock without GNSS timing	Not stated by vendors	Clock discipline over the ranging exchange	Approach defined
R3Transmitter chain against direction finding	A fixed emitter is located and engaged	Radio-silence mode (Exensor); fixed gateways	Cigarette-pack-sized transmitter nodes form a chain to the rear; transmission hops along it at random; no fixed emitter, no periodic beacon (LPI/LPD, emission control by design)	Approach defined
R4Self-healing on node loss	Attrition, capture, battery end	Mesh relay (Exensor)	Sensor nodes and transmitters take over for failed ones; the field degrades gracefully, with no single point of failure	Approach defined
R5Rapid denial: field live in 12–18 minutes	Hand-emplaced sensors and minefields take crews and time	Hand-emplaced UGS; mines laid by dedicated systems	Drone dispersal, automatic self-survey, optional announce signal	Approach defined
R6Friend-or-foe discrimination and friendly lanes	Own forces must pass without alarms; foes must not spoof friend	Not stated by vendors	Cryptographic identification token for own units; lanes opened and closed digitally	Approach defined
R7Probability of detection vs. false-alarm rate	Wind, rain, wildlife; in Latvia acoustics could not tell a drone from a moped	Seismic-acoustic fusion standard; rates rarely published	Multi-modal fusion on the node, class plus confidence; measured rates to be published	Approach defined
R8Target classification (personnel / wheeled / tracked)	Soil-dependent seismics, domain shift across terrain	Personnel/vehicle classification standard	European signature corpus, terrain-tagged	Open — talent wanted
R9Multi-sensor package in a coin-sized node	Size, power and cost per modality	Multi-sensor nodes are hand-sized and recharged	Mixed fields: base nodes plus specialist nodes	Approach defined
R10Ground coupling after air-drop	Random landing orientation versus seismic coupling	Hand-emplaced, often with a spike	Attitude sensing on landing, coupling-optimised base	Open — talent wanted
R11Drop survivability and drop mapping	Impact shock; where hundreds of nodes came to rest	UAS-deployed sensors now a procurement category (British Army, TRL 6, May 2026)	Logged release track as anchor, mesh refines	Approach defined
R12In-field track handover and forecast	Distributed multi-target tracking, data association	Tracks assembled at the far end in command and control (C2)	Track passed node to node, continued as forecast	Approach defined
R13Energy: 12–15 weeks on a primary cell	Always-listening sensing at microwatt level, cold	Rechargeable or inductive, 30 days to 3 months	Sound-level wake-up, event-only radio, inert at end of life	Approach defined
R14Low probability of intercept and detection	An emitting field reveals itself	Radio-silence mode, store-and-forward (Exensor)	Event-only bursts from changing nodes	Approach defined
R15Exfiltration through contested spectrum	Getting the track out of the field under jamming	900 MHz, Iridium, cellular	Chain of transmitter nodes carrying findings 3–5 km to the rear	Approach defined
R16Root of trust, attestation, anti-tamper	Captured nodes must not compromise the field	AES-256 and anti-tamper standard	Open RISC-V secure element made in Germany (Fraunhofer)	Building block proven
R17Robustness against spoofed signatures (EU AI Act Art. 15)	Adversarial inputs against a classification field	Not stated by vendors	Signed detections, versioned corpus, node attestation	Approach defined
R18Fast model update for new threats	New vehicle and drone types appear in weeks	Zvook: about a week of data, then remote update of all stations	Corpus update per production batch	Approach defined
R19Privacy for the civil line (GDPR)	Microphones near people	Not addressed by vendors	Classify and discard on the node	Approach defined
R20Record-keeping and end of life	Fields must be inventoried and declared spent	Nodes recovered and reused	Position, health and end of life reported; announce signal finds the field	Addressed by design
R21Lawful status and investability	Ottawa Convention; Article 9 funds exclude anti-personnel mines	Mines procured by PL, CZ, NL, IT, FR	No explosive, no initiating train, no effector in the node	Addressed by design
R22C2 interoperability	Units will not adopt another screen	Cursor on Target (CoT) for TAK clients, SAPIENT	Disclosed on request; integration path to D-LBO (Digitalisierung landbasierter Operationen) to be assessed	Approach defined
R23Environmental and EMC qualification	Military acceptance	MIL-STD-810, MIL-STD-461, IP68	Target qualification	Open — talent wanted
R24Unit cost and European supply chain	Attritable only at low cost; over 55 % of high-sensitivity sensing elements sourced outside Europe	Multi-sensor nodes EUR 2,500–15,000; Sky Fortress USD 400–1,000	Disclosed on request	Open — talent wanted

Market column from vendor pages and public reporting: Exensor Mini Mk3, McQ RANGER, Textron, L3Harris, ARA, British Army call, Latvia incident (vendor blog), Fraunhofer RISC-V secure element.

07 Design targets

What the field is being built to do

Targets, not measurements. Each will be published with its test conditions once it has been measured.

Design targets
Parameter	Design target
Sensor node	Coin-sized, as in the mock-up
Transmitter node	Cigarette-pack-sized; a few per field
Mass (sensor node)	≤ 30 g
Endurance	12–15 weeks on a primary cell, then inert
Sensing	Seismic, acoustic, sound level, pressure, magnetic, thermal, CO₂; specialist nodes with exhaust gas and passive RF
Positioning	No GNSS; inter-node ranging anchored to the release track; ≤ 1 m relative within the field
Exfiltration	3–5 km to the rear via a chain of transmitter nodes, hopping at random
Field size	From 100 nodes; 300 recommended
Time to live field	12–18 minutes
Dispersal	Any UAS with a payload container
Friend or foe	Cryptographic token, digital lanes
Temperature	−30 to +60 °C
Qualification (target)	MIL-STD-810, MIL-STD-461, IP68
Trust anchor	Open RISC-V secure element, fabricated in Germany
Effector · explosive	None
Interfaces	Disclosed on request
Unit cost	Disclosed on request

08 Competitive landscape

Everyone else places sensors. Nobody scatters them.

Public specifications of the established ground-sensor vendors, read on their own product pages. Empty cells mean the vendor does not state it.

Comparison with established vendors
Vendor	Modalities	Positioning	Emplacement	Power	Transmitter	Friend or foe
Bertin Exensor (Mini Mk3)	Seismic, acoustic	GPS	Hand-emplaced, 0.5 kg	Rechargeable, ~30 days	Radio-silence mode
McQ (RANGER)	Seismic, acoustic, magnetic	GPS	Hand-emplaced	Inductive, 1–3 months
ARA (Pathfinder)	Footstep and vehicle detection		Hand-emplaced	6–24 months
Textron (MicroObserver)	Seismic, imaging	GPS	Hand-emplaced
L3Harris (ground sensors)	Seismic, acoustic, magnetic		Hand-emplaced, 4.5 oz	30 days, inductive
Sky Fortress (UA)	Acoustic, air targets	Fixed mast	Installed
GROUNDTRUTH (concept)	Seismic, acoustic, pressure, magnetic, thermal, gas, RF	No GNSS, mesh ranging	Air-dropped, 12–18 min	Primary cell, 12–15 weeks	Transmitter chain, 3–5 km, hopping	Token and lanes

Sources: Exensor, McQ, ARA, Textron, L3Harris, Militarnyi. GROUNDTRUTH row: design targets.

A

Multi-modal, air-dropped and attritable do not exist in the same ground sensor

Expendable nodes are single-modality and event-based. Multi-modal nodes are built to be recovered, priced accordingly, and fitted with rechargeable or inductively charged cells. Cheap sensors at scale exist for the sky (Sky Fortress), not for the ground. No vendor publicly offers a multi-modal ground node designed to be scattered and never to come back.
B

Every documented node localises by satellite

The two clearest public statements in the category are a "GPS receiver for self-localization" and a "GPS Receiver for Location". Nobody advertises position holding without GNSS, and nobody describes a transmitter chain that changes the emitting link. In a theatre defined by electronic warfare that is not a feature gap, it is a category-wide single point of failure.
C

Nobody hands a track over inside the field

Detection is documented per node; assembling those detections into a trajectory is left to a command-and-control application at the far end of a link. A field that maintains the track itself degrades gracefully when the link does not — and it is the precondition for movement logs and forecasting rather than alarms.
D

There is no European signature corpus and no European supply chain

No vendor publicly claims a European-held library of vehicle and weapon signatures. On the hardware side, over 55 per cent of the high-sensitivity seismic and magnetic elements in European systems are sourced outside Europe. Both halves of the sovereignty problem are unaddressed.

09 Applications

Three channels, one field

Twelve to fifteen weeks of endurance is long enough to hold a sector through a rotation or a season, and short enough to leave nothing permanent behind.

Channel 01

Counter-mobility without mines

Approach routes, gaps and flanks that no unit can occupy and no camera can see. The field denies ground the way a minefield does — by uncertainty — and removes the attacker's assumption of being unobserved. It leaves nothing behind that has to be cleared afterwards.

Channel 02

Critical infrastructure and temporary perimeters

Pipelines, substations, airfields, ammunition depots — and perimeters that only need to exist for a few weeks. Scattered in minutes, inventoried continuously, declared spent at the end. A separately marketed civil product line, for the reasons set out under regulation.

Channel 03

Border surveillance and homeland security

Green-border sections without infrastructure, where the requirement is not a single alarm but a pattern: which corridor, at which hour, at what rhythm. Movement logs answer that question; trip-wire sensors never have.

Sensor discs on the gravel outside the fence of a remote substation at night, faint rings around two of them. — IllustrationTemporary perimeterCritical infrastructure, civil product line.

Sensor discs half-embedded in thin snow on a cleared strip between birch and pine forest at first light. — IllustrationEastern flank, winterCounter-mobility and green-border surveillance.

10 Market environment

A small market with a broken price structure

All figures below are published third-party estimates. The spreads in this segment are severe and one of them is a direct contradiction — both are shown rather than reconciled.

Why now: capital is moving

USD 7.4 bn

European defence tech venture funding in 2026 to early October, projected USD 10.5 bn for the full year (2025: USD 2.6 bn)

Dealroom × Resilience Media, Oct 2026

USD 3.5 bn

raised by defence start-ups in Germany in 2026, more than any other European country

Dealroom × Resilience Media, Oct 2026

USD 4.7 bn

for drone start-ups in Europe in 2026, across 40 rounds

Dealroom × Resilience Media, Oct 2026

Source: Dealroom × Resilience Media, The State of Defence Tech 2026 (5 October 2026), via Resilience Media. 2026 figures to early October.

556 – 579

USD m · unattended ground sensors, global, 2026

Straits Research · Fortune Business Insights

812 – 843

USD m · same market, 2034, at CAGR 4.8 to 4.9 %

Straits Research · Fortune Business Insights

7.5 – 9.0 %

CAGR · Europe through 2035, strongest in NATO frontline states

IndexBox

2.500 – 15.000

EUR · price per node in the multi-sensor class today

IndexBox

> 55 %

of high-sensitivity seismic and magnetic elements sourced outside Europe

IndexBox

5

NATO states outside the mine ban treaty since December 2025

Arms Control Association, January 2026

Unattended ground sensors, global: two publishers, 2026 to 2034

Published endpoints only — no interpolated intermediate years.

Source: Straits Research and Fortune Business Insights. Note the contradiction that follows.

What the numbers do and do not support

The contradiction, stated plainly: IndexBox puts the European market alone at USD 580 to 650 m for 2026, which is at or above what Straits and Fortune assign to the entire world market in the same year. These cannot both be right. Absolute levels in this segment should be treated as indicative; growth rates and structural statements are the more reliable part.

The number that actually matters is the price per node. At EUR 2,500 to 15,000, a sensor is a capital asset that is hand-emplaced and retrieved. Expendable devices do exist — a US supplier has shipped tens of thousands of coffee-cup-sized seismic units to the US military since 2010, at an order of magnitude of a few hundred dollars each — but they are single-modality and report isolated events. There is no ground sensor in the middle: multi-modal, air-dropped and attritable at once. That gap, not the market size, is the opportunity.

Demand and funding. The EU Action Plan on Drone and Counter-Drone Security of 11 February 2026 names Horizon Europe, EDF, EDIP and SAFE as channels; the European Drone Defence Initiative is backed by EUR 6 bn and Eastern Flank Watch is to reach full operational capability by the end of 2028 (DG DEFIS, EPRS).

Sovereign silicon is no longer aspirational. Three Fraunhofer institutes have presented a RISC-V Secure Element developed entirely in Germany, open-source, post-quantum capable, fabricated on GlobalFoundries' 22 nm FDSOI line in Dresden, usable as a discrete chip or as a block inside a system-on-chip (heise). That is the trust anchor this node requires, and it exists.

11 Regulatory position

Why a device called a mine is not one

The name is a claim about function, not about law. The distinction is not rhetorical — it is written into the definitions, and it is what makes the system fundable where a munition would not be.

Legal anchors and their consequence for the concept
Instrument	Specific anchor	Consequence
Anti-Personnel Mine Ban Convention Art. 2(1), 2(2)	A mine is defined as a munition designed to be placed on or near the ground and to be exploded by the presence, proximity or contact of a person or vehicle. An anti-personnel mine is one that will incapacitate, injure or kill.	GROUNDTRUTH contains no explosive, no initiating train and no effector. It falls outside the definition on its face, not by interpretation. Germany chairs this convention — that is an argument for the concept, not against it.
CCW Amended Protocol II recording obligations	Requires the recording and marking of emplaced minefields and their eventual clearance.	Not applicable — and adopted voluntarily anyway. Each node reports its position and its end of life, so a field is recorded from the moment it lands and can be declared exhausted. Norm-positive by construction.
EDF Regulation (EU) 2021/697 Art. 10(6)	Excludes from funding the development of lethal autonomous weapons without the possibility of meaningful human control — expressly without prejudice to early-warning systems.	A system with no output that acts on the world is not an autonomous weapon under any reading. It is named in the carve-out rather than the exclusion.
EU AI Act (EU) 2024/1689 Art. 2(3), Recital 24	Systems placed on the market exclusively for military, defence or national security purposes are fully exempt; anything also marketed for civil or law-enforcement use is pulled back in.	Two separately marketed product lines from the outset, as a design decision rather than a legal afterthought.
EU AI Act Art. 12 and 15	Automatic event logging; robustness against data poisoning, model poisoning and adversarial examples.	Signature spoofing is the obvious attack on a classification field. Signed detections, a versioned signature corpus and node-level attestation are the answer, and they are also what an evidential chain requires.
SFDR Article 9 · ESG exclusions	Anti-personnel mines are treated as controversial weapons and are categorically excluded from Article 9 funds and from most investor exclusion lists.	A field that denies ground by uncertainty, with no explosive and no effector, is not a controversial weapon. The name describes the function; the investment case does not depend on a waiver.
GDPR civil product line	Acoustic sensing near populated areas engages data protection as soon as speech could be captured or retained.	Classify and discard at the node. No raw audio is stored or transmitted, so the recording never exists. This is a hardware property, not a policy promise.

Primary sources: Anti-Personnel Mine Ban Convention, Regulation (EU) 2021/697, Regulation (EU) 2024/1689, Federal Foreign Office on the German chairmanship.

12 Roadmap & funding path

From components to an air-dropped field under jamming

Each stage ends in evidence that can be checked — a measurement, a demonstration, a requirement owner who was there.

Technology readiness stages
Stage	Content	Evidence produced
Now	Sourcing talent, testing hardware components, developing the core software	Component test reports
TRL 3–4	Ranging, handover and transmitter-chain hopping in simulation and on breadboard nodes	Measured ranging accuracy without GNSS, with conditions
TRL 5	Hand-placed field of 100 nodes on a test range	Detection and false-alarm rates per target class
TRL 6	Air-dropped field of 300 nodes, friendly lanes, operation under jamming	Demonstration with a requirement owner

Funding instruments in view: EDF 2026 call topic “Multidomain sensors demonstrator and test” (call topic descriptions); NATO DIANA challenge “Multidomain Sensing and Advanced Data Processing” (DIANA); Eastern Flank Watch, full operational capability by the end of 2028.

13 Team & unfair advantage

Operational field experience, dual-use ecosystem, procurement literacy

Founder · Initiator

Arne Eilers

Munich

Lieutenant Colonel (Reserve), multi-domain operations

Credibility in this market rests on knowing what a unit actually does with a sensor report at three in the morning. Experience as a Lieutenant Colonel in the reserve in multi-domain operations supplies that; access to the European dual-use ecosystem and an understanding of how defence procurement actually decides supplies the rest.

Status of the venture

Concept stage. No company and no pilot contract yet. We are sourcing talent, testing hardware components and developing the core software. We are looking for:

Technical co-founders — low-power mixed-signal and RF design, mesh ranging and time synchronisation, low-probability-of-intercept radio, embedded classification on microwatt budgets.
Signature and modelling partners — institutes and units able to build and validate a seismic and acoustic signature corpus with defensible provenance.
Pilot partners — infrastructure operators and public-sector requirement owners willing to run a temporary perimeter under real conditions.

14 Contact

A conversation about physics, silicon or piloting

For investors, requirement owners, funding bodies and prospective co-founders. Investor brief, requirements briefing, interfaces and cost model on request.

golf3tango@protonmail.ch

Analysis on request · English or German · Munich, European Union

GROUNDTRUTHDigital mines & sensors.Deny ground in 15 minutes.Watch it for weeks.

The eastern flank is re-mining. Germany cannot.

Ground sensors are assets, not consumables

What is cheap is blind in every sense but one

The whole category hangs on satellite navigation

Seeded in 12 to 18 minutes

Twelve to fifteen weeks of tracks, not alarms

Coin-sized, twelve to fifteen weeks, then inert

Ranging in the mesh replaces the satellite

Dispense

Self-locate without GNSS

Sense and classify on board

Hand the track over

Fuse and forecast

Multi-modal, air-dropped and attritable do not exist in the same ground sensor

Every documented node localises by satellite

Nobody hands a track over inside the field

There is no European signature corpus and no European supply chain

Counter-mobility without mines

Critical infrastructure and temporary perimeters

Border surveillance and homeland security

Why now: capital is moving

Unattended ground sensors, global: two publishers, 2026 to 2034

What the numbers do and do not support

Arne Eilers

A conversation about physics, silicon or piloting