SM-ANAT-1.0 · Sentinel Mesh

Node Anatomy

Five sensors. One pole. Every threat type covered simultaneously.

🎙

Acoustic Array

4× MEMS microphones + DSP preamp

STAGE 01

Detection range80–120m radius

Frequency band20 Hz – 20 kHz

Latency< 80ms

Unit cost~$48

Detects gunshot muzzle blast, ballistic crack, IED detonation, drone rotor signature, vehicle engine harmonics.

🌡

Thermal IR

InfiRay LWIR 256×192 @ 25fps

STAGE 02

Detection range50–80m effective

Sensitivity< 50mK NETD

Works inFull darkness, smoke

Unit cost~$160

Detects heat bloom of explosion, engine thermal signature, drone motor heat, human body. Day or night.

📳

Seismic MEMS

ADXL355 triaxial accelerometer

STAGE 02

Detection range30–60m ground wave

Sensitivity±2g, 20-bit ADC

Sample rate4kSPS

Unit cost~$12

Detects ground shockwave from IED/explosion. Confirms acoustic event. Hard to spoof — requires physical energy.

~$802

per node · BOM

📡

RF / SDR

RTL-SDR v4 · 500kHz – 1.75GHz

STAGE 01

Detection range100–300m (drone RF)

Frequency2.4GHz / 5.8GHz

Latency< 200ms

Unit cost~$35

Detects drone control signals (DJI, Autel, FPV), identifies make and model from RF fingerprint.

📷

Optical Camera

IMX219 · 8MP · 1080p60 CSI

STAGE 03

Detection range40–60m effective

Resolution3280×2464 raw

Frame rate60fps @ 1080p

Unit cost~$28

Visual confirmation at Stage 3. On-device object detection via Jetson inference. Evidence capture for human alert.

🖥

Edge Compute

Jetson Orin Nano 8GB · 40 TOPS

BRAIN

AI performance40 TOPS INT8

InferenceAll sensors parallel

Power draw7–15W

Unit cost~$299

Runs the full 5-stage cascade locally. No cloud required. All five sensor streams processed in parallel under 500ms.

Sensors per node

~$802

BOM cost per node

40T

AI TOPS on-device

≤ 90s

Alert latency Y1

±25m

Location accuracy

Detection Cascade →

↓

SENTINEL // DETECTION ENGINE

SM-VIZ-1.3

MULTI-SCENARIO LIVE SIM

--:--:--Z

SYSTEM ACTIVE

Detection Cascade

One platform · Every threat type · All stages complete in seconds

—

ACTIVE DETECTION SCENARIO

—

AUTO

Event Elapsed

0.0s

INITIALISING

01👁 Suspicion

—

▶

02✓ Confirmation

—

▶

03📍 Location Fix

—

▶

04🛤 Trail & Tracking

—

▶

05🔔 Human Alert

—

▶

NODE GRID — LIVE DETECTION SPREAD

Inactive

Detecting

Confirmed

Target

EVIDENCE CHAIN

ALERT FEED

—

Awaiting first detection event...

SYSTEM LOG

SM-VIZ-2.0Platform Comparison

Sentinel Mesh vs existing single-purpose detection systems

CAPABILITY MATRIX

Capability	ShotSpotter	Dedrone	DroneShield	CCTV Network	Sky Fortress 🇺🇦	Sentinel Mesh
🚁 Drone Detection	✗	~RF only	✓	~line of sight	✓acoustic, combat-proven	✓4-sensor fusion
💥 Gunshot / Ballistic	✓single sensor	✗	✗	✗	✗	✓dual-sig TDOA
🔥 Explosion / IED	~acoustic only	✗	✗	~post-event only	~acoustic only	✓acoustic + seismic + thermal
🚗 Vehicle Surveillance	✗	✗	✗	~visual only	✗	✓engine + thermal + optical
📍 GPS Localisation	~±50m	~bearing only	~bearing only	✗	~bearing, multi-node	✓±25m CEP90
🧠 Multi-Sensor Fusion	✗	✗	~2 sensors	✗	✗acoustic only	✓up to 5 sensors
🕸 Mesh — No Single Point of Failure	✗	✗	✗	✗	✓14,000+ nodes	✓distributed mesh
📡 Offline / No Cloud Required	✗	✗	~limited	✗	~central net required	✓edge inference
⚡ Alert Latency	~45s	~30s	~20s	Minutes	~15–30s	≤ 90s Y1 → ≤ 30s Y2

✓ Full capability

~ Partial / limited

✗ Not supported

🇺🇦 Sky Fortress — Ukrainian combat-proven distributed acoustic network. 14,000+ sensors. $400–$1,000/node. Intercept rate 80/84 drones in one documented engagement. Acoustic-only — validated proof of the distributed sensor model that Sentinel Mesh extends with multi-sensor fusion, offline edge inference, and multi-threat coverage.

SM-COV-1.0Deployment Coverage

Drag the slider to see how coverage scales with node count

CITY GRID — ILLUSTRATIVE DEPLOYMENT

Nodes deployed

—

Area covered (km²)

—

Grid coverage %

—

Est. BOM cost

≤ 90s

Alert latency

Detection range

Node position

Each node covers ~50m detection radius. Overlap provides redundancy — 2+ nodes enable TDOA localisation to ±25m CEP90.

SM-IG-1.0

Timeline of a Threat

From first sensor trigger to human alert — every second mapped

T+0s

T+15s

T+30s

T+45s

T+60s

T+90s

Suspicion

Confirmation

Location fix

Trail

Human alert

SM-IG-2.0

Cascade Confidence Flow

How raw sensor noise becomes a verified, actionable human alert

01 Raw Sensor Events

—

All acoustic, RF, thermal, seismic and optical triggers — including noise, wind, vehicles, birds

↓ AI filters by signature

02 Suspicion Events

—

Events that match a known threat signature on at least one sensor. False positives still present.

↓ Cross-sensor agreement required

03 Confirmed Detections

—

2+ sensors agree. Location fix computed via TDOA. Threat classified and tracked.

↓ Evidence packaged + operator notified

05 Human Alerts Sent

—

Actionable alert with GPS coords, sensor evidence, threat type and confidence score. Human decides.

Per 24-hour period
100-node deployment

—

raw triggers

—

suspicion events

—

confirmed detections

—

human alerts

≤ 15 false alerts
per node per day
(Y1 target)

SM-IG-3.0

Mesh Resilience

Click nodes to take them offline — watch how the mesh adapts

CLICK ANY NODE TO TOGGLE OFFLINE

Nodes online

Nodes offline

Coverage retained

100%

System fully operational

vs Single-Point Systems

ShotSpotter —

Dedrone —

DroneShield —

Online — active coverage

Offline — failed / removed

Overlap — redundant coverage