ECHONETv1.0
>Initializing distributed sensor array.....
>Establishing network coordination layer...
>Loading C2 interface protocol.............
>Verifying node synchronization............
>System ready..............................
DISTRIBUTED SENSING NETWORK0%
TRL 4-5 · CUAS Sandbox 2026 participant · Synthetic-only

 

Radar was never built to cover the low, passive layer. EchoCore is the system that does, in three parts: the EchoNet sensing mesh, the EchoCore AI fusion engine, and the Valhalla synthetic testbed.

Passive
Three sensing modalities. Zero RF emissions. No emitter to jam or geolocate.
Distributed
No single point of failure. Coverage survives node loss.
Cross-Validating
Acoustic + thermal + RF fusion catches what single sensors miss.
Complementary
Complements radar and EO layers. Operates as a standalone network where needed.

Canadian-built, grounded in simulation. Every performance figure here is synthetic and labelled. Our detection ranges come from sensor-physics models we own; we don't quote vendor envelopes.

TRL 4-5Canadian-owned IPSoftware & synthetic data only
Scroll
System Design

Sensing Architecture

Three passive modalities. One fusion layer. This is the system design doctrine.

System overview

Where terrain blinds radar, passive acoustic still hears

Concept schematic. Illustrative terrain and geometry, not a fielded measurement or a real node layout.

Radar needs line of sight, so terrain ridges leave a shadow over the low ground behind them: valleys, rivers, canyons.

A low and slow RF-silent drone flies into that blind zone, with no emitter to detect and no radar return.

Passive acoustic does not need line of sight, so the distributed nodes still hear the track and confirm it.

Layer 01

Acoustic Layer

Primary Detection
50-100 m (modeled envelope)SyntheticPassive

A MEMS microphone array with CNN-based drone-signature classification. The always-on primary passive layer, terrain-tolerant, for environments where radar line of sight is obstructed.

50-100 m modeled detection envelope (FPV-class; synthetic, to be field-calibrated)
CNN-based drone-signature classification (design); envelope is a modeled estimate, not measured
Independent of RF emissions
Modeled for urban-canyon and forested terrain
Layer 02

Thermal Layer

Close-Range Confirmation
20-30 m (modeled, close-range)Passive

Close-range infrared confirmation layer. Works in coordination with the acoustic layer to verify presence and bearing of suspected threats at short range.

20-30 m modeled close-range envelope (extensible with higher-grade IR; synthetic)
Acoustically cued for directional threat verification
Filters vehicles, birds, and static heat sources
Operates day and night, independent of lighting
Layer 03

RF Layer

High-Altitude Awareness
3-5 km (modeled envelope)400MHz-6GHzPassive

A wideband passive SDR receiver covering 400MHz-6GHz. It holds awareness beyond the acoustic and thermal envelope, including drones at longer ranges and higher altitudes where acoustic detection drops off.

3-5 km modeled envelope (emitter-dependent; synthetic)
Passive receive-only, zero RF emissions
Band classification: 433MHz / 2.4GHz / 5.8GHz / sub-6GHz emitters
No interference with existing CUAS assets
Layer 04

EchoCore AI

Trust- and compliance-bounded fusion
2 non-bypassable gatesDeterministic172 tests passing

The fusion layer turns three imperfect sensors into a decision you can audit. It reads which sensors stay silent as carefully as which ones fire, and it bounds every claim by the trust of the evidence behind it.

Provenance trust gate caps a track's confidence to the weakest source feeding it
Compliance gate runs policy at ingress and egress: permit, restrict, downgrade, segregate, block, sealed in a hash-chained bundle
Claim lattice orders presence < bearing < position < track < identity; a physics-contradicting RID raises a spoofing alert
GNSS degradation carries through to a lower claim ceiling. 172 tests passing, deterministic.
Open the Fusion Engine →
Fusion logic

What EchoCore AI decides from agreement and silence

Which sensors stay silent matters as much as which ones trigger. Each evidence combination maps to a typed disposition with a capped claim ceiling, not a raw alert.

ACOUSTIC
THERMAL
RF
VERDICT
ACOUSTIC✓ Detected
THERMAL✓ Detected
RF✓ Detected
Confirmed UAS, high ceiling
ACOUSTIC✓ Detected
THERMAL✓ Detected
RF✗ Silent
RF-silent platform, confirmed by physics, elevated
ACOUSTIC✗ Silent
THERMAL✗ Silent
RF✓ Detected
Distant emitter, cue only
ACOUSTIC✓ Detected
THERMAL✗ Silent
RF✗ Silent
Single-node bearing, cue only, low ceiling
ACOUSTIC✓ Detected
THERMAL✓ Detected
RF⚠ RID conflict
RID contradicts physics: spoofing alert

Each row maps to a maximum claim type on the lattice presence < bearing < position < track < identity, with a capped confidence ceiling. A Remote ID broadcast that contradicts the measured physics never becomes a confident track; it raises a spoofing alert.

Where EchoNet is strongest

Radar, RF, and EO/IR each go blind in a specific way: low and slow targets, RF-silent platforms, bad line of sight. EchoNet is the passive layer built for exactly those cases. The gaps below are where it does its strongest work.

RADAR LIMITATION

Ground clutter and low-altitude blind zones

Radar degrades against low, slow, small targets. Ground clutter and beam geometry leave gaps, and hostile UAS fly low on purpose to sit inside them. The acoustic layer does not rely on reflected signals, so it is meant to cover the cases where radar return is unreliable.

RF-DETECTION LIMITATION

RF-silent and tethered platforms are invisible

RF detection is blind to platforms that emit nothing. Fibre-tethered, autonomous, and pre-programmed UAS carry no RF signature, and more sensitivity does not close that gap. The acoustic and thermal layers aim to detect physical presence regardless of emission state.

EO/IR LIMITATION

Line-of-sight and lighting dependencies

EO/IR sensors need clear line of sight, adequate light, and precise aiming; they are suited to confirmation more than wide-area search. Without a cue from another sensor, their coverage is narrow. EchoNet is designed to work day and night, through terrain masking and weather, with omnidirectional acoustic coverage per node.

ARCHITECTURAL LIMITATION

Centralized sensors are single points of failure

A centralized sensor is a high-value target. Jam, spoof, or knock out one node and entire coverage zones go dark. The distributed mesh is meant to hold coverage under partial node loss.

Where EchoNet Fits

EchoNet does not replace radar. It is the persistent passive layer for the gaps radar cannot see, and it cues the systems that respond.

What Radar Covers
  • High-speed fixed-wing threats (>200 km/h)
  • High-altitude fixed-wing aircraft
  • Long-range volume search (5-15 km)
  • Wide-area volume search
Where EchoNet Operates
  • Large-scale border surveillance: northern frontier and remote terrain
  • Forward operating bases and critical infrastructure perimeters
  • Low-cost, low-flying RF-silent UAS threats
  • Forested and unmonitored geographies where conventional sensors cannot scale
What EchoNet Cues
  • Radar slew-to-cue on threat bearing
  • EO/IR camera pointing and tracking
  • Operator alert with threat classification
  • C2 handoff with sensor confidence score

Invited to CUAS Sandbox 2026

CUAS Sandbox 2026 is a military-organized counter-UAS field event at CFB Suffield. Operators fly real drones against our RF, acoustic, and IR nodes, and EchoCore AI runs on the live feed. We are invited and actively preparing.

valhalla · tactical replay
Tactical replay rendered in Valhalla. Synthetic scenario, not a recording of the field event.
Event
CUAS Sandbox 2026
Host
DND / CAF
Location
CFB Suffield, Alberta
Timeline
September 2026
OBJ-01

Live capture

Record real drone runs across our RF, acoustic, and IR nodes during the live event.

OBJ-02

Fusion on the live feed

Run EchoCore AI on the captured feed and show trust-bounded verdicts: RF-silent fusion, RID spoofing caught, multi-target separation.

OBJ-03

Reverse demo into Valhalla (goal)

Reconstruct the captured runs inside Valhalla as a deterministic replay. This is a preparation goal, not a delivered capability.

Maturity & Path Forward

EchoNet is a developmental program. Today it is a TRL 4-5 software testbed and fusion engine with a rigorous synthetic-evaluation methodology. The account below separates what is built and demonstrated in simulation, what is in active development, and what we are preparing for the CUAS Sandbox field event at Suffield.

Current TRL
4 → 5
TRL 1TRL 4-5TRL 9
Built & demonstrated (synthetic)
Deterministic synthetic testbed (Valhalla): four sensor-physics models, byte-identical output
Golden loop: synthetic datasets run through the fusion engine unmodified
Trust + compliance gates, claim lattice, RID-spoof and GNSS handling (172 tests passing)
Scorer grades fusion output against simulation ground truth, wired as a CI regression gate
Confirmation-time placement optimizer, exercised against a synthetic Monte-Carlo red team
C2 mapping artifacts: Cursor-on-Target + SAPIENT (BSI Flex 335) stub, lab level
In active development
RTS-style offline tactical visualization for the CUAS Sandbox field event
Detection-envelope calibration framework (calibration-profile slot for future field data)
Expanded adversarial scenario set: RID spoofing, GNSS suppression, raids, clutter, node dropout
Deployment-planner seed: node placement and coverage as reusable, deterministic functions
CUAS Sandbox 2026 (field event)
Live capture of real drone runs across the RF, acoustic, and IR nodes
EchoCore AI run on the live feed: RF-silent fusion and RID spoofing caught
Reverse demo into Valhalla as a deterministic replay (preparation goal)
Development Roadmap
2026 Q3
CUAS Sandbox (Suffield)
Military-organized counter-UAS field event at CFB Suffield. Real drones flown against our nodes, with EchoCore AI on the live feed.
2026 Q4
Post-Sandbox
Integrate evaluator feedback. Refine the scenario set and the fusion engine.
2027
Sim-to-real calibration
Hardware-in-the-loop calibration through the calibration-profile slot: field data tunes a parameter, not the code.
2028
Toward operational pilots
Subject to validation: linear border, forward operating base, and critical-infrastructure perimeter scenarios.

Founder-led, judged by the work

EchoCore AI Inc. is a founder-led company in Calgary, Canada. We would rather be judged by what runs than by a roster of names: a trust-bounded fusion engine, a deterministic synthetic testbed, and a golden loop that re-runs byte-for-byte from a seed. Everything on this site is something we built, and evaluators can re-run it from a seed. The full reproducibility package is available on request.

Founder-ledCalgary, CanadaReproducible on request
Simo · Founder & CEO, EchoCore AI Inc.