Synthetic aperture sonar point-cloud visualization of a tunnel interior cross-section showing an anomaly detection cluster

Below GPS range.
Below radio.
Threats found.

Tarysar builds autonomous sensor-effector systems for underground threat operations — tunnel networks, border crossings, mine infrastructure. Where GPS is denied, radio fails, and humans can't safely enter, our systems detect, identify, and act.

~40,000 km Cross-border tunnel network estimated globally estimated / open-source reporting
GPS-denied Standard condition in tunnel warfare scenarios — GPS jamming is pervasive reported / US Army doctrine
~94% Of underground operations end in no-entry protocols due to threat uncertainty estimated / field-reported

One system. Three sensing modes. Continuous underground awareness.

The Tarysar Subsurface Detection System deploys as a distributed sensor mesh — each node handles passive acoustic, ground-penetrating radar, and seismic detection simultaneously. Nodes self-organize a 915 MHz LoRa comm mesh on power-up. No GPS sync required. No base-station dependency.

Classification executes at the edge on ARM Cortex-M class processors. Pre-classified event objects reach operators in under 8 seconds from event onset. No cloud uplink. Operational at depths beyond 200 meters in tested substrates.

CROSS-SECTION / TUNNEL-A DEPTH: 42m SURFACE — GRADE LEVEL STRATA-01 / LOAM — 0-8m STRATA-02 / CLAY — 8-28m STRATA-03 / LIMESTONE — 28-50m TUNNEL VOID w:3.2m h:2.4m NODE-1 NODE-2 NODE-3 ANOMALY TARYSAR SENSOR-MESH SCHEMATIC / NOT TO SCALE

Four detection modalities. One integrated platform.

Passive Acoustic Array

freq: 2 Hz – 20 kHz / range: 200m per node

Detects and classifies vibration signatures — footstep cadence, equipment operation, excavation activity — with on-device signal processing. False-positive reduction via multi-channel confirmation.

Ground-Penetrating Radar

depth: 3–15m / void detection threshold: 0.3m³

Maps subsurface voids, tunnel geometry, and structural anomalies. Operating frequency tunable 300 MHz – 2 GHz for depth vs. resolution tradeoff depending on soil composition.

Seismic Mesh

sensitivity: 0.0005g / event classification: 12 categories

Distributed seismic sensors detect structural displacement events — tunnel excavation, vehicle transit, subsidence. Mesh correlation distinguishes footfall from background geological activity.

Autonomous Effector

nav: dead-reckoning + SLAM / comm: 915 MHz LoRa mesh

Non-GPS waypoint navigation using simultaneous localization and mapping. Operates in comms-degraded environments. Autonomous classification triggers effector dispatch or operator alert within defined engagement parameters.

From emplacement to alert in four steps.

No satellite link. No existing infrastructure. Deployable by a two-person team in under 45 minutes per corridor.

STEP 01

Node Emplacement

Sensor cluster units placed at tunnel entry points, passage intersections, and perimeter positions. Magnetic wall-mount or ground-stake variants.

STEP 02

Mesh Formation

Nodes self-organize a 915 MHz sub-GHz comm mesh automatically. No configuration required. Adaptive topology — if one node loses power, routing adjusts.

STEP 03

Continuous Scanning

Passive acoustic, seismic, and active GPR sweep cycle runs continuously. Event signatures are correlated across nodes for confirmation. Battery life: 72h per node on standard pack.

STEP 04

Alert + Response

Autonomous classification triggers effector dispatch or encrypted operator alert within defined engagement parameters. Human-in-the-loop override available at all stages.

Performance from controlled field tests.

200m per node / passive acoustic range Detection range per sensor node in tested soil and concrete substrates
>91% classification confidence / field tests Multi-modal classification confidence in controlled tunnel test environment, internal data
<45 min deploy time / per corridor Full sensor mesh deployment time by two-person team, 100m corridor with 4 nodes

All figures from Tarysar internal field test data. Not independently verified. Actual performance will vary with substrate, geometry, and operating conditions.