AeroT

All-Domain Electromagnetic Radio Operator Trainer Tool

A COTS-based RF training system that compressed electromagnetic discipline feedback from delayed after-action review into near-real-time terrain-aware instruction.

December 2024 to MVP delivery

Senior Software Engineer / Project Lead

Patrick Sherlund

Thesis Originator / Stakeholder Sponsor

Major William Oblak

The Team Behind AeroT

AeroT came together through close collaboration between engineering, product, stakeholder leadership, and field validation.

Patrick Sherlund

Senior Software Engineer / Project Lead

Led the product, engineering, RF workflow, and field validation across AeroT.

Stakeholder Leadership

Major William Oblak

Thesis Originator / Stakeholder Sponsor

Additional Contributors

Greg Gomez

Software Engineer

Jeremy Huynh

Software Engineer

Evan Stewart

Software Engineer

Carlos Flores

Product Manager

My role

I led AeroT from thesis concept to field-tested MVP, shaping the product and engineering direction across the RF sensing device, instructor map interface, low-bandwidth radio messaging, terrain-based signal modeling, ATAK field integration, and live validation. Our work turned an open-ended research idea into a usable training system that helped instructors show Marines, in near real time, whether their transmissions could be detected by simulated adversarial receivers.

The Stack

Device Manager

C++, Python, Raspberry Pi, SoapySDR, RTL-SDR, LimeSDR

Command View

React, TypeScript, MapLibre, Drogon, PostgreSQL

ATAK Client

Java, ATAK 5.1, MobileService, Room

Transport

Digi XBee, protobuf, WebSocket, MQTT

Modeling

Longley-Rice ITM, DTED terrain, OpenCV

Workflow

Agile

The Challenge

Marine instructors teaching electromagnetic discipline during live training had no way to show students the tactical consequence of a transmission while the lesson was still fresh. Feedback often took up to 48 hours, so students had already moved on before they understood whether their radio activity would have been detected by an adversary.

The Vision

Make the invisible electromagnetic battlefield visible in the moment by capturing live RF activity, modeling how it carries across terrain, and showing students whether an adversary could have seen them before the training moved on.

Unclassified training system

COTS RF hardware

Self-contained network

Terrain-aware RF modeling

Command View plus ATAK

The MVP proved...

48 hrs -> ~5 sec

Feedback loop compressed from delayed review to live instruction

10 devices

Connected during final field validation

~10 miles

Validated across hot, sandy, mountainous 29 Palms terrain

01 | Discover

Thesis concept, users, and field constraints

Why this started

Major Oblak's thesis asked whether inexpensive RF hardware could capture live signal activity, reconstruct it inside a virtual training environment, and show whether a student transmission would likely be detected by adversarial receivers.

What we learned from instructors

From December through February, I led discovery with training units from 29 Palms. Those conversations narrowed AeroT to one interaction: helping an instructor show a Marine whether a simulated enemy receiver could detect their transmission.

We used that interaction to scope the MVP. The tool needed to stay unclassified so instructors and students could use it, and work in training areas without reliable internet or power. That meant bringing our own network and power, with terrain data available offline.

What Was Holding Training Back

The core problem was feedback latency. Instructors could teach EM discipline, but they could not reliably show a student, within the same training moment, that a transmission would likely have been seen by an adversarial receiver.

What Success Needed To Look Like

Feedback

~5 sec

from signal capture to instructor-visible analysis

Scale

10 devices

connected during final field validation

Range

~10 miles

of 29 Palms terrain during validation

Access

ATAK + Web

field and command-node analysis surfaces

The MVP became a three-part system

We split the product by operational boundary: a field device that collected RF activity, a Command View that configured devices and ran command-node analysis, and an ATAK workflow backed by MobileService for instructors standing in the field.

Collect

AeroT Field Device

Raspberry Pi 5, SDRs, GPS, health telemetry, local persistence, XBee transport, and hotspot support.

Patrick configuring an AeroT device in the field

Analyze

Command View

React and MapLibre command-node app receiving devices, health, captures, heatmaps, receiver links, and event timeline.

Patrick adjusting an AeroT device during setup

Teach

ATAK + MobileService

ATAK-native plugin backed by on-device APIs so field instructors could view analysis without moving heatmaps over low-bandwidth radio.

Patrick discussing AeroT feedback with a Marine in the field

Engineering decisions

Keeping heatmaps off the radio network

The radio network could carry compact signal events, but heatmaps would consume too much bandwidth. We moved the analysis onto the device with MobileService and served ATAK over its Wi-Fi hotspot. That kept radio messages small, while instructors connected locally for the detailed view.

Making transmit power an instructor input

A passive receiver couldn't reliably tell us the original transmitter's power without calibration beyond the MVP's scope. We made transmit power an instructor input. Instructors could model likely detection against the terrain, with results tied to the power they configured.

The training loop we built toward

Configure devices

Capture RF activity

Model terrain

Brief students

What we proved

What we believed

We believed off-the-shelf RF hardware could help instructors show students whether a simulated enemy receiver was likely to detect their transmission, while training was still happening.

The result

At 29 Palms, we connected ten devices across roughly ten miles and delivered feedback within about five seconds of capture. Instructors could walk over to students and show them which simulated enemy receivers were likely to detect their transmission.

Iteration 1

Prove the device and Command View slice

Narrowband RTL-SDR collection, early LillyGo LoRa transport, baseline collection, GPS, telemetry, command configuration, and a first map-centered instructor UI.

Final system in the field

The final validation connected ten AeroT devices across roughly ten miles of 29 Palms terrain with ATAK clients in instructors' hands and Command View running at the command node. Signal feedback arrived within about five seconds of capture.

Patrick field testing AeroT equipment at 29 Palms

Key Learnings

Hardware failures become software symptoms

Power regulation, airflow, USB behavior, antenna placement, and enclosure choices directly affected collection reliability.

Bandwidth changes architecture

Compact signal messages could move over XBee, but ATAK heatmaps needed local Wi-Fi and on-device analysis.

Modeling must stay honest

Configured transmit power kept the product truthful about what a passive receiver can and cannot infer.

Reflection

AeroT succeeded because we kept the architecture modular, the message contracts explicit, and the product goal narrow enough to validate in the field.

The work was not just writing services. It was keeping a hardware, software, RF, geospatial, and training system coherent while every constraint was moving.

The smallest field-validatable behavior became the whole point: detect a relevant signal, model adversary detectability against terrain, and show the instructor fast enough to change the student's behavior.

Next case study

02 Bishop

Role: Co-Founder / Lead Software Engineer
Timeline: 2025 - 2026
Stack: C++, TypeScript, React, Drogon, ONNX

Search and rescue video intelligence built around edge computer vision and rapid mission review.

View details

AeroT

The Team Behind AeroT

Patrick Sherlund

Stakeholder Leadership

Additional Contributors

My role

The Stack

The Challenge

The Vision

The MVP proved...

01 | Discover

Why this started

Discovery needed field truth, not lab optimism

What we learned from instructors

What Was Holding Training Back

What Success Needed To Look Like

The MVP became a three-part system

AeroT Field Device

Command View

ATAK + MobileService

Engineering decisions

Keeping heatmaps off the radio network

Making transmit power an instructor input

The training loop we built toward

What we proved

What we believed

The result

Iteration 1

Iteration 2

Iteration 3

Iteration 4

Final system in the field

Key Learnings

Hardware failures become software symptoms

Bandwidth changes architecture

Modeling must stay honest

Reflection

02 Bishop