RESILIENT VISUAL COMMUNICATIONS

Maintain the
visual link.

The visual link for UAVs & UGVs.
Built for the moments video breaks.

MAINTAIN THE VISUAL LINK. COMPLETE THE MISSION.SCROLL TO EXPLORE
~2 kbps

Semantic stream

1.74 kbps

Measured prototype result

TRL 4

Software prototype

A smaller stream.
A more resilient picture.

BUILT WITH
INDUSTRIAL CONVICTION

IAI / CATALYST

Investor · IVFOS pilot completed

Culver Aerospace

Investor & commercial agreement

Lockheed Martin

Technical endorsement for government proposals

01 / THE CHALLENGETHE PLATFORM IS ONLY AS USEFUL AS ITS LINK.

A lost picture.
A platform at risk.

You built a capable machine.
Don't let the video link become its limit.

Electronic warfare and constrained channels can cut off the picture an operator needs. More video compression still leaves the mission dependent on a bandwidth-hungry stream.

Semantic Link is designed to preserve visual context on the bandwidth that remains—helping operators make decisions, recover assets and protect the platform.

A DIFFERENT DATA BUDGET1,000× less data*
Conventional video~2 Mbps
Semantic Link~2 kbps

*Reference comparison: a 2 Mbps video stream versus a 2 kbps semantic stream. Adaptive camera images require additional bandwidth. A functioning data channel is still required.

02 / MEET SEMANTIC LINK REAL PROTOTYPE OUTPUT

Send understanding.
Keep awareness.

Understand the scene onboard. Send its essential geometry, objects and changes. Rebuild a visual view for the operator—without transmitting the full video.

PROTOTYPE / MOTOCROSS SCENE
ONBOARD CAMERA NOT TRANSMITTED
Real camera footage of a motocross rider on an open dirt trackSOURCE / 01
GROUND RECONSTRUCTION 1.74 KBPS MEASURED
Actual prototype reconstruction of the motocross scene, showing terrain geometry and the tracked objectSEMANTIC LINK / 02

The same scene. Camera video stays onboard; the operator view is rebuilt from the semantic stream.

1.74 kbps

Measured link rate

136 B

Per scene update

1.5 / sec

Scene updates

ACTUAL PROTOTYPE. MEASURED OUTPUT.

Actual TRL 4 software output on real-world footage. Embedded integration and flight validation are the next steps.

03 / INSIDE THE LINKRADIO-AGNOSTIC BY DESIGN

The scene changes.
That's what we send.

Keep the heavy perception onboard. Transmit a compact description of what matters through the platform's existing radio link.

Concept visualization of a vehicle and terrain represented as a semantic wireframeWORLD MODEL / CONCEPT VISUALIZATION
GEOMETRY · OBJECTS · CHANGES
01Scene geometry

Terrain, structures and spatial context. A compact representation gives the operator a view of the environment.

02Detected objects

Mission-relevant vehicles, people and assets—represented as objects rather than entire video frames.

03Meaningful changes

Movement and changes in the scene are prioritized, instead of repeatedly sending every pixel.

04Real camera snapshots

Adaptive, prioritized images complement the reconstructed scene. Image frequency and quality depend on the available link budget.

ONBOARD

Camera → scene understanding

EXISTING RADIO

Ultra-low-bandwidth stream

GROUND STATION

Reconstructed visual context

04 / WHERE IT MATTERSAIR. GROUND. BEYOND.

Different platforms.
The same critical link.

UAV reconnaissance concept from the investor presentationILLUSTRATIVE APPLICATION
AIR / RECONNAISSANCE

Keep context.
Extend awareness.

Visual situational awareness for reconnaissance platforms operating over long-range and bandwidth-constrained links. Keep mission-relevant context available as conventional video degrades.

05 / BUILT TO INTEGRATEYOUR PLATFORM. YOUR RADIO. SEMANTIC LINK.

Build it in.
Make it yours.

A path from evaluation to platform integration, designed around UAV and UGV manufacturers.

Current stage: TRL 4 software prototype

Embedded integration targetsTECHNICAL BRIEF
ComputeJetson Orin Nano
or equivalent
Power target7–15 W
Weight target<250 g
Latency target<300 ms

These are integration targets from our development plan, not measured specifications of a production module. Total link usage depends on semantic updates, transport overhead and adaptive images.

06 / UNSOLVED MACHINESLOS ANGELES · UKRAINIAN OPERATIONS

Hard problems.
Machines that matter.

Computer vision. Autonomy.
Engineering with a purpose.

Unsolved Machines is the U.S. brand of Codective LLC. We build technology for autonomous platforms, with experience spanning visual-inertial navigation, onboard perception and semantic reconstruction.

Our IVFOS navigation pilot with IAI is complete. Semantic Link builds on that engineering foundation, with a focused goal: resilient visual communications that help operators preserve awareness and protect valuable machines.

Anton GerasymenkoFounder & CEO
CONTACT TECHNICAL DEMOS · INTEGRATIONS · PARTNERSHIPS

Keep your machines
connected.

Get in touch.

Tell us about your platform, integration, or partnership.

We use your contact details to respond to your inquiry.

ACTUAL SOFTWARE PROTOTYPE OUTPUT

One scene. A smaller stream.

ONBOARD CAMERA · NOT TRANSMITTED
Expanded source camera footage of a motocross rider
GROUND RECONSTRUCTION · 1.74 KBPS MEASURED
Expanded actual reconstruction of terrain and the motocross rider from the semantic stream

Real prototype output from the supplied investor presentation. The camera footage is shown for comparison; it is not transmitted in the semantic stream.