Orion's three traces, as on its training screen: what the Liberandum receives, the rhythms Orion predicts, and what's left over.

Built into Vestarax

Orion · Artificial intelligence

Orion

Copernix's AI for the RF background. Its goal is to detect a new RF signal among the well-established ones already in the environment: Orion learns their regular rhythms and takes them out of the passive-mode graph, so a transmitter that doesn't fit the pattern stands out.

A Wi-Fi access point beacons about ten times a second; a device advertises every few seconds. Orion learns rhythms like these, from 0.2 s up to 3 minutes apart, and attributes each new pulse to one of them. What it can't attribute stays on the graph. Before training, it measures the signal and says whether there is anything to learn.

ModelsOne per frequency

Rhythms0.2 s to 3 min apart

Training1 to 60 min of live signal

LearningCan keep learning while it runs

Before trainingA learnability verdict

WhereLiberandum passive mode

Try Orion live

Why Orion

Responders keep their technology. The victim still stands out.

Orion’s goal is to detect a new RF signal against the background: the well-established signals that are already in the environment, starting with the ones the search team brings in with it.

WITHOUT ORION

Responder’s phoneSmartwatchVictim’s phone

Which peak is the victim’s?

WITH ORION

Responder’s phone · learnedSmartwatch · learnedVictim’s phone · new

The victim’s phone stands out.

01

The problem

A first responder carries a phone and a smartwatch. Both transmit, on the same bands as the victim’s phone. In passive mode, without Orion, their peaks look just like the victim’s.

02

Orion learns the background

Before the search, Orion learns the rhythms of what is already on the air: the team’s own phones and watches, and the established transmitters around them.

03

The victim stands out

During the search, Orion takes those known rhythms out of the passive-mode graph. A new signal, like the victim’s phone, is what’s left.

A filter that saves lives, without asking responders to give up the technology that keeps them capable.

Try Orion · Live

Teach Orion a background. Then hide a victim in it.

This is the real Orion, running on Copernix’s server. Give it one to three devices with a regular rhythm, let it learn them, then add a victim’s phone it has never heard. Watch the two graphs: the same signal, without Orion and with it.

01
Background devices
The team’s own devices, each transmitting on a regular rhythm.
02
Train Orion
Orion records 2 minutes of this background, shown at 6× speed, then learns its rhythms. About a minute in all.
03
Add the victim
An irregular device Orion has never heard, on the same band.

WITHOUT ORION

Every device

WITH ORION

Learned · taken outVictim · left overBackground · left over

Background pulses taken out—

Victim pulses still visible—

Same model as the Liberandum’s passive mode. Here it trains on 2 minutes of signal, sent at 6× speed, and learning while running is off so the victim stays new. On the device, training uses 1 to 60 min of live signal. Nothing you do here is kept: the model is deleted when you leave the page.

How Orion works

Learn the band. Recognise its rhythms. Show what’s new.

01

Learn

Orion collects 1 to 60 minutes of live signal on the band, then fits its model in a few minutes. Each frequency has its own model, and Orion can keep learning while it runs.

02

Recognise

Each new pulse is matched to a rhythm Orion knows, from 0.2 s up to 3 minutes apart: a beacon, a device advertising, a transmitter on a schedule.

03

Reveal

In passive mode, Prediction shows what Orion’s model did not predict. Known transmitters drop away and a new or irregular one stands out.

Before training, Orion measures the signal and gives a verdict: learnable, partly learnable, not learnable, or no pulses to learn from.