§ RESEARCH TRACK · SPACE DOMAIN AWARENESS NEURA ORBITAL LEO · RESIDENT SPACE OBJECTS
ENTERED · SEP 2026

One architecture.
A second
domain.

Neura Orbital is a research track, not a product. In September 2026 Volantyx entered the Technology Innovation Institute’s OrbitSight Challenge: AI/ML that processes raw neuromorphic vision-sensor data to detect, track and visualise resident space objects. It is the first test of the Volantyx detect-track-decide architecture outside the atmosphere. There is no fielded deployment, contract or partner; the entry is under evaluation and the result is TII’s to announce.

Status Research track
Entry TII OrbitSight Challenge · 2026
The entry → Talk to the team →
Scroll
§ 01 — THE PROBLEM Orbit is crowded, and the sensors that watch it are changing. Event cameras see motion, not frames. The question we entered to answer: does the detect-track-decide loop built for RF-silent aircraft hold against objects in orbit?
§ 02 The stakes Why the domain matters

More objects, new sensors, the same decision problem.

The catalogue of tracked objects keeps growing, and neuromorphic sensors are arriving that produce sparse event streams rather than images. Somebody has to turn those streams into detections, tracks and eventually decisions. That is the shape of the problem Volantyx already works on in the air.

Sensor shift
Events

Neuromorphic vision sensors emit an event per pixel change, at microsecond timing and with high dynamic range. Faint, fast objects against a star field are exactly what they are good at — and exactly what frame-based pipelines miss.

Catalogue growth
Rising

Tens of thousands of tracked objects and many more too small to catalogue. Detection and custody, not storage, are the constraint.

Alert reality
Noise

Most conjunction warnings resolve without real collision risk, yet each demands a decision. Better tracks feed better screening — that is the direction this track points.

Cost of error
Total

A single missed object can end an asset and cascade debris across an orbital regime for decades. The decision has to be defensible, not just fast.

Neura Orbital research track — resident space objects tracked from event-sensor data
§ 03 · The entry

Raw events in. Tracked objects out.

The OrbitSight entry applies the Volantyx pipeline to event-camera data: raw neuromorphic sensor events are separated from background and star motion, clustered into candidate resident space objects, associated into tracks across the stream, and rendered for an operator with a confidence per track. It is the detection-first discipline of the counter-UAS work, pointed at a harder sky.

What it is not: a fielded system, a product, or a claim of performance. The entry is evaluated by TII against their data on their terms. We will state the outcome here when they announce it, and nothing before.

HeritageVolantyx core
InputNeuromorphic events
OutputRSO detections · tracks
StatusUnder evaluation
§ 04 How it works Events → Objects → Tracks → Decisions

From an event stream to a defensible track.

Three stages were submitted. The fourth is the direction the track points, and is stated as direction, not as work done.

STAGE 01
01 / 04

Ingest.

Raw neuromorphic vision-sensor events, timestamped per pixel, with sensor noise and star-field motion modelled so they can be removed rather than chased.

STAGE 02
02 / 04

Detect.

Event clusters that move against the background become candidate resident space objects, each with a confidence, so faint and fast objects are kept and noise is not.

STAGE 03
03 / 04

Track.

Candidates are associated across the stream into tracks consistent with orbital motion, then visualised for an operator with the evidence behind each one.

STAGE 04
04 / 04

Decide — direction.

Better tracks feed conjunction screening and fuel-aware manoeuvre decisions. This stage is where the track is headed; it was not part of the entry and is not claimed.

§ 05 Why it holds in orbit Same architecture · Harder sky

Built for the place you can't send a technician.

The disciplines behind the counter-UAS work are the ones orbit demands: detection first, evidence with a confidence attached, and decisions that can be audited afterwards.

Principle · 01Latency
Edge
Edge-first intelligence
Detection and tracking logic designed to run where the data is produced, so contact windows and ground-loop delays never gate a time-critical track.
Principle · 02Perception
Fusion
Multi-source fusion
Event-sensor tracks, catalogue data and ephemerides combined into one picture per object, with the provenance of each input kept visible.
Principle · 03Compounding
Loop
The flywheel
Every tracked object sharpens the models, and the same federated learning that serves the platform can move that intelligence between operators without moving their data.
Neura Orbital · A Volantyx research track

Research first. Claims when there is something to claim.

This page describes work submitted, not a product offered. When TII announces the OrbitSight result we will state it here, whichever way it goes. Operators and researchers who want to talk about the approach are welcome to write.

Contact the team → See the platform →