THE PLATFORM / MODULAR INTELLIGENCE

A common understanding. Not a single sensor.

Different sensors observe different parts of the maritime environment. Proteyus is building toward a consistent intelligence interface without erasing the limitations of those observations.

MODULAR PERCEPTION PROTOTYPE · MULTI-SENSOR FUSION PLANNED

One sea, multiple perspectives

Cameras observe appearance; radar provides surface contacts and motion; AIS reports participating vessel identities; acoustic systems inspect designated underwater regions; satellites contribute intermittent wide-area context. The platform direction is to preserve the origin, timing, coordinate system and uncertainty of each observation before any meaningful association.

How the intelligence chain is designed

01 / INPUTS

Separate sensor adapters

Each modality retains original timestamps, calibration references, source rights and quality indicators.

02 / INTERPRETATION

Per-sensor analysis

Specialized models produce observations. No single general model is assumed to understand all sensor physics.

03 / INTELLIGENCE

Cross-sensor reasoning

Future association and risk assessment use compatible times, coordinates, confidence and appropriate validation.

Sensor architecture and readiness

DATA SOURCEPURPOSESTATUS
Camera / image / recorded MP4Water, sky and obstacle segmentation with advisory evidenceAPI prototype; real inference performance unverified
EK80 fisheries acousticsWater-column and seabed research, scientific echogramsLegacy research archive; independent validation pending
Marine radarSurface contacts, range, bearing and motionFuture integration
AISReported vessel identity and movementFuture integration
Forward-looking sonarSubmerged hazards ahead of a vesselDistinct future model/sensor work
Satellite imagery & weatherRegional context and environmental informationFuture integration

EDGE / FORESIGHT

Time-critical observations

The eventual edge runtime prioritizes local sensing, bounded latency, accurate timestamps and explicit degraded-mode behavior. Camera live-stream and embedded validation are upcoming tasks.

HISTORICAL / FORENSICS

Evidence-rich review

The proposed historical path prioritizes source integrity, human annotations, repeatable reconstruction and separation of facts from algorithmic interpretation.

Clear technical boundaries

Satellite observations are not continuously available for immediate collision avoidance. Scientific fisheries echosounder classification is not forward-looking sonar. Existing API-level image outputs are not validated vessel control commands; physical distance, time-to-collision and automated maneuver selection require additional sensing, modeling and independent safety validation.