AUV

Autonomous Underwater Vehicle

An autonomous underwater vehicle concept for underwater object detection, hazard assessment, explainable decision support and inspection prioritization.

Try the interactive demo About the project

The AUV: a gold, torpedo-shaped hull with a propeller and tail fins at one end.

Target operating depth: 1–30 m below the water surface (concept)
Sensing: Multibeam echosounder, water-column acoustic data and onboard sensors
Output: Risk-assessed targets with an explained decision and a recommended action

About the AUV

An autonomous underwater vehicle built around the team's own design. It surveys an area underwater, detects objects, assesses whether they may be hazardous, explains each decision and tells the operator which targets to inspect first.

The problem

Coastal and offshore waters hold marine life, natural features and man-made objects side by side: shipwrecks, debris, lost equipment and, in some areas, old ordnance. Some of these are hazards to navigation or to people working in the water; most are not.

Different sensing methods provide different types of information about underwater targets. Acoustic sensing can remain useful in poor visibility, while target interpretation still requires analysis of multiple characteristics. No single sensing source answers the question “what is this object, and is it a hazard?” on its own.

What the system does

The AUV concept surveys an area in parallel lines and combines acoustic and onboard sensor data to detect underwater targets. Each target is analysed, classified and given a risk level, a risk type and a decision with an explanation. When the evidence is not enough, the system reports the object as unknown and asks for verification instead of forcing a class.

Targets are marked and reported to the operator at the surface.

Underwater photograph of a large ship propeller and hull structure lying on the seabed. — A wreck site on the seabed: the kind of object the AUV is meant to find, assess and flag. Reference photograph.

Operating environment

Target operating depth for the concept: 1–30 metres below the water surface. This is a conceptual range; the physical vehicle has not been tested or certified across it.

The conceptual operating range of the AUV, shown against the water column. A target for the concept, not a tested rating.

How it works

From the water to the operator: the AUV senses its surroundings, detects targets, classifies them, assesses their potential risk, explains each decision and recommends an action.

Underwater Environment
The survey area, 1–30 m below the surface (concept).
Sensor / Acoustic Data
Multibeam echosounder data, water-column acoustic data and onboard sensor data.
Target Detection
Candidate objects are found in the sensor data and given a target ID and a detection time.
Target Feature Analysis
Shape, geometry, structure, size category, movement and difference from the background are examined.
Object Classification
Marine life, man-made structure or potential hazardous object — or unknown when the evidence is not enough.
Risk Assessment
Non-hazardous, potential hazard (or high potential hazard), or unknown and requiring verification.
Risk Type Identification
For example navigation / obstruction risk or potential explosive hazard.
Explainable Decision
A decision together with the evidence and the rule that led to it.
Target Marking
The target is marked at its position on the survey map.
Recommended Action
Continue the survey, inspect the target, or verify it.
Local Data Storage
Target records and raw sensor data are stored on board.
Data Transfer / Operator System
Priority target records reach the operator at the surface.

Sensing and communication

Multibeam Echosounder

Seafloor mapping, bathymetric information and acoustic target detection.

Water-Column Acoustic Data

Acoustic returns within the water column, used to detect potential targets.

Onboard Sensors

Sensor data used to support target detection and analysis. Exact sensor specifications are TBD.

Underwater Acoustic Communication Conceptual

Links the AUV with the operator and the surface system. It is separate from the sensing layer: sensors observe the environment, while communication carries the results. Communication specifications: TBD.

Classification, risk and decision

Each target is analysed from its shape, geometry, structure, size category, movement where applicable, acoustic characteristics and difference from the surrounding background. The classification, the risk assessment and the recommended action are kept as three separate results.

Non-Hazardous

Marine life, and objects with no immediate hazard. Decision: continue survey / log observation.

Potential Hazard

Wrecks, obstructions and other man-made structures. Decision: requires inspection. A potential mine is a high potential hazard; it always requires verification and is never reported as confirmed.

Unknown — Requires Verification

Objects without enough evidence for a reliable class. The system says “unknown” instead of guessing. Decision: requires verification.

How a decision is explained Simulation example

Feature	Evidence
Rigid / geometric structure	Detected
Non-biological appearance	Detected
Biological morphology	Not detected
Size category	Large
Acoustic return differs from surrounding background	Detected

Classification: Man-Made Structure
Risk: Potential Hazard
Risk type: Navigation / Obstruction Risk
Decision: Requires Inspection

Potential risk is expressed as a priority for the operator to review; the system does not determine danger with certainty. The decision engine in this prototype is rule-based and explainable. Trained recognition models are future development and are not part of it.

Verification and background comparison

Multi-pass verification Conceptual

First pass
Target detected
Additional pass, from another angle
Evidence comparison
Decision update

A suspicious target is passed again from the neighbouring survey line, so it is seen from another angle. The two observations are compared before the decision is updated. In the demo this appears as a conceptual multi-angle consensus; no trained model is involved.

Environmental / acoustic background Conceptual

A detected return is compared with the surrounding seafloor and water column. A return that stands out from its background is stronger evidence that an object is present. In the demo this comparison is a simulated evidence item; no trained anomaly-detection model is claimed.

Data transfer Conceptual

Underwater links carry little data, so the concept keeps large raw sensor data on board and sends compact target records first.

Sensors
Onboard Processing
Target Detection
Target Analysis
Target Record
Local Storage
Data Prioritization / Compression
Underwater Acoustic Communication
Surface / Operator System

Sent first: priority event data

Target IDClassificationRiskRisk typeLocationDepthDetection timeDecisionExplanationAlert

Kept on board

Raw and large sensor datasets stay in onboard storage for later retrieval. Communication specifications (range, bandwidth, modem, protocol): TBD.

Selected Red Sea fish

Marine-life recognition is shown with selected Red Sea species. It separates marine life from man-made objects and potential hazards; it does not attempt to recognise every species. The demo uses the lionfish as its marine-life example.

LionfishClownfishGrouper

Underwater photograph of a red-and-white striped lionfish with long fan-like fins among corals. — A lionfish on a reef. Reference photograph.

Underwater photograph of several lionfish swimming around a rusted, coral-covered metal structure. — Lionfish around a submerged structure: marine life and man-made objects share the same scene. Reference photograph.

AUV design

The AUV is built around the team's own hull design: a streamlined, torpedo-shaped body with tail fins and a propeller. Move over the image to inspect the details.

Platform: Autonomous Underwater Vehicle
Target operating depth: 1–30 m — concept
Sensing: Multibeam echosounder, water-column acoustic data, onboard sensors
Processing: Onboard data processing — concept
Communication: Underwater acoustic communication — concept; specifications TBD
Detection: Underwater object detection
Decision: Explainable risk assessment
Survey: Grid / parallel-line survey

Interactive demo

Software simulation of the proposed AUV mission: a grid survey that detects underwater objects, assesses their potential risk and explains every decision.

Scenario Ready

Scanning
Detection
Analysis
Classification
Risk assessment
Decision
Alert & marking
Verification

No active alertsSimulation alert

Survey area — grid

Simulated navigation data

AUV Completed track Covered area Non-hazardous Potential hazard High potential hazard Unknown

Survey coverage

Simulation estimate

Distance covered0 m

Area covered0 m²

Coverage0.0%

Remaining area0 m²

Coverage values are simulated estimates and not validated physical performance. Area covered = distance along survey lines × the smaller of swath and line spacing; turns add no new area. Demo time is compressed and does not represent vehicle speed.

Target Observation

Reference image — not a live feed

Reference image

Analysis

Simulation / concept demonstration

Target log

Demo valuesTargets 0Demo time 00:00

Target	Detected	Identification	Classification	Risk	Decision	Detected by	Line, bearing (Sim.)	Position (Sim.)	Depth (Sim.)

Demonstration Video

A short walkthrough of the AUV interactive prototype.

[Video will be added here]

Project materials

Supporting material for the AUV project.

A0 Scientific Poster

[Poster will be added here]

Project Photo

[Photo will be added here]

Additional Materials

[Google Drive link will be added here]