Buy the AUV or the Capability? The HUGIN Superior Question

A sovereign mapping asset enters service

Norway has put its first state-owned deep-sea mapping AUV into the water. The HUGIN Superior, procured by the Norwegian Offshore Directorate (NOD) with 2025 funding from the Ministry of Energy, began its first expedition in the northern Norwegian Sea. The vehicle will be operated by NORMAR (the Norwegian Marine Robotics Facility), affiliated with the University of Bergen, and it was christened in Bergen before departure, with NOD, the Ministry of Energy, the Norwegian Institute of Marine Research and Kongsberg Group represented at the ceremony. It is rated to 6,000 m and collects seabed data using sonar, transmitting acoustic energy and interpreting the returns.

One line in the announcement carries more weight than the rest: this is the first time Norway has held this class of capability itself, having previously sourced it through private contractors. That is a procurement decision as much as a scientific one, and it is the decision we want to examine here. Whether you sit inside a national body, an oil company survey group or a contractor’s asset-planning team, the move from buying deep-water AUV services to owning the platform reshapes where your risk sits and what you have to prove.

Why the ownership question lands on your desk

Buying survey services and owning survey capability are not the same business. When you charter a spread, the contractor carries the mobilisation risk, the spares holding, the crew competency, the calibration records and – critically – the data-quality assurance case. When you own the asset, all of that transfers to you, along with the utilisation risk. A deep-rated AUV that flies 60 to 90 days a year is an expensive way to hold a capability; the same asset flying 200-plus days a year against a real programme changes the arithmetic entirely.

The attraction of ownership is straightforward. You control the schedule, you own the raw and processed data outright, you build institutional knowledge that does not walk off the vessel at demobilisation, and for a state actor you hold a sovereign capability that does not depend on a contractor’s availability. Against that, you inherit a standing cost base and, more importantly, an assurance obligation. The contractor’s quality management system, its S-44 acceptance procedures and its positioning verification were part of what you were paying for. Bring the asset in-house and you have to reconstruct all of it, or the data you produce will not withstand scrutiny.

This is the core of the client decision. The vehicle is the easy purchase. The harder acquisition is the workflow, the competencies and the verification regime that turn sonar returns at 6,000 m into a defensible bathymetric product.

What actually drives a deep-water AUV programme

Three engineering realities govern whether owned AUV mapping delivers usable data. None of them is solved by buying the vehicle.

The navigation error budget dominates everything. An AUV’s absolute position is only as good as its aiding. The core is an aided inertial navigation system, with a Doppler velocity log (DVL) providing bottom-lock velocity, and acoustic aiding from USBL or long-baseline (LBL) transponders. With DVL bottom lock and a high-grade inertial unit, position drift is typically held to a small fraction of a percent of distance travelled. Lose bottom lock – fly too high above the seabed for the DVL’s range, or transit through the water column – and the solution reverts to unaided inertial drift, which grows without bound.

At depth the surface aiding weakens exactly when you need it. USBL positioning uncertainty scales with slant range; over several kilometres of water the surface fix contributes metres, not centimetres, of horizontal uncertainty. That is acceptable for reconnaissance but not for tying together adjacent survey blocks or re-occupying a target. The mitigations are well understood: pre-laid LBL arrays for the survey box, terrain-relative navigation against a prior bathymetric grid, and disciplined surfacing for GNSS resets where the surface is open. Every one of those has to be planned before mobilisation, because none can be improvised on site.

Sound velocity is the silent source of error. Bathymetric accuracy from a multibeam or interferometric sonar is a two-way travel-time measurement converted to depth and position through the sound-velocity field. High-latitude and deep water present a difficult column: a cold, relatively fresh surface layer over warmer, saltier Atlantic-origin water produces strong gradients, and the sound-speed structure varies over short spatial and temporal scales. Get the sound-velocity profile wrong and the outer beams smile or frown, cross-lines disagree, and the error propagates into everything downstream. A sound-velocity sensor at the sonar head handles the launch-angle correction; the through-column refraction still demands a disciplined cast regime and continuous checking against beam behaviour.

Endurance, launch and recovery set the real productivity. A vehicle rated to 6,000 m spends hours in descent and ascent alone, before a metre of line is run. Energy budget, dive planning and – above all – launch and recovery in a seaway determine how many productive bottom-hours you actually get. In northern waters the weather window, not the vehicle, is usually the limiting factor. Owners who model productivity from the sonar swath width and vehicle speed, and forget the sea-state limits on the launch and recovery system, consistently overstate what a season will deliver.

Where owners get this wrong

1. Treating the platform as the capability

The most common error is equating the vehicle with the deliverable. A HUGIN-class AUV is a mature, capable platform, but a platform produces nothing without a calibrated sensor suite, a verified navigation solution, a processing pipeline and people who can run all three. The binding constraint on a new in-house programme is almost never the vehicle’s reliability; it is the maturity of the surrounding workflow and the depth of the operating team. Budget and plan for the workflow as the primary acquisition, with the vehicle as one component of it.

2. Specifying accuracy loosely

“High-resolution seabed data” is a marketing phrase, not a specification. Resolution and accuracy are different quantities, and both must be tied to a recognised standard. IHO S-44 Edition 6 (2020) defines the orders – Exclusive Order, Special Order, Order 1a, 1b and Order 2 – each with explicit total vertical uncertainty (TVU) limits. The a and b coefficients define the TVU model specifically, combining as TVU = √(a² + (b×d)²), where d is depth; the separate total horizontal uncertainty (THU) limits are specified differently, as a fixed allowance plus a percentage of depth. Deep-ocean reconnaissance may only justify Order 2, while near-bottom AUV work over a target can approach Special Order. If the acceptance criteria are not written against a specific S-44 order, with a total propagated uncertainty (TPU) model behind them, there is no objective basis to accept or reject a survey.

3. Under-resourcing the positioning ground truth

Operators new to owning AUVs frequently accept the vehicle’s own navigation solution as ground truth. Operators new to owning AUVs frequently accept the vehicle’s own navigation solution as ground truth, but it is not. The navigation solution needs independent verification – a USBL or LBL cross-check, agreement between reciprocal survey lines, and re-occupation of known features. Without an independent positioning reference, systematic offsets go undetected and only surface when a later survey or a construction contractor cannot reconcile your grid with theirs.

4. Ignoring the sound-velocity regime

We see sound velocity treated as a box-ticking cast at the start of a dive. In stratified high-latitude water that is not enough. The refraction correction depends on the profile the vehicle actually flew through, and that changes with position, tide and time of day. A single cast per dive is a false economy that shows up as unresolved cross-line disagreement.

5. Leaving data management until after the season

A 6,000 m AUV generates large volumes of bathymetry, imagery and navigation logs. Without an archive structure, metadata standard and processing turnaround agreed before the first dive, raw data accumulates faster than it can be turned into products, and provenance is lost. For a national dataset intended to be authoritative and reusable, the metadata and lineage are part of the deliverable, not an afterthought.

6. Underestimating the assurance transfer

When the work was contracted, the contractor’s quality system, competency records and calibration history came with it. In-house, that entire audit trail has to be built and maintained. Class-society and industry guidance on positioning and survey operations – IMCA and IOGP survey and positioning guidance, alongside class-society rules such as DNV, among them – provides a ready framework, but it has to be adopted and enforced, not assumed.

How to make the call

If you are weighing owned deep-water AUV capability against continued contracting, the following will keep the decision honest and the delivery defensible.

• Size the decision on utilisation, not novelty. Model the annual productive bottom-hours realistically, net of transit, descent and ascent, and weather downtime at your latitudes. Below roughly 100 to 120 productive survey days a year, contracting usually remains cheaper once the standing crew, spares and assurance overhead are counted. Own the asset when the programme genuinely fills it.

• Write acceptance criteria against IHO S-44 Edition 6 by order. State the target order for each survey type, publish the TVU and THU limits, and require a TPU model and cross-line agreement within those limits as a condition of acceptance. Make a patch test (roll, pitch, yaw, latency) a mandatory calibration deliverable at mobilisation and after any sensor change.

• Specify the navigation aiding plan explicitly. Require DVL bottom lock throughout survey lines, set a maximum aided-INS drift as a fraction of distance travelled, and define the acoustic aiding – USBL slant-range limits, LBL array geometry, or terrain-relative navigation – for depths and legs beyond DVL range. For any under-ice or open-water autonomous transit, require terrain-relative navigation against a prior grid and a documented reset strategy.

• Impose a sound-velocity regime, not a single cast. Fit a sound-velocity sensor at the sonar head, mandate a minimum cast frequency tied to the water-column variability, and require outer-beam refraction to be checked against reciprocal lines during processing, not after the season.

• Separate acquisition from quality control. Give QC to a function independent of the acquisition team, with authority to reject a line or a block. Require every gridded product to carry lineage and metadata to a recognised standard so the dataset is auditable and reusable years later.

• Plan launch and recovery around the sea state, not the swath. Define operational and survival significant wave-height limits for the launch and recovery system, build the weather-downtime allowance into the schedule, and rehearse recovery procedures before the vehicle is committed to a deep dive far from support.

• Buy the workflow with the vehicle. Whether through the platform vendor, a knowledge-transfer arrangement or an experienced operating partner for the first campaigns, resource the transfer of processing procedures, calibration discipline and competency development. The vehicle will arrive ready; the capability has to be built.

Norway’s move gives it control of its own deep-sea data and the ability to task the asset against national priorities. That is a sound strategic position. The lesson for everyone else considering the same step is that ownership relocates the hard problems rather than removing them. The vehicle you can buy off a proven production line. The navigation error budget, the sound-velocity discipline, the quality-assurance regime and the people who hold it together are what actually determine whether the data is worth having.

---

Based on: Norway Deploys New Deep-Sea Mapping AUV on First Expedition