Executive Summary
PXGEO has signed a one-year setup agreement with Equinor to trial autonomous subsea inspection using Saab's Sabertooth Underwater Intervention Drone. The vehicle docks and inspects infrastructure in AUV mode using onboard sonar and cameras. Our team examines what this signals about operator readiness, where validation programmes typically fail, and what survey managers should demand before resident drones replace tethered ROV spreads.
What Happened
PXGEO has signed a one-year contract with Equinor to conduct autonomous subsea inspection trials using the Saab Sabertooth unmanned underwater vehicle (UID). The goal is to test and validate the autonomous inspection capabilities – essentially, to ensure these vehicles deliver on their promises. The key is that these capabilities will be tested in the open sea. The agreement includes field trials to validate the entire system under real-world conditions, docking capability tests, and inspections using Sabertooth sensors. This is a step into a market where unmanned operations are gradually moving from a future technology to standard practice – but not without significant effort. Here’s our assessment of what’s truly needed for such a program. We’ve seen enough similar tests to know where the biggest challenges lie.
Why This Matters
These autonomous inspection agreements aren’t just marketing hype. They represent concrete strategies from industry giants like Equinor, aimed at minimizing risk before making major investments. A one-year trial with specific tasks? That’s a clear indication they’re in the testing phase, not operational. We need to demonstrate Sabertooth’s effectiveness before it’s even remotely deployed in real-world tasks. We mention this because the industry has been talking about stationary and unmanned systems for years, but you know what? Most inspections still rely on ROVs on dynamically positioned vessels, creating a significant gap between technological capabilities and established operational practices. These trials are designed to close that gap. Sabertooth is no joke. It operates with both wired and wireless connections, docks with underwater garages or charging stations, and carries a substantial number of sensors for visual and sonar surveys. Imagine a hybrid autonomous underwater vehicle (AUV) and ROV (ROV), stationary yet deployable on demand – the dream of many operators for years. This presents a significant challenge; the hardware is manageable, but what about trust in autonomous decision-making? That’s the crux of the matter. And that’s precisely what these test programs are designed to address.
The Reality on Deck
“Autonomous behavior”? Honestly? That hides a mountain of tedious technical challenges. Docking or autonomous inspections require multiple elements to interact, and each one presents a potential problem. First, there’s positioning. An autonomous vehicle near structures needs precise positioning – and this is where we run into difficulties – typically using INS with DVL position sensor and acoustic signals from USBL or LBL arrays. But what about near steel and concrete? Acoustic multipath propagation and shadowing compromise positional accuracy precisely where it’s needed most. A major problem. The vehicle must pinpoint a precise location that meets surveying requirements, while the structure distorts its signals. Then there’s docking – underwater docking with currents and limited visibility? What seems simple in a calm demonstration becomes increasingly difficult off the coast of Norway – choose your operating environment wisely. This is a controlled experimental setup to prove the concept before real-world field trials. And finally: data is crucial. Spectacular autonomous tests are worthless if the images and sonar don’t meet the standards of a manned underwater vehicle. Imaging requires image overlay, illumination, and distance measurement between objects; sonar must detect objects critical to integrity. Autonomy doesn’t lower the requirements. It eliminates human error in correcting it during the imaging process.
Where Clients Get It Wrong
We’ve seen it before: customers and contractors make mistakes when using autonomous inspection systems. The situation repeats itself.
1. Treating the trial as a tech demo instead of an acceptance process
The goal of this system is not to demonstrate its capabilities. That much is common knowledge. Rather, it’s about documenting what constitutes “acceptable autonomous behavior” and gathering corresponding evidence. Customers who engage with impressive demonstrations may receive engaging videos, but they won’t have any comparative data. Therefore, before you begin, define success/failure criteria: positioning accuracy, docking success rate, scope of self-inspection, and data quality relative to integrity requirements.
2. Assuming autonomy removes the survey QA chain
No; autonomous vehicles deliver the same survey results and therefore require the same quality control. Visual inspection, sonar mosaics, dimensional data – all of this must be verified. In my opinion, the S-44 standards regarding uncertainties remain relevant. Furthermore, autonomy necessitates more stringent quality control, as these critical aspects are not monitored. Verification mechanisms should be implemented to uncover gaps and errors before the work is completed, not only after demobilization.
3. Underestimating the exception case
Autonomous driving works as planned, but what happens when deviations occur? That’s the real test. An obstacle, an algal bloom blocking the cameras, unexpected anomalies – all of these put the behavioral patterns to the test. Proof of operational readiness lies in the availability of reliable alternative options: waypoint abort, return to berth, remote control. Not just assuming, but proving it.
4. Confusing “unmanned” with “uncrewed and unsupervised”
Misleading marketing contributes to confusion – does this apply to most businesses today? Controlled autonomy means the vehicle is remotely controlled and a pilot can intervene at any time. Keep in mind: the risk is not comparable to that of full autonomy. Be transparent about the level of control: communication channels, acceptable delays, and intervention conditions.
Standards, Competence, and the Validation Gap
Standards for this work are still being developed, so validation testing is essential. The IMCA guidelines (ROV operation, competence, and project execution) continue to form the basis for wireline operations and can be adapted for autonomous systems. The IMCA is currently working on remote and autonomous operation, and DNV, among others, has published relevant recommendations. Operators must compare the data collected during testing with these recommendations rather than adapting them later. Competence is a less obvious but no less important aspect. Operating an underwater drone is different from operating an ROV. You need specialists with expertise in autonomous control, mission planning, acoustic positioning, data backup, and standard subsea construction. This is not easy – there is a shortage of skilled personnel in all these areas. A framework agreement is essentially a training program, whether it is recognized as such or not. Treat the test as a learning experience and document the lessons learned. And don’t forget the purpose of the inspection: Asset integrity dictates the scope of work, and integrity assurance specialists are concerned with fault detection, not ultra-autonomy. If an autonomous flight fails to detect what a manned inspection would – corrosion, cracks, shielding displacements – then the autonomy is meaningless. To verify this, the results of the autonomous flight must be compared with reference data collected by a manned vehicle on identical infrastructure.
What We Would Want to See Before Scaling
What makes this year a milestone on the road to success? We recommend the following to research directors and facility managers: - A baseline comparison. Conduct autonomous and piloted inspections of identical assets. Meet or exceed baseline performance before replacing assets. - A quantitative assessment of autonomous control performance. During testing, capture docking success, non-intervention workload, and positional accuracy, rather than selectively eliminating these. Set targets in advance. - A proven remote intervention pathway. Repeatedly demonstrate that the remote pilot can intervene immediately in the event of a vehicle failure. Measure latency and recovery success. - Ensuring data quality for autonomous control. Implement automated quality control of roadway, lighting, and positioning before leaving the site. Refer to the IHO S-44 standard for bathymetric data. - A documented control model. Detailedly describe the levels of human control, the communication structure, and the triggers for immediate intervention. Define responsibilities before scaling. Consider the impact of weather and terrain. Coastal trials serve as proof of concept. Acceptance requires realistic depths, currents, and visibility. Plan phased interventions. Don’t jump from calm waters to active areas. The path is clear. Manned and autonomous systems will handle much of the inspection work, and operators like Equinor are paving the way with disciplined inspection methods. The value lies not in flashy technology, but in rigorously demonstrating that an autonomous system can achieve the integrity of a manned system. It’s a classic problem. If you achieve acceptance, the technology follows. If you fail, you might have a nice video, but no basis for operation.
Based on: PXGEO Secures Equinor Deal for Autonomous Subsea Inspection Trials
