Choosing an inspection ROV: five questions that actually matter

Almost every ROV enquiry opens with depth. It is the wrong place to start, and starting there is the most reliable way to buy a vehicle that technically meets the specification and disappoints in service. These are the five questions that change the answer, roughly in order of how much they change it.

1. What is the visibility, on a normal bad day?

Not on the day of the demonstration. On the day you will actually need it.

Below roughly a metre of visibility, camera specification stops being the deciding factor. No amount of 4K resolves water you cannot see through, and a vehicle chosen on imaging alone will underperform in exactly the conditions that made you buy it. Acoustics become the primary sensor at that point, and the vehicle needs to be able to carry one.

Conversely, if you work in genuinely clear water, paying for sonar you will rarely switch on is a real and avoidable cost.

2. Who is going to operate it?

This decides more purchases than any specification line, and it is the one most often left out of the requirement.

Modern inspection-class vehicles are designed for a single non-specialist operator - that is the change that made routine ROV inspection realistic rather than a contracted event. Coastguard crews, customs officers, fish farm staff and university researchers all run these systems without dedicated pilot training.

But "can be operated by non-specialists" varies across the range. If the person who will use it weekly cannot deploy and recover it without help, the programme quietly stops after the novelty wears off. That is the most common failure mode in this category and it has nothing to do with the vehicle's specification.

3. Do you need to measure, or only to look?

Looking is a camera problem. Measuring is a vehicle problem.

Quantified measurement - wall thickness, corrosion extent, dimensional change - needs a stable platform to measure from, which means station-keeping, which in practice means a Doppler Velocity Log or equivalent. It also frequently means a measurement payload such as a laser scaler or a dedicated measurement tool.

If your output is "here is a video showing a problem", you may not need any of that. If your output is "the wall has lost 3 mm since last year", you need all of it, and adding it later is expensive or impossible depending on the platform.

4. Does it have to be repeatable?

Closely related to measurement, but distinct enough to ask separately.

A one-off inspection tells you today's condition. A programme that returns to the same place with the same framing tells you the rate of deterioration, which is what actually drives a maintenance plan and a capital budget.

Repeatability needs positioning - acoustic or inertial - not just good piloting. Without it, two surveys of the same structure are two anecdotes rather than a trend, and you will not know which of the two is the outlier.

5. How does it get to the job?

The least glamorous question and frequently the binding constraint.

• Case dimensions and weight. Can one person move it? Can it fly as accompanied baggage?
• Batteries. Lithium cells have transport restrictions that can make an otherwise ideal vehicle impractical for a fleet that mobilises by air.
• Launch point. Deploying from a quayside, a small boat and a vessel's deck are three different problems.
• Tether management. The tether is what limits real-world work far more often than the vehicle does.

Where depth actually comes in

Last, and usually oversubscribed. Broadly: entry-level systems reach around 100–200 m, mid-range professional systems 200–300 m, and advanced systems 350 m and beyond.

Depth rating by class, against where the work is

For hull inspection, aquaculture, ports, inland water and most infrastructure work, 100–300 m is comfortably sufficient. Buying 500 m of depth rating for work in 20 m of water is the most common way to overspend in this category, and the money would almost always have been better spent on sonar or positioning.

And the question behind all five

What happens to the data afterwards? A vehicle that produces impressive live footage and an unusable pile of files the next morning has solved the wrong problem. Ask what the report looks like, who has to accept it, and how long it takes to produce - before the trial, so you can test it during.