Make the question inspectable
“Inspect the harbour wall” is a location and an activity, not a sufficiently defined task. A useful brief might ask whether an identified joint has opened since the previous survey, whether scour is approaching a foundation, or whether a suspected impact mark warrants closer examination. Each question creates a different requirement for access, resolution, repeatability and qualified review. The vehicle should be selected after these requirements are written, not before.
Assemble drawings, asset identifiers, previous findings and the reference coordinate system. Record where those inputs are uncertain. A legacy drawing may describe the intended construction rather than the present structure. A previous image may lack a reliable scale. Planning should preserve those uncertainties instead of silently turning them into facts. Agree who can resolve inconsistencies and who has authority to approve a change in scope.
Translate purpose into acceptance criteria
Specify what constitutes usable evidence for each inspection zone. Depending on the task, this may include a documented view direction, a validated optical scale, acoustic coverage from more than one angle, or a thickness measurement made under an approved procedure. Avoid a universal “high definition” requirement: pixel count alone says little about defect visibility at a given distance, in turbid water, behind marine growth.
State how incomplete coverage will be reported. An obscured connection should not disappear from the report because the surrounding wall was photographed successfully. Separate planned area, traversed area and usable observed area. Define the circumstances requiring a repeat pass, an alternative sensor or a formal exception accepted by the asset owner. This turns quality control into an operational instruction rather than a retrospective argument.
The mission is complete when its evidence requirements are met—or its unmet requirements are made explicit.
Write the operating envelope
Depth, current, visibility and seabed clearance interact. A vehicle may be pressure-rated for the site while unable to hold the stand-off needed for a stable measurement. A tether can snag on protruding steelwork. A navigation solution suitable for open water may degrade beneath a deck. Assess interactions with moorings, intakes, vessel movements, acoustic interference and sensitive habitats, including permissions that may restrict where a vehicle can operate.
Environmental context can come from the European Marine Observation and Data Network, the British Oceanographic Data Centre and the UK Hydrographic Office. Their products can inform preparation, but regional bathymetry or tide information does not establish present conditions at an individual structure. Check product resolution, vertical datum, survey age and local restrictions. Validate critical assumptions on site.
Plan recovery before departure
Recovery is not an afterthought to autonomy. Establish response actions for lost communications, navigation degradation, flooding alarms, excessive current and inability to reach the intended recovery point. “Surface on fault” may be inappropriate beneath a structure or in a shipping channel. The safe response depends on the location, vehicle configuration and credible failure modes. Rehearse the applicable sequence with the people who will execute it.
Distinguish an autonomous underwater vehicle from a remotely operated vehicle with automated functions. Tethered operations introduce cable management and vessel-interface hazards; untethered operations introduce different communication and recovery constraints. Neither architecture eliminates the need for competent supervision. If commercial diving becomes necessary, it requires its own properly planned arrangements; guidance from the UK Health and Safety Executive is relevant to UK diving work, not a substitute for local legal assessment elsewhere.
Use standards within their scope
For offshore structural integrity, ISO 19901-9 and API RP 2SIM provide useful frameworks in their respective scopes. DNV-ST-N001 addresses marine operations and marine warranty considerations; it is not a generic certificate that makes a robotic inspection safe or sufficient. Establish the applicable edition, asset class, contractual requirements and jurisdiction with the responsible engineers. Standards inform the inspection basis; they do not choose the evidence threshold automatically.
Finish the brief with named responsibilities: who controls the mission, who judges data quality, who accepts deviations and who reviews findings. Include a small trial area before committing to the full route. A successful trial demonstrates that the proposed observation is achievable under actual conditions. A failed trial is useful evidence too: it identifies a limitation while there is still time to change the plan.