Spatial Intelligence
A process unit has hundreds of valves. Many of them are the same make, the same size, the same colour, mounted at the same height on lines that run parallel for forty metres. The technician knows exactly which tag number they need. The problem is working out which of the physical objects in front of them carries it.
In most industries that is an inefficiency. On a live hydrocarbon plant it is the front end of an incident. Operating the wrong valve, isolating the wrong line, or signing off an isolation that was never made are not productivity problems, and they are why identification errors sit in so many incident reports as a contributing cause.
That is the specific gap augmented reality closes in oil and gas, and it only closes if the overlay lands on the correct unit rather than in the correct area.
The strongest case in the industry, and the one to build a first deployment around.
The technician holds up a device, and the tag number, line, service, last inspection and current status appear anchored to the physical object rather than in a list they have to reconcile. The value is not the seconds saved reading a schedule. It is that the ambiguity between two identical valves is removed at the point of work, by the person doing the work, without asking anyone.
A Fortune 100 industrial customer running this in a private cloud measured 2.5x technician productivity on asset finding and a 4x reduction in mean time to repair. The repair itself did not get faster. The time spent establishing which asset was the right one collapsed.
Lock-out tag-out is a procedure that depends entirely on the isolation points being correctly identified, and it is verified today mostly by a second person walking the same route with the same paperwork.
With isolation points anchored in space, the verification step becomes: stand here, look at this, confirm the state, capture it. The by-product is a timestamped record of what was confirmed, from what position, against which revision of the isolation plan. Plants pay for that evidence today, in paperwork filled in afterwards. Getting it as a by-product of the check changes the arithmetic more than the time saving does.
Turnarounds are short windows with very expensive minutes and hundreds of contractors who do not know the site. Most of the lost time is not the work. It is contractors locating scope, walking the wrong route, waiting on a permit for an item that was misidentified, or discovering an interference nobody modelled.
A shared spatial map used in planning and then in execution means the scope walked in the office is the same scope on the deck, in the same coordinate frame, for every crew. The same argument in a commissioning context is on the QA and commissioning page.
Worth stating plainly, because a lot of industrial AR marketing quietly assumes a GPS fix.
Inside a steel process structure there is no usable satellite position. Between vessels, under decks, in pipe racks, inside a column: none of it works, and the partial positions you do get are worse than none, because they are confidently wrong by tens of metres. On an offshore facility this is the entire working environment, not an edge case within it.
Everything above therefore has to work with no satellite signal at all. A visual positioning system does, because it does not use one. The device camera matches what it sees against a prebuilt 3D map of the facility and returns 6-DoF pose: exactly where the camera is and exactly which way it is facing, in the shared coordinate frame of that map. MultiSet returns that to sub-5 cm.
Two properties make it usable for this work. The pose is absolute, so a tag placed on a valve stays on that valve across sessions, devices and shifts, unlike the relative tracking built into every phone and headset, which resets whenever the app closes. And the map is shared, so the planner, the technician and the verifier are all working in the same frame.
This is the question that decides whether any of this reaches the deck, and it is usually asked far too late.
Hazardous areas are classified by how likely an explosive atmosphere is. Under the ATEX and IECEx framework, gas atmospheres are graded Zone 0 (present continuously), Zone 1 (likely in normal operation) and Zone 2 (unlikely, and short-lived if it happens). North America uses the Class and Division system to describe the same risk. The classification is a property of the location, set by the operator, and it dictates what equipment may be carried into it.
Ordinary consumer hardware does not qualify. A standard phone or headset is not certified for Zone 1, and site rules on most facilities will stop it at the gate regardless of what it is being used for. Equipment for these areas comes from a small set of specialist manufacturers building Ex-certified devices, including families from Ecom (Pepperl+Fuchs), RealWear, Aegex and Bartec. Certification is granted per model and per zone, so it has to be checked against the specific device and the specific area rather than assumed from a manufacturer's reputation.
The practical point, and the reason this is not a blocker. A camera-based visual positioning system is software. It needs a camera and enough compute, both of which certified industrial devices have. It does not require new hardware into the hazardous area, does not require anything mounted on the plant, and does not require anything powered near the process. That last part matters more than it first appears: beacon and anchor-based positioning means installing powered transmitters throughout a classified area, and every one of them becomes a certification question, an installation permit and a maintenance visit into a live zone.
So the sequence is: the operator decides which devices are approved for the area, and the positioning software runs on those. Not the other way round. Any vendor whose answer depends on their own hardware entering a Zone 1 area is proposing a much longer project than they are describing.
Read across the row for what each approach demands of a classified area.
| Approach | Works with no GNSS | New hardware into the zone | Accuracy | The blocker |
|---|---|---|---|---|
| P&ID and tag plates | Yes | None | Human | Reading a tag still requires reaching the tag, and plates corrode, get painted and go missing |
| Markers and QR | Yes | Printed targets on every asset | Exact at the marker | Placement and upkeep scale with the number of assets, in an area every visit needs a permit |
| Beacons and UWB anchors | Yes | Powered transmitters throughout the area | 1 to 5 m, or 10 to 30 cm for UWB | Every unit is a certification question, an installation permit and a maintenance visit into a live zone |
| Handheld RFID | Yes | A tag on every asset | Proximity only | Tells you what is near, not which one you are looking at |
| Visual positioning | Yes | None. Runs on devices already certified | Sub-5 cm | Needs a scan of the facility, and the map has to be updated after structural change |
Operators in this sector do not send interior imagery of a live facility to a third-party public cloud, and that position rarely moves. It is driven by security classification, by partner and licensing agreements, and on some assets by national requirements about where operational data may be held.
Treat it as a first-conversation constraint. MultiSet runs in public cloud, private cloud, self-hosted on-premises and fully on-device, and the on-device path exists for exactly the case where nothing leaves the facility. Deployment options are set out on the pricing page.
Almost every operating facility has already been laser scanned, usually to a high standard, for brownfield engineering, as-built verification or a revamp study. It is typically treated as an engineering deliverable and then archived.
That scan can become the localization map directly. MultiSet accepts E57 point clouds from Leica, Faro, NavVis, Matterport and XGRIDS, meshes, metric-scaled Gaussian splats, 360 video and phone LiDAR. The input list and what each produces is on the 3D mapping page.
On a live plant this matters more than anywhere else. Sending a capture crew into a classified area is a permit, an escort, a toolbox talk and often a production impact. Using the scan that already exists removes all of it, and the facility can be localizing before the first site visit is scheduled.
The case for AR in oil and gas is not productivity in general. It is identification in particular: getting the right valve, the right isolation point, the right piece of scope, in an environment where hundreds of objects look the same and getting it wrong is an incident rather than an inconvenience.
That requires absolute positioning to a few centimetres, with no satellite signal, running on hardware the operator has already certified for the area, on a scan of the facility that in all likelihood already exists.
See how the VPS works, or send us one scan of a process unit and we will process it.
Related: visual positioning for oil and gas covers the same ground as a product page, including why partial GNSS denial under steel superstructures and pipe racks is more dangerous than total denial, and how turnaround routing and inspection evidence work off one plant frame.
