ROV Locator (ROVL) USBL Acoustic Positioning

Resolution is not accuracy, and Cerulean say so themselves directly beneath their own specification table. 0.1° is how finely the unit reports a bearing, not how close that bearing is to correct.

The real figure is an error budget, and Cerulean publish it in the reference documentation rather than on the product page:

• Topside GNSS position, without RTK: 3–5 m CEP
• Magnetic compass error: typically 3–5°, depending on the magnetic environment and how well the IMU was calibrated
• Acoustic bearing measurement: typically 2°
• Combined angular error: about 5.2° in the published formula
• Slant range error: small enough to ignore next to the above

Which gives, in Cerulean's own notation:

CEP ≈ √( 5² + ( sin(5.2°) × slant range in metres )² )

Worked through, that is:

Computed by SepcoTech from Cerulean's published formula. Cerulean publish the formula; the table is our arithmetic.

Two things follow. The closer the vehicle is to the topside unit the better it gets - and the floor is set by topside GNSS, so without RTK you cannot do better than about 5 m however close in you are.

Cerulean's own field test agrees with the formula. At roughly 110 m slant range the position scatter covered a band about 10 m long, with around 90% of the error coming from bearing measurement and IMU heading, and slant range error of about 0.5 m.

The ±2% figure you will also find

Cerulean's marketing FAQ offers a rule of thumb - "for a distance of 100 meters, position accuracy should be planned on +/- 2% of distance" - which works out at about 2 m. Their own reference documentation and their own field test both say roughly 10 m at that range. We publish the pessimistic pair, because they agree with each other and with the measurement, and because a system that beats its quoted figure is a better surprise than one that misses it.

Sources: the error budget, the 110 m field test and Cerulean's FAQ.

If 10 m at 100 m is not close enough, the honest answer is Omnitrack rather than a different ROVL. It runs the same acoustics behind a dual-GNSS compass, which drops the combined angular error from about 5.2° to about 2.1° - on the same formula, roughly 6 m at 100 m instead of 10.

Cerulean scope this product honestly in their own overview: low-cost, medium-accuracy localisation of underwater objects, explicitly not a survey-grade instrument. Three situations are worth knowing about before you buy, and all three come from the manufacturer's own documentation.

Pools and tanks

Cerulean state plainly that pools are acoustically hostile - highly reflective, specular, with many reflecting surfaces close to the device. Expect erratic connections and large bearing, elevation and slant range errors. Most importantly, performance in a pool does not predict performance in open water. If you are planning an acceptance test, do it somewhere representative of the job; a tank test that goes badly may tell you nothing, and one that goes well tells you less.

Multipath, everywhere else

The same physics applies outside a pool. Reflections off the ROV body, the surface, the bottom, structures, the launch vessel and thermoclines all degrade accuracy and can produce ghost fixes - spurious solutions scattered among the good ones. Some can be filtered. Aquatic vegetation between the two units causes severe attenuation and phase disturbance. Ideal conditions are deep, open water where every reflected path is longer than the direct path by at least one ping length, which Cerulean acknowledge is effectively impossible and frame as a target to approach rather than reach.

Shallow water is the weak case for USBL as a category

This one is not Cerulean's. Water Linked, who sell a short-baseline alternative, state in their own technical documentation that USBL can struggle in less than about 10 m of water, and position their system as the better choice in shallow, reflective and confined environments while conceding that USBL reaches further in open water. That is a competitor's framing, and we quote it because it corroborates Cerulean's pool caveat from the opposite direction. If most of your work is in a harbour basin, alongside a hull or in a fish pen rather than open water, weigh it before you choose.

Published by Cerulean: operation in a pool, multipath, product overview. Shallow-water limitation as stated by Water Linked.

An acoustic pair is not a positioning system on its own. These are the parts that go around it.

• A topside GPS. The ROVL gives you bearing and range relative to the topside unit. Turning that into a latitude and longitude needs a position for the topside unit itself. Any third-party serial GPS on the topside computer will do, and Cerulean list USB receivers they have used. We supply a GPS Module for it.
• Something to hold the topside unit in the water. It has to be in the water, below the hull, and away from the vessel's own noise and reflections. That is what the ROVL Topside Float is for.
• For Mk V only, a network at both ends. A spare Ethernet switch port inside the ROV and a hub at the topside - the Network Hub topside, or the Manifold 5L as a subsea hub. A wireless topside is also an option, with the Wi-Fi Upgrade Kit.
• A supply that can take the surge - see the note above on 5 V rails and the Mk V's ping current.

SonarView is free with every Cerulean product: no licence, no subscription, and it runs on Windows, macOS, Linux and in-browser under BlueOS. On MAVLink and ArduSub vehicles the position fusion happens in the vehicle itself, so the ROVL works without SonarView and the position appears on the QGroundControl map. The open API is NMEA-style: $USRTH carries receiver-to-transmitter angles and slant range, which is what a baseline ROVL sends.

We supply the floats, GPS modules, hubs and cables alongside the unit, or they can go on the same enquiry.