1. The short version / what to know first / where this fits in the series

• Underwater, communication and position are the hardest problems. GPS and radio signals fade quickly in water, so acoustics is used, but sound travels at only about 1,500 m/s and the bandwidth is narrow Sourced[Ref. 7]. In our calculation, sound takes about 6.7 seconds to reach something 10 km away Our calculation.
• In the deep sea, pressure sets the limits of materials. At the deepest point (about 10,935 m) our calculation gives about 110 MPa, or about 1,100 atmospheres Our calculation. Nereus, which went beyond 10,000 m, used light ceramic spheres for flotation and a hair-thin optical-fiber tether Sourced[Ref. 9]. A soft robot that embedded its electronics in silicone, with no rigid vessel, has also worked in the deep sea Sourced[Ref. 10].
• The latest research is still at the tank, pool or lake stage. The MIT and EPFL robot that leaps out of the water weighs under 300 g and was tested in a tank and a lake Sourced[Ref. 14]. Purdue University’s three-mode robot (drifting, gliding, thrusting) is at the pool-test stage Sourced[Ref. 15, 16].

2. Three types: ROVs, AUVs and gliders

Underwater robots fall into three main types. The points that tell them apart are whether they are tied to a ship by a cable and how much a human is involved.

• ROV (remotely operated vehicle): an uncrewed underwater robot tethered to a ship. The cable carries the pilot’s commands, and the vehicle sends back live video and other data Sourced[Ref. 2]. NOAA says the cable carries power and signals (its example is an ROV connected by an eight-kilometre steel cable) Sourced[Ref. 2]. Vehicles range from laptop-sized to the size of a small truck.
• AUV (autonomous underwater vehicle): a robot cast off from the ship that carries out a pre-planned mission by itself, with no cable. When the mission ends it returns to a set location, and the data are collected afterward Sourced[Ref. 1, 3]. NOAA says AUVs can be smaller and lighter than ROVs and can be run from smaller ships Sourced[Ref. 3].
• Glider: a kind of AUV that moves not with a propeller but by changing its buoyancy. It adjusts buoyancy to dive and rise, and uses wings to turn that motion into horizontal progress Sourced[Ref. 5].

Original SVG, detailed explanation and sources

Original SVG reference: Concept diagram (vector drawing). Differences among ROVs, AUVs and gliders. The classification follows NOAA descriptions and the glider principle follows Rudnick et al. and APL-UW; this is a concept summary.

An ROV is close to “a drone a person flies”, an AUV to “a drone that flies itself”, and a glider is like a paper plane that slowly glides down while moving forward (limits of the analogy: a glider can change its buoyancy to rise again after diving).

The glider’s long journey

Gliders use little power, so they can keep going for a very long time. The University of Washington Applied Physics Laboratory (APL-UW) describes its Seaglider as using buoyancy for thrust, working to a maximum depth of about 1,000 m, with records of a mission over 3,800 km and one lasting more than seven months Sourced[Ref. 6]. NOAA’s IOOS lists uses of gliders such as hurricane intensity forecasting, ecosystem monitoring, fisheries management and ocean acidification monitoring Sourced[Ref. 4].

A hybrid: Nereus, both ROV and AUV

WHOI’s Nereus was designed as a “hybrid” that can work as an ROV or as an AUV. In ROV mode it is connected to the ship by a hair-thin optical-fiber tether; in AUV mode it is cut loose and moves autonomously Sourced[Ref. 9]. The type of a vehicle is therefore not fixed; it can change with how it is operated (explanation).

3. Communication underwater: radio does not reach, sound is slow

Ground robots talk to distant partners by Wi-Fi or mobile radio. Underwater, this is almost impossible.

Checking with numbers: how far does radio reach?

Seawater contains salt, so it conducts electricity. In a conducting medium an electromagnetic wave fades quickly. The rate can be expressed as the depth at which the wave falls to about 37% (1/e). Assuming a seawater conductivity of 4 S/m, the distance gets shorter as frequency rises: about 8.0 m at 1 kHz, about 0.80 m at 100 kHz, about 0.25 m at 1 MHz and about 0.08 m (8 cm) at 10 MHz Our calculation. Wi-Fi frequencies (GHz) are shorter still (not calculated here).

What about sound? Sound travels well in water, but it is slow. At 1,500 m/s, reaching something 10 km away takes 10,000 / 1,500 = about 6.7 seconds Our calculation. If a radio signal could cover the same distance at the speed of light, it would take about 33 microseconds (0.000033 seconds) Our calculation. This “delay of about 200,000 times” makes real-time piloting underwater difficult.

A common language for underwater links is still growing

For underwater acoustic communication, modems (devices that communicate by sound) from different makers use different methods and cannot talk to each other. JANUS, developed by the NATO research centre CMRE, is a “common base” for the coding of underwater acoustic communication. A trade publication reported in May 2017 that it was recognized as a NATO standard (a STANAG) Sourced[Ref. 19] (the NATO text itself was not checked here).

4. Position underwater: estimating without GPS

GPS signals do not reach underwater either. An AUV has to estimate where it is by other means.

1. Dead reckoning: calculating position by adding up heading and speed. Speed is measured with equipment such as a Doppler velocity log (DVL), which uses sound to measure motion. It is like counting steps to know how far you walked (limits of the analogy: real errors grow in more complicated ways than errors in stride length), and the error accumulates and keeps growing as the vehicle goes farther Sourced[Ref. 7].
2. Acoustic references: USBL measures range and bearing with a device on the ship, and LBL measures from devices placed on the seabed. Both have limited range, and the devices must be installed beforehand at cost and effort Sourced[Ref. 7].
3. Surface for GPS: gliders and others rise to the surface and take a new GPS fix with their antenna Sourced[Ref. 6].

The review by Paull et al. says the error of inertial navigation (accelerometers and gyroscopes) is about 0.1% of the distance travelled at best and 2-5% for modestly priced units Sourced[Ref. 7]. Let us calculate for 10 km. At 0.1% it is 10 m, at 2% it is 200 m, and at 5% it is 500 m Our calculation.

5. Pressure and materials: surviving 1,000 atmospheres

The weight of water presses on a machine harder as it goes deeper. At depth h the water pressure is P = rho x g x h, where rho is the density of seawater (assumed about 1,025 kg/m3) and g is gravitational acceleration (9.81 m/s2).

According to NOAA, the average depth of the ocean is 3,682 m and the deepest point, Challenger Deep, is about 10,935 m Sourced[Ref. 8]. Our calculation gives about 10 MPa at 1,000 m, about 37 MPa at 3,682 m, about 60 MPa at 6,000 m and about 110 MPa at 10,935 m Our calculation. 110 MPa is about 1,100 atmospheres, which works out to about 1.1 tonnes-force on a single square centimetre, about the area of a fingernail Our calculation. WHOI says Nereus withstood pressure about 1,000 times that at the surface Sourced[Ref. 9].

Three design approaches to pressure

There are three broad ways of thinking about pressure (the figure organizes concepts; it does not rank them).

• 1. Resist: a thick metal or ceramic rigid shell keeps the interior at 1 atmosphere. Nereus used ceramic hollow spheres for flotation (about 9 cm in diameter, about 800 in each hull) and is described as lighter than the traditional flotation, the syntactic foam used on vehicles such as Alvin and Jason Sourced[Ref. 9].
• 2. Balance: “pressure compensation”, in which the interior is filled with oil or similar so that it matches outside pressure. The Nature soft-robot paper lists rigid vessels and pressure compensation as the conventional ways of protecting deep-sea machinery Sourced[Ref. 10].
• 3. Embed softly: the robot in the Nature paper embedded its electronics in silicone, used no rigid shell, and moved with dielectric-elastomer fins. To reduce stress between parts it decentralized the electronics (increasing the spacing between components and separating them from the circuit board) Sourced[Ref. 10].

Original SVG, detailed explanation and sources

Original SVG reference: Concept diagram (vector drawing). Three design approaches to pressure. Examples come from WHOI’s Nereus announcement, APL-UW’s Seaglider page and the Nature soft-robot paper. It does not rank the approaches.

6. Corrosion: metals become a “battery” in seawater

Seawater is a liquid that conducts electricity (an electrolyte). When two different metals touch in it, a small “battery” forms. A NOAA document calls this galvanic corrosion and explains that the less noble metal (the anode) gives up electrons and dissolves, while the more noble metal (the cathode) is protected Sourced[Ref. 12]. Think of the zinc in a dry cell dissolving to produce electricity (limits of the analogy: in a robot nothing useful is drawn out; it shows up as corrosion damage).

Original SVG, detailed explanation and sources

Original SVG reference: Concept diagram (vector drawing). Concept diagram of galvanic corrosion and sacrificial anodes, following a NOAA document. Real corrosion depends on the material pair, area and water chemistry.

The NOAA document’s example is a boat’s seawater piping in which an aluminum hose barb was fitted to a bronze sea strainer and the aluminum corroded. The same document says that zinc “sacrificial anodes” are sometimes fitted to protect aluminum hulls and other parts. Zinc is less noble than aluminum, so it dissolves first and protects the aluminum Sourced[Ref. 12].

7. Research examples: swim and fly, three modes, and soft endurance

Here is how far recent research has come on the problems above (communication, position, pressure, corrosion). All of it is at the research stage.

(1) From water to air: the MIT and EPFL flapping robot (Science, July 2026)

MIT and EPFL developed a “flapping-wing aerial-aquatic vehicle” (FAAV) that swims, then leaps from the water and flies, and published it in Science on 9 July 2026 Sourced[Ref. 13, 14]. It weighs under 300 g and consists of a fuselage, two flexible wings and a steerable tail. The body contains a battery and a waterproof electric motor, and a crankshaft moves the wings up and down at preset frequencies Sourced[Ref. 14].

Tests were done in a water tank and in Lake Geneva (Switzerland). With the robot placed about 0.5 m underwater, the researchers varied flapping frequency and tail angle to find the conditions in which it breaks through the surface. They made three wing sets: small (60 cm wide), medium (80 cm) and large (100 cm), and the medium wings let the robot swim, fly and transition reliably. A wing needs both flexibility, to limit flapping amplitude in water, and stiffness, to hold the robot aloft in air. At about 5 Hz it moved at about 1 m/s in water and about 6 m/s in air, and it left the water at a 70-degree angle. It did not need the “paddling with the feet” that birds use Sourced[Ref. 14]. The paper’s abstract says frequency adaptation, flexible wings and powerful actuation enable transitions without folding wings or legs Sourced[Ref. 13].

Limits: the flapping frequency and tail angle were set in advance; the team says it will next work on wings that can turn and will test in turbulence and wind Sourced[Ref. 14]. Sampling in the open sea was described only as the researchers’ “dream vision” (a plan) Not yet confirmed.

(2) Drifting, gliding and thrusting: Purdue University’s three-mode robot (ASME Letters, September 2026)

Athar et al. of Purdue University report a compact underwater robot that integrates drifting (drifter), gliding (glider) and thrusting (thruster) in one vehicle. It has a buoyancy control module, pitch regulation by a moving mass, foldable wings and a rear propulsion and steering assembly. The paper reports that underwater tests showed reliable buoyancy control, stable gliding, thruster-driven maneuvering and transitions among the three modes Sourced[Ref. 15].

A press report says the current prototype switches modes manually, that autonomous switching is planned for later versions, and that pool tests succeeded Sourced[Ref. 16]. Quantitative performance of battery endurance, range and maneuverability is left as future work Not yet confirmed.

(3) Dropping the shell: a soft robot in the deep sea (Nature, 2021)

Li et al., inspired by the structure of a deep-sea snailfish, developed a soft robot with electronics embedded in silicone. It uses no rigid shell and carries its battery, control and actuation onboard, moving with dielectric-elastomer fins. Its fins actuated at 10,900 m in the Mariana Trench and it swam freely at 3,224 m in the South China Sea, the authors report. To reduce shear stress between components, the electronics were decentralized Sourced[Ref. 10]. These were small-scale, short tests; they do not show long-term operation or scaling to large vehicles (explanation). The basics of soft actuation are covered in our explainer on soft actuators and artificial muscles.

8. Where things stand (checked 25 October 2026)

This field changes quickly, so here is what could be confirmed from primary sources, with check dates. University and institution announcements are written as their claims.

Checked 25 October 2026. University and institution plans and targets are separate from achieved facts.

9. How to read the yardsticks and standards

The quality of an underwater robot cannot be measured by one number. There are several yardsticks: “how deep can it go (depth rating)”, “how long can it operate (endurance, days of operation)”, “how far off is its position (navigation accuracy, in % of distance)”, “how well does communication work” and “how good is the data”. Their meaning changes with the waters, season, vehicle and conditions in which they were measured. The table shows what each measures and does not measure, as far as could be confirmed.

The arrangement of the table is Our calculation (organization). The text of the ABS rules was not read, so only titles and structure are given. No ASTM International standard or website was consulted.

Cautions when reading (for materials engineers)

• Demonstrations, benchmarks and operations are different. Tank and pool results do not show performance under waves, currents or long continuous operation. Check “how many times, for how long, in what environment”.
• Institution announcements and third-party verification are different. A university or laboratory announcement describes its own results. No independent replication could be found among the sources checked.
• Check whether a standard is a recommendation or a requirement. A classification society’s rules are required to obtain that class, and a communication-coding standard is an agreement to connect. Their legal force differs (inference).
• Plans and achievements differ. Autonomous mode switching, open-sea operation and sampling are all not yet confirmed.

10. Glossary

11. References

1. NOAA National Ocean Service “What is the difference between an AUV and a ROV?” (checked 25 October 2026) — oceanservice.noaa.gov/facts/auv-rov.html
2. NOAA Ocean Exploration “Remotely Operated Vehicles (ROVs)” (checked 25 October 2026) — oceanexplorer.noaa.gov/technology/subs-rovs/
3. NOAA Ocean Exploration “Autonomous Underwater Vehicles” (published 22 June 2020, updated 14 October 2026; checked 25 October 2026) — oceanexplorer.noaa.gov/technology/subs-auvs/
4. U.S. IOOS (NOAA) “Underwater Gliders” (uses of gliders and a list of glider quality-control documents; checked 25 October 2026) — ioos.noaa.gov/project/underwater-gliders/
5. Rudnick D. L. et al. “Underwater Gliders for Ocean Research”, Marine Technology Society Journal 38(2), 73-84 (2004) — doi.org/10.4031/002533204787522703
6. Applied Physics Laboratory, University of Washington “Seaglider” (project page: buoyancy propulsion, depth to about 1,000 m, missions over 3,800 km and 7 months; checked 25 October 2026) — apl.uw.edu/project/project.php?id=seaglider
7. Paull L. et al. “AUV Navigation and Localization: A Review”, IEEE Journal of Oceanic Engineering 39(1), 131-149 (2014) — doi.org/10.1109/JOE.2013.2278891
8. NOAA National Ocean Service “How deep is the ocean?” (average 3,682 m; Challenger Deep about 10,935 m; updated 23 September 2026; checked 25 October 2026) — oceanservice.noaa.gov/facts/oceandepth.html
9. Woods Hole Oceanographic Institution “Hybrid Remotely Operated Vehicle Nereus Reaches Deepest Part of the Ocean” (2 June 2009; the institution’s announcement) — whoi.edu/press-room/news-release/hybrid-remotely-operated-vehicle-nereus-reaches-deepest-p
10. Li G. et al. “Self-powered soft robot in the Mariana Trench”, Nature 591, 66-71 (2021) — doi.org/10.1038/s41586-020-03153-z
11. Song X. et al. “Effect of Rise and Fall Time on Dwell Fatigue Behavior of a High Strength Titanium Alloy”, Metals 9(8), 914 (2019) — doi.org/10.3390/met9080914
12. NOAA Office of Marine and Aviation Operations “Classic Galvanic Corrosion” (explanatory document using a NOAA boat as the example; checked 25 October 2026) — omao.noaa.gov/sites/default/files/documents/ClassicGalvanicCorrosion.pdf
13. Zufferey R. et al. “Leaping out of the water: Aerial-aquatic locomotion with flapping wings”, Science 393(6807), 207-211 (9 July 2026) — doi.org/10.1126/science.aeb6744
14. MIT News “New flapping robot swims and flies like a diving bird” (9 July 2026; the institution’s announcement: wing sizes, speeds, 70 degrees, hydrophobic coating, test method; checked 25 October 2026) — news.mit.edu/2026/new-flapping-robot-swims-and-flies-like-diving-bird-0709
15. Athar S. et al. “Design and Development of a Multimodal Underwater Robot: An All-in-One Drifter, Glider, and Thruster”, ASME Letters in Translational Robotics 2(3), 031002 (1 September 2026) — doi.org/10.1115/1.4072705
16. Tech Briefs “Purdue Engineers Create Multi-Mode Underwater Robot for Smarter Ocean Missions” (10 September 2026; an article relaying the university’s announcement: mode switching is manual, pool tests; checked 25 October 2026) — techbriefs.com/component/content/article/55847-purdue-engineers-create-multi-mode-underwat
17. ABS Rules for Building and Classing Underwater Vehicles, Systems and Hyperbaric Facilities, Notices and General Information (January 2026 edition; checked 25 October 2026) — ww2.eagle.org/content/dam/eagle/rules-and-resources/RuleManager2/notices/january-2026/7-uw
18. ABS Rules for Building and Classing Underwater Vehicles, Systems and Hyperbaric Facilities (January 2021 edition; table of contents with Section 18 ROVs and Section 19 AUVs checked; requirements not read) — ww2.eagle.org/content/dam/eagle/rules-and-guides/archives/special_service/7_rulesforbuildi
19. Ocean News & Technology “JANUS Underwater Communication Protocol Becomes a NATO Standard” (12 May 2017; a trade-press article; the NATO text itself was not checked; checked 25 October 2026) — oceannews.com/news/subsea-cable/janus-underwater-communication-protocol-becomes-a-nato-sta

12. Claim-to-source audit

Last updated: 25 October 2026. Sources are limited to primary material (NOAA, WHOI, APL-UW, MIT and ABS documents, and papers in Science, Nature, IEEE JOE, ASME Letters and Metals). Institution and university announcements are labeled as announcements and kept separate from independent verification. Trade-press articles (Tech Briefs, Ocean News & Technology) are treated as reports of a university announcement and of a NATO standard, and the originals were not checked. The worked examples (electromagnetic penetration depth, sound delay, dead-reckoning error, water pressure) are our own explanatory calculations and not specifications of any vehicle. Military equipment is not covered. Pressure-hull materials by model, service lives in real operation, open-sea operation, the cause of Nereus’s loss and the applicable international standard numbers could not be confirmed and are therefore not given. All figures are explanatory concept diagrams or drawings with our own calculations, except Figs. 1 and 3, which are ImageGen-generated conceptual illustrations. They do not depict real vehicles, ships or appearances.