Underwater Robotics — AUV, ROV, USV — Robotics Reference

Scope. This is the robotics-applied counterpart to [[Engineering/fluid-mechanics]] (potential flow, Reynolds-number regimes, boundary-layer separation) and [[Engineering/electromagnetics-engineering]] (acoustic propagation, EM attenuation in conductive media). The full 6-DOF rigid-body-in-fluid derivation — Fossen’s $M\dot{\nu }+C(\nu )\nu +D(\nu )\nu +g(\eta )=\tau$ — lives in Fossen’s textbook (cited in §13); here we treat it as a tool. This note covers what to buy, what physics breaks the vehicle, what sensors tell it where it is when GPS is gone, and which named platforms (HUGIN, REMUS, Echo Voyager, Saildrone) define the 2026 production envelope.

1. At a glance

An underwater robot is a vehicle that operates partly or fully submerged, where the dominant physics — incompressible high-density fluid, exponential EM attenuation, hydrostatic pressure scaling linearly with depth, and acoustic propagation at ~1500 m/s — make every design choice diverge from terrestrial, aerial, or surface robotics. Four families dominate in 2026:

• AUV (Autonomous Underwater Vehicle) — battery-powered, untethered, pre-programmed or behavior-driven. Hours-to-days endurance. Examples: Kongsberg HUGIN (1000–6000 m class), HII Mission Technologies REMUS 100/600/6000/620, MBARI Tethys, Bluefin-21, IQUA Sparus II, L3Harris Iver4.
• ROV (Remotely Operated Vehicle) — tethered to a ship or platform via an umbilical carrying power + fibre-optic data. Effectively unlimited endurance, high bandwidth, human-in-the-loop control. Examples: Schilling UHD-III work-class, Forum Energy FCV-3000, Saab Seaeye Cougar, Oceaneering Millennium Plus, Blue Robotics BlueROV2 (small/research/hobby).
• USV (Unmanned Surface Vehicle) — operates on the air-water interface. Solar/wind/diesel hybrid power; satellite + line-of-sight RF comms. Examples: Saildrone Explorer/Voyager (wind+solar, multi-month missions), Liquid Robotics Wave Glider (wave-powered propulsion), ABB-OceanInfinity Armada 78 (crewless survey), L3Harris C-Worker, Sea Machines SM300-converted vessels.
• HUV (Hybrid Underwater Vehicle) — extra-large autonomous platforms that bridge AUV and submarine classes. Boeing Echo Voyager / Orca XLUUV (US Navy, ~26 m, 6500 nmi range), Anduril Dive-LD/XL, HII REMUS 620 (long-endurance medium-class).

Two specialised mobility classes round out the field:

• Underwater gliders — pure-buoyancy-driven; wings convert vertical buoyancy excursions into horizontal motion. Average ≤ 0.5 W shaft power, months-long missions, < 0.5 kt forward speed. Teledyne Slocum G3, Bluefin Spray, Kongsberg Seaglider.
• Bio-mimetic / soft — oscillating-tail (MIT SoFi, Festo BionicFish), undulating fin (FAU robotic fish, EvoLogics Manta), bladder-actuated jellyfish. Niche but growing in covert and bio-friendly surveys.

The 2026 wave in commercial underwater robotics: (a) Saildrone fleet exceeding 100 vehicles for NOAA/EU climate monitoring; (b) ABB-OceanInfinity Armada operating crewless seagoing survey ships; (c) Boeing/USN Orca XLUUV entering operational service; (d) Blue Robotics BlueROV2 as the de-facto research/hobby ROV; (e) Kongsberg HUGIN with synthetic-aperture sonar (SAS) replacing manned survey at full ocean depth.

First ask before specifying:

1. Depth class — 50 m littoral, 500 m continental shelf, 3000 m abyssal plain, 6000 m hadal? Each step doubles pressure-hull cost.
2. Endurance — hours (ROV), 24–72 h (medium AUV), weeks (Seaglider), months (Saildrone), years (next-gen autonomous platforms)?
3. Tethered or free? — Tether buys power + fibre bandwidth + zero localisation drift but kills mobility and range.
4. Payload — multibeam, sidescan, SAS, manipulator, sub-bottom profiler, CTD, ADCP, camera, environmental sampler?
5. Comms tolerance — can the mission accept hours of silence between surface fixes (AUV), or does the operator need live video (ROV)?
6. Recovery & launch — over-the-side from RHIB (small AUV), A-frame from research vessel, dry dock (XLUUV), or self-recovering surface return (Saildrone)?

2. First principles

The vehicle-level physics every underwater-robot designer uses. Detailed derivations in [[Engineering/fluid-mechanics]] and Fossen 2021.

Hydrodynamic drag. For a fully-submerged streamlined body at moderate Reynolds number,

F_drag = ½ · ρ · v² · C_D · A

with ρ the fluid density (saltwater ρ = 1025 kg/m³ at 4 °C, freshwater ρ = 1000 kg/m³), v the relative flow velocity (m/s), C_D the drag coefficient (~0.05 for a slender body of revolution at the design Reynolds, ~0.3 for a “blunt” payload-carrying torpedo, ~0.8–1.2 for a box-shaped ROV with thruster cages), A the reference area (frontal area for a torpedo; wetted area times C_F if using friction-coefficient form). Drag scales quadratically with speed — doubling speed quadruples drag and octuples power. This is why AUVs cruise at 1.5–2.5 m/s and gliders at 0.3 m/s.

Hydrostatic buoyancy.

F_buoy = ρ · V · g

where V is the displaced volume (m³). A neutrally-buoyant vehicle has F_buoy ≈ m · g; “neutral” is achieved at the working depth via ballast tanks, syntactic foam, lead trim weights, or active variable-buoyancy engines (oil bladder pumped between an internal reservoir and an external bladder). Compressibility of the hull and trapped air pockets makes buoyancy depth-dependent: a vehicle trimmed neutral at 100 m may be slightly negative at 1000 m, requiring active trim or accepting a depth-dependent thrust budget.

Added mass. A body accelerating through a fluid drags surrounding water along with it; the equations of motion gain a virtual-mass term

(m · I + M_A) · a = F_applied

where M_A ∈ ℝ³ˣ³ for translation is the added-mass matrix (also called virtual mass or hydrodynamic mass). For a sphere M_A = ½ · ρ · V · I; for a slender torpedo the axial added mass is ~5–10% of displaced mass and the lateral 50–100%. The vehicle feels significantly heavier laterally than axially, which is why AUVs are dynamically far more responsive in surge (forward) than in sway (sideways) or heave (vertical). Control design must account for this anisotropy.

Coriolis and centripetal forces in fluid. The full 6-DOF rigid-body-in-fluid equation (Fossen form):

M · ν̇ + C(ν) · ν + D(ν) · ν + g(η) = τ

with ν = [u, v, w, p, q, r]ᵀ the body-frame linear+angular velocities, M = M_RB + M_A the total mass matrix (rigid-body + added-mass), C(ν) the rigid-body + added-mass Coriolis-centripetal matrix, D(ν) the hydrodynamic damping (linear + quadratic), g(η) the restoring force from buoyancy-weight separation, and τ the actuator wrench. This is the canonical motion equation for AUV/ROV control synthesis.

Hydrostatic pressure.

P = ρ · g · h

In seawater, dP/dh ≈ 1 bar (100 kPa) per 10 m depth. At full-ocean 11 000 m (Mariana Trench): 110 MPa, an order of magnitude more than the internal pressure of a fire hose. Pressure-hull design becomes the dominant mass + cost driver beyond ~1500 m. Compressibility of seawater means a 1000 m-deep ocean column is ~0.4 % denser at the bottom than at the surface.

Optical visibility. Limited by both absorption (water absorbs red wavelengths first; chlorophyll-a absorbs blue-green) and scattering off particulates. Clear open ocean: 30–60 m vertical visibility; coastal turbid: 1–5 m; harbour silt: < 0.5 m. Practical consequence: vision is a near-field sensor underwater. Below 200 m, no sunlight reaches, and all imaging needs artificial illumination — which itself excites backscatter.

Sound propagation. Speed of sound in seawater:

c ≈ 1448 + 4.6·T − 0.055·T² + 1.34·(S − 35) + 0.016·z

with T in °C, S salinity in ppt, z depth in m (Wilson 1960; Mackenzie 1981 form). Typical c ≈ 1500 m/s; varies by ±50 m/s with temperature and depth. The depth-dependence creates the SOFAR (Sound Fixing And Ranging) channel at ~1000 m depth — a sound-speed minimum that traps long-range acoustic waves; submarines and whales exploit it for hundreds-of-km communication. Refraction off the SOFAR makes ranging at shallow grazing angles range-dependent.

EM attenuation. Seawater is a conductive medium (σ ≈ 4 S/m); skin depth δ = 1/√(π·f·μ·σ). At 1 MHz, δ ≈ 0.25 m; at GPS L1 (1.575 GHz), δ ≈ 0.005 m. Radio and GPS are unusable below ~10 cm submergence. This single physical fact dictates that underwater comms are acoustic, that localisation requires inertial dead-reckoning or acoustic positioning, and that surface vehicles or buoys are needed to relay data to satellites.

3. Worked examples

A — AUV drag and propulsion power

Sizing a torpedo-class AUV: length L = 2 m, max diameter D = 0.3 m, design speed v = 2 m/s. Assume seawater ρ = 1025 kg/m³, C_D = 0.3 (frontal-area basis, includes appendages), frontal area A = π · (0.15)² = 0.0707 m². Drag:

F_drag = ½ · 1025 · (2)² · 0.3 · 0.0707 = 43.5 N

Mechanical thrust power at vehicle:

P_mech = F · v = 43.5 · 2 = 87 W

Propeller efficiency at this operating point η_prop ≈ 0.65–0.75 (rim-driven or ducted axial); motor + ESC efficiency η_drive ≈ 0.85; battery DC-DC overhead ≈ 0.95. Shaft power demand:

P_shaft = 87 / 0.70 = 124 W

Battery draw: P_bat = 87 / (0.70 · 0.85 · 0.95) ≈ 154 W. A 1500 Wh Li-ion pack (e.g. 14S2P 18650, ~95 kWh/m³ packaged) supports 1500 / 154 ≈ 9.7 h propulsion. Hotel load (computer, sonar transmit, lights) adds 30–80 W, dropping endurance to ~7–8 h. Match against: HUGIN 1000 cruises at 4 kt (~2 m/s) for 24 h on a 15 kWh pack at the same hydrodynamic class but a lower C_D (0.07 frontal) — the difference is hull-form and appendage minimisation.

B — Pressure-vessel sizing at 6000 m

Target operating depth h = 6000 m; design pressure P_des = 1.1 · ρ · g · h ≈ 1.1 · 1025 · 9.81 · 6000 ≈ 66 MPa (round to 60 MPa nominal + 10 % proof margin). Pressure hull a cylinder, internal radius r = 0.15 m. Aluminium 6061-T6, σ_y = 276 MPa. Design factor of safety FoS = 2.

Thin-cylinder approximation (valid for t/r < 0.1):

σ_hoop = P · r / t ⇒ t = P · r / σ_allow = P · r · FoS / σ_y

t = (60 × 10⁶ · 0.15 · 2) / (276 × 10⁶) = 0.065 m = 65 mm

The thin-cylinder assumption is violated (t/r ≈ 0.43); use the thick-wall Lamé solution and check for buckling under external pressure (the dominant failure mode for deep-submergence hulls — Bryant 1954 / Windenburg-Trilling 1934 formulas). For a thinner shell of titanium Ti-6Al-4V (σ_y = 880 MPa, ρ = 4430 kg/m³ vs Al at 2700 kg/m³):

t_Ti = (60 × 10⁶ · 0.15 · 2) / (880 × 10⁶) = 0.020 m = 20 mm

Mass per metre length: Al → m_Al ≈ 2700 · π · (0.215² − 0.150²) ≈ 201 kg/m; Ti → m_Ti ≈ 4430 · π · (0.170² − 0.150²) ≈ 89 kg/m. Titanium is ~2.3× lighter for the same depth rating despite being 1.6× denser. This is why HUGIN 6000, Boeing Orca, and most full-ocean-depth platforms use Ti-6Al-4V or syntactic-foam-supported composite hulls. Carbon-epoxy composites (HII REMUS 6000 nose cone, Cellula Solus-XR) push depth ratings further at lower mass but with higher fabrication cost and impact-damage concerns.

C — DVL + INS dead-reckoning at depth

A medium-class AUV at 2000 m, no GNSS, no acoustic surface aiding. Onboard: tactical-grade ring-laser-gyro IMU (Honeywell HG1700: gyro bias stability 1°/h, accel bias 1 mg) and 600 kHz Doppler Velocity Log (Teledyne Workhorse Navigator: σ_v = 2 mm/s per axis, bottom-lock range 200 m). Bottom-lock is held throughout (depth-to-seabed < 200 m).

Velocity error from DVL is σ_v = 0.002 m/s per axis; over a 1-hour transit at 2 m/s, range = 7.2 km. Position error contribution from velocity:

σ_pos,DVL = σ_v · t = 0.002 · 3600 = 7.2 m

Heading error from gyro bias drifts as 1°/h = 0.0175 rad/h. Over 1 hour, cross-track contamination:

σ_pos,heading = (1/2) · v · t² · ω_bias (small-angle) ≈ ½ · 2 · 3600 · (0.0175/3600) ≈ 18 m for the worst-case linear-acceleration variant

A loosely-coupled EKF (DVL velocity as direct measurement, IMU as the propagation model) gives steady-state σ_pos ≈ 0.1–0.3 % of distance travelled with bottom-lock — so ~20–30 m at 7.2 km. Match against: Kongsberg HUGIN datasheet quotes 0.08 % of distance travelled with HiPAP USBL aiding and 0.4 % unaided; iXBlue PHINS INS quotes 0.06 % of distance travelled with DVL aiding. See [[Robotics/bayesian-estimation]] §3 for the EKF formulation.

When bottom-lock is lost (deep water > 200 m above seabed, or transit over soft mud), the filter falls back to water-mass tracking (DVL velocity relative to a layer of water, not ground): adequate for short transits but drifts with current. Without DVL of any kind, the filter degrades to pure INS dead-reckoning at 1–10 km/h drift for a tactical-grade IMU — survey-grade unusable beyond a few minutes.

4. Vehicle classes

Class	Depth	Endurance	Speed	Example platforms	Typical mission
Micro AUV	≤ 100 m	4–8 h	1–2 m/s	Iver3 (L3Harris), Sparus II (IQUA), Riptide UUV-1	Coastal survey, education
Lightweight AUV	100–600 m	8–24 h	1.5–2.5 m/s	REMUS 100 (HII), Bluefin-9	Mine countermeasures, MCM
Medium AUV	600–3000 m	24–72 h	2–3 m/s	REMUS 600, HUGIN 1000, Bluefin-12	Hydrography, military survey
Heavyweight AUV	3000–6000 m	60–100 h	2 m/s	HUGIN 6000, Bluefin-21, REMUS 6000	Oil-gas, telecom cables, abyssal
XLUUV	3000+ m	weeks–months	2.5–4 kt	Boeing Echo Voyager, Boeing/USN Orca, Anduril Dive-XL	Long-range ISR, payload delivery
Glider	200–1000 m	months	0.25–0.5 kt	Slocum G3, Spray, Seaglider, SeaExplorer	Climate, oceanography
Observation ROV	100–300 m	tether-limited	1–2 m/s	BlueROV2 (Blue Robotics), VideoRay Defender	Inspection, hobby, research
Inspection ROV	300–3000 m	tether-limited	1.5 m/s	Saab Seaeye Falcon, Tiger	Pipeline, hull, decom survey
Work-class ROV	3000–6000 m	tether-limited	1.5 m/s	Schilling UHD-III, Forum FCV-3000, Oceaneering Millennium Plus	Subsea construction, oil-gas
USV	surface	days–months	1–10 kt	Saildrone, Wave Glider, Armada 78, C-Worker 5/7	Climate monitoring, survey, escort
Hybrid HOV/HUV	6000+ m	hours-days	1 m/s	WHOI Nereus (lost 2014), Cellula Solus	Hadal exploration

5. Propulsion and actuation

The actuation choices that define an underwater platform.

• Brushless DC thrusters — the dominant choice. Open-impeller (Blue Robotics T200, fully-flooded magnetic-coupled), ducted (Tecnadyne, MOOG, Sub-Atlantic), rim-driven / hub-less (Voith VRD, Copenhagen Subsea). Rim-driven thrusters eliminate the central shaft and seal, giving silent operation, no entanglement risk, and higher reliability — adopted in modern USVs and emerging AUVs.
• Variable-buoyancy engine (VBE) — pumps oil between an internal compensator and an external bladder, changing displaced volume by ±100–500 cm³. The propulsion mechanism for all gliders; also used for fine depth-keep on hover-class AUVs (MBARI Tethys). Average shaft power < 1 W for a glider.
• Sail + wind harvest — Saildrone uses a 5–7 m hard wing-sail with a trim-tab elevator; vehicle steers by altering wing AoA, wing pulls vehicle through water at 3–8 kt. Solar panels recharge onboard batteries for hotel load. Multi-year missions demonstrated.
• Wave-driven propulsion — Liquid Robotics Wave Glider uses surface-float-and-submerged-fins linked by an umbilical; surface heave pumps the fins, which propel the system at 0.5–2 kt indefinitely. Solar tops up batteries for payload.
• Pump-jet (water-jet) — silent military propulsion (Boeing Orca, Anduril Dive-XL variants); no exposed propeller, lower acoustic signature.
• Bio-mimetic actuators — oscillating tail flukes (MIT RoboTuna, SoFi soft fish), undulating fins (FAU robotic fish), bladder-pumping jellyfish. Niche: silent, fish-friendly, low-Reynolds advantageous in coastal/reef environments.
• Manipulator arms (ROVs) — 5- to 7-DoF electro-hydraulic or all-electric arms; Schilling Titan 4 (7-DoF hydraulic, 250 kg lift), Kraft Predator (7-DoF), Schilling Conan (electric).

Linear motors for axial thrust exist as research niches (silent military, micro-AUV) but are not mainstream.

6. Sensing

The sensor stack that closes the underwater-localisation gap.

Sensor	Principle	Range / resolution	Example parts	Use
Multibeam echosounder	Fan of acoustic beams; bathymetry	50–500 m altitude; cm-class	Kongsberg EM2040, R2Sonic 2024, Norbit iWBMSe	Seafloor mapping
Sidescan sonar	Towed/AUV-mounted backscatter imager	50–500 m swath; ~0.1 m	EdgeTech 2200/4205, Klein 5000, Marine Sonic	Pipeline, shipwreck, MCM
Forward-looking imaging sonar	High-frequency 2D sonar “camera”	1–30 m; cm-class at short range	Sound Metrics ARIS / DIDSON, Sonardyne SOLSTICE, Oculus M1200d/M750d	Dark-water vision, dockside inspection
Synthetic-aperture sonar (SAS)	Coherent integration along track	50–300 m swath; 3–5 cm	Kraken AquaPix, HII REMUS SAS, KMS-SAS	High-resolution survey, MCM
Sub-bottom profiler	Low-frequency penetrating sonar	10–100 m below seabed	EdgeTech 3300, Innomar SES-2000	Geology, buried-cable survey
DVL (Doppler Velocity Log)	Acoustic Doppler vs seabed or water	Bottom-lock 100–500 m; σ_v ≈ 2 mm/s	Teledyne RDI Workhorse, Nortek DVL1000/500, Cerulean DVL	Velocity aiding for INS
ADCP	Doppler velocity vs water column	column up to 1 km	Teledyne RDI Sentinel V, Nortek Signature	Current profiling
INS (subsea-grade)	Fibre-optic or ring-laser gyro	0.01–0.1°/h drift	iXBlue PHINS, Kearfott T24, KVH 1750, Sercel SEAPATH	Dead-reckoning
INS (tactical/MEMS)	MEMS gyro + accel	1–10°/h drift	Honeywell HG1700/HG4930, Analog Devices ADIS-16500, Bosch BMI088	Small AUV, ROV
Pressure / depth	Strain-gauge or quartz	±0.01–0.05 % FS	Druck PTX, Paroscientific Digiquartz, Sensitec Mikrotest	Depth control
CTD	Conductivity, temperature, depth	Oceanographic-grade	Sea-Bird SBE 19plus, RBR Concerto, Idronaut OS320	Water-column survey
USBL (ultra-short baseline)	Ship-mounted transducer array; phase-difference	10 m to 7 km; 0.1–0.3 % slant range	Sonardyne Ranger 2, iXBlue GAPS, Kongsberg HiPAP	Surface-to-AUV localisation
LBL (long baseline)	Pre-deployed seabed transponder network	1–10 km; 0.1–1 m	Sonardyne Compatt 6, EvoLogics LBL	Survey + ROV
Magnetometer	Fluxgate or proton-precession	nT-class	Geometrics G-882, Marine Magnetics SeaSPY	UXO, geophysics
Camera	CMOS in pressure housing	< 10 m visibility	Cathx Aphrodite, SubC Imaging, GoPro housings	Inspection, AI perception
Underwater LiDAR	Blue-green laser scanner	10–60 m at high water clarity	2G Robotics ULS-500, Voyis Insight Pro	High-res 3D survey

See [[Robotics/sensors-perception]] for the camera/LiDAR fundamentals shared across robotics, and [[Robotics/sensors-pose-motion]] for INS / IMU specifics.

7. Communications

Channel	Medium	Bandwidth	Latency	Range	Example hardware
Acoustic modem	sound	0.1–25 kbps	seconds (range/c)	5–10 km	EvoLogics S2C, Sonardyne BlueComm 200/5000, Teledyne Benthos ATM-900
Optical (subsea)	blue-green laser	1–100 Mbps	µs	10–200 m	Sonardyne BlueComm, Hydromea LUMA, Aquatec AQUAmodem
Tether (copper/fibre)	wired	up to multi-Gbps	µs	tether length	Subconn, MacArtney, all major ROV manufacturers
Inductive coupling	near-field magnetic	kbps–Mbps	µs	< 1 m	Teledyne Inductive Modem; emerging
RF (surface only)	radio	Mbps	µs	line-of-sight	Standard 2.4/5/900 MHz; Microhard P900
Iridium SBD (surface)	satellite	~2.4 kbps	minutes (1.5 s msg + queue)	global	Iridium 9603 module
Iridium Certus	satellite	up to 700 kbps	sub-sec	global	Iridium 9770/9810

Acoustic-modem range × bandwidth is roughly a constant: 25 kbps × 1 km, 1 kbps × 10 km, deep-ocean SOFAR links < 100 bps × 100s of km. Latency is fundamental — at 5 km range, a single round-trip is ≈ 6.7 s. Implication for autonomy: AUVs cannot be teleoperated; they must execute mission scripts and report by exception when they surface or come within USBL range.

Optical comms in clean water can hit 100 Mbps at 50 m, enabling near-real-time video from a docked AUV to a base-station. Adoption is limited by alignment (the link is line-of-sight) and water clarity.

8. Mission planning and autonomy

The autonomy stack underwater is shaped by the comms reality: long stretches with no operator contact, soft real-time deadlines, and the consequence of any path that hits the seabed or a buried cable.

• Mission scripts — REMUS Mission Manager (HII), Kongsberg MUNIN/Mission Planner, Bluefin Mission Manager. Operators author a waypoint sequence with conditional behaviours (depth-keep, terrain-follow, lost-comms surface).
• Backseat-driver abstraction — the primary autopilot (“frontseat”) handles vehicle stabilisation and waypoint execution; a separate “backseat” computer running higher-level autonomy can request changes via a published API. Both MOOS-IvP (MIT, Newman & Benjamin) and DUNE-IMC (LSTS, FEUP Porto) implement this pattern, as does the more recent ROS 2 + MAUV integration.
• MBARI Tethys behaviour-based control — adaptive sampling: the vehicle re-plans transects to chase chlorophyll or temperature gradients, modelling the ocean as a stochastic field. Demonstrated multi-day in-situ phytoplankton studies.
• Adaptive sampling — NASA Ocean Worlds programme (Europa/Enceladus surrogates) studies plume-tracking and informative-path-planning under severe comms constraints. Maps to algorithms in [[Robotics/path-planning]].
• Cooperative AUV (CAUV) — multi-AUV survey with periodic acoustic rendezvous; intermittent comms drive consensus and data-mule patterns. ONR ASUW programmes, NATO CMRE GLINT/MANEX trials.
• Lost-comms safety — every fielded AUV has a “lost frontseat” → “ascend to surface, transmit Iridium” fallback; many also have a “drop weight” emergency surface.

See [[Robotics/path-planning]] and [[Robotics/bayesian-estimation]] for the algorithmic underpinnings.

9. Edge cases and failure modes

• Bottom-strike at sonar shadow. Forward-looking sonar has a near-field blind spot below the transducer; a vehicle terrain-following close to a steep slope can drive into the bottom before the down-sonar registers altitude. Mitigation: forward-looking imaging sonar + dedicated altimeter + minimum-altitude floor.
• Net and gear entanglement. Fishing nets, mooring lines, kelp forests. Mitigations: hot-knife / pyrotechnic cutters, snag-resistant fairings, “no-go” geofence around known fisheries. Acoustic releases that drop a weight if entanglement is detected (manometer + accelerometer agreement check).
• Pressure-hull leakage. A single O-ring or connector failure can flood a hull in seconds at depth. Mitigations: leak sensors (conductivity probes at low points), dual O-rings, redundant compartments, “abort-on-leak → blow ballast → surface” reflex.
• Battery thermal/pressure failure. Lithium chemistries are pressure-tolerant (no air gap to crush) but vent toxic gases on thermal runaway. Mitigation: pressure-tolerant Li-polymer or oil-compensated cells (HUGIN), thermal-fuse cell-level protection, hard separation between battery and avionics compartments.
• Marine fouling. Barnacles, algae, biofilm increase drag by 10–100 % and blind sensors. Mitigations: copper-based antifouling paint, silicone foul-release coatings, periodic dock-side cleaning, UV-LED biofouling (emerging).
• Lost-comms. Default safety mode: ascend at safe rate (avoid decompression-style bubble formation in elastomers), surface, GPS-fix, Iridium transmit, hold position until commanded.
• GPS recovery delay. Surfacing AUV needs 30–120 s for cold-start GPS fix and 5–60 s for Iridium queue clearance. Mission timeline must budget surface-time.
• Currents exceeding vehicle speed. Gliders cannot make progress against a > 0.5 kt current; planners must route along current eddies. Even AUVs at 2 m/s lose to Gulf Stream cores (2.5 m/s).
• Saltwater connector corrosion. Galvanic corrosion at dissimilar-metal joints, especially when grounded to seawater. Mitigation: sacrificial Zn or Al anodes, marine-grade alloys (316 SS, 6Al-4V Ti, AlBronze), pressure-balanced Subconn / MacArtney connectors.
• Inter-AUV acoustic interference. Two AUVs in the same survey area using the same DVL frequency confuse each other’s bottom-lock. Mitigation: frequency-multiplexed bands, TDMA scheduling, FH/SS DVLs.
• DVL bottom-lock loss. 600 kHz Workhorse: bottom-lock to ~100 m altitude; 300 kHz to ~200 m; 150 kHz to ~500 m. Beyond range, falls back to water-mass track (drifts with current).
• INS-only dead-reckoning drift. Tactical-grade IMU: 1–10 km/h position drift uncertainty; survey-grade FOG: 1–5 nm/24 h. Always pair INS with at least DVL or USBL underwater.
• Acoustic refraction errors. USBL slant-range assumes straight-line propagation; thermocline refraction causes 0.1–1 % range error. Mitigation: ray-tracing through measured sound-speed profile.

10. Tools and software

Hardware. SubseaTech ROV electronics, Insurv (Subsea Technologies) chips, Subsea-grade FOG IMUs (KVH 1750, iXBlue Octans/PHINS), DVL (Teledyne RDI Workhorse, Nortek DVL1000, Cerulean Sonar TrackLink), Subconn / MacArtney / SEACON pressure-balanced connectors, Blue Robotics electronics + thrusters (the standard for sub-USD-20k builds).

Open-source software stacks.

• MOOS-IvP (Newman, Benjamin; MIT) — autonomy middleware specifically designed for marine vehicles; behaviour-based interval-programming arbitration. Long-standing AUV community standard.
• DUNE / Neptus / IMC (LSTS, FEUP Porto) — full unmanned-vehicle stack: DUNE on-board, Neptus consoles, IMC (Inter-Module Communication) protocol. Active development through 2026; multiple AUV manufacturers ship with DUNE pre-installed.
• ROS 2 + MAUV — emerging convergence; ROS 2 Jazzy/Iron robotics stack with marine-specific packages (DAVE simulator, BlueRobotics ROS drivers, UUV Simulator successors).
• MATLAB Marine Systems Toolbox / Fossen MSS — academic Simulink models, Fossen-form rigid-body dynamics, classical and modern controllers.
• Stonefish (Patryk Cieslak) — GPU-accelerated underwater simulator; realistic hydrodynamics, sonar, camera, current.
• Gazebo + UUV Simulator / Project DAVE (DSCS/NPS) — Gazebo-based plugins for buoyancy, drag, thrusters, currents, sonar. The de-facto open simulation environment.
• MORSE — modular open-robots-simulation-engine; legacy but used in ROS-1 era.

Mission planning. Kongsberg MUNIN/SeaTrack, HII REMUS Mission Manager, Blue Robotics QGroundControl-marine fork, Greensea OPENSEA.

Datasets and bathymetry. GEBCO global bathymetry grid (15 arc-sec), NOAA NCEI / IFREMER Sismer survey archives, MBARI EARTH data portal, USGS coastal LiDAR-bathymetry merged products, EMODnet Bathymetry.

11. Case studies

OceanInfinity Armada 78 (commercial USV fleet, 2024+)

OceanInfinity / ABB operate 78 m crewless seagoing Armada survey ships — the largest commercial uncrewed vessels at sea in 2026. Hybrid diesel-electric, satellite + 4G/5G shore links, onboard launch-and-recovery for AUVs. Used for offshore-wind site survey, oil-gas decom inspection, hydrographic charting. Each Armada launches and recovers multi-AUV swarms autonomously over multi-week deployments. Demonstrates the crewless mothership + AUV swarm architecture that’s likely to dominate commercial subsea survey by 2030.

Saildrone Atlantic & Pacific monitoring (2024–2026)

Saildrone has demonstrated multi-month autonomous transatlantic and Pacific crossings with USVs carrying CTD, ADCP, biomass acoustics, and meteorological sensors. The 2024 Tropical Atmosphere Ocean array augmentation deployed > 20 USVs into hurricane formation zones, returning real-time data during named-storm passages — capability that was previously the domain of crewed research vessels at orders-of-magnitude higher cost. Wing-sail propulsion, solar payload power, Iridium Certus for data return. Multi-year endurance is the headline.

Kongsberg HUGIN 6000 with synthetic-aperture sonar (2010s–present)

HUGIN 6000 is the workhorse heavyweight AUV for full-ocean-depth survey: 6000 m depth, 70 h endurance at 4 kt, payloads up to ~200 kg. Integrates Kongsberg EM2040 multibeam, EdgeTech sidescan, and Kraken AquaPix or HISAS synthetic-aperture sonar. SAS achieves 3–5 cm along-track resolution at 200 m range — an order of magnitude better than conventional sidescan and the reason SAS-equipped AUVs displaced manned survey for telecom-cable route inspection, abyssal mining baseline studies, and military mine-countermeasures. Localisation by iXBlue PHINS INS + RDI Workhorse DVL + Sonardyne HiPAP USBL aiding from the surface vessel; 0.04–0.08 % distance-travelled accuracy.

12. Cross-references

• [[Robotics/multirotor-design]] — analogous 6-DOF under-actuated vehicle (air-based)
• [[Robotics/sensors-perception]] — camera + LiDAR fundamentals (shared substrate)
• [[Robotics/sensors-pose-motion]] — FOG / ring-laser INS, IMU mechanisation
• [[Robotics/bayesian-estimation]] — INS + DVL + USBL Kalman / EKF fusion
• [[Robotics/slam]] — subsea SLAM (sonar-based, acoustic landmarks)
• [[Robotics/path-planning]] — terrain-following, adaptive-sampling
• [[Robotics/pid-control]] and [[Robotics/state-space-lqr]] — depth + heading control
• [[Robotics/power-systems]] — pressure-tolerant Li chemistries
• [[Robotics/safety-standards]] — IMO MASS, USCG NAVSAC for USVs
• [[Engineering/fluid-mechanics]] — hydrodynamics, drag, added mass
• [[Engineering/materials-aluminum]] and [[Engineering/materials-composites]] — pressure-hull design
• [[Engineering/electromagnetics-engineering]] — EM attenuation, acoustic propagation physics

13. Citations

• Fossen, T. I. Handbook of Marine Craft Hydrodynamics and Motion Control, 2nd ed., Wiley, 2021. The canonical reference for 6-DOF marine-vehicle dynamics and control.
• Antonelli, G. Underwater Robots, 4th ed., Springer Tracts in Advanced Robotics, 2018.
• Christ, R. D. & Wernli, R. L. The ROV Manual: A User Guide for Remotely Operated Vehicles, 2nd ed., Butterworth-Heinemann, 2013.
• Stutters, L., Liu, H., Tiltman, C., Brown, D. J. “Navigation Technologies for Autonomous Underwater Vehicles.” IEEE Transactions on SMC-C, 2008.
• Paull, L., Saeedi, S., Seto, M., Li, H. “AUV Navigation and Localization: A Review.” IEEE Journal of Oceanic Engineering, 2014.
• Whitcomb, L. L., Yoerger, D. R., Singh, H., Howland, J. “Advances in Underwater Robot Vehicles for Deep Ocean Exploration: Navigation, Control, and Survey Operations.” Robotics Research, Springer, 2000.
• Bellingham, J. G., Zhang, Y., Kerwin, J. E., et al. “Efficient Propulsion for the Tethys Long-Range Autonomous Underwater Vehicle.” IEEE AUV 2010.
• Newman, P. M. & Benjamin, M. R. “The Strategy of MOOS.” MOOS-IvP technical reports, MIT CSAIL, 2010+.
• Pinto, J., Calado, P., Braga, J., et al. “Implementation of a Control Architecture for Networked Vehicle Systems.” IFAC NGCUV 2012 (DUNE / Neptus / IMC, LSTS Porto).
• Sonardyne technical notes on USBL/LBL acoustic positioning, 2020–2024 editions.
• Kongsberg Maritime HUGIN AUV specifications and operational reports, 2022–2025 editions.
• Saildrone Inc. whitepapers on autonomous Atlantic / Pacific surveys, 2023–2025.
• MBARI Tethys long-range AUV publications (Bellingham et al., 2010+).
• Mackenzie, K. V. “Nine-term equation for sound speed in the oceans.” J. Acoust. Soc. Am., 1981.
• Bryant, A. R. “Hydrostatic pressure buckling of a ring-stiffened tube.” NCRE Report R-306, 1954.
• IMO Maritime Autonomous Surface Ships (MASS) Code, IMO MSC.428(98) and successors, 2018–2026.