PLATFORMS: AUTONOMOUS UNDERWATER

VEHICLES
J. G. Bellingham, Monterey Bay Aquarium Research number of companies develop subsystems and sen-
Institute, Moss Landing, CA, USA sors for AUVs.
& 2009 Elsevier Ltd. All rights reserved. The most common class of AUV in use today is a
torpedo-like vehicle with a propeller at its stern, and
steerable control surfaces to control turns and ver-
tical motion (see Figure 1). These vehicles are used
when speed or efficiency of motion is an important
consideration. Such torpedo-like vehicles range in
Introduction weight from a few tens of kilograms to thousands of
Autonomous underwater vehicles (AUVs) are un- kilograms. Most typical are vehicles weighing a few
tethered mobile platforms used for survey operations hundred kilograms, with an endurance of about a
by ocean scientists, marine industry, and the military. day at a speed of c. 1.5 m s1. Often they have par-
AUVs are computer-controlled, and may have little allel mid-bodies, which allow the vehicle length to be
or no interaction with a human operator while extended without large hydrodynamic consequences.
carrying out a mission. Being untethered, they must This is useful when it is necessary to add new
also store energy onboard, typically relying on bat- sensors or batteries to a vehicle. A disadvantage of
teries. Motivations for using AUVs include such torpedo-like vehicles is that, like an aircraft, they
factors as ability to access otherwise-inaccessible must maintain forward motion to generate lift over
regions, lower cost of operations, improved data its control surfaces, and thus are not controllable at
quality, and the ability to acquire nearly synoptic very low speed through the water.
observations of processes in the water column. An Gliders are a class of vehicles that use changes in
example of the first is operations under Arctic and buoyancy rather than a propeller for propulsion.
Antarctic ice, an environment in which operations of Gliders use their ability to control buoyancy to gen-
human-occupied vehicles and tethered platforms are erate vertical motion. Vertical motion is translated
either difficult or impossible. Illustrating the next into horizontal motion with lifting surfaces, usually
two points, AUVs are becoming the platform of wings mounted in about the middle of the vehicle (see
choice for deep-water bathymetric surveying in the Figure 1). Several types of gliders weighing about
offshore oil industry because they are less expensive 50 kg are in use today. These comparatively small
than towed platforms as well as produce higher- vehicles are designed to move slowly, about
quality data (because they are decoupled from mo- 0.25 m s1, and operate sensors consuming a watt or
tion of the sea surface). Finally, the use of fleets of less. By minimizing power consumption, these gliders
AUVs enables the rapid acquisition of distributed can operate for periods of months using high-energy-
data sets over regions as large as 10 000 km2. density primary batteries. Disadvantages of this class
AUVs are a new class of platform for the ocean of system are that they are limited to vertical profiling
sciences, and consequently are evolving rapidly. The flight tracks, and can be overwhelmed by ocean cur-
Self Propelled Underwater Research Vehicle (SPURV) rents, especially in the coastal environment or within
AUV, built at the University of Washington Applied boundary currents. However, larger gliders in devel-
Physics Laboratory, was first operated in 1967. opment and testing will operate at higher speeds, and
However, adoption by the ocean sciences community thus not suffer from this limitation.
lagged until the late 1990s. Adoption was spurred on A final class of AUVs uses multiple thrusters to
by two developments: AUV development teams provide capabilities similar to that of a helicopter or a
started supporting science field programs with AUV ship with dynamic positioning (see Figure 1). The
capabilities, and AUVs that nondevelopers could additional thrusters enable maneuvers such as hover-
purchase and operate became available. The first ing, translating sideways, and moving vertically. These
served the purpose of building a user base and vehicles are used when maneuverability is needed, for
demonstrating AUV capabilities. The second enabled example, when operation near a very rough bottom is
scientists to obtain and operate their own vehicles. a necessity. The disadvantage is that the additional
Today, a wide variety of AUVs are available thrusters reduce efficiency for moving large distances
from commercial manufacturers. An even larger or at high speeds.

473
474 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES

Figure 1 From top-left corner clockwise: the Hugin AUV, the Spray glider (courtesy, MBARI), the ABE AUV (courtesy, Dana
Yoerger, WHOI), and the Dorado AUV (courtesy, MBARI). Hugin and Dorado are examples of propeller-driven vehicles optimized for
moving through the water efficiently, and have a torpedo-like configuration. Spray is an example of a glider, which is a buoyancy-driven
vehicle that has no propeller. ABE is a highly maneuverable vehicle capable of hovering or pivoting in place, and of moving straight up
or down. Also illustrated are different handling strategies. The large Hugin vehicle is launched and recovered from a ship using a stern
ramp. Spray is hand-launched and recovered. ABE is launched and recovered with a crane. Dorado is shown being launched and
recovered with a capture mechanism suspended from a J-frame.

While the vehicles described above are represen- propulsion have also been tested. A hybrid vehicle is
tative of the most commonly used systems, a wide being developed for reaching the deepest portion of
range of other vehicles are in development or are in the ocean. The hybrid vehicle operates as a tethered
limited use. AUVs that come to the surface and use platform via a disposable fiber optic link for tasks
solar panels to recharge batteries have been demon- requiring human perception, in other words as a
strated in seagoing operations. Gliders which extract remotely operated vehicle (ROV), but operates as an
energy from thermal differences in the ocean for AUV when that link is severed. These are just a few
 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES 475

of the diverse AUVs being developed to answer the can operate with hotel loads on the order of a watt or
needs of ocean science. less, but at the cost of operating a few very simple
sensors.
The power relationship provides insight into a
Terminology
variety of AUV design considerations. For example,
The military term for an AUV is unmanned under- the speed at which the vehicle will consume the least
water vehicle, or UUV. This phrase is ambiguous in energy per unit distance traveled (the energetically
that it can also refer to ROVs. While ROVs are optimum speed) can be computed from observing
unmanned, they are tethered and designed to be that power divided by vehicle speed equals energy
operated by human, and thus are not considered per unit distance. Finding the minimum of P/v with
autonomous. However, common usage is to employ respect to v yields the optimum speed from an energy
UUV as a synonym for AUV. Military terminology conservation perspective:
is relevant as many AUVs in use by the scientific
 
community were developed under navy funding, ZH 1=3
and the military continues to be the largest single vopt ¼ ½2
CD Ar
investor in AUV technology. Consequently, the
technical literature on AUVs also employs military For the example vehicle values given above, the op-
terminology. timum vehicle speed is approximately 1 m s1.
What can we say about vehicle performance at the
energetically optimum speed? Substituting eqn [2]
Basics of AUV Performance into eqn [1] we find that the power consumed at the
optimum speed is (3/2)H which for our example
Energy is a fundamental limitation for underwater vehicle is 45 W. If the total energy capacity of the
vehicles. Thus, energy efficiency is a fundamental battery system is Ecap, then the maximum range of
driver for vehicle design and operations. This section the vehicle will be:
outlines the relationship between vehicle speed and
endurance, and its dependence on factors such as  1=3
2Ecap Z
power consumed by onboard systems. dmax ¼ ½3
3 CD ArH 2
A simple but useful model for power consumption
P of an AUV is as follows: If our example vehicle carries 10 kg of high-energy-
density primary batteries providing a total of
1 CD Arv3 1.3  107 J, it will have an endurance of 80 h, and a
P ¼ Pprop þ H where Pprop ¼ ½1
2 Z range of 280 km. The same vehicle with 10 kg of
rechargeable batteries, with one-third the energy
Here the total electrical power consumed by the ve-
capacity of the high-energy-density batteries, will have
hicle, P, is equal to the sum of propulsion power,
its endurance and range reduced proportionately.
Pprop and hotel load, H. Hotel load is simply the
There are caveats to the above discussion. For
power consumed by all subsystems other than pro-
example, propulsion efficiency, Z, is typically a strong
pulsion. Propulsion power is a function of the drag
function of speed as the electrical motors used tend
coefficient of the vehicle, CD, the area of the vehicle,
to have comparatively narrow ranges of efficiency. In
A, the density of water, r, the speed of the vehicle, v,
practice, a vehicle’s propulsion system is optimized
and the efficiency of the propulsion system, Z.
for a particular speed and power. Also, vehicles are
What are the typical values for the coefficients
often operated at higher speeds than the energetically
in eqn [1]? Consider an ‘example vehicle’ which is a
optimum speed given by eqn [2]. For example, op-
12 3/400 (0.32 m) diameter torpedo-like AUV. Note
erators of an AUV attended by a ship will be more
that this is a standard for a mid-size class AUV.
sensitive to minimizing ship costs than to optimizing
For such a vehicle, parameter values might be
energy efficiency of the AUV.
CD ¼ 0.2 (based on frontal area), A ¼ 0.082 m2, and
Z ¼ 0.5. Hotel load would depend on sensors, but an
overall value of 30 W might be representative,
AUV Systems and Technology
although mapping sonars would consume much
more power. We use r ¼ 1027 kg m–3. Note that these AUVs are highly integrated devices, containing a
numbers, except seawater density of course, can variety of mechanical, electrical, and software sub-
differ greatly from vehicle to vehicle. For example, systems. Figure 2 shows internal and external views
gliders optimized for low speed and long endurance of a deep-diving vehicle equipped with mapping
476 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES

Doppler velocity logger

and inertial navigation system

Sub-bottom Batteries
profiler
Fluid intake
Sonar electronics
pressure vessel Computer
enclosure
Water
sensor

Acoustic
transducers
Flotation

Antenna

Lifting eye

Sonar receive

Fairing Thruster and duct

Figure 2 A propeller-driven, modular AUV with labeled subsystems. The top view shows the interior in which an internal mechanical
frame supports pressure vessels, internal components, and the propulsion system. This AUV, called a Dorado, is a ‘flooded’ vehicle
because the fairings are not watertight, and thus the interior spaces fill with water. Consequently internal components must all be
capable of withstanding ambient pressures. Joining rings are visible between the yellow fairing segments on the lower figure. These
allow the vehicle to be separated along its axis, allowing replacement or addition of hull sections. This allows reconfiguration of the
vehicle with new payloads, and if desired, with new batteries. The propulsion system on this vehicle is a ducted thruster capable of
being tilted both vertically and horizontally, to steer the vehicle in the vertical and horizontal planes. Courtesy of Farley Shane, MBARI.

sonars. The anatomy of an AUV typically includes • navigation sensors to determine the vehicle
the following subsystems: position;
• software and computers capable of managing • communication devices to allow communication
of human operators with the AUV;
vehicle subsystems to accomplish specific tasks
and even complete missions in the absence of • locating devices to allow operators to track the
vehicle and locate it for recovery or in the case of
human control;
emergencies;
• energy storage to provide power;
• devices for monitoring vehicle health (e.g., leaks
• propulsion system;
or battery failure);
• a system for controlling vehicle orientation and
• emergency systems for ensuring vehicle recovery
velocity;
in the event of failure of primary systems.
• sensors for measuring vehicle attitude, heading,
and depth; The mechanical design of an AUV must address
• pressure vessels for housing key electrical issues such as drag, neutral buoyancy, the highly
components; dynamic nature of launch and recovery, and the need
 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES 477

to protect many delicate electrical components from supporting a steady increase in software capabilities
seawater. The desire to operate large numbers of for AUVs.
sensors for long distances encourages the con- Most AUV missions involve sequential tasks such
struction of larger vehicles to hold the necessary as descending from the surface to a set depth, then
equipment and batteries. However, the need for ease transiting to a survey location at a set speed,
of handling and minimizing logistical costs en- and then conducting a survey which might involve
courages the design of smaller vehicles. Operational flying a lawn-mower pattern. The vehicle may be
demands will also create constraints on vehicle de- commanded to maintain constant altitude over the
sign; for example, launch and recovery factors will bottom if the vehicle is mapping the seafloor. A
impose the need for lift points and discourage ex- water-column mission might require the vehicle
ternal appendages that will be easily broken. The profile in the vertical plane, moving in a saw-tooth
need to service vehicle components imposes a re- pattern called a yo-yo. The mission will likely include
quirement that internal components be easily ac- a transit from the end of the survey to a recovery
cessible for servicing and testing, and when location, with a final ascent to the surface for re-
necessary, replacement. covery. During the mission, the vehicle will monitor
In addition, a host of supporting software and the performance of onboard subsystems, and in the
hardware are required to operate an AUV. De- event of detection of anomalies, like a low battery
pending on the nature of the operations, supporting level, or a failed mission sensor, may abort the
equipment will include: mission and return to the recovery point early. A
more catastrophic failure might lead to the vehicle
• software and computers for configuring vehicle
shutting down primary systems, and dropping a drop
mission plans, and for reviewing vehicle data;
weight so as to float to the surface, and calling for
• systems for communicating with the vehicle both
help via satellite or direct radio frequency (RF)
on deck and when deployed;
communications.
• systems for recharging and monitoring vehicle
More complex vehicle missions can involve cap-
batteries;
abilities such as adapting survey operations to obtain
• handling gear for transporting, deploying, and
better measurements, or managing tasks such as
recovering AUVs;
AUV docking. An example of the first might be as
• devices for detecting locating devices on the AUV;
simple as a yo-yo mission which cues its vertical in-
• acoustic tracking systems for monitoring the lo-
flections from water temperature in order to follow a
cation of the vehicle when in the vicinity of a
thermocline. Also in the category of adaptive, but
support vessel.
more demanding, surveys, is the capability of fol-
lowing a thermal plume to its source, for example,
when an AUV is used to search for hydrothermal
AUV Mission Software
vents. The docking of an AUV with an underwater
Functionally, AUV software must address a variety structure encompasses yet a different type of com-
of needs, including: allowing human operators to plexity, created by the large number of steps in the
specify objectives, managing vehicle subsystems to process, and the high likelihood that individual steps
achieve mission objectives, logging data for sub- will fail. For example, docking involves homing on a
sequent review, and ensuring safety of the vehicle in docking structure, orienting for final approach, en-
the event of failures or unexpected circumstances. gaging the dock, and making physical connections to
The software must be capable of managing vehicle establish power and communication links. Any one
sensors and control systems to maintain a set head- of these steps might fail due to external pertur-
ing, speed, and depth. The software might also need bations; for example, currents or turbulence in the
to support interacting with a human operator during marine environment might cause the vehicle to miss
a mission. In addition to software on the vehicle it- the dock. The vehicle must be able to detect failures
self, AUV operators rely on a suite of software ap- and execute a process to recover and try again.
plications to configure and validate missions, to Docking is representative of the increasingly com-
maintain vehicle subsystems such as batteries, to re- plex capabilities AUVs are expected to master with
view data generated by the vehicle, to prepare mis- high reliability.
sion summaries, and when possible, to track and
manage the vehicle while underway. The exponential
Navigation
growth of computational power available for both
onboard and off-board computers, as well as the The ability for an AUV to determine its location on
increasingly pervasive nature of the Internet, are the Earth is essential for most scientific applications.
478 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES

However, navigation in the subsea environment is error with ancillary measurements of velocity and
complicated by the opacity of seawater to all but position. For example, combining an INS with a
very low frequency electromagnetic radiation, ren- DVL for constraining velocity can result in a system
dering ineffective the use of commonly used tech- which provides navigation accuracies better than
nologies such as the Global Positioning System (GPS) 0.05% of distance traveled.
and other radio-based navigation techniques. Con- The two acoustic navigation methodologies most
sequently, navigation underwater relies primarily on frequently used in AUV operations are ultrashort
various acoustic and dead-reckoning techniques and baseline navigation (USBL) and long baseline (LBL)
the occasional excursion to the surface where radio- navigation. A USBL system uses an array of hydro-
based methods can be used. There is no single phone separated by a distance comparable to the
method of underwater navigation that satisfies all wavelength of sound to measure the direction of
operational needs, rather a variety of methods are propagation of an acoustic signal. Most often, a
employed depending on the circumstances. USBL system is mounted on a ship, and used to track
Dead-reckoning methods integrate a vehicle’s ve- a vehicle relative to the ship. With knowledge of the
locity in time to obtain an updated location. In order ship’s location and orientation, the location of the
to dead-reckon, the vehicle must know both the AUV can also be determined. In contrast, LBL
direction and speed of its travel. The simplest navigation acoustically measures the range between
methods use a magnetic compass to determine dir- the vehicle and an array of widely separated devices
ection, and use speed through the water as a proxy of known location. A common LBL approach is to
for Earth-referenced speed. However, the large place transponders on the seafloor, and let the vehicle
number of error sources for magnetic compasses range off the transponders. The process of deter-
make measurement of heading to better than a de- mining location using ranges from known locations
gree accuracy technically challenging. Currents pose is called spherical navigation, as the vehicle should
even more of a problem, as they may be comparable be located at the intersection of spheres with the
to the vehicle speed in amplitude, yet are not sensed measured radius, centered on the respective trans-
by a water-relative measurement. Dead-reckoning is ponders. An alternative LBL navigation method is to
improved by measuring velocity relative to the sea- track a vehicle which pings at a preset time to an
floor, for example, using a Doppler velocity log array of hydrophones at known locations. If the time
(DVL) or a correlation velocity log. A DVL is com- of the ping is not known, the problem of solving for
monly used by AUVs to measure velocity by meas- the vehicle location is called hyperbolic navigation,
uring the Doppler shift of sound reflected off the as only the difference in time of arrival of the
seafloor. Correlation velocity logs are more complex ping at the various hydrophones can be determined,
in concept, involving measurement of the correlation and this knowledge constrains the vehicle to be on a
of two pulses of sounds transmitted by the vehicle, hyperbola between the respective receivers. If the
reflected off the seafloor, and received by a hydro- time of the ping is known, perhaps triggered at a
phone array. In practice, DVLs are used when a ve- preselected time by a carefully calibrated clock, then
hicle operates close to the seafloor, perhaps within the problem reduces to spherical navigation. In
200 m, while correlation velocity logs are used when practice, a wide variety of USBL and LBL systems
the vehicle is operating in mid-water columns or near have been implemented for underwater navigation.
the surface in deep water. They must all address the challenges of acoustic
Inertial navigation system (INS) technology is well propagation in the ocean, which include the ab-
developed, as it is widely used for platforms like sorption of sound by seawater, diffraction by speed
aircraft and missiles. However, INS units appropriate of sound variations in the underwater environment,
for underwater use are expensive enough that they scattering by reflecting surfaces, and acoustic noise
are used only when navigation requirements are generated by physical, geological, biological, and
stringent, for example, for producing high-accuracy anthropogenic processes.
maps. A modern INS includes an array of acceler- Other methods of navigation include using geo-
ometers for measuring acceleration on three axes and physical parameters, for example, water depth, to
a laser or fiber optic gyroscope for measuring chan- constrain the vehicle location in the context of
ges in orientation. Additionally, an INS will include a known maps. These geophysically based navigation
GPS for initializing the unit’s location and orien- methods, similar to terrain contour mapping (TER-
tation, and a computer for acquiring and processing COM) navigation used by cruise missiles, depend on
data from INS component sensors. The position re- having good maps ahead of time. There are software
ported by an INS will have an error which will grow approaches in development that simultaneously
in time, and thus it is important to constrain INS build maps and use those same maps for navigation.
 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES 479

These methods are called SLAM for simultaneous towed body. If ultrashort baseline acoustic navi-
localization and mapping. gation techniques are used to determine the vehicle
position, then layback of the towed body behind the
ship introduces significant errors as compared with
Using AUVs for Ocean Science having the ship directly over the sonar platform. For
this reason, some commercial use of towed sonar
Mapping the Seafloor
platforms use two ships, one to tow the sonar
AUVs are becoming the platform of choice for high- platform, and one positioned directly over the
resolution seafloor maps. Obtaining high-resolution platform to determine its precise location. Finally,
maps requires operating mapping sonars near the surface motion of the ship will be efficiently coupled
seafloor. Alternatives to an AUV include crewed to the tow body by the tow cable. Thus, even near
submersibles and tethered platforms. Crewed sub- the seafloor, the tow body will be subject to sea state
mersibles are too valuable for routine mapping, and experienced by the ship. Consequently, attraction of
are reserved for other uses which require the pres- the use of AUVs includes more economical oper-
ence of humans. Towed vehicles are used for sonar ations and high data quality. Figure 3 shows a cost
mapping, but have disadvantages as compared with comparison of a commercial deep-water towed sur-
AUVs, especially in deeper water. The principal vey and an equivalent AUV survey.
problem is the high drag of the cable used for a tow Sonar systems used on AUVs for mapping include
sled, which in water depths of several thousand multibeam sonar, side scan sonar, and sub-bottom
meters will limit speeds to approximately half a profilers. Multibeam sonars, operating at frequen-
meter per second. Even at these slow speeds, a cies of hundreds of kilohertz in the case of
towed platform will stream behind the towing ship, AUV-mounted systems, allow measurement of
creating several problems. Controlling the position range to the seafloor in multiple sonar beams and
of the towed vehicle over the bottom is very dif- are used to build up three-dimensional maps such as
ficult, even when running on a constant heading. that in Figure 4. Side scan sonars used by AUVs
When surveying a defined area on the seafloor in a also typically operate at frequencies of hundreds of
series of passes, the turns between passes may take kilohertz, and are used to image seafloor features.
longer than the actual survey passes themselves, as it Side scan sonars are particularly useful for finding
is necessary to turn slowly to maintain control of the objects, for example, looking for a shipwreck resting

Report preparation

100% Tracking boat

90%
Cost (normalized to total towed survey cost)

Time spent turning to

return to survey site
80%
Survey
70%
Transit to work site
60%
Mobilization demob
50%

40%

30%

20%

10%

0%
Towed survey AUV

Figure 3 A comparison of the economics of deep survey taken from costs of a survey with a towed vehicle, and projected costs of
the same survey with an AUV. The principal cost saving derives from the ability of the AUV to turn much faster than a deep-towed
vehicle, reducing the total survey time. Also, the AUV can be acoustically tracked by its mother ship, while a towed vehicle requires a
second ship for tracking because the towed vehicle will trail far behind the tow ship. Finally, mobilization and demobilization costs for
the AUV can also be lower, although this depends on the size of the AUV employed.
480 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES

122° 00' 50" W 122° 00' 40" W 122° 00' 30" W 122° 00' 20" W

km
36° 46' 50" N 36° 46' 50
0 0.1 0.2

36° 46' 45" N 36° 46' 45

36° 46' 40" N 36° 46' 40

36° 46' 35" N 36° 46' 35

36° 46' 30" N 36° 46' 30

122° 00' 50" W 122° 00' 40" W 122° 00' 30" W 122° 00' 20" W

−1012.5 −1000.0 −987.5 −975.0 −962.5 −950.0 −937.5 −925.0 −912.5 −900.0 −887.5
Topography (m)

Figure 4 A bathymetric survey produced by an AUV at a depth of about 1000 m. Note the very small size of the survey area and high
resolution of the bathymetry. Courtesy of Dave Caress, MBARI.

on the seafloor. Sub-bottom profilers use lower- ocean. The smaller, buoyancy-driven gliders are
frequency sound, ranging from 1 kHz to tens of unique in their combination of mobility and en-
kilohertz in the case of an AUV-mounted system, to durance, moving at about a quarter of a meter per
penetrate into the seafloor. Depending on the bot- second for periods of months. Larger vehicles carry
tom type (e.g., sandy, muddy, or rock), a sub- more comprehensive payloads at higher speeds, but
bottom system might penetrate tens of meters. for shorter periods. Such vehicles might operate at
Cumulatively sonar payloads will consume com- 1.5 m s1 for a day. A common flight profile is to fly
paratively large amounts of energy, perhaps hun- the vehicle on a constant heading, while moving
dreds of watts. Mapping also requires high-fidelity between two depth extremes in a saw-tooth pattern.
navigation, and thus sonar-equipped AUVs will Often the upper depth extreme will be close to the
often also use more sophisticated navigation ap- surface. This strategy allows the production of ver-
proaches, like inertial navigation. Consequently, tical sections of ocean properties, such as those in
mapping AUVs of today are larger, more sophisti- Figure 5. Variations of this strategy might have
cated AUVs. the vehicle moving in a lawn-mower or zigzag pat-
tern in the horizontal plane, to develop a full three-
dimensional map of ocean properties. Figure 6 shows
Observing the Water Column
a visualization of an internal wave interacting with a
AUVs provide a relatively new tool for observing the phytoplankton layer using such a three-dimensional
physical, chemical, and biological properties of the mapping strategy.
 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES 481

Temperature
0 16

14
10
12
20 10

30 8

0 1000 2000 3000 4000 5000 6000 7000

Fluorometer
0

10 1500

1000
20
500
30
0
0 1000 2000 3000 4000 5000 6000 7000
Optical backscatter
0
1000

10 800

600
20
400
30 200
0 1000 2000 3000 4000 5000 6000 7000
Downtrack distance (m)

Figure 5 Vertical sections of water properties obtained by an Odyssey AUV operating in Massachusetts Bay. The y-axis of each
figure is depth, in meters, and the x-axis is horizontal distance in meters. The top section shows temperature in degrees Celsius, the
middle shows chlorophyll fluorescence in arbitrary units, and the bottom shows optical backscatter, also in arbitrary units. The path of
the vehicle is shown as a white line, and the interpolated values of the measured property are plotted in color. The vehicle alternated
between obtaining high-resolution observations of the thin layer of organisms at the thermocline with full water column profiles.

All AUVs are limited by the availability of sensors. operation has been carried out with AUVs in both
Temperature, salinity, currents, dissolved oxygen, ni- the Arctic and Antarctic. Sea ice poses special oper-
trate, optical backscatter properties, and chlorophyll ational challenges for seagoing ocean scientists. For
fluorescence are examples of the growing in situ sens- example, ships with ice-breaking capability can op-
ing capabilities available for AUVs. However, many erate in the ice pack, but will typically not be able to
important properties, for example, pH, dissolved car- hold station, or even assure that tethers and cables
bon dioxide, and dissolved iron, cannot be measured deployed over the side will not be severed. AUVs are
reliably from a small moving platform. Furthermore, attractive in that they provide horizontal mobility
detection of marine organisms is usually accomplished under ice, and the ability to conduct operations near
by proxy; for example, chlorophyll fluorescence pro- the seafloor without the complications intrinsic in
vides an indicator for phytoplankton abundance. tether management. Challenges of operating AUVs
In situ methods which directly detect, classify, and under ice revolve around the need to assure return of
quantify marine organism abundance are not avail- the AUV to the ship for recovery, the process of re-
able, yet are increasingly important for understanding covering the AUV through the ice onto the ship, the
the structure and dynamics of ocean ecosystems. potential for having an AUV fail and become trapped
under ice, and the difficulty of carrying out tasks that
Operations in Ice-covered Oceans
would normally be accomplished having an AUV
AUVs offer unique operational capabilities for sci- surface (e.g., obtaining a GPS update). Most safety
ence in ice-covered oceans. Successful under-ice strategies for AUVs in ice-free oceans default to bring
482 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES

Monterey Bay

Front

5m

10 m Phytoplankton
layer

Internal
15 m wave

20 m
2 km

1 km
0 km
2 km 0 km
4 km 6 km

Figure 6 Interaction of layer of phytoplankton with an internal wave in Monterey Bay. Both physical and biological properties were
measured by an Odyssey AUV, which moved in a horizontal zigzag pattern across the survey volume, while profiling constantly in the
vertical plane. The phytoplankton layer, shown in green, was detected by a chlorophyll fluorescence sensor on the AUV. The cyan
surface shows deflection of a level of constant density of seawater by a passing internal wave. Courtesy of John Ryan, MBARI.

the vehicle directly to the surface, for example, by time-constrained surveys are most often encountered
dropping a weight. In the Arctic or Antarctic, this when surveying a dynamic process. For example, if
strategy could result in the vehicle becoming trapped the temporal decorrelation of ocean fields associated
under very thick ice, making the vehicle harder to with upwelling off Monterey Bay is about 48 h, at-
find and potentially impossible to recover. The usual tempts to map the ocean fields need to be accom-
surface location devices such as RF beacons, strobes, plished within that time frame. Scales of spatial
and combinations of RF communication and satellite variability will also determine acceptable separation
navigation will not work. Clearly AUV operations of observations: for example, decorrelation lengths
within the ice pack entail higher risk and a more in Monterey Bay are on the order of 20 km, so ob-
sophisticated vehicle. servations need to be spaced significantly closer to
minimize errors in reconstructing the ocean field.
How does this relate to the number of vehicles re-
Observation Systems, Observatories, and AUVs
quired to accomplish such a survey? Consider a grid
An understanding of power consumption of AUVs survey of a 100 km  100 km area with a resolution
provides insight to the attractiveness of employing of 10 km. A single vehicle would have to
multiple vehicles for certain ocean observation travel c. 1000 km at a speed of nearly 6 m s1,
problems. In some circumstances a survey must be traveling in a lawn-mower pattern. Using the ex-
accomplished within a set period. In oceanography, ample vehicle values from the ‘Basics of AUV
 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES 483

performance’ section, the AUV would consume Autonomous Ocean Sampling Network (AOSN) II
about 3500 W if it were capable of operating at such field program (Figure 7). Gliding vehicles, with an
a high speed. In contrast, six of the same vehicles endurance of weeks to months, provided a con-
operating at their optimum speed would consume a tinuous presence with a minimal sensor suite. A few
total of 270 W. In other words, the six vehicles would propeller-driven vehicles provided observations of
consume 12 times less energy for the complete 48-h chemical and biological ocean parameters, allowing
survey. tracking of ecosystem response to the upwelling
Autonomous mobile platforms are making ob- process. Observations were fed to two oceano-
servation of the interior of the ocean more afford- graphic models, which provided synoptic realization
able and more flexible, enabling the practical of ocean fields and predicted future conditions.
realization of coupled observation–prediction sys- Among the many lessons are an improved know-
tems. For example, in late summer 2003, a diverse ledge of the scales of variability of upwelling pro-
fleet of AUVs was deployed to observe and predict cesses, an understanding of how to scale observation
the evolution of episodic wind-driven upwelling in systems to these processes, and insights to strategies
the environs of Monterey Bay. Over 21 different for adaptive sampling of comparatively rapidly
autonomous robotic systems, three ships, an air- changing processes with comparatively slow ve-
craft, a coastal ocean dynamics application radar hicles. These lessons are particularly relevant today,
(CODAR), drifters, floats, and numerous fixed given the present emphasis on developing ocean-
(moored) observation assets were deployed in the observing systems.

Figure 7 Example of a distributed observing system using AUVs. This diagram depicts an AOSN deployment in Monterey Bay.
484 PLATFORMS: AUTONOMOUS UNDERWATER VEHICLES

See also IEEE (2001) Special Issue: Autonomous Ocean Sampling

Networks. IEEE Journal of Oceanic Engineering 26(4):
Gliders. Remotely Operated Vehicles (ROVs). 437--446.
Jenkins SA, Humphreys DE, Sherman J, et al. (2003)
Further Reading Underwater glider system study. Scripps Institution of
Oceanography Technical Report No. 53. Arlington,
Allmendinger EE (1990) Submersible Vehicle Systems VA: Office of Naval Research.
Design. New York: SNAME. Rudnick DL and Perry MJ (eds.) (2003) ALPS:
Bradley AM (1992) Low power navigation and control for Autonomous and Lagrangian Platforms and Sensors,
long range autonomous underwater vehicles. Workshop Report, 64pp. http://www.geo-prose.com/
Proceedings of the Second International Offshore and ALPS (accessed Mar. 2008).
Polar Conference, pp. 473–478.
Fossen T (1995) Guidance and Control of Ocean Vehicles. Relevant Website
New York: Wiley.
Griffiths G (ed.) (2003) Technology and Applications of http://www.mbari.org
Autonomous Underwater Vehicles. London: Taylor and – Monterey Bay 2003 Experiment, Autonomous Ocean
Francis. Sampling Network, MBARI.