Underwater Gliders: Recent Developments and Future

Applications
(Invited Paper)
R. Bachmayer∗ , N. Ehrich Leonard† , J. Graver† , E. Fiorelli† , P. Bhatta† and D. Paley†
∗ National Research Council - Institute for Ocean Technology
Kerwin Place, St. John’s, A1B 3T5 NL, Canada
Tel.: +709-772-7986; Fax: +709-772-2462
E-mail: Ralf.Bachmayer@nrc.ca

† Department of Mechanical and Aerospace Engineering

Princeton University, Princeton, NJ 08544, USA
E-mail: naomi{jggraver, eddie, pradeep, dpaley}@princeton.edu

Abstract— Autonomous underwater vehicles, and in particular au-

tonomous underwater gliders, represent a rapidly maturing technology
with a large cost-saving potential over current ocean sampling technolo-
gies for sustained (month at a time) real-time measurements.
In this paper we give an overview of the main building blocks of an
underwater glider system for propulsion, control, communication and
sensing. A typical glider operation, consisting of deployment, planning,
monitoring and recovery will be described using the 2003 AOSN-II field
experiment in Monterey Bay, California.
We briefly describe recent developments at NRC-IOT, in particular
the development of a laboratory-scale glider for dynamics and control
research and the concept of a regional ocean observation system using
underwater gliders.

I. I NTRODUCTION
Sampling the oceans has traditionally been conducted from ships,
with the first global oceanographic research cruise by Sir Wyville
Thomson on the HMS Challenger from 1872-1876, Figure 1, which
led to numerous discoveries such as the mid-Atlantic ridge and the
Challenger Deep in the Mariana Trench to name only a few. It took
over 23 years to compile the results from this cruise.
Today with increasing use of remote sensing techniques from
satellites and airplanes more and more data becomes available and
needs to be processed. Current remote sensing technologies, airborne
or from space, do not penetrate very far below the ocean’s sur-
face. In order to gain more insight into the temporal and spatial
processes below the surface we were until recently still depending
on ship based measurements and moorings. Over the last decades
alternative technologies such as subsurface floats, remotely operated
vehicles (ROVs) and autonomous underwater vehicles (AUVs) have
emerged to complement the existing sensing techniques. Visions of
Fig. 1. Dredging and sounding arrangements on board
autonomous platforms roaming the oceans as described in [9] and [1] the Challenger. Photo Credit: NOAA Photo library, source:
have not come true yet, but technological advances pushed by these http://oceanexplorer.noaa.gov/history/breakthru.html.
visions brought us a long way from the Challenger cruise.
In this paper we report on a special type of autonomous underwater
vehicle, an underwater glider, and on the implementation of coupled in-house expertise in AUV development, control and operations, is
ocean observation and modeling systems. In Section II we describe described Section IV-B.
the current glider technology and their mode of operation. Section
III-A briefly describes the AOSN-II field experiment in Monterey II. U NDERWATER G LIDERS
Bay and Section III-B highlights the approach and implementation Autonomous underwater gliders, represent a rapidly-maturing tech-
of multi-glider operations for the AOSN-II effort. In Section IV-A nology with a large cost-saving potential over currently-available
we describe NRC-IOT’s role in developing an asset management tool ocean sampling techniques, especially for sustained, month at a time,
for a regional ocean observation modeling and prediction facility in real-time oceanographic measurements.
Newfoundland. Current efforts at NRC-IOT to develop a laboratory- Underwater gliders move efficiently through the water-column
scale glider to support the above described effort and to enhance by exploiting their ability to change their weight in water. As a
 SLOCUM Electric Glider
frequent communications and shallow dives, which imply frequent
changes in buoyancy, is an increase in power consumption and
OAL: 2m Embedded therefore a reduction in mission length. Currently the operational
Span: 1m Antennae
endurance of the gliders varies from 3 to 4 weeks for the shallow
Hull OD: 0.2m
Fixed Wings SLOCUM glider (max. depth ≤ 200m) to several months for the
Rudder deeper diving gliders Seaglider (max. depth ≤1000m) and Spray
(max. depth ≤1500m). All three gliders are comparable in size and
Buoyancy engine handling requirements. Their weight in air is approximately 50 kg and
Inflatable their total volume change capacity is between 0.5 and 1% of their
Bladder total displacement. The horizontal speed relative to the surrounding
water is typically around 35 cm/s. For more detailed information on
the specific performance of the gliders the reader is referred to [3],
[8], [11] and [4].
CTD
III. A PPLICATIONS

 Webb Research Inc., Falmouth, MA, USA

A. Autonomous Ocean Sampling Network II - Monterey Bay 2003
(AOSN-II)
Fig. 2. Rendering of a SLOCUM electric glider. Built by Webb Research
Inc., Falmouth, MA, USA; http://www.webbresearch.com.

result there is an upward/downward force acting on the glider.

Successive weight changes combined with a change in attitude result
in a concatenation of up/down glide cycles. The combination of
upward/downward force with the change in attitude (i.e. pitch) allow
the wings and body to generate the hydrodynamic lift and drag
forces which propel the gliders horizontally and vertically through the
water. The mechanism to achieve this change in weight is referred to
as a buoyancy engine (see Figure 2). Currently operational gliders,
such as Seaglider [3], Spray [8] and the electric SLOCUM glider
use an electromechanical displacement actuator, pump or piston, to
change their weight. A prototype glider using an alternative thermally
driven buoyancy engine is currently under development [11]. The
closed-loop control of attitude and depth is performed by an on-
board computer that also executes a pre-programmed mission while
submerged. At the surface the gliders acquire their location using
a GPS receiver and compare that position to the desired position
from the mission plan. The position error is used to compute
an estimate of the average current flow encountered between two
surfacings. The current estimate is then used to correct the dive Fig. 3. Bathymetric map of Monterey Bay, California, USA (depth in meters).
parameters (i.e. heading) for the next dive cycle. At the surface
the gliders are able to communicate globally using an IRIDIUM The Autonomous Ocean Sampling Network-II [1] field experiment
satellite connection (datarate≈2400 baud) or, for local line-of-sight was conducted during the summer of 2003 in Monterey Bay, Cali-
communication, some gliders (i.e. SLOCUM) are equipped with a fornia. This bay was chosen for its accessibility, resident research
high bandwidth RF-modem (datarate≈115.2 kbaud). An ARGOS institutions with on-site hardware (i.e. ships, airplanes, AUVs) and
transmitter is implemented as a fall-back solution. The antennae are its interesting bathymetry, Figure 3. Since the region is well studied
integrated into the gliders such that while the glider is at the surface, there is a large amount of historic data available for intercomparisons.
the antennae are at a maximum height above the water surface The objective of the experiment was to demonstrate the feasibility
for reliable communications. In the case of the SLOCUM glider, of an integrated ocean observation, modeling and prediction sys-
the antennae for communication and GPS are embedded within the tem. This experiment differs from previous efforts because of its
rudder assembly, Figure 2 and, by means of an inflatable bladder in high degree of system integration, allowing for real-time adaptation
the tail cone, can be brought out of the water. Once communication to based on ocean model predictions. The sampling patterns of mobile
a control center has been successfully established, the current glider observational assets, such as ships, airplanes, underwater gliders
mission can be updated and/or data recorded during previous missions (i.e. Spray, SLOCUM) and propeller driven AUVs (i.e. REMUS,
can be downloaded from the vehicle. DORADO) were planned and, in some cases, adapted using the
Besides the vehicles’ position, attitude and other internal states, numerical modeling and prediction capabilities of two independently-
the gliders collect data from their scientific sensors. Typically the running numerical modeling codes developed by two groups from
gliders carry a conductivity, temperature and depth sensor (CTD), but Harvard University (HOPS [7]) and the Jet Propulsion Laboratory
more recently additional instrumentation such as Photosynthetically (ROMS). Those models in turn were supplied with data coming from
Active Radiation (PAR) sensors and fluorometers have successfully the mobile assets, as well as other sources such as CODAR data
been operated. The drawback due to additional sensors as well as (COntinental raDAR), satellites, fixed moorings and surface drifters.
Fig. 4. Diagram of Real Time Operations Plan for AOSN-II. LCS stands for Lagrangian Coherent Structures, ROMS for Regional Ocean Model (JPL/UCLA)
and HOPS for the Harvard Ocean Prediction System [7].

Figure 4 gives an overview of the interactions between the different their more frequent inflections and higher sensor load, the SLOCUM
parts of the system as well as an idea of the different time-scales gliders had to be recovered during the course of the experiment.
involved in the experiment. The gliders were either directed to a designated recovery area close
to Moss Landing or were directed to an area to rendezvous with
B. Glider Operations a surface vessel for recovery. When in range of the surface vessel,
The core observational assets of AOSN-II were autonomous un- the gliders were able to directly communicate with the vessel and
derwater vehicles and in particular a fleet of underwater gliders. were controlled using the high bandwidth RF-link. After recovery
Two types of gliders were available, five Spray gliders [8] operated the gliders’ battery packs were replaced, the systems re-ballasted,
by Jeff Sherman and Russ E. Davis of the Scripps Institution of checked out and readied for redeployment.
Oceanography and 10 SLOCUM gliders [11] operated by David In order to manage the number of different assets in the water, as
Fratantoni from Woods Hole Oceanographic Institution. Prior to the well as to provide a quick overview of the last available positions, a
experiment all gliders were shipped to the Monterey Bay Aquarium realtime operational display was designed and made available in the
Research Institute (MBARI) in Moss Landing, California. On site control center at MBARI, Figure 5. The display was developed in the
the gliders were assembled, ballasted and tested in MBARI’s test- beginning of the experiment and was continuously improved during
tank. Since the gliders were deployed for long periods of time, the course of it. The display was automated and ran continuously
special attention was given to sensor calibration; the sensor data were during the experiment, which enabled the control room staff to
closely monitored during the course of the experiment. After initial closely monitor progress of the gliders and if necessary intervene. On
shakedown dives close to shore, the gliders were directed towards several occasions the operators noted that the gliders were advancing
their operational area. To take full advantage of the different depth only marginally over the course of several hours. This behavior was
capabilities of the gliders, (see Section II), the five Spray gliders associated with strong head currents close to the southern end of
were deployed in the deep water further outside the bay while the the bay (i.e. Monterey); those currents were on the order of the
SLOCUM gliders were flown closer to the bay. gliders’ horizontal velocity. On other occasions the gliders’ progress
Figure 5 shows a snapshot of the glider tracking display from 25 was far above its theoretical limits and continued slightly on shore.
August 2003, with a three day position history plotted behind each This behavior was observed three or four times and was attributed to
glider. The two large dots represent fixed moorings in the Bay (M1, fisherman “recovering” the equipment. The vehicles were retrieved
M2). The Spray gliders were flying on straight lines almost perpendic- from the recovery teams, checked out and re-deployed if necessary.
ular to the shoreline, while the SLOCUM gliders were either flown on The AOSN-II experiment successfully integrated all the above
a fixed “racetrack” (a,b,c,d) or operated in an adaptive sampling mode mentioned components and collected a valuable data-set for eval-
where the trajectories of several gliders were coordinated and adjusted uation of various sampling strategies and modeling efforts. The
in a real-time experiment [2]. Communication to and from the gliders performance of several multi-vehicle experiments during the course
during regular operations used the IRIDIUM satellite system. Due to of AOSN-II show the potential for added value by using coordinated
Fig. 5. Snapshot of real-time display for asset location. The length of the tail behind each asset corresponds to the positions during the last three days. Large
dots represent moorings M1 (right) and M2 (left). In the figure legend, WExx stands for Woods Hole Oceanographic Institution Electric Glider (SLOCUM
gliders), SIOxx for Scripps Institution of Oceanography (Spray gliders).

control strategies [2]. New tools are under development that allow for This facility will also provide a unique testing ground for future
improved planning and monitoring of the observational assets which developments in sampling strategies and technologies as well as a
will provide a higher degree of autonomy during future deployments. possibility to benchmark future improvements in the modeling and
prediction of the ocean environment.
IV. F UTURE D EVELOPMENTS AND A PPLICATIONS AT THE
I NSTITUTE FOR O CEAN T ECHNOLOGY
Observations will be based on a suite of different sampling
A. Newfoundland Ocean Observation, Modeling and Prediction Fa- platforms. We will perform conventional observations based on time
cility (NOOMPF) series from moorings, weather stations and ships. In addition to
A team of researchers from NRC-IOT, Memorial University of those measurements we will utilize available data-products from
Newfoundland and Labrador (MUN) and the Department of Fish- satellites including AVHRR and Radarsat. Besides these assets sev-
eries and Oceans (DFO) (R. Bachmayer and C. Williams from eral autonomous mobile platforms, such as autonomous underwater
NRC-IOT, B. de Young, L. Zedel, N. Bose and S. O’Young from gliders and propeller driven AUVs will be deployed for extended
MUN and F. Davidson from DFO) is currently developing a plan periods of time. NRC-IOT’s role in the development of such a
to implement a regional coupled ocean observation and modeling system is to develop the control and communication infrastructure
system in Newfoundland (NOOMPF). Possible sites for implementa- necessary to direct and monitor the observational assets. We are
tion are Conception Bay, Trinity Bay and Placentia Bay as shown going to develop an asset management tool (ASMT), Figure 7,
in Figure 6. The goal is to develop a capability for automated that will serve as the main control and monitoring interface. The
coupled ocean observations and model predictions on a regional modular design of the tool will allow us to sequentially develop
scale. NOOMPF will integrate novel approaches to ocean sampling, and improve individual components of the system. In the baseline
modeling and prediction. The potential improvements in the modeling version the ASMT will provide a basic display of asset locations
and prediction capabilities of the ocean will significantly enhance in an area of interest. Together with the first display module (Asset
our ability to predict and manage the ocean as a resource for food Location) a data interface will be developed that will allow us to
production, transportation (e.g. ice-drift predictions), and exploration. access position information and collected data from a selection of
 Placentia Bay
Trin
ity B
ay

Avalon Peninsula
Co
nce
pti
on
B ay

Fig. 7. Schematic of Asset Management Tool for NOOMPF.

Fig. 6. Photo Credit: Earth Sciences and Image Analysis, NASA-Johnson

Space Center. Candidate areas for implementation of an ocean observation,
modeling and prediction system in Newfoundland. Note: North is to the right.
Trinity Bay is approximately 140 km by 40km in size.

platforms (i.e. gliders, AUVs, buoys, ships, etc.) as soon as they

become available. Other parts to be developed include a planning
module, a vehicle health monitor and a general asset status module.
Some of these components, such as the planning module, require
access to meteorologic and oceanographic databases and models. The
vehicle health monitor will analyze data coming from the vehicles
to provide an automated early fault detection mechanism to warn
the operators of possible failures. The ASMT can also be used as
a simulation and practice environment using real-time, recorded or
generated data as inputs into the system.

B. Laboratory Scale Glider Fig. 8. Conceptual drawing for a laboratory-scale glider.

In order to complement the efforts described in Section IV-A,

NRC-IOT is considering developing a laboratory-scale glider, Figure tests in order to evaluate the performance the buoyancy engine.
8. The purpose of the laboratory glider is to conduct experiments After the completion of the design of the buoyancy engine we
for hydrodynamic testing and control and to provide a test-bed for are moving towards the hydrodynamic design of the glider. The
new actuation and flow sensing technologies. We are planning to design philosophy is to be able to build several glider hulls with
perform a complete system identification using recently developed significantly different hydrodynamic characteristics and reuse the
mathematical models [6], [10], [5] and [4] and experiments. The
electromechanical “internals” of the glider. This approach allows for
glider’s mathematical model will be used to develop various parts
experiments with uncommon designs such as flying wings or hybrid
of the ASMT, such as a health monitor and a planning module. The
gliders (added propeller propulsion, see Figure 8) at a reasonable
performance of those modules can then be evaluated using the data
cost.
coming from the glider operating in our test facilities.
As a first step towards the design of a laboratory scale glider,
V. ACKNOWLEDGEMENTS
we are currently investigating the design alternatives and constraints
of a buoyancy engine. We decided to design the buoyancy engine The authors would like to thank the Office of Naval Research
for operations in up to 20 m of water-depth and a size such that it for supporting our research for AOSN-II under research grants
fits into a cylindrical housing of 10 cm in diameter. The particular N00014-02-1-0826 and N00014-02-1-0861. We would also like to
characteristics of the engine, such as volumetric rate and absolute acknowledge the whole AOSN-II team for their support and insightful
displaced volume are to be designed such that the glider reaches a discussions that made the experiments such a success. The authors
steady state glide within 2/3 of the depth of IOT’s towing tank (tank would also like to thank Christopher Williams, Brad de Young and
depth: 7m). In a next step we are going to perform vertical motion Fraser Davidson for their roles in carving out a plan for NOOMPF.
 R EFERENCES
[1] T. Curtin, J. G. Bellingham, J. Catipovic, and D. Webb. Autonomous
ocean sampling networks. Oceanography, 6(3):86–94, 1993.
[2] N. E. Leonard E. Fiorelli, P. Bhatta and I. Shulman. Adaptive sampling
using feedback control of an autonomous underwater glider fleet. In
13th International Symposium on Unmanned Untethered Submersible
Technology, Durham, NH, USA, August 2003.
[3] C. C. Eriksen, T. J. Osse, R. D. Light, T. Wen, T. W. Lehman, P. L. Sabin,
J. W. Ballard, , and A. M. Chiodi. Seaglider: A long-range autonomous
underwater vehicle for oceanographic research. IEEE Journal of Oceanic
Engineering, 26(4):424–436, October 2001.
[4] Joshua G. Graver, Ralf Bachmayer, Naomi Ehrich Leonard, and
David M. Fratantoni. Underwater glider model parameter identification.
In 13th International Symposium on Unmanned Untethered Submersible
Technology, Durham, NH, USA, August 2003.
[5] Joshua G. Graver and Naomi Ehrich Leonard. Underwater glider
dynamics and control. In 12th International Symposium on Unmanned
Untethered Submersible Technology, Durham, NH, USA, August 2001.
[6] N.E. Leonard and J. Graver. Model-based feedback control of au-
tonomous underwater gliders. IEEE Journal of Oceanic Engineering,
26(4):633–645, October 2001.
[7] A. R. Robinson. Forecasting and simulating coastal ocean processes
and variabilities with the Harvard Ocean Prediction System. In C.N.K.
Mooers, editor, Coastal Ocean prediction, AGU Coastal and Estuarine
Studies Series, pages 77–100. American Geophysical Union, 1999.
[8] J. Sherman, R. E. Davis, W. B. Owens, and J. Valdes. The autonomous
underwater glider ”Spray”. IEEE Journal of Oceanic Engineering,
26(4):437–446, October 2001.
[9] Henry Stommel. The Slocum Mission. Oceanography, 2:22–25, 1989.
[10] Andrey Vladimirovich Ulanov. Optimization criteria and hierarchy
of mathematical models of an underwater gliding vehicle. In Third
International Shipbuilding Conference ISC’2002, St. Petersburg, Russia,
October 2002.
[11] D. C. Webb, P. J. Simonetti, and C. P. Jones. SLOCUM: An underwater
glider propelled by environmental energy. IEEE Journal of Oceanic
Engineering, 26(4):447–452, October 2001.