This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE JOURNAL OF OCEANIC ENGINEERING 1

Launch and Recovery of an Autonomous Underwater

Vehicle From a Station-Keeping Unmanned
Surface Vehicle
Edoardo I. Sarda and Manhar R. Dhanak

Abstract—In-water tests of automated launch and recovery TABLE I

(L&R) of a Hydroid REMUS 100 autonomous underwater vehicle PRINCIPAL CHARACTERISTICS OF REMUS 100
(AUV) from a station-keeping 16-ft wave adaptive modular vehicle
unmanned surface vehicle (USV) have been conducted to deter-
mine the feasibility of developed concepts of operation. The USV
is a catamaran with a custom-configured propulsion system that
enables it to maintain position and heading on the surface. AUV
launch is accomplished through lowering the AUV from the USV
top tray to the water surface and releasing it. Recovery is initiated
through requiring the AUV to navigate toward an acoustic homing shore without surfacing. A USV is a good mobile platform for
beacon on a taut line suspended from the USV. The taut line serves launch and recovery (L&R) since it can perform autonomous
as docking target and as a connecting link between the two vehi-
cles during L&R. During these operations, the USV approximately navigation in an unknown marine environment.
keeps station on the water surface. Once the docking is complete,
the USV moves forward, towing the AUV as it is extracted onboard A. Vehicles
the USV via a customized winch mechanism. It was found that for
appropriate environmental conditions, L&R of the AUV from a Hydroid REMUS 100 AUV and the 16-ft wave adaptive mod-
station-keeping USV is an effective alternative to the mobile L&R ular vehicle (WAM-V 16, Advance Marine Research, Inc., Rich-
method using the same vehicles, with the same launch and recovery mond, CA, USA) USV were selected as vehicles of choice for
system. this study.
Index Terms—Autonomous underwater vehicle (AUV), launch REMUS 100 is a compact, lightweight platform suitable for
and recovery (L&R), unmanned surface vehicle (USV). operations in coastal environments. Its compact size and low
weight make this AUV suitable for autonomous L&R operations
from mobile unmanned platforms. The principal characteristics
I. INTRODUCTION
of REMUS 100 are provided in Table I.
UTONOMOUS underwater vehicles (AUVs) and un-
A manned surface vehicles (USVs) are currently being uti-
lized for a variety of applications where manned operations
A 3-bladed propeller, a rudder, and two vertical fins provide
for the trust and maneuverability of the AUV. The vehicle is
depth rated for 100 m and can achieve a maximum speed of
can endanger lives. The different characteristics of an AUV 2.5 m/s. The vehicle configuration does not permit pure vehicle
and an USV make each more suitable for certain applications motion in all six degrees of freedom used in its dynamic model,
than the other. The two vehicles, collaboratively working to- therefore it is considered as an under-actuated system. The AUV
gether in a complementary fashion, enable a greater capability uses GPS to determine its position when it is on the free sur-
of the combined system. Examples of applications that benefit face. It navigates underwater by dead reckoning, utilizing the
from such collaborations include surveying and surveillance in inertial navigation system to estimate its geoposition. REMUS
remote unknown waters, mine-countermeasures, and port secu- 100 is equipped with an HG1700AG58 inertial measurement
rity. In these cases, an AUV alone may not be sufficient since unit (IMU) from Honywell for measurement of accelerations
it typically lacks the ability to travel long distances at high and the rate of change of the orientation angles [1]. As acoustic
speeds, and it is unable to communicate with a surface ship or Doppler current profilers, REMUS 100 mounts a TD Explorer
R100 [1] to aid the dead reckoning navigation and to measure
Manuscript received April 10, 2018; revised August 10, 2018; accepted Au- the water currents. The acoustic modem on REMUS 100 allows
gust 22, 2018. This work was supported by the Office of Naval Research un- the vehicle to transmit and receive basic messages underwater.
der Grant N000141512724 [Program Manager: Kelly Cooper]. (Corresponding In addition to the package described, REMUS 100 is equipped
author: Edoardo I. Sarda.)
Guest Editor: W. Kirkwood. with digital ultrashort baseline–acoustic positioning system
The authors are with the SeaTech—The Institute for Oceans and Systems (DUSBL-APS). This is composed of an external transponder
Engineering, Department of Ocean and Mechanical Engineering, Florida At- array, and a receiver array with onboard electronics processing.
lantic University, Dania Beach, FL 33004 USA (e-mail:, [email protected];
[email protected]). The DUSBL-APS is essential for docking the AUV, as it en-
Digital Object Identifier 10.1109/JOE.2018.2867988 ables REMUS 100 to determine the position of the USV on the
0364-9059 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE JOURNAL OF OCEANIC ENGINEERING

surface. The receiver array is mounted on the nose of AUV and it

consists of a single transceiver configuration composed of four
transducers or acoustic microphones. To complete the DUSBL-
APS, a transponder needs to be mounted on the target dock. The
DUSBL-APS on REMUS 100 is based on estimating the dis-
tance between the target, where a transponder is mounted, and
the transceiver on the AUV, and the time it takes for a specific
acoustic signal from the source to reach the target. In particu-
lar, an acoustic pulse is transmitted by the AUV and when the
transponder detects the signal, it responds with its own acoustic
pulse. When this return pulse is detected by the transceiver on
the AUV, the DUSBL-APS is able to convert the time of trans-
mission into a range, representative of the distance between the
AUV and the USV line on which the transponder is mounted.
In addition, the four transducers on the transceiver are able to
measure the vertical and horizontal bearings, based on the direc-
tion from which the acoustic signal reached them. The system Fig. 1. WAM-V 16 USV during on-water tests in the Intracoastal Waterway.
on the AUV can interrogate the transponder every three sec-
onds. The time delay necessary for the signal to loop is range TABLE II
dependent. The major advantage of the DUSBL-APS is its ac- PRINCIPAL CHARACTERISTICS OF THE WAM-V 16
curacy in the estimates, which enables REMUS 100 to reach the
desired target within 1 m. Numerous sea trials have been car-
ried out to evaluate the performance of the DUSBL-APS. These
results are described in details in [2] and summarized in [3].
Although the ultrashort baseline (USBL) performance matched
the system specifications, sea trials highlighted some major lim-
itations of the system. The main limitation of the DUSBL-APS
is that REMUS 100 only utilizes its USBL system as a sec-
ondary instrument to correct a predetermined cross target path.
The DUSBL-APS on REMUS 100 therefore does not serve as
the main sensing device to determine the vehicle’s path. For this
reason, if the AUV is programmed to reach a target location that
is not within 25 m of the acoustic estimate, the information of
the DUSBL-APS is assumed to be erroneous and is ignored by
the main guidance navigation and control (GNC) system on the
AUV. In such a case, the AUV simply dead reckons to the prede-
termined target location without any aid from the DUSBL-APS,
leading to large docking errors [2]. This limitation implies that
in order for the DUSBL-APS to be utilized for L&R from a USV,
the docking target needs to be stationary as the AUV performs
its docking maneuver. Another limitation of the DUSBL-APS capability, a tilt-compensated digital compass, a RF transceiver,
on REMUS 100 is its inability to consistently detect the acoustic a pulse with modulation signal generator, and a custom-built
response, when the vehicle is located further than 50 m from the printed circuit board for power distribution and communica-
transponder. This can become a major problem if the vehicle is tions between the computer and instrumentation. A detailed
operating in an environment heavily disturbed by currents and description of this GNC system can be found in [5]. For the
waves, which can compromise the dead reckoning navigation purpose of this project, the key components of the GNC system
of the AUV. are the sensor suite (IMU/GPS and digital compass), computer,
The WAM-V 16 (see Fig. 1) is a twin hull, pontoon style and RF transceiver. The IMU/GPS is an Xsens MTi-G sensor,
USV. The vessel structure consists of two inflatable pontoons, a which is used to estimate the position and orientation of the
payload tray connected to the pontoons by two supporting arches USV during operations. The GPS is wide area augmentation
and a suspension system. This USV is designed to mitigate the system (WAAS) enabled and can provide up to 1 m accuracy
heave, pitch, and roll response of the payload tray when the in both latitude and longitude, depending on cloud cover and
vehicle operates in waves. The vehicle’s physical characteristics satellite availability. The digital compass is used to monitor the
are given in Table II. vehicle heading and has a resolution of 0.1◦ . The lightweight
A modular GNC system was developed to enable USV’s au- communication and marshalling (LCM) system [6] is utilized as
tonomy [4]. The GNC system is housed in a plastic, waterproof the underlying architecture for the GNC software. Sensor data
box, which contains a single board computer, an IMU with GPS was transmitted using drivers incorporating the LCM system to
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

SARDA AND DHANAK: LAUNCH AND RECOVERY OF AN AUV FROM A STATION-KEEPING UNMANNED SURFACE VEHICLE 3

Fig. 2. LARS design.

the control architecture, and logged at 4 Hz. A handheld remote may be station-keeping or mobile during launch, depending on
control is used for operating the thrusters and the actuators in the operating conditions. In calm conditions, the USV keeping
manual mode and for initiating autonomous operations. station until the AUV is launched may be desirable, while in
An AIRMAR Weather Station 100WX was installed on the the presence of waves, wind, and currents, a mobile USV, con-
WAM-V 16. This ultrasonic anemometer is used to measure the trolled by the adaptive heading and speed controller presented
apparent wind speed and direction at a sample rate of 1 Hz. The in [10] better ensures that the AUV is safely clear downstream
dynamic range of the anemometer is 0–40 m/s with a resolution of the USV once launched. Furthermore, launching the AUV
of 0.1 m/s. The anemometer is located in an elevated position from a station-keeping USV may be a necessity, if the operating
at the aft end of the payload tray to avoid the effects of wind environment does not guarantee adequate space for the surge
blockage and interference from other structures. Owing to the motion of the USV, which is required for a mobile launch. Re-
relatively small size of the USV, it is assumed that the wind speed covery is initiated through requiring the AUV to navigate toward
and direction measured by a single point sensor is representative an acoustic homing beacon on a taut line suspended from the
of the wind flowing past the entire vessel. USV. The taut line serves as docking target and as a connect-
ing link between the two vehicles during L&R. Throughout the
entire operation, the USV approximately keeps station on the
B. Launch and Recovery System (LARS)
water surface. Once the docking is complete, the USV moves
An automated LARS has been designed and implemented forward towing the AUV as it is extracted onboard the USV via
to accommodate transporting a REMUS100 AUV onboard a a customized winch mechanism. During recovery, requiring the
WAM-V 16 USV. A flexible line was mounted on the USV to USV to station-keep may be desirable under certain operating
provide a connection tether to the AUV during L&R. This line conditions to maintain more stable communication between the
can be extended into the water during recovery or stay retracted two vehicles. Station-keeping of the USV also facilitates im-
on board the USV otherwise. A weighted hydrofoil at the end plementation of a simplified control scheme for identifying a
of the line is used to keep the line taut underwater. The latch desired state for recovery and for driving both systems toward
on the AUV is necessary to allow the connection to take place. this desired state through acoustic positioning.
A winch is necessary to allow the line assembly to be moved Since this LARS enables two alternatives, using either a mo-
in and out of the water. A second winch is used to move the bile or station-keeping USV to launch and recover an AUV,
carriage structure in and out of the water surface, to deploy the an algorithm that can formulate an intelligent decision on an
AUV during launch, and to extract it onboard the USV during optimal method for L&R based on sensor data and simula-
recovery. The set up for this type of system is shown in Fig. 2. tion estimates was also developed and implemented on the
The possibility of automated L&R of an AUV from a USV systems. More precisely, data collected during in-water field
platform has been explored previously through simulations [7], testing of the station-keeping controllers described in [4] and
[8] and subsystem testing [9]. This capability has been used to [5] is utilized to train a predictive model for position and head-
enable the USV to deploy or intercept the AUV, while both ve- ing error of the USV. Based on these predictions, the USV
hicles are in motion. An alternative method, which is described can then dictate the methodology for L&R that maximizes
here, is letting the USV station-keep, while the AUV is launched the probability of success. While the effort of this algorithm
and subsequently recovered. In either case, launching is accom- was essential to test these two alternatives for L&R, the focus
plished through lowering and subsequently releasing the AUV of this paper is on the sea trials of L&R of an AUV from a
from the top tray of the USV onto the water surface. The USV station-keeping USV.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE JOURNAL OF OCEANIC ENGINEERING

II. BACKGROUND proved by adding a vision system on either the dock or the
In the past decade, much research has been devoted to- vehicle itself. In [26], the use of an electromagnetic field as
ward AUV L&R [11], [12]. Most previous studies suggested alternative to acoustics is explored, for locating the docking sta-
that the most effective way to overcome the challenge of re- tion underwater. The use of a vision system to recover an AUV
covering an AUV autonomously is by letting it home to a at the surface from a USV is discussed in [27]. Specifically, a
docking station, which may be anchored to the seafloor [13]– hybrid coordinated maneuver for docking a USV on a torpedo
[16] or towed by the mother vessel [17], [18]. In addition, sev- shaped AUV is introduced. This controlled maneuver is formu-
eral previous works provide manned options to overcome the lated based on visual information to estimate the AUV position
challenge [19]–[21]. and attitude relative to the USV, with the intent of guiding the
In [22], the main aspects that need to be considered in de- USV to dock on the free-floating AUV at the surface. The ap-
signing a device for launching an AUV from a USV are briefly proach here is to let the AUV complete its mission by sending
explained. This launching system consists of a USV mounted its GPS position at the surface to the USV. The USV would
cradle that carries REMUS 100. Upon command, a servomech- then drive to that location and attempt to localize the AUV via
anism lowers the aft end of the cradle, and the AUV slides vision. The hybrid maneuver is then formulated and performed,
off into the water. Once in the water, REMUS 100 waits for enabling the USV to dock onto the AUV. Docking on the sea
a command to begin its mission. Overall, this method repre- surface can be challenging in the presence of waves.
sents a very robust and secure way to deploy an AUV, but the It is evident that none of the systems described in this lit-
proposed design does not allow for AUV recovery from the erature review can be utilized to launch and recover REMUS
same USV. 100 from the WAM-V 16. However, the difficulties encoun-
Another approach to launch an AUV is described in [20]. tered in the manned L&R of an AUV are somewhat similar to
The intent of this project was to allow one man to easily and the autonomous process and were given appropriate attention.
safely launch and recover a Gavia class AUV from a small The same can be said about the similarity between designing a
inflatable boat. The LARS developed for this project was not docking station and implementing a LARS.
autonomous, but it required minimum human interaction. The
final design selected for this purpose consisted of a crane that
III. L&R SEA TRIALS AND RESULTS
allowed launching the AUV from either side of the boat. A
combination of pulley wheels and a winch were used to lower The novel approach to L&R an AUV consists of utilizing a
the AUV into the water. station-keeping USV to run operations. Testing, via sea trials,
Hydroid, a Kongsberg company, has developed a mobile dock of L&R of an AUV from a USV, utilizing the custom designed
for REMUS 100 [18], as evolution of their previous designs [23], LARS shown in Fig. 2, were conducted in closed waters at North
[24]. The idea was to allow REMUS 100 to intercept and mate and South Lake, Hollywood, FL, USA. These tests were neces-
with a submerged towed docking cone, using USBL navigation sary to validate the theoretical analysis presented in [7], utilizing
to refine the position of the moving dock underwater. This mo- the subsystems developed for this research and introduced in [3],
bile docking system, consisting of a depressor wing, a capture [9], [10], and [28]. The results of the sea trials presented in this
system, and a transponder, was designed with the intent of being section provide verification that a station-keeping USV can be
towed behind a manned surface vessel. A REMUS 100 AUV was utilized to L&R an AUV autonomously. Station-keeping was
programmed to approach the docking cone, which was towed accomplished utilizing the station-keeping controller presented
at approximately 1 m/s. The docking cone then locked onto the in [28] and [29] on the USV and the USBL system presented in
vehicle and, once sensed, the surface craft was instructed to [3] and [9] on the AUV.
recover the AUV and dock assembly. The progress made by Before the initiation of the any of the tests, the USV was
Hydroid represented a major advance toward unmanned AUV commanded to autonomously station-keep at different headings
recovery from a mobile manned platform. However, the tow- within the area where the sea trials were conducted. Doing so
able station proposed by Hydroid [24] still represents a type of allowed the system to acquire knowledge about the operating
docking device that may not always be ideal for L&R. Towing environment, enabling it to autonomously deduce the most ap-
an entire docking station affects the performance of the USV, propriate approach for L&R: mobile versus station-keeping. If
in terms of increased payload and drag. Small USVs, such as the high level planner on the USV deduced that the ideal method
the WAM-V 16, have limited payload capacity and can only for L&R was station-keeping, a specific protocol was initiated
output limited thrust. In addition, the physical size of a docking to begin the desired autonomous operation. L&R were tested
station is usually too large to be carried on board a USV, mean- independently and their results are described in Section III-A
ing that it would have to be dragged during the entire mission. and III-B, respectively.
Towing a system also puts limitation on operating in congested
environments.
Improvements to the standard designs for LARS or dock- A. Autonomous Launch of an AUV From
ing station have been proposed throughout the years [25], [26]. a Station-Keeping USV
Specifically, Park et al. [25] demonstrates that the precision at For this test, the AUV was mounted on the LARS on board
which an AUV approaches the dock can be dramatically im- the USV, as shown in Fig. 3.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

SARDA AND DHANAK: LAUNCH AND RECOVERY OF AN AUV FROM A STATION-KEEPING UNMANNED SURFACE VEHICLE 5

ployment, highlights that the external disturbance was almost

irrelevant. As a result, the station-keeping controller was able to
perform very well and REMUS 100 was able to settle steadily
right in-between the USV’s pontoons without colliding. The ex-
cellent performance of the station-keeping controller can also
be observed in Fig. 7, which shows the vehicles’ heading during
the sea trials. The USV’s heading during launch is in fact kept
within 10◦ of its desired value, while the AUV’s remains almost
constant after deployment.
The results from the launch test confirmed that in the presence
of disturbance, it becomes very difficult to effectively launch the
AUV with the USV station-keeping. This is because, once the
Fig. 3. WAM-V 16 transporting REMUS 100 to launch site. AUV is deployed, there is no way to control it until it starts its
mission.
TABLE III B. Automated AUV Docking to a Station-Keeping USV
MEAN AND STANDARD DEVIATION OF APPARENT WIND SPEED, AND RELATIVE
WIND DIRECTION DURING LAUNCH The docking test started with the vehicles separated in the
water, as shown in Fig. 8. Since the REMUS 100 pressure vessel
developed as part of the LARS, as shown in Fig. 2, had not been
implemented at the time the sea trials were conducted, a pair
of brackets was used in place to enable the AUV to establish a
connection with the docking line underwater.
To quantify the disturbance sensed by the USV during the
sea trials, mean and standard deviation of wind speed and wind
angle of attack, over the day during which recovery was tested,
are given in Table V (recovery duration is 50 000 s).
To quantify the disturbance sensed by the USV during the The estimates and the subsequent decisions generated by the
sea trials, mean and standard deviation of wind speed and angle high-level planner on the USV prior to starting the sea trials are
of attack, over 140-s duration launching operation, are given in given in Table VI. These included: ideal heading for recovering
Table III. the AUV ψi , expected average station-keeping position error  ēr ,
The estimates and the subsequent decisions generated by and heading error  ēψ for the USV during launch, ideal station-
the high level planner on the USV prior to starting the sea keeping controller and preferred method for recovery.
trials are given in Table II. These included: ideal heading for Once autonomous mode was initiated, the USV was com-
launching the AUV ψi , expected average station-keeping posi- manded by the high-level system to station-keep at a specific
tion error ēr , and heading error 
ēψ for the USV during launch, position (see HOME in Fig. 9) with the deduced ideal heading.
ideal station-keeping controller and preferred method for The depressor wing, docking line, and transponder assembly
launch. (see Fig. 2) were then lowered into the water at a depth of 4 m
Once autonomous mode was initiated, the USV was com- by the USV. The transponder on the docking line periodically
manded by the high-level system to station-keep at a specific po- transmitted acoustic pings, constituting a homing beacon. In the
sition, with the ideal heading identified. After the USV reached meantime, the AUV was commanded to attempt docking by fol-
steady state, the launching routine was started within the LARS. lowing the preprogrammed path shown in Fig. 9, at a constant
After a safety check, the AUV was therefore lowered into the speed of 1 m/s, and altitude of −1 m.
water and held at the surface for a few seconds [see Fig. 4(a)], For this test, the AUV was manually launched at the deploy-
allowing for a fault check, and finally released [see Fig. 4(b)]. ment site marked by the green buoy in Fig. 9. The AUV initiated
A few seconds after release, the AUV carriage structure was re- its docking routine by acquiring a GPS fix, using it as a refer-
tracted and the USV commanded to leave the deployment site, ence point to dead reckon to waypoint 1 (WP1) in Fig. 9. When
leaving the AUV behind [see Fig. 4(c)]. the AUV reached WP1, more accurate navigation was enabled
To easily identify the time at which the AUV was launched, through engagement of the DUSBL-APS, which monitors and
the USV pitch angle θW is plotted in Fig. 5. A 5◦ offset can in homes in on the acoustic pings transmitted by the transpon-
fact be spotted at the instant when the AUV is deployed. der mounted on the docking line connected to the USV [see
In the controlled environment with low winds (< 2 m/s) and Fig. 10(a)]. While the AUV navigated, the USV kept station
no waves, REMUS 100 was effectively launched from the maintaining heading and position on the surface at HOME in
station-keeping WAM-V 16, without running in any failure. Fig. 9. Using the DUSBL-APS, REMUS 100 was therefore able
x, y, and z position of the systems during launch are shown to adjust its trajectory, correcting all accumulated errors and fi-
in Fig. 6. As it can be seen, the AUV always sits within 1 m nally home to the docking line connected to the WAM-V 16 [see
of the USV, even after it is deployed. The z position of the Fig. 10(b)]. The docking attempt ended when the AUV reached
AUV, which is maintained almost constant before and after de- HOME in Fig. 9, thus rising to the water surface in-between
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE JOURNAL OF OCEANIC ENGINEERING

Fig. 4. Snapshots from station-keeping launch. (a) AUV is lowered in the water. (b) AUV is deployed. (c) AUV carriage structure is retrieved.

Fig. 7. Heading (◦ ) during launch.

Fig. 5. WAM-V 16 pitch angle (◦ ) at launch.

Fig. 8. WAM-V 16 station-keeping with depressor wing assembly deployed,

and REMUS 100 equipped with docking brackets, attempting to dock to the
Fig. 6. x position (m), y position (m), and z position (m) during launch. line underwater.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

SARDA AND DHANAK: LAUNCH AND RECOVERY OF AN AUV FROM A STATION-KEEPING UNMANNED SURFACE VEHICLE 7

TABLE IV the USV’s pontoons, connected to the docking line and ready to
STATION-KEEPING ERROR ESTIMATES, PREFERRED STATION-KEEPING
CONTROLLER, AND OPERATING METHOD
be extracted from the water. For a successful run, REMUS 100
needed to be connected to the line [see Fig. 10(c) and (d)]. The
missed connection between the two vehicles represented a failed
attempt. A sequence of eight runs were completed, leading to 3
successful docks and 5 failed attempts.
During each run, REMUS 100 attempted to dock to a station-
keeping WAM-V 16 as described before. The path of the AUV
for each run, and the position of the station-keeping USV
recorded throughout the entire day are shown in Fig. 11. As
can be seen from the figure, REMUS 100 consistently followed
the preprogrammed path (see Fig. 9). In addition, the WAM-V
16 successfully maintained position over the course of the sea
TABLE V trials. Furthermore, the AUV achieved almost the same heading
MEAN AND STANDARD DEVIATION OF APPARENT WIND SPEED, AND RELATIVE as the USV anytime it attempted to dock, as can be seen from
ANGLE OF ATTACK DURING RECOVERY
Fig. 12.
The recorded surge velocity of the AUV throughout each
run shown in Fig. 13 also confirms that the AUV performed as
expected, reaching the desired speed of 1 m/s and maintaining
it until it reached WP1. Since the vehicle is then forced to
make a steep turn of more than 90◦ , its surge velocity drops
momentarily to ∼0.6 m/s. After that, the AUV picks up speed
again, reaching a steady speed of 0.9 m/s, which is less than the
desired value. This is because when the DUSBL-APS engaged at
TABLE VI
STATION-KEEPING ERRORS ESTIMATES, PREFERRED STATION-KEEPING
WP1, it resulted in the control effort on the AUV being devoted
CONTROLLER, AND OPERATING METHOD DEDUCED FOR RECOVERY to making heading and pitch adjustments at the cost of the speed.
Overall, the speed controller performance of the AUV can be
considered sufficient for purpose of this test; however, the same
cannot be said about the AUV altitude control. Fig. 14 shows
the depth of the AUV throughout each run. The goal was for
the AUV to dock at z = −1 m under water. In Fig. 14, it can
be seen that the AUV does not maintain the desired depth of
−1 m. Thus, the depth at which the AUV attempts to dock is
very unpredictable, since it constantly oscillates between −0.5
and −1.5 m. Unfortunately, the poor controller performance
cannot be compensated in anyway, since the control system
on the REMUS 100 AUV is not accessible by the user, and it
cannot be modified. The inability of REMUS 100 to maintain
a steady depth when docking is the likely reason why the AUV
was not able to consistently dock. Apart from the deficiency in
maintaining steady depth zR and correspondingly a steady pitch
angle θR (see Fig. 15), the AUV performance was consistent
in all runs for all other parameters. Since a few differences are
indefinable between successful and unsuccessful runs, two plots
are shown: RUNs 1, 5, 6, 7, and 8 where the vehicle did not dock
(see Fig. 15, top subplot) and RUN 2, 3, and 4 where REMUS
100 successfully docked (see Fig. 15, bottom subplot).
As can be seen, the AUV pitch angle variations are similar
throughout each run, with deep troughs at the start and at the
end. While the troughs at the beginning are associated with the
diving motion of the AUV, the final troughs were unexpected;
for the runs when the docking did not occur, the final AUV
pitch angle just before reaching the docking line was ∼ −28◦
Fig. 9. AUV predefined path for recovery. The deployment site was marked and may have been the cause for the AUV not docking. This is
with a green buoy, WP1 was marked with a red buoy, and the AUV homing because the significant negative pitch of the AUV at the end of
location (HOME) was marked with a yellow buoy.
the run creates a scenario where the vehicle attempts to establish
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE JOURNAL OF OCEANIC ENGINEERING

Fig. 10. Snapshots from docking tests. (a) Line and transponder assembly viewed from GoPro mounted on the depressor wing underwater. (b) REMUS 100
docking to the line underwater. (c) and (d) REMUS 100 docked.

Fig. 11. Vehicles’ path (m) during docking. The AUV trajectories for at-
tempted docking Runs 1–8 are shown; the station-keeping USV position is Fig. 13. REMUS 100 surge velocity (m/s) during each run of recovery.
marked with an “x.”

Fig. 12. Heading (◦ ) during each run of recovery. Fig. 14. REMUS 100 z position (m) during recovery.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

SARDA AND DHANAK: LAUNCH AND RECOVERY OF AN AUV FROM A STATION-KEEPING UNMANNED SURFACE VEHICLE 9

ery pose for the USV, where the AUV will be recovered.
Furthermore, communication takes place only after the
vehicles are physically connected or, in case of failure,
to initiate new recovery attempts. In the case of mobile
recovery, communication between the vehicles is a major
bottleneck, since the two vehicles need to continuously
share their state in real time. This is extremely difficult,
since the USBL performance is dependent on the distance
between the vehicles, the water depth, and the proximity
to the sea bottom.
4) L&R of an AUV from a station-keeping USV also en-
ables the operation to be carried out over a confined re-
gion, which may be important when space for docking
maneuvers is limited.
5) The presence of waves during launch can compromise the
operations when utilizing a station-keeping USV as plat-
form for L&R. The alternative approach [2] involving the
Fig. 15. REMUS 100 pitch angle (◦ ) during recovery. USV in motion during the launching process is preferred
under such operating conditions.
6) Presence of cross currents can compromise launching op-
a connection with the docking line at an angle. It is very probable erations when utilizing a station-keeping USV as platform
that the changes in pitch angle at the end of the run are caused by for L&R. Currents aligned at an angle to the vehicle can
the AUV attempting to rapidly correct for the error in its depth induce rotation of the AUV after it is deployed, possibly
from the desired value. The runs during which REMUS 100 was forcing it to collide with the USV, if the latter is keeping
able to successfully dock are the ones that show minimal change station.
in pitch angle at the end of the run. 7) The USV needs to be in motion once the AUV docks,
towing the AUV behind it as it is extracted onto the USV.
8) Identifying an ideal vehicle heading under given environ-
IV. CONCLUSION mental conditions can dramatically improve the perfor-
At sea testing, and demonstration of an automated L&R of mance of the low-level controllers, thereby reducing the
an AUV from a station-keeping USV have been carried out. A possibility of failure during L&R.
methodology for L&R of a REMUS 100 AUV from the WAM- The results of the sea trials presented here show that three out
V 16 has been identified. According to the approach considered of eight recovery attempts were successful, thus demonstrating
here, the USV station-keeps heading and position at the surface, that using a station-keeping USV to recover an AUV can be a
while the AUV is lowered in the water during launch, and as the strong alternative to the mobile approach. Sea trials [3], [10]
AUV approaches the USV to dock underwater during recovery. and simulations [8] of AUV recovery using a moving USV have
The docking system is composed of a line connected to an in fact shown that the recovery task can be compromised if each
acoustic transponder beacon, and a depressor wing that keeps individual subsystem (e.g., USBL, high level planner, trajectory
the assembly taut. tracking controller, and LARS) does not perform perfectly.
Automated L&R sea trials of an AUV, using a station-keeping Since two distinct methodologies for L&R can be carried out
USV, highlighted important aspects of the concept of operations. using the same systems, a new algorithm could be developed
Specifically, the results lead to six major conclusions. for making an appropriate selection based on the prevailing
1) L&R of an AUV from a station-keeping platform can environment. The sea trials described here were conducted in
be a useful alternative to a mobile approach, when the a protected area in the absence of any significant waves. Ad-
operating conditions allow it. Furthermore, cutting down ditional data, representative of more adverse ocean conditions
the number of autonomous systems in motions from two would improve the L&R capability as well as the performance
to one, the architecture scheme is simplified, thus reducing of low-level controllers under such conditions. In addition, use
room for error. of the latching system on the AUV shown in Fig. 2 would signifi-
2) A complex motion planner is no longer needed for the cantly improve the success rate for docking. Finally, the vehicles
USV, since the vehicle simply needs to station-keep at a can be programmed to reattempt the task, whenever the AUV
desired heading and position at the point of AUV recovery. misses the docking line.
The AUV is therefore the only vehicle required to plan and
follow a trajectory for homing onto the docking line. REFERENCES
3) Reliable communication at a defined update rate, which is
[1] F. J. Ruud, “Autonomous homing and docking of AUV REMUS 100 -
very difficult to achieve via acoustics, is no longer crucial, Homing and docking guidance algorithm and relative localization,” M.S.
since only one single message is necessary to initiate the thesis, Dept. Mar. Eng., Norwegian Univ. Sci. Technol., Trondheim, Nor-
recovery routine. This simply consists of a desired recov- way, 2016.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

10 IEEE JOURNAL OF OCEANIC ENGINEERING

[2] M. I. Miranda, “Mobile docking of REMUS-100 equipped with USBL- [23] R. Stokey et al., “A docking system for REMUS, an autonomous under-
APS to an unmanned surface vehicle: A performance feasibility study,” water vehicle,” in Proc. MTS/IEEE Oceans Conf., Halifax, NS, Canada,
M.S. thesis, College Eng. Comput. Sci., Florida Atlantic Univ., Boca 1997, pp. 1132–1136.
Raton, FL, USA, 2014. [24] B. Allen et al., “Autonomous docking demonstrations with enhanced
[3] M. I. Miranda, P.-P. Beaujean, E. An, and M. Dhanak, “Homing an un- REMUS technology,” in Proc. OCEANS Conf., Boston, MA, USA, 2006,
manned underwater vehicle equipped with DUSBL to an unmanned sur- doi: 10.1109/OCEANS.2006.306952.
face platform: A feasibility study,” in Proc. OCEANS Conf., San Diego, [25] J.-Y. Park, B.-H. Jun, K. Kim, P.-M. Lee, J.-H. Oh, and Y.-K.
CA, USA, 2013, doi: 10.23919/OCEANS.2013.6741227. Lim, “Improvement of vision guided underwater docking for small
[4] M. Caccia, M. Bibuli, R. Bono, and G. Bruzzone, “Basic navigation, AUV ISiMI,” in Proc. OCEANS Conf., Biloxi, MS, USA, 2009, doi:
guidance and control on an unmanned surface vehicle,” J. Auton. Robots, 10.23919/OCEANS.2009.5422241.
vol. 5, no. 4, pp. 349–365, 2008. [26] M. Feezor, Y. Sorrell, P. Blnkinship, and J. Bellingham, “Autonomous
[5] J. G. Marquardt, J. Alvarez, and K. D. von Ellenrieder, “Characterization underwater vehicle homing/docking via electromagnetic guidance,” J.
and system identification of an unmanned amphibious tracked vehicle,” Ocean. Eng., vol. 26, no. 4, pp. 515–521, 2001.
IEEE J. Ocean. Eng., vol. 39, no. 4, pp. 641–661, Oct. 2014. [27] A. Martins et al., “Autonomous surface vehicle docking manoeuvre with
[6] S. A. Huang, E. Olson, and D. C. Moore, “LCM: Lightweight commu- visual information,” in Proc. IEEE Int. Conf. Robot. Automat., Roma,
nications and marshalling,” in Proc. IEEE/RSJ Int. Conf. Intell. Robots Italy, 2007, pp. 4994–4999.
Syst., 2010, pp. 4057–4062. [28] E. I. Sarda, H. Qu, I. R. Bertaska, and K. D. von Ellenrieder, “Station-
[7] E. I. Sarda, M. D. Dhanak, and K. D. von Ellenrieder, “Concept for a USV- keeping control of an unmanned surface vehicle exposed to current and
based autonomous launch and recovery system,” in Proc. ASNE Launch wind disturbances,” Ocean Eng., vol. 127, pp. 305–324, 2016.
Recovery Symp., Linthicum, MD, USA, 2014, pp. 1–9. [29] E. I. Sarda, H. Qu, I. R. Bertaska, and K. D. von Ellenrieder, “Development
[8] E. I. Sarda and M. R. Dhanak, “A USV-based automated launch and of a USV station-keeping controller,” in Proc. OCEANS Conf., Genova,
recovery system for AUVs,” IEEE J. Ocean. Eng., vol. 42, no. 1, pp. 37– Italy, 2015, doi: 10.1109/OCEANS-Genova.2015.7271425.
56, Jan. 2017.
[9] E. I. Sarda and M. D. Dhanak, “Unmanned recovery of an AUV from a
surface platform,” in Proc. OCEANS Conf., San Diego, CA, USA, 2013,
doi: 10.23919/OCEANS.2013.6741386.
[10] W. B. Klinger, I. R. Bertaska, and K. D. von Ellenrieder, “Experimental
testing of an adaptive controller for USVs with uncertain displacement
and drag,” in Proc. Oceans Conf., St. John’s, NL, Canada, 2014, doi:
10.1109/OCEANS.2014.7003032. Edoardo I. Sarda received the B.S. degree in me-
[11] J. W. Nicholson and A. J. Healey, “The present state of autonomous chanical engineering from Lake Superior State Uni-
underwater vehicle (AUV) applications and technologies,” Mar. Technol. versity, Sault Ste. Marie, MI, USA, in 2012, and
Soc. J., vol. 42, no. 1, pp. 44–51, 2008. the Ph.D. degree in ocean engineering from Florida
[12] H. Singh et al., “Docking for an autonomous ocean sampling network,” Atlantic University, Dania Beach, FL, USA, in 2016.
IEEE J. Ocean. Eng., vol. 26, no. 4, pp. 498–514, Oct. 2001. He is currently an R&D Engineer in the field of col-
[13] S. Pai, P. Guerrini, J. Potter, A. Maguer, M. Chitre, and S. Biagini, “Au- laborative and mobile robotics. His research interests
tonomous initial capture system for AUV recovery,” 3rd Int. Conf. Exhi- include autonomous vehicles, launch and recovery,
bition Underwater Acoust. Measurement: Tech. Results, 2009, pp. 1–8. human robot collaboration, automated systems and
[14] J. Lambiotte, R. Coulson, S. Smith, and E. An, “Results from mechanical control.
docking tests of a morpheus class AUV with a dock designed for an OEX Dr. Sarda is a member of the IEEE Oceanic Engi-
class AUV,” in Proc. MTS/IEEE OCEANS Conf., Biloxi, MS, USA, 2002, neering Society.
pp. 260–265.
[15] L. A. Gish, “Design of an AUV recharging system,” M.S. thesis, Dept.
Ocean Eng., Massachusetts Inst. Technol., Cambridge, MA, USA, 2004.
[16] T. Zhao, “Launch & recovery of UUVs,” ATR. [Online]. Available:
www.atrcorp.com/landr_uuvs, Accessed on: Aug. 10, 2018.
[17] P. McGillivary, “Design considerations for launch and recovery of au-
tonomous systems from ships, including coast guard icebreakers,” in Proc. Manhar R. Dhanak received the B.Sc. degree (hon-
ASNE Launch Recovery, Arlington, VA, USA, 2010, pp. 1–22. ors) in mathematics from Imperial College, Univer-
[18] A. S. Kongsberg Maritime, “Underwater mobile docking of autonomous sity of London, London, U.K., in 1976, and the Ph.D.
underwater vehicles,” in Proc. OCEANS Conf., Virginia Beach, VA, USA, degree in applied mathematics from the University of
2012, pp. 1–15. London, London, U.K., in 1980.
[19] M. J. Doble, A. L. Forrest, P. Wadhams, and B. E. Laval, “Through-ice He is currently a Professor of Ocean Engineering
AUV deployment: Operational and technical experience from two seasons and the Director of the Institute for Ocean and Sys-
of arctic fieldwork,” Cold Reg. Sci. Technol., vol. 56, no. 2/3, pp. 90–97, tems Engineering (SeaTech), Florida Atlantic Uni-
2008. versity, Dania Beach, FL, USA, where he has been
[20] K. J. Kristinsson, “Launch and recovery of gavia AUV,” Reykjavik Uni- since 1990, having previously been a Senior Re-
versity, 2011. search Associate with the University of Cambridge,
[21] G. Lester, “Remus launch & recovery systems,” 15th Int. Symp. Untethered Cambridge, U.K. (1989–1990) and a Research Scientist with Topexpress Ltd.,
Submersible Tech., 2007, pp. 1–8. Cambridge, U.K.
[22] D. K. Freeman, “Remote delivery of unmanned system technologies,” Prof. Dhanak is an Associate Fellow of the American Institute of Aeronautics
Naval Surface Warfare Center, Panama City, FL, USA, 2005. and Astronautics.