You are on page 1of 6

Copyright @ IF AC Manoeuvring and Control of

Marine Craft, Aalborg, Denmark, 2000

THE NAVIGATION AND CONTROL SYSTEMS OF THE SWIMMER AUV SLED

J.c. Evans, J.S. Smith, K.M. KeIler

Department of Electrical Engineering & Electronics


The University of Liverpool, Liverpool L693GJ, United Kingdom
E-Mail.·evansjon@fiv.ac.uk

Abstract: The Swimmer Vehicle is a EU-Thennie funded project to develop an


autonomous vehicle capable of transporting existing work-class ROVs around sub-sea
installations - removing the need for surface support vessels and long umbilical cables.
As part of the Swimmer project, Liverpool has developed a SRPS (Short Range
Positioning System) for the autonomous Swimmer vehicle. The system does not use
deployed ultrasonic transponders, but instead uses profile data obtained from a high-
speed, narrow-beam imaging sonar. The system is based on ASOP (Active Sonar Object
Prediction), a novel technique developed at Liverpool. It uses CAD infonnation to derive
a 3D model (or map) of the Docking Station on the seabed. This paper gives an overview
of the Swimmer vehicle, and describes the technical details of the ASOP algorithm used
for Swimmer's Short Range Positioning System. Copyright @20001FAC

Keywords: Navigation systems, Robot navigation, Virtual reality, Intelligent Control,


Artificial Intelligence, Image analysis

consequent improvement in vehicle dynamics


I. INTRODUCTION and energy usage.
• Once docked, the work-class ROV can be
1.1. The "Swimmer" Concept operated completely conventionally via a
sub-sea fibre optic communications network
As offshore exploration and production moves to and short (-lOOs metres) umbilical. This
increasingly greater water depths, the greater the allows the ROV pilot to be stationed ashore.
reliance on remote intervention using ROVs
(Remotely Operated Vehicles). This means that any
technology that can reduce the cost of deploying and
using such vehicles could produce considerable
savings.
The aim of EU-Thennie funded Swimmer project is
to construct a prototype AUV (Autonomous
Underwater vehicle) capable of carrying a standard
work-class ROV to a seabed installation, in effect an
autonomous deployment platfonn. This would be
three immediately benefits:
• Remove the need for a pennanent support
vessel, with subsequent cost savings. Fig. I: Computerised Mock-up of the Swimmer
• Remove the long tether / umbilical with Sled and Payload Vehicle on Final Approach

51
1.2. Typical Mission Profile 2. VEHICLE SUB-SYSTEMS
A typical mission scenario might be as follows. The
2.1. Overview
Swimmer AUV is to be deployed from an FPSO
(Floating Production Storage Offioading) vessel. The The Swimmer vehicle is designed to carry a work-
work-class ROV is mounted on the Swimmer AUV. class ROV as a payload. This is achieved by
The ROV has a short 200m umbilical that is wound mounting the Swimmer as a tool-skid under the
round a self-contained winch mounted on the ROV. The Swimmer has computer-controlled clamps
Swimmer. The ROV is un-powered for the duration that allow the ROV and the Swimmer to be
of the transit to the seabed. The Swimmer AUV is disengaged remotely while on the seabed. While
tetherless and powered from onboard batteries. The docked, power and communications are fed from the
combined Swimmer / ROV vehicle is launched using seabed network to allow the Swimmer to recharge
a crane on the FPSO. A low-speed acoustic data link and the ROV to be remotely controlled. This is
is established to the Swimmer, final system checks shown schematically in Fig. 2.
are done and the command to begin the deployment
mission is sent via the acoustic link. Short Umbilical
(- lOOm)
During the Transit Phase, the Swimmer uses its on winch
LRPS (Long Range Positioning System - based on a
Long Base Line network) to fix its position, and
autonomously decide on its heading, speed and dive
rate. Periodically, navigational fixes are sent via the Electrical Power
acoustic data link (part of the LRPS system). The ' - - - - - - - - - - - & Fibre Optic
Comms
depth of dive could be as high as IOOOm to 3000m,
combined with a horizontal displacement of several
kilometres. Fig. 2: Schematic of Swimmer (with payload
ROV) Docked to the Sea bed Station
The Approach Phase commences when the vehicle is
50m above the Docking Station and within IOm
radius. Based on a novel technique developed at
Liverpool, the SRPS (Short Range Positioning Other Minor
System) uses CAD information to derive a 3D map Sub-Systems
(e.g . Fail-5afi:R£:c:ovcr)")
(profile) of the Docking Station on the seabed and
automatically manoeuvres the vehicle from 50m to
within O.2m. The close quarters approach is
illustrated in Fig. 1. MCU
Mission Command Unit
The final alignment in the Docking Phase is aided by
the mechanical design of the Docking Station. The
sloping members on the top of the rig help the
Swimmer align correctly and allow the wet
connectors (carrying power and communications) to
mate.
Once docked, the Swimmer AUV is powered down
(the onboard batteries are recharged). The work-class Fig. 3: The major Swimmer AUV sub-systems
ROV is now connected, via the seabed fibre optic
communication network, to the pilot ashore Internally, the Swimmer AUV has three major sub-
somewhere in the world. The ROV can be detached systems (see Fig. 3): the MCU, LRPS and SRPS. In
from the Swimmer, and operated around the seabed addition, there are some other minor systems (such as
installation completely conventionally. Thruster Control, ROV Tether Control, and Fail Safe
Recovery). To aid modular, distributed development
The ROV could stay deployed for months without
of the various sub-systems, interaction between the
further need of a surface support vessel. The
different components is governed by a detailed
Swimmer "concept" envisages a network of Docking
interface specification. This will also aid future
Station nodes fitted across a whole production field.
development and integration of the various
If the ROV is required elsewhere, the ROV is docked
technologies with other systems. The SRPS has been
back onto the Swimmer and powered down. The
developed at The University of Liverpool. The LRPS
Swimmer is then given commands to make its way
system has been developed by Ifremer, and the MCU
autonomously to the Docking Station nearest the
(together with overall project management, vehicle
work site (using the LRPS and SRPS), recharging at
design and manufacturing) by Cybernetix. Other
immediate stations as necessary.
partners providing expertise into the consortium are
Noordhoek, Total and GERTH.

52
2.2. LRPS: Long Range Positioning System He then looks out into the world in the direction
predicted by the map of the first landmark (say,
The LRPS (developed by Ifremer) is an integrated
the mountain summit). The summit is in partial
system that produces periodical position fixes (at
cloud (i.e. the signal is noisy), and nonnallY it
I Hz) for the MCU. It uses a combination of dead-
would be impossible to tell the mountains from
reckoning (based on data coming from a doppler
each other. However, since the map predicts a
log), and Long Base-Line (LBL) infonnation
particular mountain at that heading and range, if
(refreshed every 45 secs) provided by acoustic
he can observe any mountain like shape at that
modules. The LBL system uses a CHIRP modulation
location it is highly likely that it is the correct
scheme, to offer an improved distance / precision
one. By repeating this process with the other
ratio and SNR.
landmarks, it is possible to triangulate his
The acoustic telemetry link is used to provide a low position, confmning and improving his initial
bitrate data / position uplink, and vehicle command deck-reckoned estimate based upon his inertial
down link. It uses the same hardware as the LRPS, navigation.
although functionally it is considered separate.
A system diagram of the SRPS is shown in Fig. 4. An
The Swimmer LRPS has been developed from a estimate of current position is provided by the MCU
similar (but not as advanced) system tested on the (from LRPS sub-system). The SRPS uses this
Ifremer SIRENE AUV as part of the EU-MAST infonnation to generate an index into the 3D
DESIBEL project. On the Swimmer AUV, it plays an environmental simulation and its database (map). The
essential role in the surface to seabed transit, and system then generates a predicted set of sonar
provides important feedback and monitoring data to features ("the landmarks"), and then actively seeks
the surface. out evidence for these objects in the sonar data. The
SRPS's estimate of the current location is returned to
2.3. ASap: Active Sonar Object Prediction the MCU, together with a confidence measure (based
of the triangulation geometry and the size of the
The Swimmer's SRPS system uses a novel sonar sonar search field).
object tracking technique called ASOP (Active Sonar
Object Prediction). The system works by using a 3D
map / model (derived from existing CAD SRPS
infonnation), and uses this to drive an active search INS Nay Fix
for significant sonar features ("landmarks") in the
live sonar data. The system was originally developed
as part of the TUUV (Technology Unmanned Monitor of
Underwater Vehicle) programme (Lane, et aI., 1998; -- debug interface
Lane, et aI., 2000).
The "active" or "attentive" approach is in contrast to 3D environment
the more conventional method of processing the simulation
image (or signal) to identify edges and boundaries. (dVise)
The eventual aim is to classify the detected shapes
against a database of known objects (Smith, 1999; Fig. 4: SRPS system diagram
Smith 1994; Ashraf 1994). This "image analysis"
approach is always difficult, and in ultra-noisy sonar The whole system can be monitored and controlled
images perhaps impossible. via a fast Ethernet interface. The SRPS is PC-based
running embedded WinNT. This provides a very
To understand the approach, it is useful to consider
powerful interface to the system, and allows virtually
the following illustrative example:
every parameter to be monitored, logged and altered
Imagine that someone is going walking in the (even when the system is live in the water). This
countryside. The area is unknown to him, but he includes a real-time Virtual Reality (VR) display of
does have a detailed map (in effect a model) of the current location (in effect a virtual camera
the environment. He parks the car at a point image). The 3D simulation is provided by using
clearly marked on the map. He decides on his certain backend software components from a
direction and he starts walking. After I 0 minutes, commercial VR software package (dVISE), together
he wishes to calculate his present position. with custom written code.
From his initial general heading and using his Although the SRPS specification allows for two dual
average walking speed as a guide, he can estimate frequency Tritech Profiling Sonar heads (arranged at
from the map the area he is in. Choosing the 90° to each other), the possibility of using a single
centre of this area as the most likely position, he head is still being investigated.
uses the map to predict the hearing and range of
landmarks in his vicinity. For example, a
mountain peak, the edge of a forest and bend in a
river.

53
3. SYSTEM TRIALS 3.2. SRPS Dock Trials
The section concentrates of the development work The next stage of development involved integrating
and trials undertaken by Liverpool on the Short the SRPS system onto a small inspection ROV . This
Range Positioning System (SRPS). A short allowed the system to be evaluated on a "mini"
description of the fmal Swimmer AUV deep water Swimmer vehicle (the main vehicle is still in under
trials is included at the end. construction). The system has already been fitted on
a Hyball ROV (from Hydrovision).
3.1. Tank Trials
The trials were conducted in an outside, semi-
The initial trials on the system were conducted using protected salt-water dock (roughly ISm by ISm by
a motion control rig (a large Cartesian, S-degrees-of- Srn), with stone walls and a mud / sand floor. The VR
freedom robot) mounted above an indoor tank image of the dock structure is shown in Fig. 8.
measuring 4m x 3m x 2m. Due to the restricted size Several hours of synchronised sonar and telemetry
of the tank, it was decided to scale the whole have been captured, and this is currently been used to
environment by 1/6. This created an environment reduce the effects of motion blur on the sonar data.
much closer to the nominal 20m diameter x SOm
cylindrical operating zone of the SRPS specification.
The 1/6 scale model of the Swimmer Docking
Station was placed in the centre of the tank. A series
of synchronised sonar-motion data sets were
collected, at varying speeds and approach angles. The
aim was to collect a series of representative data sets,
which could be used repeatedly to improve the
system 's performance.
For these initial tests, a single Tritech Dual
Frequency Profiling Sonar head was used operating
at 1.2 MHz. It was mounted on the robot and
arranged to scan downward creating 60°-120° fans .
Fig. S shows the VR display output for the SRPS
system for a typical scene (the vehicle's image is Fig. 5: VR display output (used for monitoring)
placed at the sonar's nominal position). In this scene, from the Short Range Positioning System
the vehicle is in fmal approach above the Docking
Station. The transparent fan shape object represents
the sonar swept volume, and is predicting a certain
pattern of interaction with the (blue) structure of the
station.
Fig. 6 shows the actually sonar data collected for the
same scene. The position and orientation of the
diagonal cross-members in the station are clearly
visible in the image. It is interesting to note some of
the internal structure can also be made out. However,
ASOP currently only uses the 3D outer profile of the
Docking Station to match against.
One of the navigational displays of the SRPS system
can be seen in Fig. 7. The 2D map output shows the Fig. 6: Real Sonar Image of the above scene
current estimated position and vehicle track (it can be captured in Tank Trials of the SRPS
overlaid on any chart, or scarmed map). The lower
display shows the live sonar image, together with the At the time of writing, the next dock trials are being
sonar targets currently being tracked. It is possible plarmed. The aim is to use the SRPS system "online".
for the system to highlight sonar targets not The Hyball will be manually piloted in various
contained within the CAD database. This could be directions over the submerged 1/6 scale model of the
used to identify potential collisions with unknown Docking Station. The ASOP system will determine
obstacles. the vehicle's position based on the 3D model,
These initial trials established the SRPS could fix the together with the live sonar and telemetry (heading,
vehicle's position to within 10cm, I m above the scale and depth) data.
model. It is hoped that further testing and refmement In this mode, the Virtual Camera monitor output of
to ASOP algorithm will improve this figure further. the system can also be used by the pilot to help in the
manual navigation of the vehicle.

S4
3.3. Full Scale Dock Trials
These are currently planned for summer 2000, at
Ifremer's docks (Toulon, France). Using a speciaUy
instrumented crane, the SRPS sonar head wiU be
"flown" across the fuU size Docking Station
(approximately 7m in length) in various trajectories.
Using the instruments, it will be possible to measure
the performance of the SRPS system, and identify
possible problems before the fuU system sea trials
later in the year.

3.4. Deep Water Field Trials


The details of the fmal deep water (200m-500m)
trials have yet to be fmalised but they will take the
form iUustrated in Fig. 10. The work-class ROV wiU
be mounted on the Swimmer AUV and launched
from a support vessel. The site, probably off the
French Mediterranean cost wiU be chosen to provide
500 metres of water depth (as close to the coast as
possible).
Fig. 7: Example of one Navigational Display The trials will use an ACHILLE ObsROV and a
available on the SRPS system - continuous REMORA 2-seat submarine to install the Docking
output of the estimated position and Station and monitor the Swimmer AUV during the
confidence trials. They wiU also be able to affect a recovery in
case of a technical failure . The trials are currently
scheduled for March 2001 .

Fig. 8: VR image of Dock Test Environment


Fig. 10: Proposed Environment for
Swimmer AUV Deep Water Trials

ASOP Profile In'rnmdhon "1l!I


4. CONCLUSIONS
The aim of Swimmer AUV is to form part of an
autonomous deployment system for work-class
ROVs in ultra-deep offshore fields . The "concept"
envisages a network of Docking Stations, an
Acoustic Positioning and Telemetry Network, and a
vehicle capable of autonomously carrying a ROV as
a payload to the seabed.
One component of the Swimmer AUV, the SRPS
(Short Range Positioning System) has been
Fig. 9: Example profile showing ASOP predicting developed at Liverpool using a novel technique
and locating objects within the live sonar data caUed ASOP (Active Sonar Object Prediction). The
system uses a 3D map, derived from existing CAD
information, to drive an active search for significant

55
features in the sonar data.
Although the prototype's deep-water tests are
scheduled for Spring 2001, the Swimmer "concept"
is a promising one. It could significantly reduce the
cost of using ROVs in deep-water production fields,
and help fonn an important step in the move towards
fully autonomous systems.

References
Ashraf, M. and I . Lucas (1994), "Underwater object
recognition techniques using ultrasonics", Proc.
Oceans 94, Vol 1, pp 170-175, 1994 (Brest,
France).
Lane, D.M. et al (1998). "Mixing Simulations and
Real Subsystems for Subsea Robot Development:
Specification and Development of the Core
Simulation Engine", Oceans '98, MTS / IEEE
International Conference, Nice, France, Sept
1998.
Lane, D.M. et al (2000). "Achieving Hardware in the
Loop Interoperability in Distributed Undersea
Robot Simulation", Undersea Weapon Based
Design Workshop, Newport RI, to be presented
June 2000.
Smith, J.S., R. Vu, I. Sarafis and 1. Lucas (1994),
"Computer vision control of an underwater
manipulator", Proc. Oceans 94, Vol 1, pp 187-
192, 1994 (Brest, France).
Smith, PJ., I.S. Smith and I. Lucas (1999),
"Localization of a UUV within structures using
range data and world modelling techniques", Int.
Journal of System Science, 1999, Vol 30, No.9,
pp 915-928.

ACKNOWLEDGEMENTS
The Swimmer Consortium is comprised of
Cybemetix, Ifremer, The University of Liverpool,
TOTAL, Noordhoek and GERTH. Thanks go to
Pierre Marty (Cybemetix) as project manager. The
Swimmer project is supported in part by the EU /
Thermie Programme: Contract No. OG/153/98.
Special thanks to the Ocean Systems Laboratory,
Heriot-Watt University, Edinburgh for the use of
their Cartesian Robot and Tank. The ASOP research
was originally supported by the TUUV (Technology
for Unmanned Underwater Vehicles), a joint industry
/ EPSRC managed programme.

56

You might also like