Further Advances in Unmanned Marine Vehicles 1849194793, 9781849194792

Citation preview

Control Engineering Series 77

Further Advances in Unmanned Marine Vehicles

Edited by G.N. Roberts and R. Sutton

IET CONTROL ENGINEERING SERIES 77

Further Advances in Unmanned Marine Vehicles

Other volumes in this series: Volume 8 Volume 18 Volume 20 Volume 28 Volume 33 Volume 34 Volume 35 Volume 37 Volume 39 Volume 40 Volume 41 Volume 42 Volume 44 Volume 47 Volume 49 Volume 50 Volume 51 Volume 52 Volume 53 Volume 54 Volume 55 Volume 56 Volume 57 Volume 58 Volume 59 Volume 60 Volume 61 Volume 62 Volume 63 Volume 64 Volume 65 Volume 66 Volume 67 Volume 68 Volume 69 Volume 70 Volume 71 Volume 73 Volume 74 Volume 75 Volume 76 Volume 78

A history of control engineering, 1800–1930 S. Bennett Applied control theory, 2nd edition J.R. Leigh Design of modern control systems D.J. Bell, P.A. Cook and N. Munro (Editors) Robots and automated manufacture J. Billingsley (Editor) Temperature measurement and control J.R. Leigh Singular perturbation methodology in control systems D.S. Naidu Implementation of self-tuning controllers K. Warwick (Editor) Industrial digital control systems, 2nd edition K. Warwick and D. Rees (Editors) Continuous time controller design R. Balasubramanian Deterministic control of uncertain systems A.S.I. Zinober (Editor) Computer control of real-time processes S. Bennett and G.S. Virk (Editors) Digital signal processing: principles, devices and applications N.B. Jones and J.D.McK. Watson (Editors) Knowledge-based systems for industrial control J. McGhee, M.J. Grimble and A. Mowforth (Editors) A history of control engineering, 1930–1956 S. Bennett Polynomial methods in optimal control and filtering K.J. Hunt (Editor) Programming industrial control systems using IEC 1131-3 R.W. Lewis Advanced robotics and intelligent machines J.O. Gray and D.G. Caldwell (Editors) Adaptive prediction and predictive control P.P. Kanjilal Neural network applications in control G.W. Irwin, K. Warwick and K.J. Hunt (Editors) Control engineering solutions: a practical approach P. Albertos, R. Strietzel and N. Mort (Editors) Genetic algorithms in engineering systems A.M.S. Zalzala and P.J. Fleming (Editors) Symbolic methods in control system analysis and design N. Munro (Editor) Flight control systems R.W. Pratt (Editor) Power-plant control and instrumentation D. Lindsley Modelling control systems using IEC 61499 R. Lewis People in control: human factors in control room design J. Noyes and M. Bransby (Editors) Nonlinear predictive control: theory and practice B. Kouvaritakis and M. Cannon (Editors) Active sound and vibration control M.O. Tokhi and S.M. Veres Stepping motors: a guide to theory and practice, 4th edition P.P. Acarnley Control theory, 2nd edition J.R. Leigh Modelling and parameter estimation of dynamic systems J.R. Raol, G. Girija and J. Singh Variable structure systems: from principles to implementation A. Sabanovic, L. Fridman and S. Spurgeon (Editors) Motion vision: design of compact motion sensing solution for autonomous systems J. Kolodko and L. Vlacic Flexible robot manipulators: modelling, simulation and control M.O. Tokhi and A.K.M. Azad (Editors) Advances in unmanned marine vehicles G. Roberts and R. Sutton (Editors) Intelligent control systems using computational intelligence techniques A. Ruano (Editor) Advances in cognitive systems S. Nefti and J. Gray (Editors) Adaptive sampling with mobile WSN K. Sreenath, M.F. Mysorewala, D.O. Popa and F.L. Lewis Eigenstructure control algorithms: applications to aircraft/rotorcraft handling qualities design S. Srinathkumar Advanced control for constrained processes and systems F. Garelli, R.J. Mantz and H. De Battista Developments in control theory towards glocal control L. Qui, J. Chen, T. Iwasaki and H. Fujioka (Editors) Frequency-domain control design for high-performance systems John O’Brien

Further Advances in Unmanned Marine Vehicles Edited by G.N. Roberts and R. Sutton

The Institution of Engineering and Technology

Published by The Institution of Engineering and Technology, London, United Kingdom The Institution of Engineering and Technology is registered as a Charity in England & Wales (no. 211014) and Scotland (no. SC038698). © 2012 The Institution of Engineering and Technology First published 2012

This publication is copyright under the Berne Convention and the Universal Copyright Convention. All rights reserved. Apart from any fair dealing for the purposes of research or private study, or criticism or review, as permitted under the Copyright, Designs and Patents Act 1988, this publication may be reproduced, stored or transmitted, in any form or by any means, only with the prior permission in writing of the publishers, or in the case of reprographic reproduction in accordance with the terms of licences issued by the Copyright Licensing Agency. Enquiries concerning reproduction outside those terms should be sent to the publisher at the undermentioned address: The Institution of Engineering and Technology Michael Faraday House Six Hills Way, Stevenage Herts, SG1 2AY, United Kingdom www.theiet.org While the authors and publisher believe that the information and guidance given in this work are correct, all parties must rely upon their own skill and judgement when making use of them. Neither the authors nor the publisher assumes any liability to anyone for any loss or damage caused by any error or omission in the work, whether such an error or omission is the result of negligence or any other cause. Any and all such liability is disclaimed. The moral rights of the authors to be identified as authors of this work have been asserted by them in accordance with the Copyright, Designs and Patents Act 1988.

British Library Cataloguing in Publication Data A catalogue record for this product is available from the British Library ISBN 978-1-84919-479-2 (hardback) ISBN 978-1-84919-480-8 (PDF)

Typeset in India by MPS Limited Printed in the UK by CPI Group (UK) Ltd, Croydon, CR0 4YY

Contents

1

Editorial: further advances in unmanned marine vehicles 1.1 Introduction 1.2 Contributions 1.3 Concluding remarks Acknowledgement Reference

2

ROV LATIS: next generation smart underwater vehicle 2.1 Introduction 2.2 Background 2.3 International collaboration 2.4 System description 2.4.1 Design objectives 2.4.2 Features 2.4.3 Operation modes 2.4.4 Frame 2.4.5 Technical specifications 2.4.6 Overall hardware architecture 2.4.7 Control architecture 2.5 Innovations 2.6 Test trials 2.6.1 CE-09-04 Cruise (ROV LATIS onboard RV Celtic Explorer) 2.6.2 CV-10-029 Cruise (ROV LATIS onboard RV Celtic Voyager) 2.7 Feedback 2.8 Implementation 2.9 Conclusions Acknowledgements References

3

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles 3.1 Background 3.1.1 System requirements 3.1.2 Design considerations and implications 3.2 A modular design concept

1 1 1 7 8 8 9 9 10 12 12 12 13 13 13 15 15 18 25 32 32 33 39 39 41 41 43

45 45 45 46 47

vi

Further advances in unmanned marine vehicles 3.2.1 The command module 3.2.2 Electrical systems 3.2.3 Cameras 3.2.4 Lasers rangers 3.2.5 Lights 3.2.6 Thrusters 3.2.7 Hydraulic motors 3.2.8 Hydraulic valve packs 3.3 Tool modules 3.3.1 Bulk sample grab 3.3.2 Tool sledge 3.3.3 Lander recovery module 3.3.4 Lander deployment module 3.3.5 Top-side control system 3.4 Results 3.5 Conclusions Acknowledgements References

4

An AUV project applied to studies on manoeuvrability of underwater vehicles 4.1 Introduction 4.2 The Pirajuba AUV 4.3 Control architecture 4.4 Investigation on modelling the dynamics of the Pirajuba AUV 4.4.1 Estimation of the AUV hydrodynamic parameters through the ASE- and CFD-based methods 4.4.1.1 4.4.1.2 4.4.1.3

Bare hull forces and moment Estimation of the hydroplane-generated normal force Body–fin interaction as a function of the angle of attack

4.5 Results and experimental validation 4.6 Conclusion References 5

Neural network–based switching adaptive control for a remotely operated vehicle 5.1 Introduction 5.2 A switching supervisory control approach for non-linear dynamical systems 5.3 Preliminaries 5.4 Neural network–based switching control 5.4.1 Preliminaries 5.4.2 Stabilizing neural network–based fix control 5.4.3 Stabilizing neural network–based adaptive control

47 49 49 50 50 51 53 54 54 54 56 59 61 62 63 64 66 66

69 69 71 74 79 80 80 82 84

85 88 88

91 91 93 95 97 97 98 99

Contents 5.5

6

Switching strategy 5.5.1 Known vehicle configurations 5.5.2 Unknown vehicle configurations 5.6 The remotely operated vehicle model 5.7 Numerical results 5.7.1 Neural network structure and validation 5.7.2 Simulation tests 5.8 Conclusions References

100 100 102 102 104 104 105 108 108

Development of dynamic positioning and tracking system for the ROV Minerva 6.1 Introduction 6.2 ROV Minerva specifications 6.3 Mathematical models 6.4 Control system architecture and platform 6.4.1 Hardware and software platform 6.4.2 Signal flow 6.4.3 Control system modules

113 113 114 115 118 118 118 119

6.4.3.1 6.4.3.2 6.4.3.3 6.4.3.4

7

vii

Signal processing Observer Controller Guidance

120 121 122 122

6.5 Experimental results 6.6 Conclusions Acknowledgements References

123 127 127 127

Port-Hamiltonian control of fully actuated underwater vehicles 7.1 Introduction 7.2 Port-Hamiltonian systems 7.3 Vectorial and port-Hamiltonian dynamic models 7.3.1 Rigid-body dynamics 7.3.2 Kirchhoff’s equations and rigid-body equations of motion 7.3.3 Fluid-force models 7.3.4 Complete dynamic models and PHS form 7.4 Motion control retaining the PHS form 7.4.1 Reference signals and control objective 7.4.2 Regulator controller design for trajectory tracking 7.4.3 Dynamic positioning (without integral action) as a particular case of tracking 7.4.4 Integral controller design for trajectory tracking 7.5 Case study 7.6 Conclusion References

129 129 130 131 131 133 135 137 139 140 140 141 142 143 145 146

viii 8

Further advances in unmanned marine vehicles Sonar-based simultaneous localization and mapping for autonomous underwater vehicles 8.1 Introduction 8.2 The SLAM problem 8.3 Sonar-based SLAM for AUVs 8.3.1 State of the art in underwater sonar-based SLAM 8.3.2 Scan matching 8.3.2.1

8.4

8.4.1.1 8.4.1.2 8.4.1.3

8.4.2

9

Comparison

Underwater scan-matching SLAM 8.4.1 ScanGrabbing algorithm Beam segmentation and range detection Relative vehicle localization Scan forming

149 149 150 151 152 154 155

156 156 156 156 157

SLAM algorithm

158

8.4.2.1 8.4.2.2 8.4.2.3 8.4.2.4 8.4.2.5

158 159 159 160 160

Map initialization Prediction Scan matching Loop closing candidates State update

8.5

Experimental set-up and results 8.5.1 ICTINEUAUV 8.5.2 The dataset 8.5.3 Results 8.6 Conclusions Acknowledgements References

161 161 161 163 165 165 165

T-REX: partitioned inference for AUV mission control 9.1 Introduction 9.2 A motivating example 9.3 Key concepts in T-REX 9.3.1 Mission, goals, actions 9.3.2 Distributing decision processes: a conceptual view 9.3.3 Interleaving planning and execution 9.4 The T-REX execution cycle 9.4.1 Synchronization: maintaining reactor state 9.4.2 Deliberation: making the agent proactive

171 171 173 175 175 177 180 182 182 184

9.4.2.1 9.4.2.2

Definitions Flexible constraint-based plan representation

9.4.3 Consistency in timely plan dispatch Experimental results 9.5.1 Augmenting traditional surveys 9.5.2 Extending Lagrangian surveys 9.6 Conclusion Acknowledgements References

9.5

184 185

188 190 191 193 195 196 196

Contents 10

Vehicle-following for unmanned surface vehicles 10.1 Introduction 10.2 Background 10.2.1 Unmanned surface vehicles: research and scientific prototypes 10.2.2 Task classification 10.2.3 Path-following 10.2.4 Experiments in cooperative and coordinate control of UMVs 10.3 System architecture 10.4 Vehicle-following: path-tracking system design 10.4.1 Virtual target–based path-following system design 10.4.1.1 10.4.1.2 10.4.1.3

10.5

Kinematics modelling Kinematics guidance Heuristics

10.4.2 Range control design Experimental set-up 10.5.1 Vehicles 10.5.1.1 10.5.1.2

Charlie USV ALANIS dual-mode vessel

201 201 202 202 205 206 207 209 210 210 210 211 213

213 214 214 214 216

10.5.2 System integration Experimental results 10.6.1 GPS performance validation 10.6.2 Concept validation experiments 10.6.3 Repeatable experiments and performance metrics definition 10.7 Conclusions Acknowledgements References

220 225 227 227

An automatic control and fault-tolerant multi-sensor navigation system design for an unmanned maritime vehicle 11.1 Introduction 11.2 Springer USV 11.2.1 Hardware development 11.2.2 Software development 11.3 Rigid body modelling and system identification 11.3.1 Steering dynamics 11.4 Navigation system design 11.5 Guidance and control systems 11.5.1 Fuzzy-LQG controller 11.5.2 GA-MPC autopilot design 11.6 Experimental results 11.7 Summary Acknowledgement

231 231 232 233 237 238 241 242 244 245 246 247 248 249

10.6

11

ix

217 218 218 218

x

Further advances in unmanned marine vehicles Appendix A Appendix B References

12

249 251 251

Cooperative control of multiple autonomous marine vehicles: theoretical foundations and practical issues 12.1 Introduction 12.2 Practical motivation and scientific mission scenarios 12.2.1 Scenario 1: the quest for hydrothermal vents 12.2.2 Scenario 2: marine habitat mapping 12.3 A general architecture for multiple vehicle cooperation 12.3.1 Multiple vehicle cooperative motion control 12.3.2 Single and multiple vehicle primitives (SVPs and MVPs) 12.3.2.1 12.3.2.2 12.3.2.3 12.3.2.4 12.3.2.5 12.3.2.6 12.3.2.7

13

Point stabilization – PS (also referred to as Go to Point) Path following – PF Trajectory tracking – TT Path planning for multiple vehicles (cooperative path planning) – CPP Cooperative path following – CPF Cooperative target following (CTF) and cooperative target tracking (CTT) Cooperative manoeuvring in the presence of tight communication constraints

255 255 257 258 259 260 260 262 262 262 263 263 264 265 265

12.4

NetMarSyS Simulator: simulation results 12.4.1 An illustrative 2D example in simulation 12.4.2 An illustrative 3D example in simulation 12.5 Experimental results 12.5.1 Tests in the Azores, 2008 12.5.2 Tests in Sesimbra, 2011 12.6 Conclusions and future work Acknowledgements References

266 266 268 270 271 273 277 278 278

Hybrid glider: motivation, design and evaluation 13.1 Introduction 13.1.1 Motivation 13.2 Glider system overview 13.3 Hybrid system concept 13.4 Component selection and design 13.4.1 Propeller geometry parametric analysis 13.4.2 Adding a motor and gearbox 13.4.3 Model verification 13.5 Evaluation 13.5.1 Flume tank drag and propulsion tests

283 283 284 285 287 288 288 290 293 298 298

Contents

14

15

xi

13.5.2 Field trials 13.6 Conclusions and future work Acknowledgements Appendix References

303 304 305 305 306

The enhanced Folaga: a hybrid AUV with modular payloads 14.1 Introduction 14.2 eFolaga design 14.3 Dynamic modelling of the eFolaga 14.3.1 Linear motion 14.3.2 Rotational motion 14.3.3 Complete standard model 14.3.4 Actuation system identification for the eFolaga 14.4 Architectures for module software integration: a case study from the UAN10 experiment 14.5 Conclusions Acknowledgements References

309 309 311 315 317 318 319 320

A biomimetic underwater vehicle design concept 15.1 Introduction 15.1.1 Biomimetics and biologically inspired design 15.1.2 Biological AUV design 15.2 Biological swimming 15.2.1 Types of aquatic swimming 15.2.2 Fish swimming

331 331 331 332 333 333 333

15.2.2.1 15.2.2.2

15.3

15.2.3 Method of biomimicry: RoboSalmon Mathematical modelling of a biologically inspired AUV 15.3.1 Vehicle dynamics 15.3.1.1 15.3.1.2 15.3.1.3 15.3.1.4 15.3.1.5 15.3.1.6 15.3.1.7

15.4

Genera of fish Swimming gait

Rigid body dynamics Hydrodynamic added mass terms Restoring forces and moments Hydrodynamic damping terms Tail manoeuvring capability Recoil motion Tendon drive system: input forces and moments

15.3.2 Vehicle hull kinematics 15.3.3 State space form 15.3.4 Model results Forward propulsion analysis 15.4.1 Tail-beat frequency and amplitude 15.4.2 Power

322 328 328 328

334 334

335 336 338 338 339 339 340 341 341 343

343 344 344 346 346 347

xii

Further advances in unmanned marine vehicles 15.5

16

17

Open-loop manoeuvring analysis 15.5.1 Turning from stationary 15.5.2 Turning at speed 15.5.3 Turning circle 15.5.4 Power consumption 15.6 Conclusion References

348 349 351 351 354 355 356

Development of robotic fish for the next generation unmanned marine vehicle 16.1 Introduction 16.2 The beginning of development 16.3 Development of flexible oscillating fin propulsion system 16.4 Development of robotic fish 16.4.1 Development of robotic sea bream 16.4.2 Fully automatic swimming system 16.5 Flapping wing-type robotic fish 16.6 New tail fin mechanism 16.7 Outline of shark ray 16.8 Conclusions Acknowledgement References

359 359 359 360 361 361 361 362 367 368 371 372 372

Nature in engineering for monitoring the oceans: comparison of the energetic costs of marine animals and AUVs 17.1 Introduction 17.2 Cost of transport 17.2.1 Hotel power/metabolic rate 17.2.2 Propulsion power 17.3 Optimum cost of transport 17.4 Discussion 17.5 Conclusions Acknowledgements References

373 373 374 381 383 391 395 397 398 398

Index

407

Chapter 1

Editorial: further advances in unmanned marine vehicles G.N. Roberts and R. Sutton

1.1 Introduction In our editorial for the previous volume Advances in Unmanned Marine Vehicles [1], we brought together 18 chapters describing research and developments in unmanned marine vehicles (UMVs). We provided a brief history of the important developments in UMVs from 1886 to date. We do not think it is necessary to repeat this here as readers of this volume will no doubt already possess a copy of Advances in Unmanned Marine Vehicles. We also reflected on the more recent increases in worldwide activity in this area and that almost without exception research groups were developing and working on real UMVs. This meant that they were able to test, evaluate and re-evaluate their designs in relatively quick succession, thereby rapidly reporting new approaches, techniques, designs and successes. This rapid designevaluation cycle is the prime mover for progress, not only for consolidating designs but also for leading to new design ideas and innovation. Since its publication in 2006, Advances in Unmanned Marine Vehicles has proven to be a useful and popular source of reference. However, the rapid design-evaluation cycle means that further advances have been made and with this in mind it was felt that there was scope for a second more up-to-date volume in order to promulgate recent important developments.

1.2 Contributions As was the case when we compiled Advances in Unmanned Marine Vehicles, the main emphasis on research and development of UMVs continues to focus on remotely operated vehicles (ROVs) and autonomous underwater vehicles (AUVs), and this is reflected in Chapters 2–9, 12, 13 and 14. Chapters 10 and 11 consider unmanned surface vehicles (USVs), whereas Chapter 12 describes cooperative behaviour of multiple AUVs and USVs. In Chapters 15 and 16 biomimetic underwater vehicle design concepts are used for the design of AUVs and robotic fish. Finally, Chapter 17 provides insight into the energetic costs of marine animals and AUVs. As each chapter in the book is a self-contained exposition of the work

2

Further advances in unmanned marine vehicles

undertaken, some introductory remarks about the contributions of each chapter are provided here. Chapter 2 describes innovations built inside the set of assistive tools and technologies for system integration, deployment, monitoring and maintenance of ocean energy devices (wave energy converters and tidal turbines). These tools comprise simulation, modelling, control and visualization tools, including modelling of ocean waves, currents, surface and submerged energy converters, marine platforms and supporting vessels (ships and ROVs). The flexible design of these tools enables their use as separate standalone modules, as well as their integration into a unique integrated system. A major component of the system, designed as a prototype platform to demonstrate system validity and operability and to prove new technologies developed in the Mobile and Marine Robotics Research Centre at the University of Limerick, is a smart, remotely operated vehicle: ROV LATIS. LATIS is a next generation smart ROV with multiple modes of operation: Surface-Tow Mode, Surface-Thrusted Mode, ROV Operation Mode and ROV Submerged Tow in Strong Currents Mode. Four main software innovations of the proposed system are 2D Advanced Topview Display, Advanced Pilot Interface, 3D Real-Time Augmented Reality Display and Advanced Control System with Fault-Tolerant Capabilities. System validation and technology demonstration were performed through a series of test trials with different support vessels during 2010 off the west coast of Ireland, in Galway Bay and in the Shannon Estuary. Selected results of these trials are presented in this chapter. The HyBIS low-cost, multipurpose, highly manoeuvrable, fibre-optic controlled survey and sampling robotic underwater vehicle (RUV) capable of diving to 6000 m is described in Chapter 3. HyBIS was built in the United Kingdom by HydroLek Ltd in collaboration with the National Oceanography Centre, Southampton. The vehicle has a modular design, with the top module being a command and power system that comprises power management, cameras, lights, hydraulics, thrusters and telemetry. The lower modules comprise a clam-shell sampling grab, a manipulator-arm and tool sled, a winch for instrument recovery and an ocean bottom seismometer deployment module. Unlike a conventional ROV, HyBIS does not have any floatation, rather it is suspended by its umbilical cable directly from the ship. The advantage of direct suspension is that HyBIS can recover or deploy a payload of up to 500 kg. The utility of HyBIS was first demonstrated in October 2008 when it located and recovered a stranded, 250 kg, benthic lander from 2200-m water depth off the Canary Island of Forteventura. Its greatest achievement to date was in April 2010, during its first scientific mission, in the Cayman Trough, Caribbean. Here, HyBIS was used in combination with an AUV (Autosub6000) to locate, image and sample two hitherto unknown hydrothermal vent sites, one of which is at a depth of 5000 m and is the deepest known on the planet. In 2011, the vehicle’s versatility and reliability were again demonstrated when it recovered and redeployed a 4000-kg lander from the Arctic seafloor, sampled methane gas from active seeps and deployed a series of geophysical instruments onto the seafloor. Chapter 4 presents an AUV project and its application to a study on dynamics of underwater vehicles undertaken at the Department of Mechatronics Engineering,

Editorial

3

Polytechnic School of University of Sa~ o Paulo, Brazil. The investigation of AUV dynamics includes captive and free model tests. The free model tests are carried out with the AUV Pirajuba. The main systems of this vehicle are presented, emphasizing its control architecture. The embedded system is low cost and based on offthe-shelf components, free and largely known software tools, like C language, and GNU compiler. The tank tests carried out with models of this AUV are related to the validation of prediction methods applied to the calculation of static hydrodynamic derivatives. Computational Fluid Dynamics (CFD) methods and the analytical and semi-empirical methods are presented. A symbiosis between both approaches is proposed in order to improve and interpret the predictions. Neural network–based switching adaptive control used for the tracking control problem for an underwater vehicle subjected to different load configurations is described in Chapter 5. The aim of this work is to propose an adaptive switching control strategy to cope with the large transient errors related to the considered mode-switch process, when a poor knowledge of the different possible vehicle configurations is available. It is shown that if there is adequate a priori information on the different possible vehicle configurations, a switching control scheme can reduce the transient response with respect to a conventional adaptive control, when the vehicle changes its operative condition. However, with a poor knowledge of the actual vehicle-operative condition, a switching control is unable to guarantee the same control performance; the precomputed controllers cannot cope all environment and load conditions and an auto-tuning mechanism is required. Therefore, the proposed switching control in this chapter consists of an adaptive control policy improved by the connection with supervised switching logic, which is composed of a bank of alternative candidate controllers that switch among them according to a suitably defined logic. This makes the approach particularly suited to deal with large parametric variations and/or uncertainties. At each time instant, the supervisor decides which candidate controller should be put into the feedback loop. Multiple neural networks are considered to compensate for the effects of nonlinearities and plant uncertainties and to store different dynamics of the vehicle for different tasks and configurations. The developed switching adaptive control is applied to the tracking control problem of a ROV developed by ENI Group (Italy), which is used for the exploitation of combustible gas deposits at great sea depths. Chapter 6 describes the development of a dynamic positioning (DP) and tracking system for the ROV Minerva. This development is motivated because ROVs are used in a variety of subsea work tasks from small hand-deployable ROVs to large work-class ROVs for heavy intervention work at subsea offshore installations. Common for all types of ROV is that many applications instead of being controlled manually are an automatic positioning control system with higher accuracy and efficiency, enabling the pilot to focus on monitoring and planning of operations that demand human intervention or decision making. In 2010 and 2011 a group of MSc students, PhD candidates and researchers at the Norwegian University of Science and Technology (NTNU), Trondheim, Norway, developed a DP and tracking system for ROVs. The DP system is designed primarily for research and educational purposes making it easy to re-design and test control strategies, sensor

4

Further advances in unmanned marine vehicles

configurations and tool packages including manipulators. Results verified experimentally on the small ROV Minerva are presented. Chapter 7 presents a novel control strategy for trajectory tracking of underwater marine vehicles undertaken as a joint project between the School of Engineering, University of Newcastle, Australia, and the Centre for Ships and Ocean Structures at the Norwegian University of Science and Technology, Trondheim, Norway. The control design is based on Port-Hamiltonian theory. A mathematical model for neutrally buoyant underwater vehicles is formulated as a Port-Hamiltonian system, and a tracking controller with integral action is designed for the horizontal plane: surge, sway and yaw. The design is approached by formulating the error dynamics as a set-point regulation Port-Hamiltonian control problem with integral action of nonpassive outputs. The global asymptotic stability of the closed-loop system is proved and the performance of the controller is illustrated using a model of an open-frame offshore underwater vehicle. Although the presentation is limited to the case of horizontal-plane control of fully actuated vehicles, it is argued that the results obtained can be extended to the case of under-actuated marine vehicles. Techniques for sonar-based simultaneous localization and mapping (SLAM) for AUVs are covered in Chapter 8. Here it is explained that solving localization for AUVs in unstructured, unconstrained and turbid areas is still a challenging problem. The ideal AUV should be able to localize itself in any conditions with its onboard sensor suite and without the need of external infrastructure. The chapter focuses on SLAM techniques that utilize sonar as their main perception sensor, as sonar can penetrate inside the water deeper than the vision systems. The chapter starts with a survey of recent advances of the underwater sonar-based SLAM and proposes a method based on scan-matching. A comparison between the most recognized scan-matching algorithms reveals their strengths and weaknesses. It is shown that the proposed sonar-based SLAM algorithm first corrects the distortions that the vehicle motion introduces in the scans. Then consecutive scans are crossregistered under a probabilistic scan-matching technique, for estimating the displacements of the vehicle. Finally, an augmented state–extended Kalman filter estimates and keeps the registered scans poses. No a priori structural information or initial pose is considered. The SLAM algorithm is tested with a real-world dataset, showing the viability of the proposed approach. Chapter 9 describes a general-purpose artificial intelligence (AI)-based control architecture that implements a variant of the sense–plan–act paradigm on AUVs. Plans are no longer scripted a priori but synthesized onboard with highlevel directives instead of low-level commands. The Teleo-Reactive EXecutive (T-REX) framework deliberates about future states, plans for actions and executes activities while monitoring plans for anomalous conditions. The architecture uses partitioned control loops for a ‘divide and conquer’ problem-solving strategy allowing for ease of vehicle adaptation, robust and focused failure recovery, incremental computational model building, simplification of software development and ability to use legacy or non-native computational paradigms. It is shown that vehicle adaptation and sampling occurs in situ with additional modules that can be selectively used depending on the application in focus. Abstraction in

Editorial

5

problem-solving allows different applications to be programmed relatively easily, with little to no changes to the core search engine thereby making software engineering sustainable. Time is represented flexibly in part to deal with environmental and operational uncertainty. T-REX is in routine use for operational oceanography, providing a new tool for science to sample and observe the dynamic coastal ocean. Both Chapters 10 and 11 consider USVs. Chapter 10 illustrates the main issues and results related to vehicle-following methodologies applied to USVs. In particular, in the framework of a path-tracking approach, separating the tasks of satisfying spatial and time constraints, a Lyapunov-based control law is designed in order to drive a follower USV to navigate along the path generated by the motion of a leader vessel, without any a priori information about the path itself. The effectiveness of virtual target–based approaches is discussed on the basis of experiments carried out with the ALANIS dual-mode vessel and Charlie USVs. Particular attention is devoted to the definition of procedures for the execution of repeatable experiments and suitable metrics for a quantitative evaluation of the performance of the proposed techniques. Chapter 11 outlines the development of the Springer USV. The focus of this chapter is on the research and design aspects of Springer’s navigation, guidance and control (NGC) systems and their implementation. The navigation system comprises an advanced fuzzy logic–based federated Kalman filter approach, whereas the autopilots include modified forms of linear quadratic Gaussian and model predictive controllers. The proposed hardware and software architecture for Springer is also presented, which is shown to be modular and which allows flexibility for other operators to modify and use their own instrumentation and NGC software. Experimental results for multiple waypoints following algorithm that demonstrates the functional integrity and efficacy of the complete system are shown. The challenges encountered in the design, development and cooperative operation of groups of autonomous marine vehicles for the execution of scientific and commercial missions at sea are the area covered in Chapter 12. Besides the obvious technological issues that must be solved to operate a group of autonomous vehicles efficiently and reliably, the subject motivates a plethora of challenging theoretical problems in the areas of multiple vehicle motion planning, guidance, navigation and control. This chapter is an overview of some of the theoretical and practical issues that arise in the process of developing advanced motion control systems for cooperative multiple autonomous marine vehicles. The presentation is focused on a number of problems that have been formulated and solved in the scope of two European projects in which the authors participated: (i) GREX – Coordination and Control of Cooperating Heterogeneous Unmanned Systems in Uncertain Environments and (ii) CO3AUVs – Cooperative Cognitive Control for AUVs. The chapter affords the reader a fast-paced, simplified presentation of the foundations of cooperative motion control and highlights related practical work. Examples of scientific mission scenarios with ASVs and AUVs acting in cooperation set the stage for the main contents of the presentation. Special attention is given to developing networked control strategies capable of yielding robust performance of

6

Further advances in unmanned marine vehicles

a fleet of heterogeneous vehicles in the presence of complex vehicle dynamics, external disturbances and severe acoustic communication constraints imposed by intermittent failures and latency. To this effect, a general architecture is introduced for cooperative marine vehicle control in the presence of time-varying communication topologies and communication losses. The results of simulations with the NetMarSyS – Networked Marine Systems Simulator at the Institute for Systems and Robotics, Instituto Superior Te´cnico, Technical University of Lisbon – are presented and show the efficacy of the algorithms developed for cooperative motion control. To bridge the gap between theory and practice, the last part of the chapter focuses on practical issues and describes the results of tests carried out with a group of marine vehicles at sea in the Azores and in Sesimbra, Portugal. During the past decade, propeller-driven AUVs and autonomous underwater gliders have become a common tool in ocean-related research. The range of applications is stretching from water column measurements using gliders to seafloor mapping, pipeline monitoring and mine detection using more and more sophisticated propeller-driven AUVs. Chapter 13 describes the motivation behind the development of a long-range underwater glider and in particular, the choice of augmenting an existing underwater glider is highlighted by a careful analysis of the vehicle’s energy efficiency and system reliability. Based on these system parameters an auxiliary propulsion module is designed. A system-level optimization is used to match the individual subsystems of the propulsion module. During the process close consideration is given to system robustness and long-term performance. The approach is evaluated using extensive tank and field trials. In chapter 14 the development of a modular payload-carrying capability on the AUV Folaga (‘eFolaga’) is described and documented. Folaga are AUVs characterized by propulsion through jet-pumps or screw propellers (as in conventional, torpedo-shaped, AUVs) and ballast/asset change (as in oceanographic gliders). The eFolaga design is in line with its predecessors: in particular, the light-weight, small dimensions and low-cost characteristics have been maintained, as well as high manoeuvrability and hovering capacities. A general methodology to derive dynamic models of eFolaga-like vehicles is described, and identification of the eFolaga buoyancy change and mass displacement actuators is analysed and reported. The software design architecture of the payload module installed in the eFolaga for the activities of the Underwater Acoustic Network (UAN) European project is also described, showing how the interface between the mission management and the eFolaga guidance, and navigation and control system can be designed to maintain modularity, at the hardware and software level. Chapters 15 and 16 explore biomimetic underwater vehicle design concepts. Chapter 15 highlights that the limiting factor to AUV mission is the efficiency of the propulsion system, which is usually based on marine propellers and proposes that the potential method of increasing the efficiency of the propulsion system would be to implement a biomimetic approach, i.e. to imitate a solution found in nature. As fish are highly efficient swimmers, greater propulsive efficiency may be possible by mimicking the fish tail propulsion system. An investigation at the University of Glasgow focusing on a prototype underwater vehicle with a

Editorial

7

biomimetic propulsion system called RoboSalmon is presented in this chapter. The aim of this work is to investigate any potential benefits of using a biomimetic system over a conventional propulsion system. The first part of the investigation covers the development of a mathematical model of the RoboSalmon vehicle. This model has been developed to improve the general understanding of the dynamics of the system. Simulation results from this model are compared with the experimental results and show good correlation. Experimental results obtained comparing the RoboSalmon prototype with the biomimetic tail system to the propeller and rudder system are then presented. These experiments include a study into the straight swimming performance, manoeuvring performance, recoil motion and power consumption. The main findings of the investigation show for forward swimming the maximum surge velocity of the RoboSalmon is 0.18/ms and at this velocity the biomimetic system has been found to be more efficient than the propeller system. Also, when considering manoeuvring the biomimetic system has a significantly reduced turning radius. Chapter 16 continues with this theme with the development of a bio-manoeuvring-type underwater vehicle. This was developed by analysing observations of marine creatures such as fish and the engineering application of their manoeuvring mechanisms. The resulting underwater robot is known as a robotic fish. This chapter describes the history and future development of robotic fish and an underwater vehicle with fin propulsion system that were created as part of research on bio-manoeuvring type underwater vehicle at The University of Kitakyushu, Japan. The final chapter, Chapter 17, is a timely exposition on the comparison of the energetic costs of marine animals and AUVs. The range of physiological adaptations possessed by marine animals allowing them to successfully operate in the marine environment is a plentiful source of inspiration for the designers of AUVs. The chapter compares the total energetic cost of straight line swimming for both marine animals and AUVs, using cost of transport (COT) as a comparative metric. COT is a normalized measure of the energetic cost of transporting the animal’s or vehicle’s mass over a unit distance. It includes non-propulsion power requirements as well as considering the energy lost by actuators and mechanical couplings and energy lost in the wake. Comparisons presented in this chapter show that marine animals typically have higher optimum COT than engineered systems of equivalent size. However, parallels may be drawn, for example, to increase range both marine animals and AUVs appear to favour reducing non-propulsion power costs and travelling slowly to ensure operating at the minimum COT.

1.3 Concluding remarks This book includes 16 new chapters covering NGC of UMVs. It is sincerely hoped that, as with Advances in unmanned marine vehicles, readers from all disciplines will find this book interesting, informative and useful in whatever aspect of UMV design they are involved with, and that it will also provide the genesis for the importation of ideas into other related fields of study.

8

Further advances in unmanned marine vehicles

Acknowledgement We would like to congratulate the contributors for producing chapters of such excellent quality and also to thank them for their support and enthusiasm for this book.

Reference 1.

Roberts G.N. and Sutton R. (eds.). Advances in Unmanned Marine Vehicles. IET, London; 2006

Chapter 2

ROV LATIS: next generation smart underwater vehicle Edin Omerdic , Daniel Toal , Sean Nolan and Hammad Ahmad 1

1

1

1

2.1 Introduction Deployment, installation and maintenance of ocean energy devices require use of underwater robots and support vessels that are also used by other offshore industry, for example, oil and gas. These vessels may be very expensive and, moreover, their costs are very volatile, depending on offshore peak demands. Thus, it is important to address the requirements for vessels to be used in ocean energy deployments and how these requirements may be configured to reduce the costs of these vessels and, simultaneously, affect technology development. According to Richard Vandervoort, chief of ROV operation and underwater robotics, Marine Institute, Memorial University of Newfoundland and Labrador, Canada, ‘the only real automatic controls present on modern work-class ROVs, used in offshore oil and gas exploration, are auto heading, auto depth and auto altitude. It really depends on pilot skills to do good piloting.’ Challenges faced by ROV pilots during deep water operations include low visibility, time-varying ocean currents and umbilical drag effects. The research team at the Mobile & Marine Robotics Research Centre, University of Limerick, has developed the MPPT Ring, a set of hardware/software components that should be installed on an ROV and inside the Control Cabin, in order to increase level of automation, to make ROV operations easier and to save expensive ship time by 20% or more. Research outputs of the project are applicable to the growing international offshore oil and gas sector, and also for future deployment, monitoring and maintenance of ocean energy devices (in particular, wave energy converters and tidal turbines). A major component of the system, designed as a prototype platform to demonstrate system validity and operability and to prove new technologies developed in the Mobile & Marine Robotics Research Centre, is a smart remotely operated vehicle ROV LATIS. It is a next generation smart ROV with unique 1

Mobile & Marine Robotics Research Centre, University of Limerick, Limerick, Ireland

10

Further advances in unmanned marine vehicles

features, including multiple modes of operation, advanced 2D and 3D displays, intuitive and easy to use pilot interface and fault-tolerant control system. System validation and technology demonstration was performed through a series of test trials with different support vessels off the west coast of Ireland, in Galway Bay and in the Shannon Estuary. This chapter highlights the main features of the system, presents selected results of test trials and discusses implementation issues and potential benefits of the technology.

2.2 Background Over the last eight years, researchers in the Mobile & Marine Robotics Research Centre (MMRRC) at the University of Limerick have been engaged in science collaborative and engineering-led seabed survey projects, including technical design, integration and offshore support (Grehan et al., 2005, 2006) and survey operations carrying out detailed survey projects acquiring high-resolution bathymetric, sidescan and video imagery/maps. The team has further developed real-time Virtual Underwater Laboratory (VUL) (Omerdic et al., 2006) and realtime high-resolution sidescan sonar simulators (Riordan, 2006) for use in laboratory testing, training and offshore operations support. The idea of integrating all these technologies into a unique system (multi-purpose platform technologies for subsea operations, MPPT Ring) has been proposed in Omerdic et al. (2008). The MPPT Ring is a set of multi-purpose platform technologies for subsea operations, including survey equipment integration, efficient planning and mission simulation, ROV pilot training, ROV fault-tolerant control, enhanced in-mission survey execution, and offline analysis and replay of acquired data. The MPPT Ring (see Figure 2.1) symbolically represents the dual character of platform technologies: the inner and outer rings can be rotated/expanded independent of each other, indicating that any technology/module can be transparently interchanged between the simulated and real-world environment. This duality of operation facilitates the application of modern control, modelling and simulation tools in marine technology development. It provides a framework for researchers to develop, implement and test advanced control algorithms in a simulated virtual environment, under conditions very similar to the real-world environment. With the MPPT Ring, a single set of hardware and software tools (System Core) is used for both lab development (Virtual Environment) and field trials (Real-World Environment) (Omerdic et al., 2010a). Since virtual and real-world components are compatible on a signal level, switching between them is easy and transparent. The Virtual Environment is used for development of new control algorithms. Based on the experiences gained in previous projects, the challenges met, the solutions developed, and the inherently high costs associated with marine technology and offshore operations, the MMRRC decided to develop a flexible prototype vehicle, ROV LATIS, which serves as a platform to test system validity and viability.

ROV LATIS: next generation smart underwater vehicle

11

Figure 2.1 MPPT Ring

The maritime sector offers a broad variety of applications for advanced computer graphics technology. Review of these applications, based on virtual/augmented reality, is given in Lukas (2010). Use of virtual reality in underwater tele-operation and training has been proposed in Lin and Kuo (2001). Recent advances in high-level simulators of underwater vehicles are described in Curtis (2010). The cost of operation and maintenance represents a significant share of the build up of overall offshore energy cost. Thus, the development of tools to assist in the design and operation of ocean energy farms has been identified as a research priority (Beaudoin et al., 2010). As highlighted in Rourke et al. (2009), the major obstacles restricting the development of wave and tidal energy devices include high deployment and maintenance costs, due to the harsh nature of locations where ocean energy devices are deployed. The assistive tools, built into the design of ROV LATIS, directly address these issues by reducing these costs in two ways: making offshore operations (including ROV operations) easier and more efficient, and saving in expensive support vessel time during deployment, ongoing servicing, inspection, maintenance and repair (IRM) on ocean energy installations.

12

Further advances in unmanned marine vehicles

2.3 International collaboration In 2009, the Marine Institute School of Ocean Technology (SOT), St. John’s, Canada, signed an MOU with the University of Limerick. This MOU recognizes shared interests in underwater robotics, and provides a framework for institutional collaboration in a number of areas, including opportunities for student and faculty exchange, joint initiatives in research and development, and sharing best practices in design and delivery of education and training programmes. Richard Vandervoort, an instructor at SOT with many years experience as an ROV operator, and two MI students, Donovan Tulk and Andrew Pomeroy, travelled to Ireland to assist with the checkout and validation of the ROV LATIS with the MMRRC team. During the CV-10-029 Cruise in August 2010 onboard the Irish Marine Institute research vessel Celtic Voyager, ROV LATIS was put through its paces testing various smart features, while at the same time investigating various physical and biological phenomena on the Irish continental shelf. Feedback from SOT collaborators about the ROV LATIS is given in section 1.7.

2.4 System description 2.4.1

Design objectives

The MMRRC have been focused on research in a number of areas, including AUV/ ROV systems integration, embedded controller development, sensor systems development and in the modelling and real-time simulation of AUV/ROV dynamics and acoustic payload instruments since 2000. Scientific sponsored zoological surveys, such as deep ocean habitat mapping and commercial geo-referenced acoustic mapping of ship wrecks, are among ancillary activities conducted by this research centre. A new remote sensor deployment platform was required to meet the immediate (shallow water) survey needs of the MMRRC and to provide a flexible test bed for research in vehicle and control system development and in sensor systems going forward. The design of this new vehicle was motivated in part by practical difficulties and predicaments encountered in pre-survey instrument system design/integration and during inshore and offshore survey work with leased ROV sensor platforms. Key design requirements for the new platform included the following: ●

●

●

It should be able to support the core survey suite, including multibeam, sidescan, Inertial Navigation System (INS), Doppler Velocity Log (DVL), pressure (depth) sensor and Ultra-Short Base Line (USBL) transponder. It should accommodate both subsea survey activities and wide area (surface) survey activities in order to maximize equipment utility. It should be deployable using small surface support vessels, such as tugs or trawlers, thus reducing operational costs, as well as larger survey vessels.

Control design requirements included the following: ●

It should have built-in auto-tuning features for low-level controllers.

ROV LATIS: next generation smart underwater vehicle ● ●

13

It should be able to detect, isolate and accommodate thruster faults. It should be able to work in full automatic mode, i.e. automatic way-points navigation.

2.4.2 Features The main features of ROV LATIS are highlighted in the following: 1. 2. 3. 4. 5. 6. 7. 8.

Modular design with multiple modes of operation Very high positioning accuracy of ROV in deep water Semi-Automatic Speed Modes enable robust, stable and accurate ROV Course Following and ROV Dynamic Positioning with simple mouse click Fully-automatic way-points navigation with auto-compensation of ocean currents and umbilical drag effects Advanced 2D and 3D real-time visualization, providing better situation awareness Built-in thruster fault-tolerance and optimal control allocation for any thruster configuration Built-in auto-tuning of low-level controllers, providing optimal controller performance, regardless of changes in ROV configuration between missions Modular software architecture and extensive interface library enable easy system adaptation to any ROV and SHIP in the market

2.4.3 Operation modes ROV LATIS is a vehicle with multiple modes of operation (Toal et al., 2010). It can be operated on the surface as a survey platform either towed (Figure 2.2(a)) or thrusted by four horizontal thrusters (Figure 2.2(b)) to allow to surge, sway and yaw. It can also operate as an ROV fully controllable in 6DoF by four horizontal and four vertical thrusters (Figure 2.2(c)) or as ROV with submerged tow/holding line for operations in submerged tow or on station in strong currents (Figure 2.2(d)). In these various modes of operation it is used in conjunction with a fibre optic umbilical and winch; the umbilical carrying vehicle power, control and data from sensors and instruments.

2.4.4 Frame ROV LATIS is designed and constructed in modules, allowing for ease of handling in inshore surveys on small boats with small crews. Each component in this modular design (buoyancy modules, upper frame, lower toolskid, etc.) is kept to a twoman dry weight lift. The upper frame mounts the eight thrusters – four horizontal and four vertical. The lower toolskid frame carries the payload sensors (Figure 2.3). In the surface-tow or surface-thrusted modes of operation, overall vehicle buoyancy is maintained strongly positive by eight buoyancy modules mounted on the vehicle upper frame. While in surface-tow mode, a ‘Quick Release’ arrangement (‘On the Fly’ reconfiguration, Figure 2.4) allows the two top most buoyancy

14

Further advances in unmanned marine vehicles

(a)

(b)

(c)

(d)

Figure 2.2 Operation modes. (a) Surface-tow mode. (b) Surface-thrusted mode. (c) ROV operation mode. (d) ROV with submerged tow/holding line in strong currents

Figure 2.3 Design concept modules to be detached from the vehicle, reducing overall vehicle buoyancy to neutral or slightly positive. The vehicle can then be operated in ROV mode for submerged survey with control in six degrees of freedom (surge, sway, heave, pitch, roll and yaw). The syntactic foam buoyancy modules are depth rated to 1000 m, while other components integrated on the vehicle including payload sensors are depth rated up to 2000 m or beyond.

ROV LATIS: next generation smart underwater vehicle

Figure 2.4

15

‘On the Fly’ reconfiguration

2.4.5 Technical specifications Technical specifications are given in Table 2.1.

2.4.6 Overall hardware architecture Figure 2.5 illustrates the overall hardware architecture. Topside components include Control Cabin and winch with umbilical. The Control Cabin provides housing for power supply and control/supervision centre with computers, monitors and auxiliary equipment. Four Ethernet networks have been implemented: CONTROL NETWORK (navigation and control data), SURVEY NETWORK (PHINS-aiding sensors and sidescan), VIDEO NETWORK (video data) and MULTIBEAM NETWORK (multibeam data). A local GPS, mounted on top of the Control Cabin, is used for precise time synchronization of system components. Navigation data for position/orientation of support vessel (used in Augmented Reality Display) are obtained from navigation devices (GPS, Motion Reference Unit (MRU), etc.). The USBL output, with estimated ROV position, is forwarded from the USBL through the Control Cabin to PHINS via the SURVEY NETWORK. At the same time, ROV depth is sent from the depth sensor via Control Cabin to USBL as aiding information to resolve multipath problems. The umbilical has six/four copper cores for AC/DC power, respectively, and eight single-mode optical fibres (currently four of them are used – each network has a dedicated fibre core). On the topside, the umbilical is terminated with a Power Box and a Signal Box, mounted inside the Control Cabin. The 11-kW Power Supply Unit provides AC/DC power for the ROV through the Power Box. The Signal Box provides housing for Ethernet–Fibre Optic (FO) converters. Wetside components include five wet bottles, eight thrusters, four cameras, six lights, multibeam, sidescan, sound velocity probe, obstacle avoidance sensors, INS and a set of aiding sensors.

16

Further advances in unmanned marine vehicles

Table 2.1 Technical specifications Base vehicle Chassis Payload Max. depth Thrusters Power supply Instruments and navigation suite Multibeam sonar Sidescan sonar Sound velocity probe INS Depth DVL USBL interface GPS (surface)

Obstacle avoidance Cameras Pan and tilt Lights

Control system Embedded

Marine grade aluminium 100 kg 1000 m Seaeye SM4 (4 H, 4 V) 11 kW, upgradable Reson SeaBat 7125 Tritech SeaKing sonar 325 kHz Reson SVP-24 ixSea PHINS CDL Microbath (Digiquartz) RDI WorkHorse Navigator 600 iXsea GAPS CSI-Wireless Seres or Submersible GPS GPRS-6015G 6 Tritech single-beam echosounders Bowtech Explorer-3K monochrome 2 LCC-600 monochrome Tritech Typhoon colour Bowtech SS-109 3 Bowtech LED-1600 1 Bowtech LED-800 2 Tritech LED lite

Control and visualization software

Digital Logic EBX945 National Instruments CompactRIO ControlPC VisualizationPC LabVIEW 8.6

Umbilical Length Diameter Core

400 m 25 mm 6 AC, 4 DC, 8 single-mode optical fibres

Topside

The Umbilical Bottle serves as a signal/power junction box: high-voltage DC power lines are forwarded to the Thruster Bottle, while the AC lines and fibre optic cores are connected with the Power Bottle. The Power Bottle provides housing for DC power supply units, Ethernet–FO converters and RS-232/485 converter (used for pan and tilt control for the Pilot Camera). The multibeam Link Control Unit (LCU) is directly connected to this bottle. Also, this bottle is interfaced with four cameras: Pilot Camera, Umbilical Camera and two downward looking cameras. Power and signal lines are further forwarded to the Survey Bottle and Control Bottle. A set of aiding sensors is connected with PHINS through the Survey Bottle, including Doppler Velocity Log (DVL), depth sensor, USBL transponder and local submersible GPS receiver. A submersible GPS is included such that best navigation

ROV LATIS: next generation smart underwater vehicle Control Cabin VIDEO ● Video Data Processing ● Overlay MULTIBEAM ● Multibeam Data Processing

Local GPS ● Time Synchronization

SURVEY ● Sidescan Data Processing ● Adaptive Multi-Sonar Controller

Support Vessel

CONTROL ● Control ● Visualization ● Simulation (optional)

USBL Navigation Devices AC PSU

DC

Power Source AC

Power Box

Winch ● Slip Ring ● Tension Control

Signal Box ● Ethernet – FO Converters Umbilical

Umbilical

Water Line DC Thrusters

Umbilical Bottle ● Power Junction Box

AC

Power Bottle ● DC Power Supply Units ● Ethernet – FO Converters ● RS-232/485 Converter DC

Control Bottle ● Ethernet Serial Server ● Ethernet Switch ● Relay Module ● PC EBX945 ● CompactRIO ● Light Interface ● Thruster Interface

DC Survey Bottle ● Ethernet Serial Servers ● Ethernet Switch ● Relay Module

PHINS

Lights

FO Cores

Cameras

Multibeam

Thruster Bottle ● DC Power Filters ● HT/VT Interface

● Sidescan ● Aiding Sensors: ○ DVL ○ Depth ○ USBL ○ GPS2

Figure 2.5 Overall hardware architecture

17

18

Further advances in unmanned marine vehicles

solution can be maintained by the INS with GPS aiding right up to the point of diving, and GPS aiding can be re-established whenever the ROV surfaces. These sensors are used by PHINS to improve accuracy of ROV position estimation. The output of PHINS is directly sent to the Control Bottle for control purposes through a serial interface. At the same time, sidescan data, raw data from sensors and navigation data processed by PHINS are sent topside via the SURVEY NETWORK. All control-related components are mounted inside the Control Bottle. The PC EBX945 (mini PC) receives and processes ROV navigation data, transforms data into system states using standard control frames (body-fixed and earth-fixed), bundles data into clusters and shares it over the CONTROL NETWORK. NI CompactRIO is used as a real-time controller and I/O interface with thrusters, lights and leak detectors. The Thruster Bottle is used as an interface with thrusters: control data is received from the Control Bottle and DC power is provided from the Umbilical Bottle. DC power filters are used to eliminate spikes from the DC power lines. Ethernet Serial Servers (ESS) inside the Survey Bottle, Control Bottle and Control Cabin provide Ethernet to serial connections for RS-232, 422 and 485 devices. The serial ports are accessed over a LAN using Virtual COM Port connection mode, i.e. software drivers installed on topside computers create virtual COM ports that provide access to any remote port of an ESS, like any other serial port on the computer. Any program running on the computer and using COM ports can access the serial devices attached to the ESS. In this way the LAN becomes transparent to the serial device and the software running on the PC. Every bottle (except the Umbilical Bottle) is equipped with a leak detector. The outputs of leak detectors are connected as digital inputs to NI 9435 DI module and shared throughout the CONTROL NETWORK. Serial relay modules, mounted inside the Control Bottle and the Survey Bottle, are used to switch on/off selected system components. A serial relay module inside the Control Bottle switches the power of the PC EBX945, CompactRIO and bottom-pointing lights, while the relay module inside the Survey Bottle switches the power for PHINS, local GPS, depth sensor and DVL. Remote power control of system components is a powerful feature that increases availability and performance of the overall system, as sometimes a power down and restart is called for by some instrument or component if settings are changed.

2.4.7

Control architecture

In contrast to most common control architectures used in the modern ROV industry (where an ROV is equipped with basic I/O modules, while control synthesis is performed topside), ROV LATIS is equipped with a full real-time embedded control system, which performs all necessary data processing and control synthesis online, aboard the vehicle in real time. The block diagram of the ROV LATIS control system is shown in Figure 2.6. The ROV LATIS control system (Figure 2.6) utilizes a control allocation approach, where the actuator selection task (mapping of the total control efforts onto individual actuator settings) is separated from the regulation task (design of

ROV LATIS: next generation smart underwater vehicle Navigation Data (ROV)

Navigation Data (Support Vessel)

19

Pilot Input Interface

Interface & Data Management

Map Builder

Mission Builder Mission Planner

Way Points Database

Trajectory Planner

Arbitration Collaborative Behaviours

Coordinator

Standard Control Cluster

Exclusive Behaviours

Standard

Exclusive Mode Switch

Emergency Control Cluster Emergency AUV

Operation Mode Switch

ROV Mode Control Cluster ROV

Fusion

Obstacle Avoidance Control Cluster

Winner Control Cluster Fault Accommodation Fault Diagnosis

HT Saturation Bounds VT Saturation Bounds

Disturbances ROV

Low-Level Controllers d LLC + VJ

+

Synthesis

Control Allocation

d

Propulsion System

Rigid-Body Dynamics

Control Allocation Faults

Figure 2.6 Control architecture

total control efforts) in the control design (Omerdic et al., 2010b). The Mission Builder module, Arbitration module and Synthesis module perform the regulation task, while the Control Allocation module performs the actuator selection task. The main objective of the Mission Builder module is to transform the mission objective, pilot inputs and measured navigation data into the desired ROV trajectory, i.e. to formulate the trajectory planning problem. A description of the Arbitration module is given in the following. To achieve the desired trajectory under real-time constraints, a set of task executors (Exclusive Behaviours (only one active at a time) and Collaborative Behaviours (many active at a time)) has been developed. Each of these task executors is competing to take control of actuators. The control buffer concept has been developed to provide transparency and easy fusion of different task executor control demands. Each task executor produces its own control cluster inside the control buffer. A control cluster consists of Virtual Joystick components (to mimic direct controls generated by a

20

Further advances in unmanned marine vehicles

FF SP Offset + SP + – PV

Error

1 τFFs+1

KFF

KP

+

Relay

+

MV

KO KI s

+ dPV/dt

KD

tanh

–

Figure 2.7 Internal structure of LLC loop virtual pilot) and a set of settings for Low-Level Controllers (set points, feedforward inputs and on/off switches to enable/disable individual controllers). The Virtual Joystick components are normalized surge, sway and heave forces and roll, pitch and yaw moments. The Coordinator performs the fusion of these control clusters into the Standard Control Cluster. The Operation Mode Switch is used to switch between AUV Mode and ROV Mode. In AUV Mode, control algorithms are exclusively used to create control actions (write values and set switches inside control clusters), while in ROV Mode the pilot has full freedom to generate these actions by writing values directly to ROV Mode Control Cluster. However, in ROV Mode the pilot should be aware not to ‘fight’ with enabled low-level controller. For example, if the low-level heading controller is enabled to keep a set point (desired heading) of 60
, any yaw moment created by the pilot using an input device, such as a joystick or gamepad, is considered as a disturbance by the heading controller. Therefore, the controller will create corresponding actions to reject this disturbance and keep the heading at 60
. The Exclusive Mode Switch is used to switch between Standard and Emergency Control Clusters in AUV Mode. In the case of any leakage (water penetration inside any bottle), two Leakage Answer Modes are available. In Auto-Answer Mode, the main state machine will activate the Emergency State setting the Exclusive Mode Switch to Emergency, setting the Operation Mode Switch to AUV Mode and initiating automatic ROV recovery to the surface. In Manual-Answer Mode, the Operation Mode Switch is set to ROV Mode and the pilot is informed of the leakage, but no other automatic action is undertaken. In future trials, the Manual-Answer Mode may be upgraded to give the pilot regular warnings, and if the pilot does not perform any action in defined time, the system will switch to Auto-Answer Mode. Finally, the output control cluster is (optionally) blended with the Obstacle Avoidance Control Cluster to create the Winner Control Cluster, the ultimate ‘boss’ with exclusive rights to control the actuators. Inside the Synthesis module the Winner Control Cluster is unbundled into Virtual Joystick components (vector tVJ ) and the Low-Level Controller (LLC) cluster,

ROV LATIS: next generation smart underwater vehicle

21

which is used as one of the inputs to the LLC loop (vectors SP and FF in Figure 2.7). Other inputs include ROV navigation data (vectors PV and dPV/dt) and other parameters (vectors SP Offset, tFF , KFF, KP, KI, KD and Relay Amplitude). There is a single controller for each degree of freedom (DOF). Surge and Sway controllers are velocity controllers, while Heave, Roll, Pitch and Yaw are position controllers. Each controller generates a manipulated variable MV to be applied to drive actuators, in order to keep process variable PV as close as possible to set point SP. Individual outputs are bundled into a vector of normalized forces and moments tLLC . If a controller is disabled (not active), the corresponding MV (element of tLLC ) is set to zero. Otherwise, the controller output is calculated as a normalized output of a modified PID controller. Feed-forward (FF) input improves the tracking performance in the case of the time-varying SP vector. Vector SP Offset is used to avoid integrator saturation problems. Finally, in special ‘Auto-tuning’ operation mode, the controller output is generated using a relay (see Figure 2.7). Between successive ROV missions, it is likely that some of the onboard instruments/sensors/equipment will be added/removed/replaced, leading to changes in dynamic properties of the ROV (mass, moments of inertia, drag properties, etc.). Controllers optimally tuned for a particular vessel configuration will not give the optimal performance in the case of a change in configuration. Auto-tuning of LLC is an advanced feature of the control system, yielding optimal controller performance, regardless of changes in configuration. It is recommended that the autotuning is performed at the beginning of a mission. Two types of auto-tuning algorithms have been developed. Velocity controllers are tuned by recording and utilizing the force-speed static characteristics. Auto-tuning of position controllers, described below, utilizes self-oscillations. Auto-tuning algorithms described in Miskovic et al. (2006) have been expanded for four DOF controllers: Heave, Roll, Pitch and Yaw. The auto-tuning process involves the following steps (see Figure 2.8): 1. 2. 3. 4.

Generate self-oscillations Wait for transient stage to finish Measure amplitude and period of steady-state oscillations Find new values of controller gains using tuning rules

Performance of controllers with new gains is shown in Figure 2.9. A novel set of tuning rules for underwater applications has been developed, which provides the optimal performance of low-level controllers in the case of configuration changes and the presence of disturbances (waves and sea currents). The final output td of the Synthesis module, which represents the total control efforts to be exerted by thrusters, is obtained by summation and normalization of vectors tVJ and tLLC . The Control Allocation module in Figure 2.6 performs the actuator selection task. The thruster fault-tolerant control system consists of two subsystems: 1.

The Fault Diagnosis Subsystem (FDS) uses fault detector units to monitor the states of thrusters (Omerdic and Roberts, 2003).

22

Further advances in unmanned marine vehicles

Figure 2.8 ROV self-oscillations during auto-tuning

Figure 2.9 Performance of Heave (Depth) and Yaw (Heading) controllers after auto-tuning

ROV LATIS: next generation smart underwater vehicle 2.

23

The Fault Accommodation Subsystem (FAS) uses information provided by the FDS to accommodate faults and perform control reallocation. The output of FAS controls the HT/VT Saturation Bounds sliders.

A hybrid approach for control allocation (Omerdic et al., 2004) is based on integration of the pseudo-inverse and the fixed-point iteration method. It is implemented as a two-step process. The pseudo-inverse solution is found in the first step. Then the feasibility of the solution is examined analysing its individual components. If violation of actuator constraint(s) is detected, the fixed-point iteration method is activated in the second step. In this way, the hybrid approach is able to allocate the exact solution, optimal in the l2 sense, inside the entire attainable command set. This solution minimizes a control energy cost function, the most suitable criteria for underwater applications. A hybrid control allocation approach is implemented inside the Control Allocation Express VI (Figure 2.10). All inputs to the Control Allocation Express VI (Figure 2.10) are normalized to standard intervals. Depending on user-defined settings inside the Configuration Dialog Box (Figure 2.11), the outputs (actuator settings) can be compensated for non-symmetrical propeller T-curves and adapted (scaled and/or rounded, if necessary) to meet requirements of thruster input interfaces. The degree of usage of each thruster is controlled by sliders HT (VT) Saturation Bounds. The position of these

VT Properties

+

1 +

HT Properties 1 +

Inputs 3 +

Control Allocation Plus VT Properties VT Thruster Configuration HT Properties HT Thruster Configuration Inputs HT Outputs HT FPI Parameters HT General Parameters HT Shapes HT Vectors VT Vectors VT Outputs VT FPI Parameters VT General Parameters VT Shapes

HT Outputs 2 +

VT Outputs 2 +

Figure 2.10 Control Allocation Express VI: Block diagram connections

Figure 2.11 Control Allocation Express VI: Configuration dialog box

ROV LATIS: next generation smart underwater vehicle

25

sliders is determined by the ROV pilot or fault accommodation system, depending on the state of thrusters (healthy, partial fault or total fault). The control allocation problem is decomposed into motions in horizontal and vertical planes. The current version of the Control Allocation Express VI covers most common thruster configurations with 1/2/3/4 thrusters. The user can choose the combination that matches with the particular thruster configurations of a given vehicle. In the fault-free case, the control system uses thrusters over the maximumsized attainable command set in an optimal way, minimizing the energy consumption, the most suitable criteria for underwater applications. In the case of thruster faults, the control system automatically accommodates faults, keeping full controllability of the ROV for low speeds and providing the opportunity for safe continuation of a mission in the presence of thruster fault(s). The physical layout of LabVIEW software modules (VI distribution over system components) is given in Figure 2.12. Arbitration, Synthesis and Control Allocation software modules are executed on EBX945. The CompactRIO system consists of two main components: ● ●

Real-time controller NI cRIO-9004 Chassis cRIO-9104 with the FPGA chip and the following I/O modules: * NI 9481 (DO, Slot 4) – Light Interface * NI 9435 (DI, Slot 5) – Leak Detectors * NI 9263 (AO, Slot 6) – Light Interface * NI 9401 (DO, Slot 7) – HT Control Interface * NI 9401 (DO, Slot 8) – VT Control Interface

The real-time monitoring of thruster control signals (Figure 2.13) and the link between the Shared Variable Engine (EBX945) and the FPGA is performed by the Real-Time Controller. The FPGA application creates thruster control signals and provides the interface for external lights and leak detectors.

2.5 Innovations Four main innovations of the proposed system include Advanced Pilot Interface, 2D Topview Display, 3D Real-Time Augmented Reality Display and Advanced Control System with Fault-Tolerant Capabilities. The Advanced Pilot Interface (Figure 2.14) presents all important control data to the ROV pilot using familiar graphic controls and indicators. The pilot is able to use a combination of joystick, gamepad, mobile device, mouse or keyboard as input devices to generate commands, switch operating modes and enable/disable lowlevel controllers. Set points can be entered numerically (e.g. using numeric control fields) or graphically (e.g. moving instrument pointers by mouse). The pilot can also easily switch between manual mode, semi-automatic modes (Follow Desired Speed & Course, Keep Current Position and Go To Position) and fully-automatic mode (automatic navigation through way points).

26

Further advances in unmanned marine vehicles Navigation Data (Support Vessel)

ControlPC ● Control Centre ● 2D Advanced Topview Display ● ROV Pilot Interface ● LLC Monitoring

VisualizationPC

Ethernet Switch (CONTROL NETWORK)

● 2D Map Builder ● 3D Real-Time Augmented Reality Display ● Thruster Management & Monitoring ● Mission Planner & Way Points Editor

Ethernet ↔ Fibre Optic Water Line Fibre Optic ↔ Ethernet

Ethernet Switch (CONTROL NETWORK)

EBX945 ● Shared Variable Engine ● Arbitration ● Synthesis ● Control Allocation

NI Compact RIO

Navigation Data (ROV)

● RT: Monitoring of Thruster Control Signals & Link between EBX945 & FPGA ● FPGA: I/O Interface with Thrusters, Leak Detectors and Lights

Figure 2.12 Software modules—physical layout

ROV LATIS: next generation smart underwater vehicle

27

Figure 2.13 Real-time monitoring of thruster control signals (software module running on RT Controller, CompactRIO)

Figure 2.14 Advanced Pilot Interface

28

Further advances in unmanned marine vehicles

There are four Speed Modes: ‘Independent Surge & Sway’, ‘Follow Desired Speed & Course’, ‘Keep Current Position’ and ‘Go To Position’. The ‘Follow Desired Speed & Course’ Speed Mode enables reliable and precise speed and course following with excellent dynamic performances (speed settling time of 3 s and course turn resolution of 1
). A simple and intuitive graphical user interface enables coupled or fully independent control of heading and course (Figure 2.15). The control system automatically compensates effects of ocean currents and umbilical drag, yielding a performance that cannot be matched by most experienced human ROV pilot in manual control mode. The ‘Keep Current Position’ and ‘Go To Position’ Speed Modes enable robust, stable and highly accurate Dynamic Positioning of the ROV in deep water (Figure 2.16). Desired ROV position (Target Marker) can be entered numerically (using control fields) or graphically (using simple mouse ‘Drag and Drop’ operation). The approach speed to the Target Marker follows a Speed Profile (Figure 2.17), which can be dynamically updated.

Figure 2.15 Speed Mode ‘Follow Desired Speed & Course’. (a) Heading automatically follows desired Course. (b) Independent control of Heading and Course

ROV LATIS: next generation smart underwater vehicle

29

Figure 2.16 Speed Mode ‘Go To Position’. (a) Select Target Marker. (b) Move to new destination. (c) Destination reached

Figure 2.17 Speed profile The 2D Topview Display (Figure 2.18) shows a top view of the working zone and includes features like auto zoom, nav info display, floating heading indicators, visualization of way points, real-time visualization of sensors measurements (INS, DVL, USBL, GPS, etc.), distance and angle measurements tools, ROV-fixed, SHIP-fixed and free Lever Arms, etc. The 3D Real-Time Augmented Reality Display (Figure 2.19) provides 3D realtime visualization of the support vessel, ROV, ocean energy device, ocean surface, seabed, etc. Virtual components (ocean energy device, ship, ROV) have the same appearance as corresponding real components (same dimension, colour, etc.). They are driven by real-time measurements obtained through component interfaces. Waves on the surface are generated from estimated sea state, while the seabed is built from previous bathymetry data. The control system includes fast auto-tuning of low-level controllers, automatic thruster fault detection and accommodation, semi-automatic and fully-automatic control modes, optimal control allocation of thrusters, etc.

30

Further advances in unmanned marine vehicles

Figure 2.18 2D Topview Display

Figure 2.19 3D Real-Time Augmented Reality Display

ROV LATIS: next generation smart underwater vehicle

31

Figure 2.20 Tracking performance in absence of ocean currents: Go To Target Method ¼ ‘LOS only’

There are two methods implemented for automatic way-points navigation: ‘Line of Sight (LOS) only’ and LOS þ Correction’. The ‘LOS only’ Go To Target Method gives good performance in situations where there is little or no ocean current (Figure 2.20). In this case there is no lateral corrective action generated by controller to bring the ROV on the desired path. Desired course is calculated as a bearing from ROV to the next way-point and heading follows course automatically. The ‘LOS þ Correction’ Go To Target Method is used in the presence of ocean currents (Figure 2.21). In this case, in addition to the standard LOS algorithm, controllers generate adaptive lateral corrective action to bring and keep the ROV on the desired path.

32

Further advances in unmanned marine vehicles

Figure 2.21 Tracking performance in the presence of northern ocean currents of 0.3 m/s: Go To Target Method ¼ ‘LOS þ Correction’

2.6 Test trials The performance of the overall system has been successfully validated through a series of test trials off the coast of Connemara, in Galway Bay and on the Shannon Estuary.

2.6.1

CE-09-04 Cruise (ROV LATIS onboard RV Celtic Explorer)

Offshore trials (survey cruise; CE-09-04) of ROV LATIS were undertaken off the west of Ireland’s Connemara coastline on 26 February 2009. The vehicle was mobilized using the ship RV Celtic Explorer. The cruise consisted of a day in Galway port integrating and testing ROV and ship systems and 6 days at sea. During the trials all of the ROV’s systems were proved, sensor interoperability was

ROV LATIS: next generation smart underwater vehicle

33

demonstrated and comprehensive vehicle diagnostics were performed. In addition, system identification was performed on the ROV and tuning of vehicle controllers was successfully carried out. A series of pre-planned survey missions was also conducted. These missions were used to trial the operation of the vehicle, the MPPT Ring, the ROVs augmented reality topside control and visualization, the multi-sonar controller and the use of vision systems for near seabed navigation. Selected survey result (INS/DVL calibration) is reported further. If a Doppler Velocity Log (DVL) is used as an aiding sensor for an Inertial Navigation System (INS), it has to be calibrated with respect to the INS, prior to being used for navigation aiding. Speed inputs received from the DVL are measured in the DVL-reference frame without any pre-compensation. These inputs have to be compensated through an INS/DVL calibration procedure by calculating misalignment and scale factor between the INS and DVL-reference frames. The INS on the ROV LATIS is an ixSea PHINS and the log sensor is an RDI Workhorse Navigator 600 kHz DVL. For the purpose of this calibration a local DGPS receiver was mounted on the ROV, which was towed behind the ship in a straight line in shallow water on the surface for approximately 2 km. ROV towing, as seen on the 3D Real-Time Augmented Reality Display from different viewpoints, is shown in Figure 2.22. ROV and ship trajectories during INS/DVL calibration are shown in Figure 2.23. After calibration the following values were obtained: Roll and Pitch misalignment: 0
(both instruments are mounted on opposite sides of the same plate); heading misalignment: –44.654
and scale factor: 0.1%.

2.6.2 CV-10-029 Cruise (ROV LATIS onboard RV Celtic Voyager) Full system test and validation of assistive tools was performed during this cruise in August 2010. ROV LATIS was mobilized using the ship RV Celtic Voyager (Figure 2.24). The objectives of this research survey were focused on collection of field data for roll out of renewable ocean energy projects (tidal and wave energy) in

Figure 2.22 3D Real-Time Augmented Reality Display: INS/DVL calibration with ROV LATIS towed behind the ship. (a) Pilot camera view. (b) View from ship’s deck

34

Further advances in unmanned marine vehicles ROV UTM Position SHIP UTM Position 5900600.00 5900400.00 5900200.00 5900000.00 5899800.00 5899600.00 5899400.00 5899200.00 Northing

5899000.00 5898800.00 5898600.00 5898400.00 5898200.00 5898000.00 5897800.00 5897600.00 5897400.00 5897200.00 5897000.00 5896800.00 461000.00

461200.00

461400.00

461600.00

461800.00

462000.00

Easting

Figure 2.23 ROV and ship trajectories during INS/DVL calibration

Figure 2.24 Smart ROV LATIS on the deck of RV Celtic Voyager

west coast of Ireland using the smart ROV LATIS and ocean sensors. Specifically, this involved the deployment of ROV LATIS equipped with payload instrumentation and use of ship-mounted sensor to acquire data in detailed energy farm site analysis and in operations/performance testing of ROVs in support of offshore energy installations. The standard suite of navigation equipment was extended with

ROV LATIS: next generation smart underwater vehicle

35

submersible GPS and leased BlueView forward looking sonar. Selected survey results are reported here. Remote Current/Wave Radar has been acquired by NUIGalway and is being installed at present. The original plans at the time of ship-time application were for the simultaneous acquisition of current and wave data in the field with ROV-mounted instruments to calibrate Remote Current/Wave Radar measurements. Delays in the delivery and commissioning of the wave radar meant it was not operational at the time of the CV-10-029 survey. Regardless of these delays, the MMRRC team performed a series of ocean currents measurements in Galway Bay, both on the surface and in the water column. Acoustic Doppler Current Profiler (ADCP) data was acquired with both ship-mounted ADCP and ROV-mounted ADCP instruments in transects across the wave radar scan sector. The ADCP work in Galway Bay (Figure 2.25) included a mix of ‘on station’ work with the ship at anchor and ‘ROV in tow’ work with the platform towed along a transect behind the ship. The envelope of ROV manoeuvring performances and operation limits in strong currents was investigated through dynamic changes in ship’s speed, ROV depth, heading and attitude. A 3D Real-Time View of the ROV towed behind the ship is shown in Figure 2.26 (‘bubbles’ represent the trajectory behind ROV). A multibeam survey of Foynes Port was performed with ROV LATIS deployed from a stationary ship in surface operation mode. Weather conditions were calm, with strong East–West tidal currents. To demonstrate the quality of semi-automatic control modes, the ROV was initially controlled manually by joystick (Figure 2.27, Transect 1). After switching to the semi-automatic Speed Mode: Follow Desired Speed & Course (Figure 2.15(a)), a much smoother trajectory and better turn control was obtained (Figure 2.27, Trajectory 2 and Turn). The Desired Speed was set to 0.3 m/s, while the

Figure 2.25 ‘ROV in tow’ work, where ROV is towed along a transect behind the ship (Real-Time Pilot Camera View)

36

Further advances in unmanned marine vehicles

Figure 2.26 Submerged ‘ROV in tow’ performing multibeam survey and collecting ADCP data simultaneously (3D Real-Time View)

Figure 2.27 ROV trajectory during multibeam survey of Foynes Port

ROV LATIS: next generation smart underwater vehicle

37

Desired Course was dynamically updated with a pointer or numerical control with resolution of 1
. As a result of switching to semi-automatic mode, the overall survey was performed in a shorter time, and the quality of raw multibeam data collected during the survey was very high. The final bathymetry of Foynes Port, obtained after post-processing of raw multibeam data, is shown in Figure 2.28. Way-point navigation test trials have been performed in Galway Bay close to the Aran Irelands. The ROV had to follow a path defined by way points with desired speed and desired depth/altitude. Comparison of manual versus automatic way point navigation modes is shown in Figures 2.28 and 2.29, respectively. The desired path consists of two straight run lines connected with semi-circular arc. In the first case (Figure 2.29), the ROV was controlled by Richard Vandervoort, an experienced ROV pilot. The ROV position estimations from PHINS and USBL are both shown on the display in real time. Regardless of pilot skills, significant tracking errors are visible between desired and actual trajectory. In the second case (Figure 2.30), the same trajectory has been followed by the ROV in the ‘LOS only’ automatic way point navigation mode (see Figure 2.20). Significant improvements are achieved in this case, including reduced tracking error and better speed/altitude following.

Figure 2.28 Foynes Port bathymetry

38

Further advances in unmanned marine vehicles

Figure 2.29 Manual way-point navigation by experienced ROV pilot

Figure 2.30 Automatic way-point navigation (without human in the loop)

ROV LATIS: next generation smart underwater vehicle

39

Other trials (not shown here) included very successful test of thruster faulttolerance (full ROV controllability in six DOF with two disabled thrusters: one vertical and one horizontal) and repeatability test (five times repeated return to the same location over 2-h period with very high positioning accuracy).

2.7 Feedback According to Donovan Tulk, MI student and future ROV pilot, ‘ROV LATIS is exceptional, cutting edge technology, and most likely the future of ROV piloting’. (International Collaboration in ROV Development, 2010). Richard Vandervoort, ROV expert, said Recently, I had the opportunity to be part of the sea trials of a prototype vehicle, the ROVLATIS, onboard the research vessel Celtic Voyager. It is a truly remarkable piece of technology, and I am convinced it represents the next generation of underwater robotics. The ability of the vehicle to navigate, its user friendly piloting controls, dynamic and static positioning capabilities and built-in malfunction corrective functions far surpasses anything that presently exists on the market and as such would greatly increase both the speed and accuracy of subsea operations. It is my estimation that a significant savings in the time required to perform underwater tasks, along with the commensurate savings in operational costs, would be achieved should this technology become available for wide scale commercial and industrial use.

2.8 Implementation To enhance operations of existing ROVs and implement assistive tools, the overall MPPT Ring suite is integrated into the ROV Upgrade Kit. The ROV Upgrade Kit is a mix of hardware and software components that should be installed on a work-class ROV and inside the Control Cabin in order to give increased levels of automation, to make ROV operations easier and to save expensive ship time by 20% or more. The ROV Upgrade Kit consists of Dry Components (installed inside the ROV Control Cabin on the ship deck, see Figure 2.31) and Wet Components (installed on the ROV frame/toolskid, see Figure 2.32). Software modules running on the ControlPC (Figure 2.31) perform the following tasks: 1. 2. 3. 4. 5.

Supervise and control overall mission Read ROV and SHIP navigation data Present ROV and SHIP navigation data to pilot using advanced 2D real-time display Provide graphical interfaces to display control data and acquire commands from the pilot Perform high-level control tasks

40

Further advances in unmanned marine vehicles SHIP

M2

ControlPC

M1

M2

ROV

M1

Network/Serial Interface

Network/Serial Interface

Dry Components (Control Cabin)

VisualizationPC

Figure 2.31 Dry Components, installed in the Control Cabin on the ship deck

INS

Existing ROV Bottle

Interface Bottle (PC104) Wet Components (ROV)

Aiding Sensors: ● GPS ● Depth ● DVL ● USBL ● LBL ● CTD ● SVP

Figure 2.32 Wet Components, installed on the ROV frame/toolskid 6. 7. 8.

Provide remote access to the onboard PC104 on the ROV to monitor low-level control modules, and to configure the INS and Aiding Sensors (Optional) If a direct link between USBL and INS is not possible, perform protocol synchronization and matching (Optional) Create auxiliary output messages (e.g. messages for Video Overlay, Global Mapper)

Software modules running on the Visualization PC (Figure 2.31) perform the following tasks: 1. 2. 3. 4.

Reading and auto-tilling of background maps Presenting ROV and SHIP navigation data to pilot using advanced 3D realtime display Providing graphical tools for mission planning Providing visualization for thruster saturation bounds that informs the pilot about the feasibility of commanded inputs

Software modules running on the PC104 (Figure 2.32) perform the following tasks:

ROV LATIS: next generation smart underwater vehicle 1. 2. 3. 4. 5. 6. 7.

41

Receive and process INS data Transform data into system states using standard control frames (body-fixed and earth-fixed) Provide shared variable engine for data distribution Perform low-level control tasks (Arbitration, Synthesis and Control Allocation) Send thruster demands to the Thruster Control Unit via network or (Optional) serial link Perform initial setup of the INS and Aiding Sensors through Remote Desktop from the ControlPC Perform data logging

2.9 Conclusions This chapter described the main features of ROV LATIS, the next generation smart underwater vehicle. The vehicle was built as a prototype platform to demonstrate system validity and operability and to prove new technologies developed in the Mobile & Marine Robotics Research Centre, UL. Field trials have demonstrated that the ultimate objective (saving ship time and making ROV operations easier) can be achieved through: ●

●

●

●

Improved user interface (advanced 2D and 3D displays – better situation awareness) Advanced control modes (enabling ROV pilots with average skills to achieve exceptional results) Use of state-of-the-art positioning and orientation instruments (fibre gyro INS), which provide high-accuracy control and navigation data (yielding easier postprocessing and better quality of bathymetry data) Improved Control Cabin – Ship’s bridge communication (better synchronization of ROV and Ship motions)

Instruments such as fibre gyro INS are expensive; however, in the overall cost of a work-class or light work-class ROV they are not so expensive. Yet, they facilitate the implementation of the advanced control systems described in this chapter by providing highly precise position and orientation information in real time. The dividend for including precision navigation instruments is unlocking the possibility of implementing the advanced control and auto pilot systems as described in this paper and implemented on ROV LATIS. Further work will include implementation and test trials of smart technologies (bundled into the ROV Upgrade Kit) on commercial work-class ROV in collaboration with MI SOT, Canada. The project is expected to start at the end of 2011.

Acknowledgements The development of ROV LATIS and of the modelling and operations support tools described has been supported by funding under the Irish Marine Institute and

42

Further advances in unmanned marine vehicles

the Marine RTDI Measure, Productive Sector Operational Programme, National Development Plan 2000–2006 (PhD-05-004, INF-06-013 and IND-05-03); Science Foundation Ireland under Grant Number 06/CP/E007 (Science Foundation Ireland – Charles Parsons Energy Research Awards 2006); HEA PRTLI 3 (MSR3.2 project – Deep Ocean Habitat Mapping using and ROV); HEA PRTLI 4 Environment and Climate Change Impacts and Responses Project/Environment Graduate Programme; and Enterprise Ireland Commercialization Fund Technology Development 2007 projects – MPPT Ring (CFTD/07/IT/313, ‘Multi-Purpose Platform Technologies for Subsea Operations’) and PULSE RT (CFTD/07/323, ‘Precision Underwater Accelerated Sonar Emulation in Real Time’). List of acronyms and abbreviations ADCP AO AUV COG DGPS DI DO DOF DVL FAS FDS FO FPGA GAPS GPS GUI HT INS INSS IRM LLC LOS MMRRC MOU MPPT Ring PHINS PSU RIO ROV RT SOG SOT SVP USBL UUV VR VT VUL

Acoustic Doppler Current Profiler Analog Output Autonomous Underwater Vehicle Course Over Ground Differential Global Positioning System Digital Input Digital Output Degree of Freedom Doppler Velocity Log Fault Accommodation System Fault diagnosis system Fibre Optic Field Programmable Gate Array USBL made by ixSea Global Positioning System Graphical User Interface Horizontal Thruster(s) Inertial Navigation System Irish National Seabed Survey Inspection, Repair and Maintenance Low-Level Controller(s) Line Of Sight Mobile & Marine Robotics Research Centre Memorandum Of Understanding Multi-Purpose Platform Technologies for subsea operations INS made by ixSea Power Supply Unit Reconfigurable Input Output Remotely Operated Vehicle Real-Time Speed Over Ground School of Ocean Technology Sound Velocity Probe Ultra-Short Base Line Unmanned Underwater Vehicle Virtual Reality Vertical Thruster(s) Virtual Underwater Lab

ROV LATIS: next generation smart underwater vehicle

43

References Beaudoin, G., Robertson, D., Doherty, R., Corren, D., Staby, B. and Meyer, L. (2010). Technological challenges to commercial-scale application of marine renewables. Oceanography, Journal of The Oceanography Society, vol. 23, no. 2, pp. 32–41. Curtis, B. (2010). Simulation reality. The Journal of Ocean Technology, vol. 5, no. 3, ISSN 1718-3200. Grehan, A., Wilson, M., Guinan, J., Riordan, J., Molnar, L., Omerdic, E., et al. (2005). ROV investigations of cold-water coral habitats along the porcupine bank margin, West Coast of Ireland. 3rd International Symposium on DeepSea Corals, Miami, Florida, USA. Grehan, A., Toal, D. and Brown, C. (2006). ROV investigations of cold water coral habitats in the porcupine/rockall off the west coast of Ireland. The Irish Scientist Yearbook, no. 14, p. 96. International Collaboration in ROV Development, The Journal of Ocean Technology, vol. 5, no. 3, 2010. Lukas, U.F. (2010). Virtual and augmented reality for the maritime sector – applications and requirements. 8th IFAC Conference CAMS 2010 (Control Applications in Marine Systems), 15–17 September. Rostock-Warnemu¨nde, Germany. Lin, Q. and Kuo, C. (2001). On applying virtual reality to underwater robot tele-operation and pilot training. The International Journal of Virtual Reality, vol. 5, no. 1. Miskovic, N., Vukic, Z., Barisic, M. and Tovornik, B. (2006). Autotuning autopilots for micro ROVs. Proceedings of 14th Mediterranean Conference on Control and Automation (MED06), Ancona. Omerdic, E., Riordan, J., Molnar, L. and Toal, D. (2006). Virtual underwater lab: Efficient tool for system integration and UUV control development. 7th IFAC Conference on Manoeuvring and Control of Marine Craft (MCMC 2006), Lisbon, Portugal. Omerdic, E., Riordan, J. and Toal, D. (2008). MPPT ring – Multi-purpose platform technologies for subsea operations. IFAC Workshop on Navigation, Guidance and Control of Underwater Vehicles, Killaloe, Ireland. Omerdic, E., Toal, D. and Leahy, M. (2010a). Assistive tools for system integration, deployment, monitoring and maintenance of ocean energy devices. Journal of Engineering for the Maritime Environment, vol. 224, no. 2, pp. 155–172. Omerdic, E., Toal, D., Nolan, S. and Ahmad, H. (2010b). Smart ROV LATIS: Control architecture. UKACC International Conference on CONTROL 2010, Coventry, UK. Omerdic, E. and Roberts, G. (2004). Thruster fault diagnosis and accommodation for open-frame underwater vehicles. Control Engineering Practice, vol. 12, pp. 1575–1598.

44

Further advances in unmanned marine vehicles

Omerdic, E., Roberts, G. and Toal, D. (2004). Extension of feasible region of control allocation for open-frame underwater vehicles. CAMS 2004, Ancona. Riordan, J. (2006). Performance Optimised Reverberation Modelling for Real-Time Synthesis of Sidescan Sonar Imagery. PhD in Computational Ocean Acoustics, University of Limerick, Limerick, Ireland. Rourke, F.O., Boyle, F. and Reynolds, A. (2009). Renewable energy resources and technologies applicable to Ireland. Renewable and Sustainable Energy Reviews, vol. 13, pp. 1975–1984. Toal, D., Omerdic, E., Riordan, J. and Nolan, S. (2010). Multi-mode operations marine robotics vehicle – mechatronics case study, Ch. 7, in: D. Bradley and D.W. Russel (Eds.), Mechatronics in Action: Case Studies in Mechatronics – Applications and Education, ISBN 978-1-84996-079-3, Springer-Verlag, London.

Chapter 3

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles Bramley J Murton , Veit Hu¨hnerbach and Jo Garrard 1

1

2

3.1 Background 3.1.1 System requirements The HyBIS concept was born out of the desire, as marine geoscientists, to survey and interact with the deep-ocean floor without recourse to expensive and complex work-class remotely operated vehicle (ROV) technology. Up until recently, the majority of deep-ocean floor sampling by research scientists has been by ‘blind’ technology: deploying passive mechanical sampling devices such as corers or grabs over the side of a ship. Yet, it has become increasingly apparent in recent years that the ocean floor is far more patchy than we had previously realized. This patchiness takes the form of small areas of biological activity (e.g. deep-water corals and cold seeps) or localized geological activity (e.g. hydrothermal vents and mineral deposits). On bare-rock outcrops, the patchiness is even more pronounced. Marine geologists are no different from their terrestrial counterparts and need to see how different rock types relate to one another in order to understand the processes that have formed and shaped them. This requires visual surveys and targeted sampling. There is also a need to interact with the seafloor to deploy, operate and recover heavy instruments. Dropping instruments over the side of a ship is no longer sufficient for high-resolution geophysical studies. Instead, there is a growing need to place instruments on solid parts of the seabed, with precise positioning and often with a preferred orientation. Similarly, deploying long-term observatories requires precise location and even the deployment of distributed, cabled sensors. Hence our specification was to develop an easy-to-use system that would operate to 6000-m depth, carry high-definition (HD) cameras and powerful lights, have some basic manoeuvrability and be able to accommodate a range of tasks including largevolume seabed sampling, delicate manipulation and the deployment, interaction and recovery of heavy payloads and seafloor instruments. As academic scientists, our budget was modest, both for the acquisition of the vehicle and for its operation 1 2

National Oceanography Centre, Southampton, UK Hydro-Lek Ltd, Berkshire, UK

46

Further advances in unmanned marine vehicles

offshore, yet its specified capability was high. The challenge, therefore, was not to repeat the design path for a conventional ROV vehicle, but to consider an alternative design concept that would meet the specific requirements of the science. We already had a number of existing parameters that the HyBIS system had to work with. Our research vessels are fitted with a combined electrical and fibreoptic armoured umbilical tether of up to 10 km long with a safe working load (SWL) of 3 tonnes at 6000 m. The vessels have full dynamic positioning (DP) and also have, as standard, a mid-ship launch and recovery system (LARS). However, none of our vessels have heave-compensated winches. Although the existing winch, umbilical and LARS reduced the need to acquire specific systems to deploy HyBIS, the lack of heave compensation is also a limiting factor.

3.1.2

Design considerations and implications

Our design-tree had to accommodate a large sample volume and instrument deployment payload capability. The vehicle’s horizontal manoeuvrability would have to encompass a radius of at least 2% of the water depth. We needed real-time control and visibility. The system had to be adaptable and have versatility for additional instrumentation and modules with real-time data telemetry. To meet the requirements to control a variety of tools, we opted for an electric hydraulic system. As a result of the payload requirements, floatation was ruled out as an option, and hence the vehicle was not able to be neutrally buoyant. Thus, a direct tether would satisfy the payload requirements, but the vertical stability of the vehicle would be affected by the ship’s heave. For example, analysis of heave from a mid-ship deployment position on our 98-mlong research vessels indicated a 5-s frequency with heave of less than 2 m in sea states up to and including F5. Similarly, the vertical length of the tether and the power of the thrusters would determine the horizontal manoeuvrability of the vehicle. However, the advantages of not having floatation are the reduced size and mass of the vehicle, a lower cost, a requirement for horizontal thrusters only and a greater payload capacity. Dynamic positioning of our ships allows metre-scale accuracy in station holding and survey line navigation modes. Given a typical working depth of 3000 m, a 2% radius of manoeuvrability would enable operations around a radius of 60 m from the ship’s position, which is sufficient to allow us to survey and sample at the scale of patchiness that we require. Direct tether would also enable us to deploy instruments of up to 500 kg onto the seafloor. However, large-volume sampling is limited by the potential for high pull-out loads rather than any sample weight limit. Working with a 3-tonne SWL at 6000 m, we restricted the sample volume to 0.25 m3 with maximum depth of penetration into soft sediment of 30 cm. Despite this restriction, modelling of potential pull-out loads in particularly clay-rich sediment was unable to rule out the possibility of ‘sticktion’, in which the plastic behaviour of some sediment types would result in effective capture of the sampling grab. Although the HyBIS system was specified to enable both repeated opening and closure of the sampling grab, we also required the capability of jettisoning any sampling tool to protect both the 10-km-long umbilical and the rest of the HyBIS vehicle.

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

47

3.2 A modular design concept Meeting the requirements for the HyBIS vehicle with a budget equivalent to the cost of the 10-km-long umbilical cable was a considerable challenge. Referring to the design-tree in which versatility was imperative, a solution was devised that would allow for a modular design configuration. This concept (Figure 3.1) resulted in a command module that would carry the power supply, controls, thrusters, hydraulic pumps, cameras, lights and telemetry. This command module would be firmly attached to the umbilical cable and have a payload capacity of 500 kg (Figure 3.2). The base of the command module would comprise a standard docking template accommodating any number of detachable modules. The most basic of these was a ‘clam-shell’ grab. Other modules would include a tool sledge with a five-function manipulator arm and retractable sample tray, an instrument payload deployment module and a lander recovery module with a passive drum carrying 600 m of 20 tonne breaking-strain recovery line. All the modules have in common the ability to be remotely jettisoned from the command module.

3.2.1 The command module The command module is an open-chassis structure with a footprint of 1200 mm  1300 mm and height of 600 mm (Figure 3.3). Fabricated from 316-grade tubular stainless steel, it has a suspension point with an adjustable centre-of-gravity that distributes the load across the upper square frame of the chassis, which is, itself, strengthened by cross-bracing to the lower square of the chassis. This cross-braced open-frame structure ensures an even load spread across each corner of four vertical

Figure 3.1 HyBIS design concept. A top Command module (pale grey) that comprises electronics, power supply, thrusters, cameras and lights attached to tool modules (dark grey) via a hydraulic release and docking system

48

Further advances in unmanned marine vehicles

Figure 3.2 Drawing of the HyBIS command module showing its major elements: EP, electronics pod; PP, power pod; L, lights; C, cameras; T, thrusters; HPP, hydraulic power pumps; TB, tilt-bar

Figure 3.3 HyBIS command module being docked with one of the tool modules on board the RRS Discovery. The command module can be deployed alone as a manoeuvrable visual survey platform. It is released from the various sampling tools via hydraulic-activated release pins. The lower frame of the command module acts as a docking template for all other sampling modules

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

49

members of the chassis that in turn form the template for the docking system to which are attached the various tool modules. The suspension point is adjustable in two horizontal axes to allow for changes in centre-of-gravity when different tool modules are attached. Attached to the command module are two electric thrusters, two hydraulic power packs, two hydraulic valve packs, two dry-space electrical pods, an oil-compensated transformer, three cameras, several lights and three laser rangers.

3.2.2 Electrical systems Within the command module is a 6-kVA step-down toroid transformer receiving single-phase 1500 Vac from the 10-km-long umbilical and transforming it into single-phase 220 Vac. The transformer is housed in an oil-filled, pressurecompensated box. Output from the transformer is split between two separate dry spaces: a power pod and an electronics pod. The power pod converts the 220 Vac into a 220-Vac three-phase supply that supplies two motor controllers. These, in turn, supply two 2.2-kW thrusters and two 2.2-kW hydraulic pumps. Due to power limitations, these motors are operable on an either–or basis: either one or both thrusters operate, or one or both pumps operate. The electronics pod contains a 220-V supply for two 1-kW lighting circuits. Each is independently switchable through micro-controllers. Three switch-mode power supplies inside the electronics pod supply 5, 12 and 24 Vdc to service electronic components and relays. All incoming telemetry is channelled through a FOCAL 907TM fibre-optic single-mode mux, enabling three 8-bit colour video channels, four RS232 and two RS485 bi-directional data channels (Bell, 1976). Two Hydro-Lek Ltd telemetry boards comprising an array of micro-controllers (including Hydro-Lek’s proprietary manipulator arm control card) receives RS232 data strings that operate up to 32 functions including the two hydraulic valve packs. A SeaEye Marine Ltd. compass board provides heading, pith and roll data from a three-component Honywell flux-gate magnetometer (model: HMR3200/HMR3300) and receives, concatenates and sends pressure data from two transducers: one indicating the ambient water pressure (i.e. depth) and the other the differential hydraulic pressure. Spare serial data channels and power connectors allow for the addition of sensors such as a conductivity, depth and temperature (CTD) instrument, a sector scanning sonar and an altimeter.

3.2.3 Cameras All three video channels are used: two are dedicated to low-light (0.1 lx), composite colour cameras with fixed focus (100 mm to infinity), auto-iris, 1/3 in charge-coupled dervice (CCD) producing 625 line phase alternating line (PAL) analogue output. These titanium-housed cameras are supplied by Bowtech Ltd (L3C-550 Aquavision) and are small, lightweight (133 mm long by 31 mm diameter, titanium housed) and low powered (130 mA). The third video channel is dedicated to an HD video/stills camera. To allow access to the most up-to-date cameras, and to minimize costs, we

50

Further advances in unmanned marine vehicles

developed a generic stainless steel pressure housing with a planar sapphire glass window to accommodate a top-range consumer digital HD (1080i) camcorder (e.g. SonyTM). The HD camera relays standard definition video to the surface while simultaneously recording up to 30 h of HD video and stills on a 250-GB micro-harddrive. The HD video is subsequently downloaded from the camera on deck after each dive through a USB2 high-speed data cable. The camera is controlled from the surface through an infrared (IR) interface (Home ElectronicsTM) with the IR command data being converted into RS232 strings (using public domain IR control software, Event Ghost, on a windows PC), relayed through the fibre-optic mux, and sent to a microprocessor where it is converted back into the SONYTM IR commands and then transmitted to the camera. This allows the user to access all functions on the camera including white balance, zoom, manual and automatic focusing, image stabilization, slow motion video and 8- and 12-Mpix stills. The SONYTM IR commands are proprietary and not made generally available to the public. To overcome this, we had to read the IR binary command strings using a sound card and hard-wired input from the camera’s IR remote controller into a PC allowing us to examine the transmitted bit-string. We then had to reproduce this IR bit-string when commanded by an RS232 message using a programmable microprocessor and IR LED. Both the HD camera and the forward directed lights are attached to a hydraulically actuated tilt-bar on the front-top of the HyBIS command module giving 60 of vertical rotation about a horizontal axis.

3.2.4

Lasers rangers

Three laser units were developed at the National Oceanography Centre and fitted to the command module to yield range and scale indication. The lasers units employ Class 1, 5-mW, 650-nm, 4.5-Vdc ImatronixTM laser diodes housed in 316 stainless steel tubular housings with optically polished quartz glass windows. Each unit is 85 mm long by 20 mm in diameter and rated to a safe operating depth of 6000 m. They are supplied via an oil-filled pressure-compensated battery box with a wetdetection switch that, for safety reasons, activates the lasers only when they are submerged. This configuration of stand-alone laser modules allows them to be easily repositioned: either forwards or downwards. Two of the lasers are mounted parallel and set apart at a distance 100 mm, while the third is set at an angle such that the three laser spots become aligned at a selectable range. This is particularly useful when downward pointing to give an indication of the close proximity of the seafloor.

3.2.5

Lights

To try and conserve the longer wavelength colour temperature of the video imagery, we opted for halogen lighting (Deep Sea Power and Light Ltd. DeepMultiSeaLightsTM). Although this is one of the least power-efficient forms of illumination, halogen lights are easy to maintain and operate. Each of the two lighting circuits supplies three lights, wired in parallel, of up to 320 W. Thus there is up to ~2 kW of lighting available in total. The quartz halogen lamps emit a colour

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

51

temperature of 3500 K and enable acceptable colour reproduction at a range of about 5 m (Duntley, 1971; O’kirk, 1991; Rea, 1993). The lights are mounted on metal frames and attached to the vehicle using high-density polyethylene clips. With a standard diameter tubing used throughout the HyBIS command module and tool modules, the lights can be easily mounted in a large variety of positions ensuring versatility and optimal illumination. Two of the lights are mounted on the front tilt-bar and track the movement of the standard definition and HD cameras (Figure 3.4). Other lights are mounted both forwards and downwards. One light is used to illuminate downwards from the centre of the command module where it is used in combination with a downward viewing camera to view the seafloor, the contents of the grab or the operations of the other tool modules.

3.2.6 Thrusters Two thrusters are located at the rear of the command module on both the port and starboard sides (Figure 3.5). These 2.2-kW, three-phase electric thrusters each provide about 50 kg of thrust in either forward or reverse motion. The propellers are shrouded and ducted inside kort nozzles for increased efficiency by reducing tip vortices and increasing directionality of thrust. Each thruster can be operated independently. For simplicity, the thruster motor controllers have a single top speed, but with a 2-s ramp-up power profile. This has two advantages: it reduces the start-up current demand and enables more subtle control by the pilot. The latter is especially important when yawing the vehicle as the thrusters provide considerable torque. The thrusters are oil filled and pressure compensated. At extreme

Figure 3.4 The HyBIS system being tested in the test-tank facility at the National Oceanography Centre. The forward lights and cameras are on a tilt-bar and can rotate 60  from the horizontal

52

Further advances in unmanned marine vehicles

Figure 3.5 View of the starboard thruster and one of the hydraulic power packs. These three-phase induction motors share components, reducing the need for duplication in field spares depth (greater than 4000 m) a low-viscosity oil is used to reduce viscous drag on the motor stator. As a general rule, the thrusters are able to provide a radius of operation of between 3% and 10% of the water depth (Figure 3.6). The difference is caused by the increasing weight of the deployed umbilical. For example, the radius of manoeuvrability is limited by the amount the thrusters can drive the mass of the vehicle and its umbilical against gravity along an arc whose length is a function of the length of the umbilical (i.e. water depth) and the angle from vertical. The amount of thrust needed approximates to the sin of the angle subtended by the cable from vertical, multiplied by the weight of the vehicle and cable combined: F ’ sina  M where F is the thrust required, a is the angle of the umbilical from vertical and M is the mass of the vehicle and cable combined. sin a approximates to the radius of manoeuvrability (r) divided by the length of deployed umbilical (d). Hence F ’ or r ’

rM d

F d M

The drag on the umbilical cable causes it to form a catenary or bend when a force is applied to its remote end. We therefore approximate the umbilical cable’s effective weight (1 kg/m in water) as half that of the length deployed. Where the vehicle

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

53

6000

Length of umbilical (m)

5000

4000

3000

2000

1000

0 0

50

100

150

200

Radius of manoeuvrability (m)

Figure 3.6 Estimates of the radius of manoeuvrability based on a maximum thruster force of 1000 N with loading approximated by the variable length and catena of the umbilical and fixed mass of the vehicle in water weighs 800 kg in water, provides 100 kg of thrust and 6000 m of cable is deployed, the radius (r) of manoeuvrability can be estimated as ^157 m. For shallower operations of, say 1000 m, r ^ 77 m. The range of manoeuvrability is illustrated in the graph shown in Figure 3.6. In practice we find the degree of manoeuvrability is greater than estimated since the degree of curvature at the bottom of the umbilical cable is greater than expected. Hence the effective weight of umbilical that the thrusters have to work against gravity is substantially less than the 50% of the total cable weight estimated earlier in this section. Experience has demonstrated that we can achieve almost twice the radius of manoeuvrability than calculated, and is further increased by attaching floatation to the lower-most 50 m of the umbilical. Provided by five oblate syntactic-foam spheres, each offering 10 kg of floatation, the bottom end of the umbilical is able to pivot about a position 50 m above its end enabling an even greater radius of operation.

3.2.7 Hydraulic motors Hydraulic power is provided by two 2.2-kW, three-phase electric pumps. The pump motors share the same model of stators as the thrusters allowing for commonality of components hence reducing the need for additional spares. Hydraulic pressure is established by both pumps operating in parallel, with each pump individually switchable. The option of two pumps is essential to ensure operation at extreme

54

Further advances in unmanned marine vehicles

depths (and hence high water pressure) where the hydraulic fluid becomes increasingly viscous (Herzog et al., 2003). A non-return valve prevents one pump from driving the other. Each pump also has an individual, pre-selected pressure setting allowing for either high or low flow rates and actuation forces. For example, when operating the grab-sampling tool, both pumps are used providing the maximum force and closure speed. When a five-function manipulator arm is used, the lower pressure pump is used initially to manoeuvre the arm, but both pumps can be used when lifting heavy objects, closing the jaw tightly or cutting warps with its cable cutter. Like the thruster motors, the pump motors are oil filled and pressure compensated. A pressure transducer indicates the pressure differential between the hydraulic fluid and the ambient seawater, and these data are displayed to the operator to monitor for leaks or other problems.

3.2.8

Hydraulic valve packs

Two valve packs are used, each providing four independent bi-directional functions. These are controlled via the telemetry cards inside the electronics pod. Each valve pack switches the fluid flow from either or both pumps to eight separate functions. These are attached to hoses with releasable fittings allowing for different hydraulic tools to be fitted. In addition, the quick release option allows the hoses to disengage from the lower tool module should it need to be jettisoned. The only hydraulic functions permanently attached to the command module are a single 100-mm-stroke cylinder to actuate the camera and lighting tilt-bar, and four 100-mm-stroke cylinders located at each corner of the lower-square-frame docking template. These four cylinders operate simultaneously to enable tool modules to be jettisoned in emergency.

3.3 Tool modules 3.3.1

Bulk sample grab

The basic tool module comprises a ‘clam-shell’ grab with a 0.25-m3 capacity and a 30-cm penetration depth. Fabricated from aluminium with stainless steel semicircular braces around the outside of each shell and cutting edges, the grab has a footprint of 0.5 m2 (1000 cm by 50 cm). It is mounted in an open chassis fabricated from 316 stainless steel tubing and stands 900 cm tall. Its upper square frame docks with the lower square frame of the command module via the hydraulic release pins allowing for it to be jettisoned if required (Figure 3.7). Four hydraulic rams drive the grab shells with a closure force of 4 tonnes. Each half of the grab closes simultaneously and two polypropylene lids close over the top enclosing the interior and protecting the sample from wash and the splash-zone as it is recovered to the surface (Figure 3.8). Closure speed depends on depth and ambient temperature (affecting the hydraulic oil viscosity) and varies from 10 to 30 s. The lids are translucent allowing for the centrally located down-light to illuminate the interior of the grab once closed. A downward pointing camera is also housed above the top

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

55

Figure 3.7 The hydraulic rams act as a release mechanism for the various tool modules

Figure 3.8 Drawing of the grab-sampling module showing the side (left) and end (right) elevations with the bucket closed and the end elevation (bottom) with the bucket open. Major components: C, camera; L, lights; HC, hydraulic cylinders; HL, hinged lids of the grab such that it can view both the seafloor and the contents of the grab after it has closed. The advantage of this configuration is that the sample can be inspected and jettisoned if required allowing for another sample to be taken.

56

Further advances in unmanned marine vehicles

During grab-mode operations, the laser rangers are pointed vertically down and set to indicate the distance from the seafloor. Two of the downward lights are also located on the lower square frame of the grab providing up to 640 W of illumination. Once on the surface the lids can be opened and the sample accessed for subsampling. If required, the grab module can be seated on a deck frame allowing sample trays to be placed under the grab while it is being opened, enabling the contents of the grab to be released and removed. The grab has been successful in obtaining clay- and sand-rich sediments as well as hard rock samples. Sessile biology, such as sponges, has been located and sampled in situ with their fibrous ‘roots’ still attached and embedded in the sediment. The surface of the sample remains relatively undisturbed, even when recovering the grab through the splashzone where waves can wash over the HyBIS instrument, although the grab is not watertight and the sediment water interface is often drained (Figure 3.9).

3.3.2

Tool sledge

The tool sledge module comprises a low-centre of gravity open-frame chassis fabricated from 316 tubular stainless steel and attached via the docking system to the HyBIS command module (Figure 3.10). It stands 600 m tall, has either 20-cmwide polypropylene skids running the full length of the sides of the module and extending a further 20 cm at each end or separate pads located at each corner. Within the module is a retractable sample tray that has rollers top and bottom on both sides that are located within tubular rails. This cantilever suspension

Figure 3.9 The grab module is fabricated from aluminium with stainless steel leading edges. It has a closure force equivalent to 4 tonnes and a capacity to recover 0.25 m3 of sediment or rock weighing up to 500 kg, shown here being deployed at sea

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

57

Figure 3.10 Hydraulic five-function manipulator arm (HLK-HRD5) fitted to the HyBIS Tool Sledge. The tool sledge has a stowable sample tray that protects the contents from waves in the splash-zone during recovery of the vehicle configuration allows for the sample tray to extend to its full length. The tray is 500 cm wide by 600 cm deep and 30 cm tall and actuated by a single long-stroke hydraulic cylinder mounted under, and the rear of, the tray. Fabricated from angular stainless steel and surrounded by stainless steel mesh, the tray retracts into a ‘drawer’ with a stainless steel mesh top. This ensures that the samples are fully secure within

58

Further advances in unmanned marine vehicles

the retracted tray to prevent their loss or damage in the splash-zone when the HyBIS is being recovered. Attached to the front lower starboard side of the Tool Sledge is a five-function manipulator arm (Hydro-Lek Ltd HLK-RHD5) with an 80-kg lifting capacity (when actuated with 160-bar hydraulic pressure) at its full reach of 943 cm (Figure 3.11). It has a continuous 360 rotating jaw with embedded 12-mm-diameter cable cutter. The arm is located on a slew plate, mounted 15 from horizontal, such that the arm can reach the seafloor in a 270 arc in front of the vehicle as well as reaching upwards to within 30 from vertical. Because our hydraulic valve packs do not have flow control, we reduce the operating hydraulic pressure supplied by one of the two valve packs to 60 bar. This enables relatively slow, and hence fine, control of the manipulator arm and jaw. The option remains to activate the second hydraulic pump, which is set to 100 bar, to increase the strength of the arm’s functions allowing higher lifting capacity and jaw closure pressure. All hydraulic hoses attached to the manipulator arm are

Figure 3.11 HyBIS (on board the RRS James Clerk Ross) configured with its tool sledge and carrying a Niskin sampling bottle and a gas collection funnel and hose for capturing free-gas released from the seafloor of the Arctic ocean

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

59

interfaced with the command module via snap releases allowing for rapid and easy exchange of the arm with other modules as required. One of the composite colour cameras can be mounted on the arm for close work if required. In Tool Sledge operating mode, up to five lights on both independent lighting circuits can be mounted forward, with one light set to illuminate the sample tray. Normally one of the composite cameras and two lights are mounted on the forward tilt-bar and another camera is fixed looking obliquely down from one side. These two viewing positions give a perspective to the scene and allow for easier manipulating operations by the pilot.

3.3.3 Lander recovery module One of our requirements for the HyBIS system was to be able to recover and deploy relatively heavy instruments on the seafloor (Figure 3.12). One of these was a 4-tonnebenthic observatory deployed in 400 m of water in the Arctic. Limitations set by the vessel operators for this task include a restriction on deployment of no more than one umbilical cable to be deployed over the side of the ship at any one time. As a result, a solution was developed that would involve the HyBIS command module deploying a lander recovery module carrying a passively driven drum of 600-m-long lifting warp (12-mm diameter DynemaTM) with a 4000-kg SWL and 20-tonne breaking strain. The lander recovery module was fabricated in 316 stainless steel tubing forming an open-chassis structure. It was fitted with the

Figure 3.12 Isometric view of the lander recovery module showing the position of the five-function manipulator arm and the passive recovery line drum. The drum carries 600 m of 20 tonne ‘DynemaTM’ lifting warp attached to a steel lifting hook and latch

60

Further advances in unmanned marine vehicles

five-function manipulator arm from the tool sledge module (Figures. 3.12 and 3.13) that carried a lifting hook, attached to the jaw of the arm via a T-bar, and spliced onto the lifting warp (Figure 3.13). This warp was then attached to the 4-tonne lander and the HyBIS instrument recovered to the ship while the lifting warp spooled off from the drum. The drum had both a friction clutch and a restriction on the spooling of the lifting warp to ensure the warp remained tightly wound on the drum. In the event of the lifting warp fouling causing the vessel to become anchored to the lander via the HyBIS vehicle, two options were retained: the capacity to cut the lifting warp with the manipulator arm’s cable cutter or to jettison the entire lander recovery module. HyBIS has had prior experience with recovering landers, but not as heavy as the 4-tonne one in the Arctic. In 2008, on its first trials cruise, HyBIS recovered a benthic lander from 2200 m of water in a location off the coast of Forteventura. This lander (Myrtel) was developed by the Proudman Oceanographic Laboratory, Liverpool, as a long-term monitoring system with CTD and acoustic Doppler current profilers (ADCP) as part of its expensive scientific payload. It was designed with two independent acoustic releases and glass sphere buoyancy such that it should have jettisoned its anchor weights and returned to the surface on command. Unfortunately, at the end of its trial mission, the lander failed to release its anchors and became stuck on the seafloor. By acoustically ranging to its sonar transponders, its position on the seafloor was triangulated to within a few tens of metres. Deployed from a non-DP ship, the HyBIS vehicle managed to sweep a path towards the lander’s position and located it after a search across 500 m of seabed. Although the HyBIS was equipped only with its grab module, a line and grapple was attached to the lander and the entire assembly recovered to the surface, thus salvaging equipment of roughly similar value to the HyBIS itself.

Figure 3.13 The HyBIS Lander recovery module on deck on the RRS James Clerk Ross in the Arctic. The large, steel hook is held by the manipulator via a T-bar and is itself attached to 16-mm dynemaTM cable stowed on the vehicle’s storage drum

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

61

3.3.4 Lander deployment module Like the other modules, the lander deployment module is an open-frame chassis comprising 316 tubular stainless steel and attached to the command module via the four-point docking release system. The module (Figures 3.14 and 3.15) was specifically designed to deploy ocean bottom seismometers (OBS) and electromagnetic receivers (EMR). These instruments have a common footprint size and rely on an anchor weight to hold them to the seafloor attached to glass floatation via an acoustic and timed release system to return them to the surface. The challenge for HyBIS was to deploy a module that would enable the landers to be deployed down to 6000 m, navigated by Ultra-Short Baseline (USBL) transponders (Jourdan & Brown, 1997), and placed on level, and well-consolidated parts of the seafloor.

Figure 3.14 Isometric drawing of the deployment module showing the position of one of the OBS instruments carried in its delivery position. The EMR instruments are taller and have four antenna extending 16 m from the mid-point of the instrument, hence the tall nature of the deployment module. A hydraulic ram and pin holds and releases the instruments from a position at the top of the module. Like the other modules, this one also attaches to the HyBIS command module via the four docking plates and release pins located at each corner

62

Further advances in unmanned marine vehicles

Figure 3.15 Deployment module loaded with an OBS (Ocean Bottom Seismometer) and ready to be deployed over the stern of the RRS James Clerk Ross in the Arctic as part of a combined high-resolution electromagnetic and seismic survey A 100-mm-stroke hydraulic cylinder was implemented at the top of the deployment module forming a retractable rod through which the bottom instruments are held. Once a suitable area of seabed is located, and the position is acceptable, then the HyBIS rotates (via its thrusters) until the landers are oriented in the required way. The hydraulic ram is then activated and the pin withdraws, releasing the instruments onto the seafloor. Post emplacement inspection then confirms the orientation and coupling of the instruments to the seafloor.

3.3.5

Top-side control system

The top-side control system comprises a portable case with four video screens, a PC, fibre-optic mux, video overlay, GPS receiver and a swappable pilot control box (Figure 3.16). The pilot control boxes are specific to the tool module used: either for the grab module or for the tool sledge. A video overlay displays depth, heading turns, and GPS or USBL latitude and longitude and is overlain on one of the composite camera feeds. We use a

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

63

Figure 3.16 Top-side control box and video recordings to its left. The system is portable and readily set up enabling ease of use ‘VideostampTM’ device that accepts serial data that includes vehicle depth, position, GPS time and heading and interlaces it onto the PAL video feeds to the monitors and video recorders. The other video screens show video feeds from the other cameras and a PC window in which a dashboard shows pilot information such as a compass ring for heading, depth, turns and position. These digital data are recorded for each dive. All video feeds can be exported externally to further video monitors and recorders.

3.4 Results The utility of HyBIS was first demonstrated in October 2008 when it located and recovered a stranded, 250-kg, benthic lander from 2200 m off the Canary island of Forteventura (Figure 3.17). In 2010 HyBIS was deployed in the Cayman Spreading Centre (Perfit, 1977) to the deepest hydrothermal sites on Earth (Murton et al., 2010). Dive sites for HyBIS were chosen on the basis of chemical signals and micro-bathymetry derived from the autonomous underwater vehicle (AUV) Autosub 6000 (Figure 3.18). The first HyBIS dive site was at a depth of 5000 m on the flanks of an axial volcanic ridge. HyBIS surveyed a 4-km-long location map showing the track, covering the region of chemical plume indications. On visually locating iron-oxide sediments, and then following a biological gradient, HyBIS located, filmed and sampled a 150-m diameter mound of active hydrothermal chimneys and sulphides. These are the deepest hydrothermal vents known and exhibit supercritical fluid emissions at temperatures of 500 C. A second vent field was also located using the same methodology,

64

Further advances in unmanned marine vehicles

(a)

(b)

Figure 3.17 (a) Image from HyBIS of the Myrtel lander on the seafloor at a depth of 2200 m and (b) being recovered to the aft deck of the RRS Discovery making the Autosub6000/HyBIS combination the world’s most successful technology for locating hydrothermal vents. During this voyage, HyBIS collected a hydrothermal chimney, rock samples, water samples and, using a slurp gun, biology samples. In the 15 days of operations, it completed 100 h of bottom time and reached a maximum depth of 5162 m (Figure 3.19).

3.5 Conclusions The philosophy behind the HyBIS instrument is to develop a technological solution that addresses the specific requirements of the user. Rather than provide a comprehensive engineering capability in a single system (e.g. an ROV), we used the specific scientific needs of the user to inform the design and development of the technology. The result is surprising: a versatile technology, modular in design, that has a low capital cost and is relatively easy to operate. We believe that this approach is both efficient and cost effective. For a fraction of the capital and running cost of a conventional ROV, the HyBIS system meets many of the users’ needs. Ensuring that

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

65

(a)

(b)

(c)

Figure 3.18 Location map of the Mid-Cayman Spreading Centre (inset) with multi-beam bathymetry (a) and high-resolution 3D bathmetry from the AUV of (b) the ~5000-m-deep hydrothermal sulphide mound and (c) the 2200-m-deep hydrothermal sulphide mound on Mt Dent

excess capacity and common formats are designed in from the start, the instrument can be easily expanded and developed to meet future requirements. Key to the success of this approach is knowing exactly what the user needs and distinguishing this from what the user thinks they want. This requires a bi-lateral process of

66

Further advances in unmanned marine vehicles

(a)

(b)

Figure 3.19 (a) Video frame-grab from the HyBIS vehicle while approaching the deepest hydrothermal system on Earth, at a depth of 5000 m in the Mid-Cayman Spreading Centre. The HyBIS vehicle was deployed after initial surveys by the National Oceanography Centre’s AUV Autosub 6000, demonstrating the efficiency of using an AUV and RUV combination in exploring and discovering active geological features on the deep-ocean floor. (b) Image of one of the actively venting chimneys taken using the HyBIS HD video and stills camera and using 1280 W of halogen lighting

information flow: educating both the engineer and the user so that each knows what can be achieved and what is actually required.

Acknowledgements The HyBIS system was developed with funding support, granted to B.J. Murton, from the Natural Environment Research Council and National Oceanography Centre, United Kingdom. We are also indebted to Hydro-Lek Ltd, and especially the Managing Director Mr Chris Lokuciewski, for their continuing enthusiasm and support for the project from its outset.

References B.J. Murton, D.P. Connelly, J.T. Copley, K.L. Stansfield, P.A. Tyler, Cruise JC044 Scientific party. Hydrothermal vents at 5000 m on the Mid-Cayman rise: The deepest and hottest hydrothermal systems yet discovered, American Geophysical Union, Fall Meeting 2010, 2010, abstract #OS33F-05, 2010. doi: 10.1109/ OCEANS.1997.634456 C.H. Bell, Fiber-optic multiplex optical transmission system, USPTO, 05/715, 485, 1976

HyBIS: a new concept in versatile, 6000-m rated robotic underwater vehicles

67

D. Jourdan, B. Brown, Improved navigation system for USBL users. OCEANS ’97. MTS/IEEE Conference Proceedings, vol. 1, Halifax, NS, Canada, pp. 727–35, 1997 J.T. O’Kirk, Volume scattering function, average cosines, and the underwater light field, Limnology and Oceanography, vol. 36, pp. 455–67, 1991 M. Rea, Lighting Handbook: Reference & Application, Illuminating Engineering Society of North America, New York, p. 957, 1993 M.R. Perfit, Petrology and geochemistry of mafic rocks from the Cayman Trench: Evidence for spreading, Geology, vol. 5, pp. 105–10, 1977 S.N. Herzog, C.D. Neveu, D. Placek, R. Max, Selecting the optimum viscosity grade of hydraulic fluid, Lubrication & Fluid Power, vol. 6, pp. 7–12, 2003 S.Q. Duntley, Underwater lighting by submerged lasers and incandescent sources, Scripps Institution of Oceanography, La Jolla, CA, Visibility Lab, unpublished report, p. 286, 1971

Chapter 4

An AUV project applied to studies on manoeuvrability of underwater vehicles Ettore A. de Barros , J.L.D. Dantas , Luciano O. Freire and Rodrigo L. Stoeterau 1

1

1

1

4.1 Introduction Predicting the autonomous underwater vehicle (AUV) manoeuvring performance is important during the vehicle’s design phase. This has impact on the design of the autopilot, and guidance systems, for instance, and influences the vehicle autonomy as well. A number of authors have applied numerical tools to predict added masses, static forces and moments, and the wake field of submarines and other underwater vehicles. Recently, methods based on computational fluid dynamics (CFD) were used to predict the stability derivatives of an AUV. For example, in the works of References 1 and 2, the forces and moments acting upon the AUVs hulls, when they are inclined to the flow (i.e. have an angle of attack), are compared with those obtained in experimental tests in the towing tank. Together with the static coefficients, Phillips et al. [3] predicted the dynamic stability derivatives of the Autosub AUV simulating a turning manoeuvre, and compared with the dynamic efforts obtained by towing tank tests with a planar motion mechanism (PMM) [4]. Tang et al. [5] used CFD methods in the prediction of hydrodynamics coefficients of AUVs with non-conventional shapes. The validation was done by comparison of the vehicle trajectory obtained by dynamic simulations and manoeuvres performed in the field. Singh et al. [6] and Lauder et al. [7] used unsteady CFD simulations in their investigation of the propulsion performance and control of flapping airfoils, mimicking the fins of fishes for the new generation of biomechanics AUVs [8]. Beyond the identification of hydrodynamic coefficients, the CFD tools are also used to optimize the shape of AUVs [8,9] or the performance of AUV propellers [10], and to examine the efforts upon AUV hull during the docking process [11] and in the cooperative missions [12]. Currently, the CFD approach to parameter estimation requires a considerable degree of expertise on grid generation and on the mechanization of the supporting

1

Polytechnic School of University of S~ao Paulo, Brazil

70

Further advances in unmanned marine vehicles

numerical models. It holds, however, a great promise for the development of expedite vehicle parameter estimation methods. Therefore, CFD methods are expected to play an increasingly important role in the design of more efficient AUVs. The economy of computational resources and expertise in numerical models make the analytical and semi-empirical (ASE) methods still an interesting alternative to the CFD approaches. The calculation of hydrodynamic derivatives yields approximate results that can be used to predict manoeuvrability characteristics, select hydroplanes and investigate control strategies at an early stage of vehicle design. ASE methods are directly focused on the estimation of parameters such as added mass and inertias (acceleration-related coefficients), linear and non-linear damping coefficients (related to velocities) and control action–related parameters. Advanced approaches to AUV design may also involve combined plant/controller optimization, where prediction methods of hydrodynamic derivatives play an important role. The ASE approach to derivative estimation provides an analytical formulation based on physical concepts that can help in interpreting experimental and CFD results as well as in defining uncertainty interval for hydrodynamic parameters in order to help in the design of robust control systems. This chapter deals with the problem of AUV modelling and hydrodynamic coefficient estimation in the context of the Pirajuba project. The Pirajuba (‘yellow fish’, according to the Brazilian Indian Tupi language) is an AUV (Figure 4.1), which is being developed as a test bed for investigations on dynamics and navigation of this class of vehicle. The original application of this AUV is the experimental support to studies on predictions of hydrodynamic derivatives. This investigation started with the application of ASE methods to predict derivatives of the Maya AUV [13–15]. In a later phase, the ASE methodology has been integrated to CFD tools in order to

Figure 4.1 The former version of the Pirajuba AUV during tests in a swimming pool

An AUV project applied to studies on manoeuvrability of underwater vehicles

71

improve the derivative predictions [16]. Up to now, most of the experiments with the vehicle are related to towing tank tests. They have been used for the validation of the static derivative prediction. The use of free model tests with Pirajuba is aimed to include the validation of the dynamic derivatives prediction as well, and to reduce the costs of the experiments. The common type of shape related to the Myring geometry [17], which is present in vehicles such as Remus and Maya, makes Pirajuba a useful test bed for studies on AUV hydrodynamics and control. In this sense, similarly to the case of joint investigations on ship manoeuvring, this shape can be adopted as a reference for tests in different institutions. The main dimensions of the vehicle are the same as those of the Maya AUV, which was developed under a joint Indian–Portuguese project that also included the cooperation with the Unmanned Vehicles Laboratory at the University of S~ao Paulo for manoeuvrability studies [13,14]. Main differences to the original Maya version include the hydroplane configuration (in Pirajuba, the cruciform-type configuration is adopted, located at the rear end of the middle body), the free-flooding construction, the customized thrusters and all the embedded hardware and software systems. This chapter presents the Pirajuba AUV project and its application to the investigation on dynamics of underwater vehicles. The chapter is organized as follows. The motivation for a test bed is presented in section 4.1. Section 4.2 presents an overview of the Pirajuba AUV. Section 4.3 presents its control architecture. Section 4.4 introduces the state of art of numerical, ASE methods that have been applied to the prediction of hydrodynamic coefficients of underwater vehicles. Section 4.5 describes the experiments using captive models of the Pirajuba AUV and analyses the results according to the methods presented in section 4.4. Finally, the chapter concludes with the next developments on the Pirajuba project.

4.2 The Pirajuba AUV Pirajuba is a cruising-type AUV designed for having an autonomy of 4 h at 2 m/s. The main particulars of the vehicle are given in Table 4.1. The vehicle has a Myring-type body defined by a nose section, a middle body cylindrical section and a tail section. The nose shape and the tail shape are described by the modified semi-elliptical radius distribution [17]: "
 #1=n 1 x  nl 2 ð4:1Þ rðxÞ ¼ d 1  2 nl and the cubic relationship " # 1 3d tan q rðxÞ ¼ d   ðx  nl  clÞ2 2 2ðlp  nl  clÞ2 ðlp  nl  clÞ " # d tan q þ ðx  nl  clÞ3  ðlp  nl  clÞ3 ðlp  nl  clÞ2

ð4:2Þ

72

Further advances in unmanned marine vehicles Table 4.1 Main parameters of the hull Features

Dimensions

Bare hull length (m) Hull maximum diameter (m) Base diameter (m) Nose length (m) Middle body length (m) Myring body parameter q (º) Myring body parameter n

1.742 0.234 0.057 0.217 1.246 25 2

respectively, where x is the axial distance to the nose tip, d is the maximum hull diameter, lp is the length of the Myring body [17], nl is the nose length, cl is the middle body length and q, the Myring parameter, is the tail semi-angle. Unlike the Myring body, which ends in a pointed tail, the Pirajuba has blunt base. Therefore, the parameter lp is used in the generation of the bare hull geometry. The hydroplanes are formed by all movable control surfaces with the same section NACA 0012 and dimensions. Details are given in Table 4.2. The vehicle is a free-flooding-type AUV, having an external hull made up of fibre glass and three aluminium pressure vessels inside: the main vessel, the manoeuvring vessel and the thruster vessel. The main vessel carries the computer units, motion sensors (rate-gyro, compass and inclinometers), depth and liquid-level sensors, lithium polymer batteries, communication and power electronics. The manoeuvring vessel includes the servo systems for moving the control surfaces and the thruster electronic driver. The thruster vessel carries a 150-W DC motor, which moves a polyurethane propeller made by a rapid prototyping manufacturing system. The propeller parameters were selected from the Wageningen series, and its geometry defined in a CAD-based numerical tool. Structure of all the vessels was calculated for operating at a depth of 100 m (Figure 4.2). Improvements achieved in the third version of the vehicle refer to the mechanical design of the main components of the AUV: hull, propulsion vessel, manoeuvre vessel and the internal structure of the main vessel. Table 4.2

Main parameters of the control surfaces

Features

Dimensions (m)

Span Height Root chord Tip chord Position*

0.554 0.016 0.090 0.060 1.373

*Distance from the leading edge to the hull nose tip.

An AUV project applied to studies on manoeuvrability of underwater vehicles (a)

73

(b)

Figure 4.2 The Pirajuba AUV. (a) A view of the main and manoeuvre vessels inside the vehicle. (b) The vehicle assembled without the down fin The AUV shape should be as close as possible to the ideal geometry of the models used in the applications of CFD and ASE methods. This is important to establish the vehicle as a reliable test bed for validating the predictions produced by these methods. The new hull is lighter than the former version, and it was laminated on a mould that was manufactured by numerical control of a robot manipulator from the ship model laboratory of the ‘Instituto de Pesquisas Tecnolo´gicas’ (IPT-SP). Attached to the inner part of the middle hull, there is a layer of flotation foam commonly employed for boat construction in the marine industry. The layer allows for a lighter structure, weight balance stability and the hull structural reinforcement (Figure 4.3). The foam also filled the inner part of the bow and stern shapes. In the latter part, the foam structure is used for supporting the propulsion vessel (Figure 4.4). The new platform for the arrangement of components in the main vessel was made of aluminium by precise electro-erosion procedures. The longitudinal beams are covered by a Teflon layer, which is also attached to the bottom of the transversal frames for allowing the easy movement of the structure when it is placed inside the vessel. Each longitudinal beam also includes a channel for allocating all the cabling system (Figure 4.5). The manoeuvre vessel was designed in order to provide the positioning accuracy of the hydroplanes so that the quality of the free model test for identification of

Figure 4.3 Hull middle body, tail and nose, and the buoyancy foam layers (in black)

74

Further advances in unmanned marine vehicles

(a)

(b)

Figure 4.4 Stern (a) and bow (b) filled with the buoyancy foams (a)

(b)

Figure 4.5 Inner platform of the main vessel (a) and the manoeuvre vessel (b)

the AUV dynamic model is assured. High-accuracy servo-motors were combined to synchronous belt-transmission systems. The connection of the hydroplanes to the vessel was also improved, so that the assembly process becomes very simple just before the beginning of experiments (Figure 4.5). The propulsion vessel was improved by a new combination of sealing, bearings and lubrication (Figure 4.6).

4.3 Control architecture The control architecture design for this AUV considered the predominance of graduate and undergraduate students in the Pirajuba project, from the system design to the implementation, upgradation and maintenance. In the Pirajuba project, one should seek for generality and extensibility in the hardware architecture. The choice for small and simple parts that can be explored and reused in new projects is highly desirable in an AUV component project. This is a way to increase the reliability of the AUV while keeping its project and implementation accessible to the students.

An AUV project applied to studies on manoeuvrability of underwater vehicles

75

Figure 4.6 The AUV thruster Concerning the hardware components, it is important to orient design and implementation to the use of cheap and locally available parts. The reason, in this case, is related to the higher risk of component damage during the work of students and the avoidance of schedule delays reducing the loss of short-term man power. Moreover, concerning the choice of CPUs and microprocessors, it is important to keep the choice close to the platforms familiar to the students, reducing training and development time. Those considerations oriented the choice of microcontroller-based boards for the implementation of control units in the embedded system. Particularly, the ARM7 microcontrollers were selected for most of the functions to be implemented. The navigation was implemented in one ARM9 microcontroller, since this function requires a more powerful CPU related to the implementation of the Extended Kalman Filter algorithm for sensor fusion. Usually, the Ethernet is preferred as a communication protocol, although CAN and RS-232 can be found in some AUV implementations. However, a number of authors report drawbacks when using the Ethernet, which are related to the space necessary for hub, and cabling, especially when redundancy should be included in the system. If one takes into account aspects such as cost, availability, volume, cabling, energy consumption, predictability and bandwidth, CAN becomes a better choice than Ethernet. The hardware architecture adopted in this project followed the general idea of Ortiz et al. [18], which proposes one CAN bus with many sense and act nodes and one central processing unit. Cabling is reduced, less volume is occupied, and the overall reliability is improved, since a reduced number of connectors are used. However, in this project the use of some powerful CPUs allows for filtering and the processing functions that make their role more sophisticated than just sense and act nodes (Figure 4.7). The choice for more powerful processing units avoids the

76

Further advances in unmanned marine vehicles

Figure 4.7 Block diagram representing part of the hardware architecture necessity of concentrating the memory and intensive computing demanding tasks into PC-based systems. That enlarges the autonomy and reduces internal heating, improving the reliability of all electronics inside the vehicle. Taking into account a low-cost architecture with availability of components in the local market, a CAN network was conceived on the basis of a number of sense/ act nodes (Figure 4.8), which are employed in the automotive industry, together with one or more main nodes using more powerful CPUs. The existence of a standard electrical interface and the use of common components across the vehicle allow adding, removing and replacing modules easily. For these operations, the

Figure 4.8 The implemented network node

An AUV project applied to studies on manoeuvrability of underwater vehicles

77

mechanical robustness of the components was taken into account, keeping the safety of assembling and handling tasks. Each node with the ARM7 microprocessor can be represented in a three-layer composition: driver, data conversion and network control (Figure 4.9). At the driver level, electrical signals from sensors or power drivers are translated to a low-current signal at a voltage level between 0 and 3.3 V. Those signals are used by the peripherals A/D and D/A converters included in the microcontroller boards. The ARM7-based nodes transform the network data to the binary code inserted in the D/A converters or the data acquired by the A/D converters to the data to be transmitted in the CAN network or through a serial port to the ARM9 processor (Figure 4.7). At this level, simple routines of data conditioning and interlocking can be included as well. Network control occurs at specific integrated circuits, responsible for implementing the CAN protocol. These circuits communicate to the microprocessors through serial or parallel protocols. From the software development point of view, tools based on C language and easily understood by undergraduate engineering students were selected. In this way, the project dependence on software engineering professionals is reduced for

Data Bus NODE Network Control CAN Controller MCP2515

Data Conversion ARM7 LPC2148

Drivers Instrumentation & Signal Conditioning

Power Converters

Sensors

Actuators

Figure 4.9 Three layers compounding each node

78

Further advances in unmanned marine vehicles

the code upgradation and maintenance. However, every development follows the structured programming and object-oriented concepts. On the basis of the same philosophy, the operational system adopted, unlike the mainstream, is the mC-OSII, like in Reference 19, which is a commercial OS FAA certified and freely available to universities. This OS is very simple to use and portable to many microprocessors. It is a very helpful resource for embedded systems learning. The software architecture as a whole can be also represented by a system of layers (Figure 4.10). Each node is developed according to the object-oriented programming. A number of objects constitute a functional layer. The objects that implement a function can be changed without affecting the whole architecture. A layer can use the objects from lower-level layers, but not from higher-level layers. Each object is independent of the node where it is implemented. All these features provide flexibility to the architecture. At the lowest abstraction level, there is a peripheral controller. It is responsible for controlling the microprocessor peripherals through configuration registers. It is where generic routines of resource use and configuration are located, such as pulse width modulation (PWM) generation, digital/analogical conversion, serial port configuration. The second layer includes the embedded operational system, i.e. the mC-OSII, where task scheduling resources are provided. The third layer is responsible for the communication among nodes and provides resources for message transmission, reception, synchronicity and communication-fault diagnosis. This layer also allows the access of each running process to the system information. That is implemented through a shared memory

Figure 4.10 Software architecture

An AUV project applied to studies on manoeuvrability of underwater vehicles

79

and a CAN data exchange module where the nodes access the necessary information for their tasks. The fourth layer is responsible for driving the AUV sensors and actuators. It is where sensors are automatically initialized and configured as soon as the system is turned on. In this layer sensory signal processing and interpretation is carried out, and reference values to the actuators are converted to binary codes to feed their drivers. In the fifth layer, there are the high-level functions such as navigation (‘EKF’), guidance, control, mission management, etc.

4.4 Investigation on modelling the dynamics of the Pirajuba AUV The application of ASE methods to the estimation of submarine and AUV model parameters has been previously reported in the literature. Most of the approaches rely on the adaptation of methods originated in aeronautics for predicting the dynamics of aircrafts. However, especially concerning non-linear dynamics, more recent progress in the field of missile aerodynamics can be explored and adapted to improve the prediction of AUV hydrodynamic coefficients. Moreover, those approaches can simplify the AUV dynamics representation compared to the traditional marine model, which is based on the Taylor series expansion of the hydrodynamic efforts as function of the vehicle motion variables. To compute the flow field around the Pirajuba’s hull, the well-known commercially available software ANSYS FLUENT was used, which uses the finitevolume CFD method in its calculations. For the calculation of the transport flow equations, the Reynolds-averaged Navier–Stokes (RANS) model was used, where the standard Navier–Stokes equations were decomposed into mean and fluctuation components and simplified to be solved as time-averaged, i.e. the solution is not instantaneous. These simplifications are done in order to reduce the computational cost of the simulation. Turbulent effects on the flow are accounted for through transport equations, known as turbulent models. Solutions with good accuracy both in the free stream and near vehicle locations could be achieved with the ‘k–w Shear Stress Transport (SST)’ turbulence model [20], that uses a robust function to blend the turbulence dissipation rate (‘e’) and specific dissipation rate (‘w’) together with the kinetic energy (‘k’), which have a better definition in the hull boundary condition. The simulations deal with the flows around the bare hull and hydroplane, taken separately, and in combination. For the first two cases the grid was composed of 1.5 million elements, whereas the combination was modelled with 4.5 million grid elements. In this section, the application of ASE and CFD methods to the manoeuvrability analysis of AUVs is analysed through the prediction of normal force and moment coefficients acting upon the vehicle as a function of the angle of attack. These so-called static coefficients usually have an important influence in the manoeuvrability of the vehicle. Even in the case of dynamic coefficients (those that

80

Further advances in unmanned marine vehicles

are functions of angular velocities), the basis of their prediction formula is the phenomena considered in the estimation of those static coefficients. Both methods were mixed in some cases for improving the ASE formulas.

4.4.1

Estimation of the AUV hydrodynamic parameters through the ASE- and CFD-based methods

The adopted ASE methodology is based on the results derived by slender body theory. However, some conditions in the theory application are relaxed for the type of AUV considered. For instance, the slender body theory assumes that a smooth transition in the hull geometry occurs, which strictly would imply that a body of revolution ends at a point in the stern [21]. Moreover, for the body–fin combination, the slender body theory assumes that no negative lift is developed behind the maximum fin span. These conditions are not followed in the ASE formula presented in this section, and a simple linear superposition of effects produced by the bare hull and the hull–fin combination, at the location of the vehicle control surfaces, is assumed. This kind of approximation has been adopted in other works related to the prediction of hydrodynamic derivatives [22,23].

4.4.1.1

Bare hull forces and moment

The bare hull normal force and moment, when produced by a change in the angle of attack, are predicted on the basis of a modification of the slender body theory, and related formula derived by Munk [24] and Hoak and Finck [22]. Results were experimentally validated for the Pirajuba AUV [15]. The proposed formula for the bare hull normal force coefficient CNðBÞ ðaÞ is given by CNðBÞ ðaÞ ¼ ðK2  K1 ÞCNP ðaÞð1  sink aÞ þ CNv ðaÞ

ð4:2Þ

The coefficients CNP ðaÞ and CNv ðaÞ denote the contribution of potential and viscous terms, respectively. In (4.2) CNp ðaÞ ¼

S sinð2aÞ Sref

ð4:3Þ

and CNv ðaÞ ¼ hCdn

Sp sin2 a Sref

ð4:4Þ

In (4.3) and (4.4), S* is the sectional area in a station at a longitudinal distance from the nose tip. The coefficients K1 and K2 are, respectively, the longitudinal and transversal apparent mass factors [24]. The factor (1 sink a) is related to the influence in the potential term of the boundary layer-thickening, vorticity and cross-flow separation in the bare hull [16]. For the hull shape considered, k ¼ 1.3 was adopted. The parameter Sp is the planform area at the xy plane (Figure 4.11), and Sref is the reference area (taken as L2 in this work). The coefficient h is a x0

An AUV project applied to studies on manoeuvrability of underwater vehicles

81

Figure 4.11 Definition of the considered coordinated system correction factor, equal to the ratio of Cdn for a finite length cylinder to that for an infinite cylinder. This parameter is a function of the fineness ratio L/d [25], where L is the body length and d is the body maximum diameter. The distance x0 is the average value between the hull length and x0. The parameter x0 is adopted by the Datcom handbook [22], as the axial distance between the nose and the station which the flow can no longer be considered as potential. The semi-empirical expression for the estimative x0 is given by x0 ¼ 0:378
L þ 0:527
x1

ð4:5Þ

where x1 is the coordinate where the body profile has the most negative slope in the aft direction. The parameter x0 is adopted as a better estimative as that of (4.5), since it takes into account that a significant portion of the potential flow is also occurring at the stern, as indicated by the flow visualization by the numerical flow simulation [15]. The moment coefficient Cm is defined by Cm ðBÞ ¼

ð x

0

cN ðxÞðxm

ð4:6Þ

xÞdx

0

where xm is the axial distance from the nose tip to the centre of rotation and cN ðxÞ ¼ ðK2

K1 Þsinð2aÞ

dSðxÞ ð1 dx

sink aÞ þ 2hCd sin2 ðaÞrðxÞ

ð4:7Þ

with r(x) denoting the cross-section radius at the axial distance x from the nose tip and S(x) is the corresponding cross-sectional area. From (4.6) and (4.7), the final expression for the moment coefficient result is    V S  ðx0 xm Þ sinð2aÞð1 sink aÞ CmðBÞ ¼ L3 þ hCdn

Sp xm xc  2 sin ðaÞ L2 L

ð4:8Þ

where S* is the cross-sectional area at the station distance x0 from the nose, xm is the axial distance from the nose tip to the centre of rotation and V* is the volume

82

Further advances in unmanned marine vehicles

between the nose tip and such station. The parameter xc is the axial distance from the nose tip to the centroid of the bare hull planform area at the xy plane. The hydrodynamic centre of the bare hull is defined as xðBÞ ðaÞ ¼

CmðBÞ ðaÞ CNðBÞ ðaÞ

ð4:9Þ

For all the AUV parts (bare hull, hydroplane and the combination), the CFD approach adopts the same mesh construction methodology, using the software ANSYS ICEM with only hexahedral elements in a block-type interface. Because the body is axisymmetric, and subjected only to a variation in the angle of attack, a grid including just half body was considered to save computational resources. The mesh space was decomposed into three regions: the outer region, the space closely around the vehicle and the boundary condition. The outer region has rectangular shape, with width, height and length of, respectively, 3, 10 and 13 times the length of the vehicle (‘L’). The vehicle is located at the same distance, 3L, from the bottom and the side walls. These dimensions were obtained empirically so that the external boundary conditions do not interfere with the vehicle flow pattern. The non-slip stationary wall was used as a boundary condition. This condition and the turbulence model are employed by the CFD solver in order to resolve the flow in the region near the body surface where the viscosity influence is the most significant inside the boundary layer. Using such approach, it is expected that a more accurate model of the velocity profile should be produced in comparison to that one generated by standard methods. Moreover, a greater number of boundary layers phenomena can be modelled as well [26]. To use this method, a minimum grid space in the body-mesh interface must be attained. This space is achieved in function of the parameter yþ, representative of the local Reynolds number, which must be close to 1. This parameter is defined as yþ ¼

yut u

ð4:10Þ

where y is the distance to the body surface, u is the fluid kinematic viscosity and ut pffiffiffiffiffiffiffiffiffiffi is the friction velocity, defined as ut ¼ tw =r, with tw being the shear stress at body surface and r the fluid density. As a starting point, the thickness of the near wall mesh was obtained dividing the maximum value of the boundary layer, calculated in Reference 27, by 40. This procedure provides a good result for the bare hull case (Figure 4.12).

4.4.1.2

Estimation of the hydroplane-generated normal force

The estimation of the forces generated by the hydroplane alone could be carried out using the formula proposed by Whicker and Fehlner [28], after testing a number of rudders in the towing tank. However, the Reynolds number considered on those experiments is higher than that one experimented by the hydroplanes at the AUV.

An AUV project applied to studies on manoeuvrability of underwater vehicles

83

Figure 4.12 Computational mesh for the bare hull

The effects of the low Reynolds numbers on decreasing the lift generated by the foils and their stall angle are well documented in the literature [29]. This limitation of the ASE approach and the relatively simple procedure to simulate numerically the flow on the specific hydroplane to be investigated lead us to investigate the combination of the CFD estimation of the efforts generated by the hydroplane and the fin–bare hull interaction calculated through the formulas derived from the slender body theory. The methodology for implementing the flow numerical simulation around the hydroplane adopted the same turbulence model and grid generation procedure assumed in the case of the bare hull. The grid volumes close to the fin surface, however, had their thickness reduced in order to achieve the desired value of the yþ parameter (Figure 4.13). The thickness value, in this case, was achieved through an iterative procedure.

Figure 4.13 Computational mesh on the Pirajuba’s hydroplane

84

Further advances in unmanned marine vehicles

4.4.1.3

Body–fin interaction as a function of the angle of attack

There are methods for predicting the interaction between the hydroplanes and the bare hull related to ASE- and CFD-based approaches. Results derived from slender body theory [30] can be applied to calculate the lift coefficients produced by the combination of the hydroplanes and the bare hull. In the Pirajuba AUV, the combination occurs at the end of the middle body. As a consequence, the moment can be calculated by the addition of the effects from the bare hull and the body–fin interference considered separately. The total normal force coefficient CNðWBÞ ðaÞ (where the notation ‘WB’ borrows from aircraft wing–body interactions) is given by CN ðWBÞ ðaÞ ¼ CNðBÞ ðaÞ þ ðKWðBÞ þ KBðWÞ ÞCN ðWÞ ðaÞ

ð4:11Þ

where CNðWÞ ðaÞ is the normal force coefficient of the exposed fin surfaces, and KBðWÞ and KWðBÞ are the interference factors from the surfaces to the body and from the body to the surfaces, respectively. Let b be the maximum span of the fins in combination with the hull, i.e. the total distance between the lift surface tips, as if they were extended inside the hull, and define k¼

d b

ð4:12Þ

Then, the interference factors can be written as KWðBÞ ¼

2 ð1 þ k 4 Þz1  k 2 z2 p ð1  kÞ2

ð4:13Þ

and KBðWÞ ¼ ð1 þ kÞ2  KWðBÞ

ð4:14Þ

where 


 1 1 1 1 p ðk  kÞ þ z1 ¼ tan 2 2 4

ð4:15Þ

and

z2 ¼ ðk 1  kÞ þ 2 tan1 k

ð4:16Þ

CmðWBÞ ðaÞ ¼ CmðBÞ ðaÞ þ ðKWðBÞ þ KBðWÞ ÞCNW ðaÞxW ðaÞ

ð4:17Þ

After calculating the normal force, the moment of the combination is calculated as a superposition of the bare hull moment and the moment generated by force component due to the interference multiplied by the hydrodynamic centre of the exposed fin as a function of the angle of attack:

An AUV project applied to studies on manoeuvrability of underwater vehicles

85

Figure 4.14 Computational mesh for the bare hull–hydroplane combination In terms of the hydrodynamic centre for the vehicle, the result above is expressed as xðWBÞ ðaÞ ¼

CmðWBÞ ðaÞ CNðWBÞ ðaÞ

ð4:18Þ

The hydrodynamic centre of the body–fin region is assumed equal to that of the fin alone calculated from the CFD simulations as a function of the angle of attack (Figure 4.17(b)). At the control surfaces region, the density of the elements was increased by reducing their size so that a good grid quality in the integration between the grids around the bare hull and hydroplanes could be achieved, considering the yþ grid requirements (Figure 4.14).

4.5 Results and experimental validation Captive model tests using real scale models of the Pirajuba AUV have been carried out in a towing tank [31]. During the static hydrodynamic tests, the models are towed at a constant depth and speed by the towing carriage. In this carriage the marine craft model is attached to a rotating platform that imposes the sideslip angles. They represented the angle of attack variation referred in section 4.2. The hydrodynamic efforts are measured during the motion by a dynamometric system. The first tests with submersible models were conducted with a typical arrangement of the dynamometric system (Figure 4.15), which is commonly used in experiments of marine crafts. In this arrangement the model is towed by a vertical strut attached at its middle body and the measuring unit is fixed in the towing carriage. The dynamometer system is composed of an assembly of six isolated load cells, reducing the influence between cells (cross-talk).

86

Further advances in unmanned marine vehicles

Figure 4.15

Vertical strut-type arrangement (a) and the sting (horizontal)-type arrangement (b), attached to the Pirajuba AUV model for effort measurements during static tests

To avoid the hydrodynamic effects of the struts on the flow and in the drag measurement, the sting-type arrangement has also been adopted. In this kind of arrangement, the dynamometric system is installed inside the model hull, measuring only the efforts on the hull, and it is fixed to the carriage through a sting attached at model stern (Figure 4.15), eliminating the flow interference generated by the vertical strut. This type of arrangements is commonly used in wind tunnels [25], but similar arrangements are also used for AUV hydrodynamics investigations [2]. The first tests, which employed the vertical strut arrangement, were preceded by experiments with a ‘calibration body’ whose experimental results are available in the literature [25]. In this way, the experimental procedure and measurements had a previous validation. Results from the sting arrangement have reproduced those obtained by the strut arrangement for the bare hull case. This fact was considered when assuming the reliability of the results observed by such arrangement. The experiments described in this section are all static tests, where the models are towed for several angles at a longitudinal velocity of 1 m/s. The axial and normal forces, and the yawing moment were measured for each angle of attack. The presentation of the results is conducted in accordance with Reference 32, where the forces are divided by dynamic pressure of the fluid times the squared body length, and the moments by the same dynamic pressure times the cubed body length. The results related to the moment coefficient are presented indirectly through the axial position of the hydrodynamic centre. Its non-dimensional form is defined as the axial distance from the bare hull nose tip, positive aft: xHC ðaÞ ¼ L




xm Cm ðaÞ  L CN ðaÞ



ð4:20Þ

The first tests involved the bare hull geometry (Figure 4.16). A good agreement between CFD and ASE predictions and the experimental results is observed. The graph of the hydrodynamic centre shows that the moment changes its signal only at high angles of attack, when the viscous term in (4.8) takes over.

Hydrodynamic Centre [m]

Normal Force Coeff. (10–2)

An AUV project applied to studies on manoeuvrability of underwater vehicles CFD EXP ASE

–0.5 –1 –1.5 –2 –2.5 –30

–20

(a)

Figure 4.16

–10 0 10 20 Angle of Attack [deg]

0.5 0 –0.5 –1 –1.5 –2

CFD EXP ASE

–2.5 –3 –30

30

87

–20

(b)

–10 0 10 20 Angle of Attack [deg]

30

Normal force coefficient (a) and hydrodynamic centre (b) for the Pirajuba’s bare hull

1

Hydrodynamic Centre [%c]

Normal Force Coeff. (10–2)

In the case of the normal force prediction by the CFD approach, two hydroplanes are tested. The fin based on the NACA0012 profile is the original one adopted in the Pirajuba AUV. It was compared with another one based on the NACA0015, which was designed such that the analytically predicted normal force coefficient derivative is the same for both hydroplanes. The usefulness of CFD predictions can be observed for the normal force prediction in the case of the hydroplanes tested (Figure 4.17). This is particularly observed in the linear angle of attack range, before the stall angle. In the non-linear region, for the range greater than the stall angle, the CFD predicts a tendency of the increase in the normal force, whereas the experiments show a saturation behaviour. The NACA0015 case exhibits a higher stall angle, indicating a better performance. The case of the combination between bare hull and hydroplanes shows that the ASE and CFD approaches provide good estimative for the normal force, with a slightly superior performance of the CFD result for the non-linear angle of attack range (values greater than 10 in this case). The same can be affirmed in the case of NACA0012-CFD NACA0012-EXP NACA0015-CFD NACA0015-EXP

0.5 0

–0.5 –1 –30

–20

(a)

Figure 4.17

10 20 –10 0 Angle of Attack [deg]

30

0.4 0.35 0.3 0.25 0.2 –30

(b)

NACA0012-CFD NACA0015-CFD

–20

–10 0 10 20 Angle of Attack [deg]

30

Normal force coefficient (a) for the NACA0012 and NACA0015 hydroplanes profile, and the dimensionless chordwise hydrodynamic centre from CFD simulations (b)

Further advances in unmanned marine vehicles 4 CFD EXP ASE

3 2 1 0 –1 –2 –3 –4 –30

(a)

1 Hydrodynamic Centre [m]

Normal Force Coeff. (10–2)

88

–20

10 20 –10 0 Angle of Attack [deg]

0.95 0.9 0.85 0.8 0.75

0.65 –30

30 (b)

CFD EXP ASE

0.7 –20

–10 0 10 20 Angle of Attack [deg]

30

Figure 4.18 Normal force coefficient (a) and hydrodynamic centre (b) for the Pirajuba’s bare hull–hydroplane combination the hydrodynamic centre prediction (Figure 4.18). In this last case, the approximate equivalence between ASE and CFD predictions in the small angle of attack range can be assumed considering the numerical uncertainties due to the fact that CN(a) is very small in (4.20).

4.6 Conclusion The Pirajuba project has achieved the third version of an AUV that is ready to be applied as a test bed for investigations on hydrodynamics and manoeuvrability. Free model tests and system identification techniques are going to be applied to the estimation of hydrodynamic parameters and transfer functions that represent the AUV dynamics. From the point of view of estimation of hydrodynamic parameters, these tests are going to complement the static tests that have been carried out in the towing tank. Considering the modelling of the hydrodynamic efforts it is important to emphasize the symbiosis of ASE and CFD methods. Numerical flow simulation can provide insights to the improvement of the ASE formulation. From the other point of view, the ASE formulas can help to understand and verify the results generated by the CFD approach. Moreover, the application of CFD methods can be simplified, saving expertise and computational resources, when combined with the ASE methods.

References 1.

2.

Tyagi, A. and D. Sen (2006). Calculation of transverse hydrodynamic coefficients using computational fluid dynamic approach. Ocean Engineering, vol. 33, pp. 798–809 Jagadeesh, P., K. Murali and V. Idichandy (2009). Experimental investigation of hydrodynamic force coefficients over AUV hull form. Ocean Engineering, vol. 36, no. 1, pp. 113–118. ISSN 0029-8018 (Autonomous Underwater Vehicles)

An AUV project applied to studies on manoeuvrability of underwater vehicles

89

3. Phillips, A., M. Furlong and S.R. Turnock (2007). The use of computational fluid dynamics to determine the dynamic stability of an autonomous underwater vehicle. In Proceedings of 10th Numerical Towing Tank Symposium, Hamburg 4. Kimber, N.I. and K.H. Scrimshaw (1994). Hydrodynamic testing of a 34 scale Autosub model. In Oceanology International 94, Brighton 5. Tang, S., T. Ura, T. Nakatani, B. Thornton and T. Jiang (2009). Estimation of the hydrodynamic coefficients of the complex-shaped autonomous underwater vehicle TUNA-SAND. Journal of Marine Science and Technology, vol. 14, no. 3, pp. 373–386 6. Singh, S.N., A. Simha and R. Mittal (2004). Biorobotic AUV maneuvering by pectoral fins: inverse control design based on CFD parameterization. IEEE Journal of Oceanic Engineering, vol. 29, no. 3, pp. 777–785 7. Lauder, G., P. Madden, I. Hunter, J. Tangorra, N. Davidson, L. Proctor, R. Mittal, H. Dong and M. Bozkurttas (2005). Design and performance of a fish fin-like propulsor for AUVs. In Proceedings of 14th International Symposium on Unmanned Untethered Submersible Technology, Durham, New Hampshire 8. Yamamoto, I. (2007). Research and development of past, present, and future autonomous underwater vehicle technologies. In Proceedings of International Masterclass AUV Technologies. Polar Science, Southampton, pp. 17–26 9. Inoue, T., H. Suzuki, R. Kitamoto, Y. Watanabe and H. Yoshida (2010). Hull form design of underwater vehicle applying CFD (Computational Fluid Dynamics). OCEANS 2010 IEEE, Sydney, pp. 1–5 10. Husaini, M., Z. Samad and M.R. Arshad (2010). Performance analysis of AUV propeller using CFD. In 3rd International Conference on Underwater System Technology: Theory and Applications 2010, Cyberjaya, Malaysia, pp. 204–208. 11. Wu, L. (2010). Applying dynamic hybrid grids method to simulate AUV docking with a tube. In IEEE International Conference on Information and Automation, Harbin, China pp. 1363–1366 12. Husaini, M., Z. Samad and M.R. Arshad (2009). CFD simulation of cooperative AUV motion. Indian Journal of Marine Sciences, vol. 38, no. 3, pp. 346–351 13. de Barros, E.A., A. Pascoal and E. de Sa (2004). AUV dynamics: Modeling and parameter estimation using analytical, semi-empirical, and CFD methods. In Proceedings of IFAC Control Applications in Marine Systems, Ancona, Italy, pp. 369–376 14. de Barros, E.A., A. Pascoal and E. de Sa (2006). Progress towards a method for predicting AUV derivatives. In Proceedings of IFAC Manoeuvring Control Marine Crafts, Lisbon, Portugal 15. de Barros, E.A., A.M. Pascoal and E. de Sa (2008a). Investigation of a method for predicting AUV derivatives. Ocean Engineering, vol. 35, no. 16, pp. 1627–1636 16. de Barros, E.A, J.L.D. Dantas, A. Pascoal and E. de Sa (2008b). Investigation of normal force and moment coefficients for an AUV at nonlinear angle of

90

17. 18.

19.

20. 21. 22. 23.

24. 25.

26. 27.

28.

29.

30.

31.

32.

Further advances in unmanned marine vehicles attack and sideslip range. IEEE Journal of Oceanic Engineering, vol. 33, no. 4, pp. 538–549 Myring, D.F. (1976). A theoretical study of body drag in subcritical axisymmetric flow. Aeronautical Quarterly, vol. 27, no. 3, pp. 186–194 Ortiz, A., J. Proenza, G. Bernat and G. Oliver (1999). Improving the safety of AUVs. In Proceedings of OCEANS’99 MTS/IEEE, Seattle, USA vol. 2, pp. 979–984 Lichuan, Z., X. Demin, L. Jun and J. Lijun (2008). Design and experiment of automatic pilot for long range AUVs. In 3rd IEEE Conference on Industrial Electronics and Applications, ICIEA, pp. 1824–1827 Menter, F.R. (1994). Two-equation eddy-viscosity turbulence models for engineering applications. AIAA Journal, vol. 32, no. 8, pp. 1598–1605 Karamcheti, K. (1966). Principles of ideal-fluid aerodynamics. New York: Wiley, pp. 568–588 Hoak, D.E. and R.D. Finck (1978). USAF stability and control Datcom. OH: Wright-Paterson Air Force Base Bohlmann, H. (1990). Berechnung Hydrodynamischer Koeffizienten von Ubooten zur Vohrhersage des Bewegungsverhaltens. Ph.D. dissertation, Institut fur Shifbau der Universitat Hamburg, Hamburg, Germany Munk, M.M. (1923). The aerodynamic forces on airship hulls. NACA Tech. Rep., NACA, USA, p. 184 Jorgensen, L.H. (1977). Prediction of static aerodynamic characteristics for slender bodies alone and with lifting surfaces to very high angles of attack. NASA Tech. Rep. TR-R-474, NASA, USA ANSYS Inc. (2009). ANSYS FLUENT 12.0 theory guide. Available from http://www.ansys.com/ ESDU Aerodynamics (1978). The influence of body geometry and flow conditions on axisymmetric velocity distribution at subcritical Mach numbers. ESDU International, London, UK (Data Item 78037) Whicker, L.F. and L.F. Fehlner (1958). Free-stream characteristics of a family of low-aspect ratio, all movable control surfaces for application to ship design. David Taylor Model Basin, Tech. Rep. 933, Naval Ship Research and Development Center, Washington, DC, USA Hoerner, S.F. (1985). Fluid dynamic lift: practical information aerodynamic and hydrodynamic lift. Albuquerque, NM: Hoerner Fluid Dynamics. ISBN: 9998831636 Pitts, W.C., J.N. Nielsen and G.E. Kataari (1957). Lift and center of pressure of wing-body-tail combinations at subsonic, transonic, and supersonic speeds. NACA Tech. Rep, 1307, NACA, USA Dantas, J.L.D., E.A. de Barros, F.V. Boas, F.A. Mutscheler and C.H. Umeda (2011). Experimental research on AUV manoeuvrability. In 21st International Congress of Mechanical Engineering, Natal, Brazil, vol. 21, pp. 1–102 The Society of Naval Architects and Marine (1952). Nomenclature for treating the motion of a submerged body through a fluid. Technical and Research Bulletin, no. 1–5, pp. 1–15

Chapter 5

Neural network–based switching adaptive control for a remotely operated vehicle Matteo Cavalletti , Gianluca Ippoliti and Sauro Longhi 1

1

1

5.1 Introduction Great effort is currently being devoted in the development of underwater vehicles with self-governing capabilities, capable of performing reliably complex tasks in different environments and load conditions. Depending on the operative condition, the different possible vehicle configurations may or may not be known in advance. However, in general, it is not known a priori when the operating conditions are changed and what the new vehicle configuration is after the change. This operative situation is considered for a remotely operated vehicle (ROV) developed by ENI Group (Italy) and used in the exploitation of combustible gas deposits at great sea depths (Ippoliti et al., 2002). Traditional control techniques have been found to be rather inadequate for the control of this kind of mode-switch process; therefore, it appears useful to develop advanced control techniques to supply the underwater vehicle with the necessary ‘intelligence’ to achieve some degree of self-governing capability. In this regard, different control strategies have been developed and efficient implementations of such controllers in real environments have been proposed. Significant solutions based on adaptive control, robust control, variable structure control, Lyapunov-based control, model predictive control and intelligent control have been recently investigated (see e.g. Aguiar & Pascoal, 2001; Conte & Serrani, 1994; Conter et al., 1989; Corradini & Orlando, 1997; Cristi et al., 1990; Fossen, 2002; Longhi & Rossolini, 1989; Naeem et al., 2004; Sutton & Craven, 2002). An extended bibliography on the control of underwater robotic vehicles has been reported in Smallwood and Whitcomb (2004). Although all these methodologies showed good performance in controlling uncertain dynamic systems, there are unavoidable large transient errors at the time of task variation. The aim of this chapter is to propose an adaptive switching control strategy to cope with the large transient errors related to the considered mode-switch process, 1 Dipartimento di Ingegneria dell’Informazione (Information Engineering Department), Universita` Politecnica delle Marche (Marche Polytechnical University), Ancona, Italy

92

Further advances in unmanned marine vehicles

when there is no adequate knowledge of the different possible vehicle configurations. In fact, if there is adequate a priori information on the different possible vehicle configurations, a switching control scheme can reduce the transient response with respect to an adaptive control, when the vehicle changes its operative condition (Ippoliti & Longhi, 2004; Ippoliti et al., 2005, 2006; Narendra & Xiang, 2000). On the contrary, with a poor knowledge of the actual vehicle operative condition, a switching control is not able to guarantee the same control performance; the pre-computed controllers cannot cope with all environment and load conditions and an auto-tuning mechanism is required. Therefore, the switching control considered in this study consists of an adaptive control policy improved by the connection with a supervised switching logic. It is composed of a bank of alternative candidate controllers that switch among themselves according to a suitably defined logic (Angeli & Mosca, 2002; Antonini et al., 2006; Cavalletti et al., 2005; Hespanha et al., 2001; Ippoliti & Longhi, 2004; Ippoliti et al., 2005, 2006; Narendra & Xiang, 2000; Xiang & Narendra, 2002). This makes the approach particularly suited to deal with large parametric variations and/or uncertainties. At each time instant, the supervisor decides which candidate controller should be put into the feedback loop with the process. Multiple neural networks (NNs) have been considered to compensate the effects of nonlinearities, plant uncertainties, and store different dynamics of the vehicle for different tasks and configurations. For each known vehicle task and configuration, a non-linear stabilizing controller based on these NNs is designed. The switching logic is driven by a specially designed supervisor that at each time instant decides which candidate controller should be put in the feedback loop with the process, making use of a Lyapunov-based falsification criteria (Angeli & Mosca, 2002). A complete description of the considered NN switching control has been reported in Cavalletti et al. (2007). To cope with a poor knowledge of the actual vehicle operative condition, an adaptive mechanism is included in the switching logic. This adaptive control is composed of two NNs with self-tuning capability. NNs with their auto-tuning mechanism can identify the non-linear part of the process and store its unknown dynamics. The weights of nets are adjusted by a Lyapunov-based adaptive mechanism, and a proof of stability is presented. Radial basis function networks (RBFNs) have been used to develop the switching control scheme. These NNs have been widely used for non-linear system identification (Cavalletti et al., 2007; Chen et al., 1991) because they have the ability both to approximate complex non-linear mappings directly from input–output data with a simple topological structure that avoids lengthy calculations (Chen et al., 1991) and to reveal how learning proceeds in an explicit manner (Yingwei et al., 1998). These nets also avoid the non-linear optimization techniques used in the learning algorithm of Multi-layer neural networks (MNNs) and the related problems of local minima (Corradini et al., 2003). Moreover, in Poggio and Girosi (1990), it has been proved that RBFNs do have the best approximation property with respect to MNNs, and this significant result provides theoretical support favouring RBFNs.

Neural network–based switching adaptive control

93

The developed adaptive switching control has been applied to the tracking control problem of the considered ROV that, during its tasks, is subjected to different load configurations, which introduce considerable variations of its mass and inertial parameters. On the basis of the performed numerical simulations, satisfactory performance of the considered control system seems to be really attainable. The chapter is organized as follows. In section 5.2, the switching control approach for non-linear dynamical systems is recalled. Some preliminaries are introduced in section 5.3. The details on the NN-based switching control are discussed in section 5.4, and the switching strategy is given in section 5.5. The ROV model is recalled in section 5.6, and the results of numerical simulations are reported in section 5.7. The chapter ends with comments on the performance of the proposed controller.

5.2 A switching supervisory control approach for non-linear dynamical systems In this section, basic definitions and concepts of a switching supervisory control approach for non-linear dynamical systems are recalled. This methodology has been developed to improve the performance of dynamical systems operating in rapidly varying environments. The main feature of switching control strategies is that one builds up a bank of alternative candidate controllers and switches among them according to a suitably defined logic (Angeli & Mosca, 2002; Chen & Narendra, 2000; Hespanha et al., 2001; Ippoliti et al., 2006; Mosca & Agnoloni, 2001; Narendra & Xiang, 2000). This makes the approach particularly suited to deal with large parametric variations and/or uncertainties. The switching logic is driven by a specially designed supervisor that uses measurements to assess the performance of the candidate controller currently in use and also the potential performance of alternative controllers. At each time instant, the supervisor decides which candidate controller should be put in the feedback loop with the process. In the proposed solution, multiple models have been used. These models are based on NNs, which are widely used for the identification and control of complex nonlinear dynamical systems in the presence of uncertainties. By assuming that N identification models fMj gNj¼1 are used in parallel with the given plant, the objective is to determine which model is the closest (according to some criterion) to the plant at any given instant and to use the corresponding controller Cj, designed for controlling the corresponding model Mj, to control the plant (Chen & Narendra, 2000; Narendra & Xiang, 2000). The structure of the proposed control scheme is shown in Figure 5.1. With this control structure, the performance of the different controllers cannot be directly evaluated in parallel, because at any time instant, only one control input can be chosen. Therefore, the ^ j ; j ¼ 1; 2; . . . ; N , of the multiple models and the plant output output responses x x are used for an estimate of the controller performance. Hence, at every instant, ^ j  x are determined with some measures of the identification errors ^e j :¼ x suitable performance indices Jj(k), j ¼ 1, 2, . . . , N. These performance indices are

94

Further advances in unmanned marine vehicles xˆ N

MN

+

εˆ N Σ –

xˆ 2

M2

CN

M1

uN u

C2 C1 xr +

+

xˆ 1

PLANT

+

Σ

εˆ 2

Σ

–

εˆ 1 –

x

u2 u1

Σ –

Figure 5.1 Switching supervisory control

used by the supervisor to direct the search to a new controller to be inserted in the control loop. This search of a new controller starts when the supervisor detects the necessity to change the actual controller that appears inadequate to the control task. Different supervisory switching approaches have been developed: those based on process estimation (Baldini et al., 1999; Borrelli et al., 1998; D’Amico et al., 2006; Hespanha, 2001; Hespanha et al., 1999, 2001; Hockerman-Frommer et al., 1998; Ippoliti & Longhi, 2004; Ippoliti et al., 2006; Karimi et al., 2001; Morse, 1995, 1996, 1997; Narendra & Balakrishnan, 1997; Narendra & Driollet, 2001; Narendra & Xiang, 2000; Narendra et al., 1995) and those based on a direct performance evaluation of each candidate controller (Ippoliti et al., 2006, 2004; Kosut, 2004; Mosca et al., 2001; Safonov & Tsao, 1997; Woodley et al., 1999). In the proposed solution, the last approach is considered and specialized by Lyapunovbased falsification criteria. If the controller is ‘falsified’ (Angeli & Mosca, 2002; Safonov & Tsao, 1997), then the supervisor replaces the formerly acting controller with the controller corresponding to the model with the minimum performance index. When the process to be controlled is non-linear, NN-based models fMj gNj¼1 can be used for the theoretical ability of NNs to approximate arbitrary non-linear mappings (Cybenko, 1989). The approach considered in this study has been applied to the control of an underwater vehicle, and details are reported in Cavalletti et al. (2007). In particular, RBFNs have been used for identification and control.

Neural network–based switching adaptive control

95

5.3 Preliminaries Consider that the class of single-input, single-output non-linear systems is expressed in the following companion form (Slotine & Li, 1991): xðnÞ ðtÞ ¼ f ðxðtÞÞ þ gðxðtÞÞuðtÞ

ð5:1Þ

where t is the time, u(t) [ < is the control input, x(t) [ < is the output of interest, x(t) ¼ [x(0)(t), x(1)(t), . . . , x(n1)(t)]T [