tjohn WIlli

DEVELOPMENTOF SUBSEA ROBOT NOMAD WIlli A MICRO COMPUTER BASED INTEWGENT CONTROL SYSTEM

By ZHONGQU N LU. P.ENG.

B.Engin Electrical Engineering M.EnginElectricalEngineering

A ThesisSubmittec:lToTIleSchoolOf Gradua teStudies In Partia lFulfillmen tOfTheRequiremen ts For TheDegreeOfPhD InElectricalEnginee ring

Facu ltyOf EngineeringAnd AppliedScience MemorialUniversi tyOfNewfoundland

August.1996

Stjohn's Newfoundland Canada

.+.

Acquisitionsand ^BNottI8queduAcquisitions^C&nadaet^nationale Bibliographic5eMces services bibliographiques 395WlIAnglDnSb_ 395.....WeIIinglDn oaa-ON K1A0N4 ou-aON K1AONot

eo- .,..,..

The authorhasgranted a non- exclusive licence allowingthe NationalLibraryof Canada to reproduce,loan, distribute or sell copies ofthisthesisinmicroform, paperor electronic formats.

The author retains ownership of the copyrightinthis thesis. Neither the thesis nor substantial extracts fromit may beprintedor otherwise reproduced without the author's permission.

L'auteur a accorde une licence nOD exclusive permettant

a

^la

Biblictheque nationale du Canada de reproduire,preter,distnbuer au vendre des copies de cettetheseSOllS

laforme de microficbe/film. de reproduction sur papier au sur format ejectronique.

L•auteur conservela proprietedu droit d'auteurqui protege cette these.

Nilathesenides extraits substantiels de celle-ci ne doivent etre imprimes ou autrementreproduits sans son autcrisation.

0-612-36207 -8

Canada

ABSTRACT

Thegoal of thisPhD projectwastodeSignand developa smallinexpensive subsearobot withamicrocomputerbasedinte lligent control system. Therobot developediscartedNOMAD. It couldbe thekey elementin aDistri but ed Marine Observat ion System( OMOS). MostEngineeri ngPhDs are researchoriented; this onehas adesignfocus.

NOMAD use s an airlwaterballasttank instead of a battery/motor system to driveitself vertically.To facilitat e missionrequirements,gre ateffortswere made to develop ahighperfonnanceonboard micro computer based controlsystem.To deal withthe uncertainty and nonlinearity ofthe robot model,inv estigation s were conduded tocheck the potentialofstrategiesbasedonNeuralNetworks andFuzzy Logic. A Realnme Kernelsoftware andanonboard microcomputer with a Z180 CPU were used toimplement aFuzzyVariabfa StructureSwitching(FVSS)control scheme and a multiple-task,multiple-layeredcontrol structure.

The designanddevelopmentprocess forNOMAD are detailedin this thesis.

Theresultsofdigitalsimulation,theoreticalanalysisandtypical datarecorded from testsin a deep water tankon therobotarepresented. Successfultestsand good agree ment between dataand analysisindicate great potential for industria l applicat ion of thetechnologies devel opedinthis project.

ACKNOWLEDGMENTS

The authorisgreatty indebtedto my supervisors.Dr.M.J.Hinchey and Profe ssorD.

A.Friis,fortheirexcelentguidance,supportandunderstanding.

Special!hamare alsoduetothe TechnicalStaffintheElectronics labOratoryand intheCenterfor Comput erAidedEngineering (CAE),whomade theirservicesreadily availa ble throughoutthehardware andsoftwa re developme n tphase of thisproject.andto thestaffin the University'sTechnical Serv ice s,wncgave greathelpduring thecons tructi on and testingphaseof the proj ect.

The author gratefully acknowledges financialsupport fromthe author'ssupervisor, Dr.M.J.Hinchey,theFacu lty of Engineering andAppliedScience andtheSChool of GraduateStudiesof MemorialUniversity of Newfoundland.

CONTENTS ABSTRACT

ACKNOWLEDGMENT UST OF FIGURES UST OFTABLES

Chapter1 INTRODUCTION

§1.1 ConceptoftheDistributedMarineObservation System(OMOS)

§1.2 Overview ofThisPhDProject

ill vii xi

§1.3 Thesis Outline .9

Chapter2 OVERALL DESIGNAND DEVELOPMENTPROCEDURES 13

§2.1 Technical Requirementsfor the Subs e aRobot NOMAD 13 p.2 The Row Chartfor the Developme ntofthe SubseaRobot Nomad

,.

§2.3 GeneralDesignoftheSubse aRobot NOMAD 17

§2 .• ArchitectureoftheSubse a Robo t NOMAD 2.

Cha pter 3 THE DYNAMIC MODEL OF THE SUBSEA ROBOT NOMAD 3'

§3.1 TheoreticalCalculationfor the Model oftheRobot Body 31

§3.2 Experime ntal Estimation for theBodyMod el 34

§3.3 Simplificationof theBallastTank Model 38

§3.4 Linearizationof the Body Model 43

C~pter4 CLASSICALCONTR OL STRATEGIESFOR ROBOTDYNAMICS 41

§4.1 ConventionalPIOControl SCheme 41

§4.2 Sliding Mode Control Scheme 50

Chapter5 NEURALNETWORK BASEDADAPTIVECONTROLSCHEME 51

§5.1 Config uratio nfOf"theController ofNeura l NeflNOoo

sa

§5.2 OptimizedDesignrcr the Linear ReferenceModel 61

§5.3 Numeric SimulationforUleNeuralNetwor1(Based AdaptiveControl 13

Chapter6 FUZZY VARIABLE STRUCTURE SWITCHINGCONTROL SCHEME 90

§6.1 GeneralOesaiptionfor theFuzzyControl SCheme 91

§6.2 FuzzyDepth Control With Triangular MembershipFuncti on 93

§6.3 FuzzyVariableStr1Jctu reSwitchi ngControll erWith ADiscon tinuous 101 Output

§6.3.1 ProblemsRaisedBythe Disco ntinuousOutputoftheFuzzy 103 Controller

§6.3.2 Solutions TotheProb lems 114

Chapter 7 SOFTWARE ARCHITECTURE FORTHE CONTROL SYSTEM 126

§7.1 ContextRela tionshipin the Software Structure 126

§7.2 Real Time Kernel (RTK)-Implementationof the Software Structure 126

§7.3 Inte rruption -TimeManagement of Computi ngCapacity 134

§7.4 SpaceManagementoftheMemory Capacity 140

§7.5 Functio nalConfig uratio nof the Contro lSoftwa re 146

Chapter8 PRACTICALDEVELOPMENT AND PERFORMANCE TEST 152

§8.1 Deta iledDigital Simulatio nAt MechanicalComponent level 153

§8.2 Real TimeDigital Analog Hybrid Simulation 159

§8.3 Tests for Subsyst emsinA SimulatedEnvironmentandIdentification 164 for the Bod yModel

§8.4 In-Water Tests forthe Dynami cPerformance 174

Chapter9 CONCL USIONSAND RECOMMENDATIONS

§ 9.1 Condusi ons

§9.2 Recommenda tion s

REFERENCES

188

188 190

192

LIST OF FIGURES

FIGURE NO. NAME OF THE FIGURE

Figure1.1 .1 The Helical Trajectory of a Torpedo-Uke Subsea Robot CanyingOut Vertica l MovementMissions

Figure1.1.2 The Oistr1buted MarineObservation SystemwiththeSubsea Robot NOMAD

PAGE

FlQ'ure2.2.1 Procedures in the Deve lopment of SubseaRobot NOMAD 16

rtg'ure2.3.1 layoutof the PneumaticSubsystem 18

Figure2.3.2 TheContinuousWater LevelTransducerandthe Water Level Controlloop 22

Figure2.3.3 Wave AdionTransducer 23

Figure 2.3.. Layoutofthe MicroComputer with Z180CPUUsed in the Onboard 24 Cont rolSystem of NOMAD

FigureVI.l Architedureofthe SubseaRobot NOMAD 29

FIQure2.4.2 The Subsea RobotNOMAD Outing the Testsinthe DeepTank 30 Figure3.1.1 TheDragCoefficientVersusthe ReyooldsNumber 32

Figure 3.2..1 TheBlock Diagramofthe Body Model 36

FlOure 3.3.1 The BallastTankModel RepresentedbyaFirstOrder Device 36 Figure 3.3.2 TheWaterLevelResponseof An IdealFirstOrder Device 37

Figure3.3.3 Calculation FortheDraw-inFlow Rate 38

Figure3.3.4 TheBlock Diagr1illmof the BallastTankUsingSolenoidValves 39 Figure3.3.5 Make Outflow Velocityv~Equal to InnowVelocityv.atthe HalfHeight of 40

the Ballast Tank

Figure3.3.6 Comparison Between servoValveand Solenoid ValveSchemes Figure3.4.1 Unearizati on of theNonlinearDrag ForceFeedback Figure3.4.2 The ApproximateUnearModel of theRobot Body Figure3.4.3 The Blodt Diagram of the Whole System Figure3.4.4 TheSimplified Model of the WholePlant Figure4.1.1 PIDControl Scheme

Figure4.2.1 TheSWitchSurface S(t)in theSliding Mode Control

vII

"

4 '

«

4 '

.. 4.

"

F"lgure 5.1.1 SChematic Representation of StaticNeuralNetworK 59 Figure5.1.2 SChematicRepresentationof DynamicNeura lNetwork 60 Figure5.1.3 Three MainParts of theNeuralNetwork BasedContrn ller. NNA.NNI.OLR 63

Figure 5.2.1 Strudureof theOptimal Linear ReferenceModel 69

Figure 5.2.2 Detailed Strudure of the OptimalUnearReference Model 70 FlQure5.2.3 Location of theEigenv alues of the Optimal Unear Reference Model 72 Ftgure5.2.4 Response of theOptimalUnearReferenceModelto a Unit StepInput 73 Figure 5.3.1 Digital SimulationSChemeofthe Training Procedure forthe Neural 76

Network Identifier(NNl)

Figure 5.3.2 Divisions for the Ranges of theTrainingVariables n

Figure 5.3.3 FlatPortion of the Sigmoid Function (Shadow Area) 78 Figure5.3.. Surn-SquaredErrorin the Trainingof the Neural Network Identifier 80 Figure 5.3.5 Comparison of theOutputs ofthe AdualRobot Model and Thatof 151

Neural NeiwOrX Identifier(NNI)

Figure5.3.6 Digital Simulation Scheme of theTrainingProcedurefOf' theNeural 82 Network Adjuster (NNA)

Figure 5.3.7 Sum-SquaredError in the Training of the Neural Network Adjuster 85 Figure5.3.8 Step Responses and DrivingForce of theSubseaRobotwith the Neural 87

Network Based COntroller Under theInitialCOndition x(D)".,v(0);;0.15 Figure 5.3.9 Step Responses and DrivingForce ofthe Subsea Robot with the Neural 88

Netwol1lBased ControllerUnder the Initial Conditionx(D);; 2.v(0)·0.05 Figure5.3.10 Response and Driving Force of the Subsea Robot witl\ the Neural Network 89

Based Controller to a SineFonnReferenceInput UndertheInitial Condition X(O):o-e,v(O):O.15.

Figure 8.1.1 SChemeortheFuzzy Logic Controller 91

Figure 6.1.2 Eq uiv alen1Block Diagram After Moving the GainpA_ Backward 91 Figure 6.1.3 Different Definitions of the Membersl'lipFund ion 92 Figure6.2.1 Definition of theMembersl'lipFunctions fortheVariables:E.,V..."Fe 94 Figure 6.2.2 TheReasoningProcessUsedin the OepthController of NOMADRobot 95 Figure8.2.3 30Plot of the Fuzzy Logic Rule Base of the Depth Controller 97

Figu re 6.2.4 The Output of the Weigh1 Average Defuzzifier 99

viii

FlQ'ure6.2.5 Mapping Results of the Rules of 4,5,9,10 100 FIgure 6.2.6 Step Responses and Driving Force of the Subsea Robot with the Fuzzy 102

Logic Controller Which Has TriangularMembership Functions and a Continuous Driving Force

Figure 5.3.1 Float Micro SWitcnes Installed in the Ballast Tank 103 Figure 6.3.2 The CrispSetsof the Output Variable FC<k) 104 Figure 5.3.3 The Stair Case Nonlinearity Introduced By the D!sconUnuous Output of the 105

Controller

Figure 6.3.4 A System with a NonUnearController 107

Figure 5.3.5 The Plot of Desaibing Function for the Siair Case Nonlinearity versusthe 109 Amplitude of the Umit Cycle

Figure 6.3.6 Nyquist Chart ofG~{SIand the Describing Function of the Stair Case 110 Nonlineatity

Figure 6.3.7 The Definitions of Nonoscillation Margins 111

Figure6.3.8 IntersectionsA"~Indicate the Oscillations Caused Bylhe Stair Case 113 Nonlinearity

Figure 5.3.9 Frequency Character of a Phase LeadCompensator 115 Flgure 6.3.10 Bode Plot oflhe Phase Lead Compensator G,,(s) 116 Figure 6.3.11 Nyquist Chart of~(S)Ge<S)and the DeSClibing Function of the Stair Case 117

Nonlinearity

Figure 6.3.12 The Rule Base of the Fuzzy Logic Controflerwith a Variable Siructure 118 Figure 6.3.13 Comparison of the Umit Cycle Occurred in a FixedStructure Fuzzy 119

Controller of Discontinuous Outputs with That in the FVSS Controller Figure 6.3.'4 Nyquist Chart ofG~(s)G,,(s)and the Describing Functionof the Retay 121

Nonlinearity' with a Small Hysteresis Loop

Figure 6.3.15 The Stable Oscillation (A... Illo)Existingin theFVSS Controller 122 Figure 6.3.16 The Depth Response and Driving Force of the Subsea Robot with the 124

FVSS Controller Which Has a Crisp Membership Function and Discontinuous Wave Form in the Driving Force

Figure8.3.17 Depth Response and Driving Force of the Subsea Robot with the FVSS 125 Controller VVhich Has a Crisp Membership Function and Discontinuous

WaveFonnIn theDrivlnqFoo:e

Figure7.1.1 Hiera1CtlicaiSoftware Architectureforthe Micro Computer Based 128 OnboardContro ller

Figure72 .1 Detail of the Softwa re Stl'\Jdure ofthe OnboardControlle r 132

FIgure7.3.1 The Priority in Interruption 137

Figure7.3.2 Proposed andLabTested RadioCommuni cation SystemfortheSubsea 140 Robot NOMAD

Figure7.4.1 Registe rsCSAR,CBR,BBRUsedtoMap Logical Spaceto Physical Spa ",

Figure7.4.2 Calcu lation for the Physical Address 1-42

Figure7.-4.3 Corres pond ence Betweenthe Logical Space andthe PhysicalSpace 1-44 Figure7.5.1 Main FlowChartofControlPrecessof Ihe Subsea Robot NOMAD 151 Figure8.1.1 Flow Chart of Digital Simulation Around theEquilibrium stat eatthe 155

Com ponent Level

Figure8.1 .2 Depth ResponsesUnder Different Initial Conditions with Paramet er 157 Variati ons

Figure8.1.3 Small LimitCycleOscillationin theD-VPhasePlanewith Parameter 158 Vartations

FlQure8.1. 4 The Accum ulated EnergyCOnsumption Under the Wor1ting Press 158 Figure8.2.1 Real TImeDig italAnalogHybridSimulationand Testfor the OnbOard 160

Contro lle r

Figure8.3.1 Durable High WaterPressureTest forthe Electro nicBoxes andCable 165 Conned ors

Figure8.3.2 Testfor MaIntainingaConst ant WorkinqPressure 16&

Figure8.3 .3 Variationof the Wol1Ung Pressure Due10 Inputand Out put Disturb ances 167

FIgure8.3.-4 Setup forthe Ballast Tank Fundion Test 169

Figure6.3.5 Testfor the TIme Delayofthe Mechanics 170

Figure8.3.6 TheRecord sof InWater Test forEstimating the Paramete rsInthe Robot 173 Body Model

Figure 8.4.1 The Errorof the Depth UsingtheFVSSControlle r Wlt ho utaPhaseLead 178 Com pensat ion Recorded in theShallow Tank Tests

Figure 8.4.2 Nyquist Chart of G,,(s)and the Describing Functionof the Relay m Nonlinearity with SmallHysteresis

Figure 8.4.3 Recorded DepthWhenthe Subsea Robot NOMAD Moved Up and Down at tat DifferentSpeeds Inthe DeepTank Tests

Ftgure8.4.4 Deep Tank Test:for the Subsea Robot NOMAD Using theFuzzYVariable '84 StructureSwitching Controller withaPhase lead COmpensation

Figure 8.4.5 Deep TankTests for NOMAD with Multiple layered COntrolSystem res COntaining Diff erent Mission Descriptions

Figure 8.4.6 The RobotWas Woken From Sleeping at the Sea Floor,then SCanned!he

, ..

WaterBody,then Hoveredat a Specified Depth

FlQure8.4.7 WaterfAirJet Helped the Robot to BreakAwayFrom beingStuck10the ret SeaFloor

LIST OF TABLES

TABLENO. NAME OF THETABL E

Table 1.1.1 The Featuresof SubseaRobot NOMAD Compared withTraditionalOnes Table 5.3.1 The Training Rangesand Divisions

Table 6.2.1 Fuzzy Knowledge RuleBaseforDepthCont rol Table 7.2.1 Inteffil ptions and Their PointefSAddresses Table 7.2.2 Prioritiesof Interruptions Table 7.4.1 BasicFund lons of 64 KbyIelogical Space

PAGE

83

"

135 tae

CHAPTER 1 INTRODUCTION

With the signifseant increaseofhuman activities on andinthesea,longterm observationof themarine environment isbecomingmoreand moreimportantin orderto controlthelevelofwater pollution,tokeep the ecologicalbalance,andto developoffshoreindu stry(BusbyandVadus,1990).

AtoffshOre oil well sites,parnccratedrilling wastes are dischargedduring explorati onand produ d iondri lling operation s.In addition largevolumes of water alongwith oilare releasedover the lifetimeof a field.Sed iment on theseafloor at different locationsneeds tobepickedupas an indication of pollution.Researcher~

in the HibemiaManage ment and Developmen t Company Ltd.(HMOC)indicate that thewaterbodyfromthe benthic boundary layer,wheredri lling wastesconcentrate, shouldbesampledormonitoredoveralongperiodoftime(http1/www.hibemiabank.

e«rJ.1995).The resultsofchemicalanalysisfromthese samplesof sedimentsand waters couldbe used topredicttheimpacton the marine ecology.

Nearfishterms,watertemperature,tidalcerrents.water turbidityand levels ofdissolvedoxygenatdifferent depths arequit edoselyre latedto ecologicalimpact onmarine species. Researchindicates that highrates of organicmatterinto sediments result inlowlevelsof dissolvedoxygenthere andthis together with anaerobicmetaboli smby sulfate reducingbacterialead toreduced numbersof speciesand biomass of benthic fauna (http:JMww.hibemiabank.coml,11195).

Tooceanographers, thepH,turbidity,velocityof sound,disturbance stressat

differentdepths are allessentialto their research.

Militarypowerspaygreatattention tosubmarineforce s Noisedetectorsare often mounted atstrategic siteson the seabed locations tofindand recordthe noisefromprope llers ofpassingsubmarines.

§ 1.1 Conceptof theDi stributedMarineObse rv atio nSystem (DMOS)

$0far many sophisticat ed underwatervenictes have been developed,such as the Autonomous Benthic Explorer (ABE) developed by WOOds Hole Oceanograp hicInstitutionintheUSA (yoerger,Bradleyand Walden, 1991)and Odyssey developed by MIT(leonllrd,J.J.1995). Usually these are equipped withLongBase LineorGPS navigationsystems,and sonar scanners,anduse batteriesastheir mainpowersource.They canbelaunchedfromaship,or from a near-by harbor from whichthey can headto theworkingsites (NomocoandHattOl1,1986).Most of the developedsubseavehideshave a torpedoShape. Whentheyarerequiredto perform adescentor ascent maneuver.theyhave todOthisby going intoahelical trajectoryasshowninFigure1.1.1.This consumes alot of energy.Obviously,itis alsonotsafe andcosteffective to leavesuchasophisticat edsubsea vehicleat the workingsitesfor alongterm observationmission,whichmay take weeksormonths, eventhough they mayhavethat capability.

Therefore,a cost effectiveand energy saving system formissions with vertical movementsis needed. This thesis describes a subsea robotknownas

NOMADwith an onboard micro computer based intelligent controller which was developed for thispurpose (Lu.HincheyandFriis.1994). It is suitablefor long term observationand alsoforobtaining samplesfromthe sea flooratdifferentsites of interest,as shown in Figure 1.1.2.

For sediment and water sampling missions,a helicopterusingits navigation system couldcarry NOMAD to the siteof interest anddrop it into the sea.When NOMADdives into the water,its onboard micro computer would triggerits mission routine.When NOMAD arrives onthe sea floor, a mechanicalhand mountedonits bottom could grab a sampleofthe sediment. On theway back to the surface, NOMAD couldcarry out a water samplingmission. Ataspecified depth,a small chambercould be openedand thenclosedunder the control of the onboard micro computer. In thiscase,the velocity of the robot would be set to rather slow, to allow the robot to get the sampleofthe water body atthe precise specified depth.If itis notrequ ired to samplethe wateron thewaybacktothe surface, NOMADcould retum to the surface at themaximum speed.

For long term marine observation missions,the robot couldberequired to stay at a workingsiteup to weeks and in somecases months. The robot would spend most ofits nme"sleeping-on the sea floor; thenatpre-programmedtimes,it wouldmove upanddownvertically and recordvarious physicaldata versus depth.

After completing these measurements, the robot wouldgo back to thesea floorto steep. The recordeddata would bestored in the memoryarea of theonboard computer,At preset intervals,the robot would surface and transferdata to a shore

=

station by radio signals.Whenthe shore station has received these data, an acknowledgment signal would be sent to the robot which would allow the onboard computer to clearits memory area.Using radio communication,there wouldalso be anopportunity herefor the control personnel atthe shore station to modify missions and pre--programmed routines stored in the onboard computer.When the robot has finishedits mission,it would retumto the water surface and emitaradio signal that a helicopter could homein on.Dischargedcolor dye could alsobeused to help a helicopter pilotlocate therobot and retrieve it.

During a mission, a number of NOMAD robots couldbe dropped from a helicopter at different spots ofinterest. Theywould form an observation andda~

collection network. Obviously,the robots used in this Distributed Marine Observation System (CMOS) wouldbethekey components in this system. As noted,they would be ableto stayatthe work sites for alongtime or scanthe sea water vertically. They would also be ableto use a rnechentcal hand tograsp sedimenton the sea floor or chambers to sample water at differentspecified depths.

The low cost and energysaving features of NOMAD makeit quite different from traditional subsea robots.

A prototypeof NOMAD with anonboard microcomputer based intelligent control system has been designed, built and tested inthis PhDproject. Inslead of using a battery, DC-motor,and propeller asthe main propulsion system,NOMAD uses a simplewater/air,ballast tank plus gravityand buoyancyto make itself move vertically up and down (Lu. Hinchey and Friis,1994)(Hinchey and Muggeridge1994). A

comparisonofNOMADwith traditiona lsubsearobotsis give nin Table 1.1.1 (All m endinge r,1990)•

Table 1.1.'TheFeatures of SubseaRobotNOMA D Compared with Traditiona lOnes

RobotNOMAD TraditionalUnderWaterVehicles

MainPower Supply Compressed Air Battery

Energy Energy Consumpti ononlyatStart SignificantEnergyConsumpti on Consumpt ion andEnd of VerticalMovement throughoutVertical Movem ent

Max imum Speed sznvs Usually<1m1s

of VerticalMov ement

Rati o of Payload Mass Higher (1:3) lower(1:5-1:10)

to Vehicle Mass

LocaUng Method Using theNav ig ation System UsingtheNavigatio nSystem

on theHelicopter on theSubsea Vehicle

Dev elopment Costs Low High

(NotIncluding Research (SingleNOMA D < 10 K CanS) (Usually >1(}() K CanS) Labor Costs)

Maintenance Expense 10S/day 10D-1000$lday

During theMission

Suitabilityfor the Task of Go'" ^NotWOf1hy

LongTerm Observation

§ 1.2 Overview of This PhD Project

NOMAD was designed by the author in the first half of 1994. Manufacture was started in June 1994. Assembly was performed by the author in February 1995.

During this period,every effort was made to develop a good computer control system. Many investigations and studies were conducted for different control strategies,including: Proportional-Integral-Derivative (PIDI.Sliding Mode(SI01ine.

198 3and1984 ).Artificial Neural Network(lshibuchi,1993) , Fuzzy Logic (Self,1990),and Nonlinear Variable Structure Control (OeYong andPolson.1992).The author's studies focused on developing a controller with Neural Networks,Fuzzy Logic and with a Variable Structure.Most of author's studies were based on digital simulation, and some on digital-analog hybrid simulation. Thetools usedin the simulations were C-,Togai TilshelJ (Fuzzy Logic) and Matrab (Neural Networks).A suitable control strategy,Fuzzy VariableStructure Switching(FVSS),and its software framework, were developed and tested during these studies.

Since January 1995. the author has concentrated on electronic aspects such as:computer interfaces, communication hardware, sensors,conditioners and drivers to physically implement the control strategies.In the process of the development of NOMAD, a great number of technical difficulties had to be overcome such as: the water proofing of the onboard electronics,the automatic maintenance of constant working air pressureandthe digital communication between the onboard micro computer and the station computer.

ByMay 1995thesubsearobotNOMADwithitsmicro computerbasedcontrol systemandFVSS control strategy was ready for testing in a water test

Tests werecarriedout duringJuneof 1995 to Dece mber of 1995,mainly for optimiZationof the control parameters.Thesetests onNOMADwereconductedin theDeepTank of theOceanEngineeringResearch Centerof MemorialUniversity ofNewfoundland. The results of these tests ind ica ted that the design and developmen tofNOMAD was successfu l. Thevast datarecordedin the tests compared well with theresults of theoreti ca lstudie sanddigital simul ations. The tests show edtha t NOMADwasreliableand dependabl e. NOMADwaspresented to thepublic onCanad ian Eng ineer ingDayof1996.It has also beendemons tratet;t for the engi neeringstudentsandnowit is be ing enhancedforindustrialapp lications.

§ 1.3ThesisOutline

Thisthesisgivesa detaileddescription ofthe de signanddevelopment process for the subsearobotNOMAD.Every subsystem ofNOMAD was designed and constructedin thisproject.However,the mainemphasiswasputon the control strategies and theircomputerimplementation.The thesisis organized asfollows.

Chapte rOne is theintrod uctory chapter.A briefreviewof thepreviouswork before this project ispresent ed. Based on the mission requiremen ts from applicati ons intheoffshore industry, a newconce pt of a• Distributed Marine Observa tionSystem-(DMOS )isputforward. Theresearch interest isdecl aredin thischapter.Itistophysicall ydevelopa prototypeof a singledegreeoffreedom

subsea robot with a micro computer based control system, that could be a critical part of the Distributed Marine Observation System.As noted above, this new subsea robot is called NOMAD.

Chapter Two gives the detailed technical requirements for the robot based on the nature of the mission.the procedures forits overalldesign and the flow chart for its development. The architectureofthe subsea robot and the design of its main subsystems are detailed in the last section of this chapter.

Chapter Three studies the dynamic behaviorsofthe mechanical setup ofthe robot. The model of the robot is dividedinto two main portions:the model of the robot body and the model of the ballast tank.The latteris the main actuator of the robot. Both of the modelsare highl ynonlinear.Linearization and simplificationare applied to the original models.Block diagrams are established for the designof the computer control system whichis detailedinthe following chapters.

Chapter Four to Chapter SIX study control strategies whichare candidates for possible usein NOMAD. Chapter Four discusses two classical control strategies PIC control and Sliding Mode Control. In Chapter Five ,a new strategy - Neural Network. Based Adaptive Controlis put forward for model Identification and depth control. Thisscheme consists of three main parts:the Neural Network for Model Identification (NNI), the NeuralNetwork for Adjustment (NNA)and an OptimalLinear Reference(OLR) model. Results of digitalsimulation for this scheme are presented.

Because there are some inherent defects in classical contro l strategies and the Neural Network Based Control Systemis computationally expensive,a more reliable

10

and economicalcontrol scheme,Fuzzy VariableStructure Switching (FVSS) Contro l,is putforward anddetailedin Chapter Six.This strategy causes the robot to undergo a self-excitedoscillationknown as a limitcycle.To enhancethe system stability and diminish the amplitudeof the limit cycle.hysteresis and phaselead compensation were introduced into thiscontrol strategy.

Digitalsimulation in the time domain and the describingfunctionmethodof analysis in thefrequency domainareapplied tothe FVSS controllerwithand without hysteresisandphaselead compensation. The studyindicated that this control schemeis not onlystable and robustbut also energy saving. Because of this,it was then adapted tobephysicallyinstalledonthe robot.

Chapte rSeven deals with thesoftware architecture. Becauseofthe complexity of the control proce ss, a multiple-task, multiple-layer software architectureis developed using Dynamic C and a microcomputer witha Z180 CPU.

The main aspects ofthe software are discussed:itsRealTime Kernelforthe implementation of the software structure;theInterruptionfor thetime management of computingcapacity;SpaceManagement ofthememory capacity. Thefunction configuration chart of the software is presented at the endof this chapter.

ChapterEight presents the process of developmentand testsin practical engineeringterms. As an importantstepinthe development,a detaileddigital simulation atthe mechanicalcomponentlevelanda digital-ana log hybr id simulation atthe system management level were carried out to confirm the suitability of the control system.In these two simulations, identical parameters and computer

11

hardware and software as that mounted on the robot were used. In this chapter. the setup and resultsofsome dry-land tests are presented.At the end of this chapter.

the process of the physical tests in the deep tank are also presented. The dynamic performance of the robot NOMAD are recorded under different test conditions.The results from theoretical analysis and calculation in previous chapters are confinned by the tests and compared withthe data records from the physical tests A number of recorded charts in different conditions are presented in this chapter.

Chapter 9 gives conclusions from this project and illustrates the experience gained.It also gives recommendations for futurework.

12

CHAPTER2OVERALLDESIGNANDOevELOPMENT PROCEOURES

This chapter gives the detailedtechnical requirements basedonthe nature of the mission.The procedureof generalsystem design for the subsea robotNOMAD andthe development schedule arepresented.The architectureof thesubsea robot andthe designof its mainsubsystems aredetailedin thelastsection.

§2.1 Techn ic alRequ irem en tsfor the Su bs ea Robo tNOMAD

Asmentioned in the first chapter, the subsea robot NOMAD withthe capability of carryingout differentinspectionmiss ionswill playacritical roleinthe Distributed MarineObservationSystem. To performsuchmissions the author concludedthatthe robotshouldhave thefollowingfeatures:

(1) It shouldbeenergysavingduringmissions with large verticalmovements.

(2) It should accuratelyindicate the presentdepthas theindepende ntvariable of therecorded marinephysicaldata.

(3) It should be ableto hoverat a specified depthwith an error notexceeding O.5m.

(4) Itshould have anonboarclcomputerwhichcontrolsthelowerleveldynamics and higherlevel behaviorsof the subsea robot.

(5) It shouldhavereliable communicationfacilitieswhich allowsthe robot to contact the shorestationcomputerwhen necessary.

(6 ) Itshouldbe easy to maintain,easy to rechargeorchangethemain energy

II

source(the gasbottleandauxilia rybatt eries),and easy to attach/detach differentpayloads.

(7) It shouldbe light inweight andhave highratioofpayload to robot self weight.

(8) Itshould have alow cost of developmentandoperation.

§2.2 The FlowChartfor theDevelopm ent ofthe Subsea Robot NOMAD

Ingeneral,underwater vehicl es are typically designed witha particular mission in mind. Thisleads to a set of requirementsthat define howvarious subsystemsshouldbe integrated.However,a system architecture foran underwater vehiclewhose functionis to act as a prototype for somenew technologiespresents uniqueproblems to the systemdesigner (HenricksonandOzielski,1995).Sincethere are no well-d efined mission s and detailed specifica tions, the subsea robot as a whole must be adaptableto systemsthat have yet to be designed.Ina conventionaldesign one has already narrowed the possible solution space intoa smallregion. Inthe innovat ive design with newtechnologies, one has a wide region of possible solutions. Themain pointis thatat the start we did not know whetherthereis a worthwhilewor1<ingsolutionto bring the conceptideato a successful prototype.

Atechnology test bed is const antlybeing usedin a system integrationmode.

Inotherwords, experi mentalordevelopmentalsubsystemsareconstantly being integratedinto the submersible.Thereis also a requirementto supportthe design and testin g ofthe new systemat varioustimesduringitsdevelopment. Thetesting may occurina pure simulatio nenvironme nt,wherevariousconceptsfor the new

"

system are evaluated,or in a hardwere-ln-the-lccpsimulation,where the new system is debugged while in operation.

Figure 2.2.1is aflowchart for the development of NOMAD.The first stage in the processis the conceptual design stage. The second stage is the main design process which consists oftheoretical design and simulationof the overallsystem together with detailed design of all the subsystems:energy supply,body layout and mechanical components,sensors and electronics,computer hardware and software, communications, and payload. Cost effectiveness analysis was carried out throughout the main design process. The searchfor suitablecomponents wasal~o involved in the process. The design requires the most effectiveandsuitable technology instead of the most sophisticated . Everypossiblealtemative was thoroughly studied.

The third stage isthe constructionstage.Thismainly consistsof subsystem assembly(electrical. mechanica land pneumaticsubsystems )and control strategy implementation (computer hardware and software). Many technical modifications are made at this stage. Thelast stage is the test stage,consisting of the dry-land tests, in-water tests and data analysis. Asseen inFigure2.2.1, many interactions occurred between the different stages and phases of development,so thai "on-line' designs and tests became essentialin the development process.

"

Mi$$lOl'lRequirements I

e...ultiClnofCostEtrectlY er'leSs

, FM •

,R,diO:

i WIW !

Commllnlellbons

!RS232i

!Cllble;

:: _,

Miero Computlll"Hl rdWalll Design

ITnb ....theSU~Robot!

~ llnCIComput_H.rd'war.

Buildu p

Oryl.llndTnt

' O _W'(K'

Oatil ArII IysIS

i

t

I I

^{P -}&.Ibs)'S' .m

... Ii

DigltaiSimull tiOll

I LJ

Figure2.2.1 ProceduresintheDevelopmentof SubseaRobot NOMAD

"

§2.3 General Design of the Subsea Robot NOMAD

Subsea robot NOMAD is the critical part of the Distributed Marine Observation System.To satisfy the requirements of its missions, it should be designedto have a modular structure inits physical layout and a layered structurein its control software.All overviewofits main physicalsubsystemsis givenbelow.

1. Main driving mechanisms

Taking advantage of the gravity and buoyancy forces,the subsea robot will consume no energy during descent or ascent. The only energy consumption will occur at the point wherethe direction of movement is changed. That makes·it possible for NOMAD toworkon long duration missions.

A ballast tankis installed on the upper partofthe robot. The water levelin the ballasttank is controllable. Asshown in Figure 2.3.1, the tankhas hares top and bottom. With thetop hole closed,letting gas bottle airflowinto the ballast tank forces water out the bottom and causes the robot to surface; lettingair out thetop and water in through the bottom causes the robot to sink. Thisis alsothe method used to adjust the subsea robot to a neutral buoyancy when the robot is commanded to hover at a specified depth. Pneumaticplugs within the ballast tank are used to open or closeits openings and thus control theflowof water into and out of the tank.

The plugs are controlled by solenoidvalves.A solenoid valve is also used to control theflow of compressed air from the gas bottle into the ballast tank.The valves are controlled by an onboard micro computer which is powered by DC batteries. When a

"

18

• 0"is sent out by the micro computer,thevalves arein the OFF position. The openings top and bottom are closed and there is no airflowfrom the gas bottle.In the OFF position the valves do not draw power fromthe batterieseither. Use of pneumatic plugs allows rapidfilling and purging of the ballasttank andthus fast response.A waterlevel sensor is installed in the tank forming a closec:lcontrolloop to adjustthe weighlbuoyanc yforce and obtain the neutral buoyanc y point. High pressure within the ballasttank couldbe usedto create strong wateror airjets to accelerate the subsea robot (Muggeridge andHinchey,1992)and it could also be used to blow away mud accumulatedon the subsearobotif it sat on thesea floorfor a long time. However.thiswould drain the compressed air bottlefast and wouldonly be used in rare cases.

Main Energy SupplySystem

A standard gas bottlewith maximum pressure 200 bar and 11liter capacity is adapted as a mainenergysource. Gas bottles in thesame serieswithhigher maximum pressure and larger capacity could be mountedon the subsearobotas well. NOMADisdesigned to havea modularstruct ure. Itonly takes acouple of minutes to change a gas bottle.

A first stage pressure reducer(SCUBA) reduces the outputair pressure from the bottle to13 bar.Asecond stage adjustablereducer(FEST a LRI/8FT)reduces the pressure from 13 bar to around3bar. Sincethe reference pressureused bythe pressure reducer isthe pressure inthe surrOUndingwaters,the stabilizedworking

19

pressure is with respect to the pressure atthe current depth. Tooperate the solenoi dvalves,the workingpressure of the pneumaticcontrolsystem~ustbe kept constantaround 3 barfor anygas consumption rateorrobot depth.A flowcontrol valve(FESTO GRO-M5)isused to adjust theflowrate when the ballasttankisbeing charged.The ballasttank has a reliefvalve(FESTO H112B)to vent pressure to the surroundingsif therelative pressure acrossits wall becomes too high.

3. Sensors

NOMADhasfour different typesof electronicsensors on board.These detect respectively: the depth of the subsea robot,thepressure inthe ballasttank,the water level in the ballast tank andthe waveactionofthesurro und ings.

(1) Depth sensor

A pressure transduce r (OMEGAPX181) is usedtodetect the depth. This electronic sensor transfers the pressure differencebetween its input and areference pressureto a voltage range of 0 - 5 V. Thenthevoltage signalis sent to a 12 bit NDconverter. Since PX181 is a relativepressure sensor,it isimportant to keep a stablereference pressureon the subsea robot. The sensoris fixed in a watertight box, inwhich the airpress ureis kept at1 bar.Theinputto the sensoris opened10 the surroundingwater.

(2) Pressure Sensorin theBallastTank

A pressure transduceris used to measure the airpressu rein the ballasttank.

The output ofthe transduceris fedto the onboard microcomputerthroughan 8bit

20

AIDconverter.The onboard micro computer adjusts the pulse width of the electrical current supplyingthe solenoidvalve which controls the air flow charging to the ballast tankso that thepressure insidethe tankcanbe charged to a preset value which is proportional tothe intensity of the water/air jets.As mentionedbefore,the water/ai rjets help to accelerate theinitial speed of the subsearobot.

(3) WaterLevelTransducer in the BallastTank

A continuous waterlevel transducer was goingtobeinstalledin the ballast tank as shown in Figure 2.3.2. For this sensor,the capacitance between the conductorinsidethe probe and the water varies with the waterlevel:the higherthe waterlevel , thelarger area of the twopolar plates, the bigger the capacitance.the wider the pulseoutput from the Mono-PulseGenerator,and the higher the voltage output from the LowerPass filter(Allocca andStuart.1984.)

This type of water leveltransduceris able to obtain continuo uswater level measurement. However,it is excenslve (1000 US$) and very sensitive to electromagnetic disturbancefrom surroundings. In theprototype of NOMAD,five micro switcheswith small floaters aremounted along the axisof the ballast tank.

They are connected to the digital1/0 ports of the onboard microcomputer. The advantageof this alternative is that it is reliable,economicand simple. Each switch canindicate two water levels by the status of"ON-and·OFF" Hence10water levels could be precisely detected.The water levels between the switches would be estimated by the software.

21

Fit;pre2.J.2 l'heContinuous Watwleve! Transducer and Ihe WattrLevIIl Contrclloll p

(4) Wave Action Sensor

This sensor is designed to detect the wave action of the surroundings. It transfers the robot rockingto a series of pulses. As shown in Figure 2.3.3,the sensor consists of two micro switches,a mass and a chamber.If the subsea robot is subjected to a big wave disturbance,the mass willmove between the left and right ends. The outputs from the switches are fed to the seriallIOportofthe cnboard computer. The time durationI"I;and the frequencyare used to scalethe"wave action",whereI,is the limerag between the back edge of phase B and the fronl edge of phase A andI;is the width of the pulse.

The wave action sensordeliversinformation about surroundings tothe higher level controllerof the robot. Thisallows the onboard micro computer totakethe state of surroundings into account when it makes decisions.Itcould also record the marine dynamic dataat different depths.

22

Figure 2.3.3 WaveActionTransducer

4. The Onboard Micro Computer

To facilitate missions of long term observation an onboard computer is necessary. It should have sufficientinput ports andoutputportsto receivethe information from different sensorsand drive the different actuators,as wellas sufficientmemory capacityind udingEPROM,battery backed RAM,andEEPROM to store the system program, mission description and recorded data. A local expandablebusis also required to supplythe power,facilitate the meansfor data exchange and access toinstrumentsin the payloadchamber.

A microcomputer named·Rugged Giant",fromthe Company of Z-World Engineering, was selected to be used as the onboardcomputer for NOMAD.It has sevenAiDinputs,twoD/A outputs,ten digitalinputs,two serialcommunication ports, andaPLC BusExpansionPort(ZwortdEngineering,1992). The layoutofthe micro computer withZ180 CPUusedin theonboard controlsystemof NOMA Dis shown in Figure 2.3.4.

Whenthe systemis underdevelopment,a serialRS232 port is usedto

23

Figure 2.3.4Layout of the MIcro Computer with Z180CPU Used in theOnboard Control System of NOMAD

connecttheoobcerdmicrocomputertothestationcomputer.Thesystem program written inDynamic Cis debugged andcompiled to machine code and then downloadedtothe EPROMofthe onboard computer.whilethe missiondesaiption and missionrelated parametersare downloadedto the EEPROMoftheonboard computer.Thedatarecorded by differentsensors arestoredin battery backedRAM on theonboard computer. Theyaretobeuploaded tothe stationcomputerlater.

\'Vhenthesubsearobot worksinautonomou s mode orinremoteautomaticmodeit willbecommandedbased onthe agend astored in theEEPROMandthe decisions madebythe higherlevel controloftheonboard computer.

A short wave radioemitter/r eceiver willbe connectedto theonboa rdcompute r and anantennawillbemountedon thetopof the robotWhentherobotemergesat thewater surface.it willbeable to communicatewith thestationcomputer to send the recorded data or receive instrudionsforagenda and missionmodification.

A set ofDC batteriesis mounted onthesubsea rcoct.with thecapaci tyof 1200 mAh,recharg eable. The biggestusers oftheDC electricalpowerarethe soleno id valves(24V,2OmA).However,they wouldnot drainany energy whenthe robo tis"asleep-except foratiny current supplyfora timecounterworkingall the time. When aspecified timein the agendaarrives,aninterruptionrequ estissent to the CPU,thesubsearobot then wakesup and entersinto missionmode.The battery is expectedtolastforalongperiod.which depends on themissionintensity.

However,if theinstruments inthe payload chamberconsumesignifi cantpower,an extra batterywould be needed.

"

5. Payload

For most missions, the payload willbeattached at the bottom of the subsea robot. To sample the sediment on the sea floor,a simple mechanical hand couldbe mounted on the robot frame. The hand could bedrivenbyasmall pneumatic cylinder using the power source in the air bottle and controlled by the onboard computerthrough solenoid valves. Electronic instruments couldbesealed in the payload chamber. An open structure could also be applied to the payloadif the instruments are watertight. When the massdensity of the payload is much greater than that of water,the buoyancychamber on thetop of the robotshould be changed.

Buoyancy chambers with different sizes havebeen madefor compensating the higherpayload mass density. Presentlythe maximum payload for NOMAD is 16 kg.

Because of using gravityand buoyancyforcesto drivetherobot up and down,the robot would not consume more energywhenit carries a heavierpayload.However, the acceleration willbesmaller and the responsetime will belonger.

§2.4 Architectureof the Subsea Robot NOMAD

Thegeneral configurationofthe subsea robot NOMADis shownin Figure 2.4.1and 2.4.2.The robot is designedto be approximatelyneutrally buoyantwhen the ballast tank contains 6 kgof sea water (Effectiveheightofthe ballast tank,Heff-

'( "HD ')

./00mm;Diameter of the ballast tank.DrmJt=ZOO mm:

'2

~.".6kg ). A full

26

ballast tank is denoted as+6kg(i.e.the SpecificGravity of the whole robot is greater than1),and an empty ballast tank isdenoted as -6kg(l.e.the SpecificGravityof the wholerobot islessthan1).These are the maximum forceswhich drive therobot up anddown.The buoyancychamber onthetop is used to balance the payload. The purebuoyancy force created by the buoyancychambercould be adjusted by putting some waterintothe chamber. The neutral status issearched by the onboard computer at the stage whenthe robot isthrown into thewater.

Four aluminumboxeswhich are completely watertight areattached tothe frame of the robot.Thesecontain thecomputer, electronicsand sensors,solenoid valves,and communicatio ns. Theconnectors for cables and pneumatic pipes that run through theseboxes arealso water tight. Onthe end of everyexhaust pipe open to the water,one-way valves are mounted. Hencethethree subsystemswith differentworkingmediums- electricity ,air,and water are totallyinsulated. That is, the solenoidvalves, electronics,and pneumaticparts are kept dryall thettrne.

The payloadandother heaviercomponents are well below the buoyancy chamber andballasttank. Due to the roughlyneutralweight of the air bottle,the large separationof the centerof buoyancyandthe centerof gravityproduces a stable platfo rmwhich can maintain verticalpostureeven in aroughsea.

Followi ng are the technical detailsofNOMADdevelopedin this PhD project:

Weight: 35kg Height 1.3 m Maximum diameter: 0.3m

27

Working airpressure: 3 -13 bar Worki ng voltage: 24V Maximum working oxrent 100mA;

• • sleep-osrent <1rnA Battery capacity: 1200 mAh

Standard Volume(under1barpressure)of gas needed for changing directionof motion once:

fromupwa rd to downward:0.02standardliters fromdownwardtoupward:0.35standard liters Airbottle capacity: 200 bar11 liter

Maximum exploration depth: 100 m (NeedsEnhancemen t) Maxim um driving force: 60N

Maximl.MTl speed: 2 mls Depthlocation error: <0.4m

"

BuoylIncy Che mber

wat~~::

- - - I --IlI--H

Tlllnsmitt«

&Receiv«"- 801( "'-,

InsidePressure Measurement Connection

DownwardOpen~__

Throttle -t-:;r-t

Payloa d Chamber

Prl!$sure

~-+l-f---+---^Release VI Ne

If-- -- BallastTank __Plugs

Fil'$lPressure _.__~ .--eeeueer

Figum2.4. 1 Architectureof the Subsea Robot NOMAD

as

F;gw,2.42 The Subsea Robot NOMADOutingthe Tests in the Deep Tank

30

r

CHAPTER 3 THE DYNAMIC MODEL OF THE SUBSEA ROBOT NOMAD

This chapter studies the dynamic behavior of the mechanical setup of the robotNOMAD.The model ofthe robot is dividedintotwomain portions:the model of the robot body and the modelofthe ballast tank.The latteris the main actuator of the robot. Both of the models are highly nonlinear.linearization and simplifications are applied to the original models.Blockdiagrams are establishedfor the design of the computer control system which will be described in the following chapters.

§ 3.1 Theoretical Calculation for the Modelof theRobot Body

It is necessary to estimatethe hydrodynamic characteristics for the robot before developingthe control considerations (Voelller.CookeandSlotine,1990). When a robot moves closeto or at a water surface,it generates waves.Because the energy in these wavesis suppliedby the robot. the generation process canbemodeled as a drag force. However,as subsea robotsusually do not operate close to the water surface, such dragis insignificant. Wind generated waves die out as one moves down from the water surface. At only one half a wave length down into deep water, the motionis only about 5% of that near the water surface. The wave length dependence means that only long wavelength waves or swells are feltwell downinto the water. This is where robots tend to operate. Swells have periods around10 seconds. Because the time it takes for a water particle to move around a robot is typically much smaller than this, the flows around the robot are approximatelyquasi-

II

steady. Thisimplies that wake drag forces are also approximatelyquasi-steady.

Thisin tumimpliesthat Reynolds Number isthe dominant parameter govemingthe load.Reynolds number canbeestimated as:

R.

=

^pU,.,DIf.J""1000x(0.05-2)x0310.001

=

1.5)(10" - 6 >110' where,

f.J=O.OO1kglmsistheviscosity of water;

p=1000kg/m'isthe density of water(testing done infresh water):

Um

=

(0.05mls -2mls)is thespeed of the robot relative to the swell;

D=O.3m is the characteristicdiameter oftherobot

10' 10' 10' 10· 10' 10' 10'

Reynolds Number

Figure3.1.1 TheOflIg Coefficient Ver=sThe ReynoldsNumber

Asshownin the Figure3.1.1,thedragcoeffici entisnearly a constant within R.e( lO^l,6x lo' ).

Accordingto Morison'sEquationthe basic dynamic modelfor a small subsea robot is

J2

(Equation3./.1)

where,M is mass of the robot, Z is vertical distance moved by the robot,F, is the inertia force on the robot.Folis the drag force on the robot, and

where, C",is inertia force coefficient,determined by the geometric shape of the submersible which is estimated to be0.5 (White, 1986); Vol","isthe volume of displacement of the robot.Uis the swell wave particle speed, and

r,

=0.5C~A""",p(dZ/dt-V)jdZ,dt -VI

where,Colis drag coefficient determined by Reynolds Number and the geometric shape of the robot;Ar.-^tis the front area of the robot(Sarpl<.ayaand Isaacson,19M).

Define:

Include the drivingforceF,intothe dynamic equation.It then becomes:

Because of the quasi-steady assumption, the termdUldtmay be ignored.Assuming x-xr,dxldt-X2,the state space equation is

where,

d:c1IdJ=x:

(M+a: )d:c:Idt-=-F~+F,-=-a1x! lx: I+F,

a_l=O.5C~Af-p

=

0.5 x(l-I O)xItx(O.lSi x 1000"" 3S-350 a:=C",PV~=0.5x1000 )(0.035-=17.5

(Equation3.1.2)

J)

§3.2 Experimental Estimation for the Body Model

Even though the robustness ofthecontro ller makes the system capable of tolerating a rather big model error,a more precsemodel is always helpful,especially inthe simulation.In the digital simulation, the real world plant will be replaced by its mathematicalmodel when evaluatingdifferent control algorithms.In order to have a more precise hydrodynamic model, experiments for identifying the coefficients in the equation

(Equation3.1. /)

were conducted.

Equation3.2.1always holds with different initialconditionsofx/(O)andxJfO)or with different driving forceFr Att"iTandj Twehave,

(M+a:) xll(i)+ al x~li)!x:li)j=F"

(M+a:Ixn/j)+a,r~ij)lr:.:(j)~=F,:

where,

F,I.F,l are different drivingforces in the experiment;

(Equation3.1.2)

x:Ji). r:lOare the accelerationand velocity atthe timet..iT,and underthe driving

x:2()).rn(^j)are the acceleration and velocity at the timet-jTandunder the driving forceF, 2 .

Properly choosing the driving forcesFp 1andFpJand letting:

gives

I

^x,,(I) x,,(O'x,, (i )1

I

x,, {J ) x,, {J )lx,, {J )1

~

⁰

I

^F" x,,(I)lx,,(I)1

I

r.; xli

))Ixn(J)

(Eq ua tio n 3.}.))

I

x "

(I) F, .

I

x

1dJj F' l

I

x,, (i )

x,,(

j)

x,, (I )lx,,(I)1

j.

x,, (J)lx,, {J )

(Equation3.2.4)

Usually "/andQ1vary with thevelocity ofthe robot. However,when the Reynolds Numberis within(tOl,6xlOs) ,itwillnot introducesignificanterrorIf they are treated as constants. Theresults of tests conducted inthedeeptank ofOcean Engine ering Research Center,Memorial University confirmed thata_landa₁haveno significant variation at differentvelocities of the robot. The detailof the experimentto identify theparameters ofu/andQ]is discussed inChapter8. The final result of the experiment shows:Q/=37.5,a1=15.

Hencethe second order dynamicmode lfor NOMADis

(Equation3.2.5)

Usingthismodel to simulatethe displacementresponse to the drivingforce Fp tandFp 2, theresults are quite close to the datarecorded from the deeptank test.

Jl

Thisisa nonlin ear model whichgives usaroughideaof the dynamic charact erof therobotbody. It is used as a part of theplant model inthelower control level design,Theblock diagramofthe body modelis shownin Figure3.2.1.

F, -I

! ^!

~

^I

~

FiglHe3.2.1The Block Diagramofthe Body Madel

§3.3 Simplificationof the BallastTankModel

As shown inFigure3.3.1,the water levelh inthe ballast tankis fed back to the cnbcam microcomputerandcompared witha desired valuehJ•The errore~is used to drivethe valvesopen and closed, so thathapproacheshJ

Figure 3.3.1 The BallastTank ModelRepresent edby a First Order Device

Ifservo valveswere used tocontrol theflowinto and outoftheballast tank, the modelofthe tank couldbe quite precisely represented by a first orderdevice

16

whiCh mainly consistsofanintegral path and a feedback path asshownin Figure 3.3.1, where T_ isthetime constant and A....is the cross sectionalareaofthe ballas t tank.Assuming thatthereferenceinputis astepwithheigh th~,theresponse ofthefirst orderdevice to thisinputis

and its derivativeand slope are

where

k.,.

is the slope of the responsecurveat the startingpoint. It corresponds10 the maximum flow rate.This time response is shown in Figure3.3.2. Theabo....e calcul ation is basedonusing servo valvesto control the flowinto orout ofthe ballast tank. Servovalvesarequite sensitiveto the impuritiesinsea water.The servoparts inthesevalvesare easily dogged.They31"9alsoexpensive,especiallythosethat havehighflow rates andcorrosi~roofcharacteristics

"~ .- ^{- .. 1 / -- .}

^.

O.6JIr~

.. - . .

, __ . .un-,.!!L.. .

T_

T_

Time

Figure3.3.2 The Water Lev elResponse ofAnIdealFirst Order Device

17

To enhance the reliabil ityandreduce the cost an openi ~s ing med"lanism driven by soleno id valves is adopted.

Whenthewaterinthetankisto be purged.

theplug at thebOttomofthetank isopened.

The water outwardflOW" rateis equalto the flowrate ofthecompressedair chargedinto thetank.This is adjustableby athrottle (seeFigure 2.3.1). Whenthewateris drawn into the tank.the plugsat bOth the topand the bottomof the lank are opened.

Figure3.3.3 Calculati on For theOraw-tn Flow Rate

1hetnfIOW" rate depends onthewaterlevel inthe tank. To explainthis, picktwo points in Figure 3.3.3: PointAis atthemiddleofthenozzle.pointBis an arbitrary pointoutsidethetank.. Ignoring the effectofvortices,a virtua l streamtubecould be imaginedtoconnectpointsAandB.Thelaw ofconservationofenergy requires that the water partideatpointA containthesame amountofenergy as at pointB.Here theBernoulliEq uation for one-dimensionalincompress ible flow,s valid (White,1980).

Thenwehave:

where,

p..,p, arethepressures atpointsA and 8;

"

h~.hllare the elevations of points A and B relative to a common reference:

V~'v1Iare theveloci ties of water particles at points A and B;

Substitution of the variablesIn Figure 3.3.3into Bernoulli's Equation gives

(pdg+ phg)+pgh~+~pv;^{=( p dg+p Hg+ plg)+pgh}ll +'iPV_II:

...hll+ l=h~, ^{:.ipv; -}tPV; =p g ( H-h) , where H ISthe height of the ballast tank,O.47m.

If point B IS far enough away from the nozzle,the velocityofthe water particle at B is nearly zero,so that we have V~=J2g ( H-h) .

The block. diagram of the ballast tank mOdelusing solenoid valves IS shownin Figure3.3.4,where

V.v is the water outflow velocity.It ISadjustable bythe airinfl ow rate whichis set by the throttle.

vp is the water inflow velocity.It equalsV~= J2g ( H-h);

A-.,IS the cross sectionalarea of the nozzle,A-.Io=If"C.02^l=12 5" 10-¹m~;

Q is the water flow rate into or out of the ballasttank.

Figure3.3.4 TheBlock Digram of the Ballast Tank Using SolllnoidValves

"

To make the water flow rate into and out of the tanksymmetrical at the neutral buoyancypoint, the air inflowrateis adjustedto makethe water outflowvelocity v.v roughly equal to the waterinflowveloc ity

v ,

^{when thewaterlevel}is at the halfheight ofthe ballast tank,as shownin Figure3.3.5,

V, =V, =,J2(H-h igl ' =,fHi .

.. i"

i ' .

'

^o

- ~

^{O.2H 0.' "}

i~l

^{OI!lH O,SH H}

FigunJ 3.3.5Making OulllowVeloaty"Equ:ailto InIIow VeloeitylaltheHaltHeogllto'th eB8HastTank

Substitutionofthe heightH ofthe ballast tank,H""O.47m,gives

VN

=.[iii

=J0.47x9 8=2.l4mls

Thetimeneededto drainthe amountofwater in thetank which a!lowsthe robot to shiftfrom fullydownwardmoving state tothe neutral stateis

t _ F,_ _ 6kg

- - p.vNoA..-.- IOOOkglml_{x 2.I4(ml}s) xI.256 xlO-'r^mz ) 223sec (Equation3.3.1) where,Fp-is the maximumdriving force in eitherdirection,F,..-""'6kg.

The time needed to draw in water 'Ntlich allows therobo ttoshift fromthe