Full Text 22

Embed Size (px)

Citation preview

  • 8/11/2019 Full Text 22

    1/108

  • 8/11/2019 Full Text 22

    2/108

  • 8/11/2019 Full Text 22

    3/108

    ThesisD

    escription:Thepurposeofthe

    thesisistoinvestigatedifferen

    tmethodsforunmannedaerial

    vehicles

    (UAV)autopilotdesign.

    T

    his

    includes

    path-generation,path-following

    controland

    regulation.

    Computersimulationsshouldbeusedtoevaluatethe

    performanceofthedifferent

    guidance

    -controllersystems.

    Thefollo

    wingitemsshouldbeconsidered:

    1.

    L

    iteraturestudyonUAVflightcontrolsystems.Giveanoverviewofdifferentmethodsfound

    in

    theliteraturefortakeoff,altitudecontrolandturningcontrol.BothdecoupledandMIMO

    designtechniquesshouldbereviewed.

    2.

    D

    efinethescopeofthethesisan

    dclarifywhatyourcontributionsare.

    3.

    C

    hooseanUAVmodelforcomp

    utersimulationsandgraphical

    visualizationinX-plane.

    4.

    D

    esignautopilotsystemsforautomatictakeoff,altitudecontrol

    andturningcontrol.

    5.

    D

    esignguidancesystemforpath

    -following.

    6.

    In

    vestigatehow

    goodtheautop

    ilotsystem

    isforcoupledmaneuvers,varyingpayloadand

    w

    inddisturbances.

    7.

    C

    oncludeyourresults.

    Startdate:

    2012-01-1

    6

    Duedate:

    2012-06-1

    1

    Thesispe

    rformedat:

    Departmen

    t

    of

    Engineering

    Cybernetics,

    NTNU

    Supervisor:

    Professor

    Thor

    I.

    Fossen,

    Dept.

    of

    Eng.

    Cybernetics,

    NTNU

  • 8/11/2019 Full Text 22

    4/108

  • 8/11/2019 Full Text 22

    5/108

    Thisthesiswillpresentadesignofaguidanceandcontrolsystemtouseon

    aircra

    fts,primar

    ilyon

    UAVs.

    On

    econtrolmet

    ho

    dfor

    head

    ingcontrolan

    d

    twoforp

    itchan

    daltitu

    decontro

    lw

    illbe

    invest

    igated

    .Thecontrolmet

    h-

    odsar

    eProportional-I

    ntegral-D

    erivat

    ive

    (PID)an

    dsl

    iding

    mo

    decontrol

    .

    PIDw

    illbetestedon

    bot

    hhead

    ingan

    dp

    itchan

    daltitudecontrol

    ,w

    hile

    sliding

    mo

    dew

    illon

    lybeapplie

    dtop

    itchan

    daltitu

    de.

    Th

    erew

    illbepresente

    dapat

    h-f

    ollow

    ingmet

    ho

    d,

    Lin

    eof

    Sight,

    for

    hea

    din

    ggu

    idancean

    da

    kinemat

    iccontrol

    ler

    foraltitu

    dereference

    .

    Th

    epresente

    dmet

    ho

    dsareim

    plemente

    dinMat

    labSim

    ulin

    kw

    hilethe

    aircra

    ftmo

    deluse

    dcomes

    from

    the

    flightsimu

    lator

    X-Pla

    ne.

    X-P

    lane

    is

    alsouse

    dtov

    isual

    izetheper

    formanceoftheautop

    ilot

    design

    .

    PIDan

    dsl

    idingmo

    decontro

    laretested

    infour

    differentscenar

    ios

    toinvestigatew

    hichcontrol

    lerw

    ho

    hasthe

    bestper

    forma

    nce

    .Afterthe

    simula

    tions,

    itwasobservedtha

    tthe

    PIDhad

    betterperf

    ormancesthan

    sliding

    mo

    decontrol

    .

    iii

  • 8/11/2019 Full Text 22

    6/108

  • 8/11/2019 Full Text 22

    7/108

    Idenn

    eoppgavenvildetblipresentertetstyrings-ogkon

    trol-systemfor

    bru

    ki

    enautop

    ilot

    for

    fly,

    frstog

    fremst

    foru

    bemannedefl

    y.Detv

    ilvre

    enregu

    latormeto

    de

    forpos

    isjonogtometo

    der

    for

    hy

    dereg

    ulering.

    Regu-

    leringsmeto

    denesomv

    ilblipresenterter

    Proporsjonal-I

    nteg

    ral-Der

    ivas

    jons(

    PID)-

    regu

    leringogsl

    idingmo

    de-reguler

    ing.

    PIDv

    ilblibru

    ktfor

    badepos

    isjon

    oghy

    deregu

    lering,menssl

    idingmo

    de

    kunv

    ilblibru

    ktfor

    hy

    de.

    Et

    styringssystemer

    des

    ignet

    forap

    lan

    leggeen

    baneogen

    hy

    dereferanse

    soma

    utop

    ilotenskal

    flge.

    Lineof

    Sighter

    bru

    kt

    forap

    lan

    leggeen

    bane

    idet

    hor

    isonta

    lep

    lan

    ,nor

    d-

    st,

    mensen

    kinematis

    kkontrol

    lergir

    hy

    dere

    feranse

    .

    Meto

    denev

    ilsa

    bliimp

    lemen

    tert

    iMat

    labSimu

    lin

    k,m

    ens

    flymo

    del

    len

    kommer

    fra

    flysimu

    leringsprogra

    mmet

    X-P

    lane.

    X-P

    lane

    blirogsa

    bru

    kt

    for

    av

    isual

    isereopp

    frselentila

    utop

    iloten

    .

    PIDogsl

    idingmo

    dev

    iltils

    lutt

    blitestet

    ifireu

    likes

    cenar

    ioer

    for

    a

    sehvilkenregu

    latorsom

    har

    den

    besteopp

    frselen

    .Ettersimu

    leringene

    kommer

    det

    framat

    PIDharen

    bedreopp

    frselennsl

    idin

    gmo

    de.

    v

  • 8/11/2019 Full Text 22

    8/108

  • 8/11/2019 Full Text 22

    9/108

    Thist

    hesis,

    andtheworkitpresents,

    istheculmination

    ofmymasters

    degree

    atthe

    Departmentof

    Engineering

    Cy

    bernet

    icsof

    the

    Norweg

    ian

    Un

    ivers

    ityof

    Sciencean

    dTec

    hnology

    (NTNU).Iwou

    ldlik

    etothan

    kmy

    superv

    isor

    Thor

    I.Fossenatth

    eDepartmentof

    Engineering

    Cy

    bernet-

    ics

    for

    hispat

    iencean

    dhisgui

    dance

    forth

    isreportand

    inthe

    fiel

    dof

    Gu

    idance

    ,Nav

    igat

    ionan

    dContro

    lSystems.

    Alsoagreatthan

    kyouto

    Kj

    etilHope

    Tu

    ftelan

    dan

    dKare

    Vistnes

    for

    goodd

    iscuss

    ions,

    inputan

    dfeedb

    ack

    .Than

    ksgoesalsotoa

    llthemem

    bers

    ofthe

    unmannedve

    hiclelaborat

    orythe

    lastcoup

    leofmon

    ths

    formak

    ing

    agoo

    dwor

    kingenv

    ironment.F

    inal

    ly,

    Imustthan

    ktoM

    orten

    Wol

    lert

    Nygren

    forcorrectiveread

    ing,a

    ndTor

    leifHagen

    for

    bein

    gsuchagoo

    d

    helpw

    hen

    itcomesto

    flight

    detai

    lsfor

    Cessna.

    vii

  • 8/11/2019 Full Text 22

    10/108

  • 8/11/2019 Full Text 22

    11/108

    ThesisDescription

    i

    Abstract

    ii

    Samm

    endrag

    iv

    Acknowledgments

    vi

    Listo

    fAbbreviations

    xiii

    Listo

    fFigures

    xvii

    Listo

    fTables

    xix

    1Introduction

    1

    1.1

    Mot

    ivat

    ion

    ......

    ...............

    ......

    1

    1.2

    What

    has

    been

    done.

    ...............

    ......

    3

    1.3

    My

    Contr

    ibution

    ...

    ...............

    ......

    6

    2GuidanceTheory

    7

    2.1

    Lineof

    Sight

    Gu

    idance

    for

    Pat

    h-F

    ollow

    ing

    ....

    ......

    7

    2.2

    Kinemat

    icControl

    forA

    ltitu

    de

    Gu

    idance

    ....

    ......

    11

    3Co

    ntrolTheory

    13

    3.1

    PIDControl

    ......

    ...............

    ......

    14

    3.2

    Sliding

    Mo

    de

    Control.

    ...............

    ......

    16

    ix

  • 8/11/2019 Full Text 22

    12/108

    5

    CaseStudy

    35

    5.1

    Without

    disturba

    ncesan

    dpay

    load

    ....

    ..........

    37

    5.1.1

    PID

    .................

    ..........

    38

    5.1.2

    Sliding

    Mo

    de

    ............

    ..........

    41

    5.2

    Withpay

    load

    ................

    ..........

    43

    5.2.1

    PID

    .................

    ..........

    44

    5.2.2

    Sliding

    Mo

    de

    ............

    ..........

    47

    5.3

    WithWind20kn

    ots

    ............

    ..........

    49

    5.3.1

    PID

    .................

    ..........

    50

    5.3.2

    Sliding

    Mo

    de

    ............

    ..........

    53

    5.4

    WithWind40kn

    ots

    ............

    ..........

    56

    5.4.1

    PID

    .................

    ..........

    56

    5.4.2

    Sliding

    Mo

    de

    ............

    ..........

    59

    5.4.3

    Discuss

    ion

    ..............

    ..........

    62

    6

    ConclusionsandFurtherWork

    65

    6.1

    Conclusions

    .................

    ..........

    65

    6.2

    Furt

    her

    Wor

    k................

    ..........

    66

    Bibliography

    69

    A

    DefinitionsandLem

    mas

    71

    B

    AttachmentDescrip

    tionandMatlabCode

    73

    B.1

    Attac

    hment

    Description

    ..........

    ..........

    73

    B.2

    Mat

    labCo

    de

    ................

    ..........

    75

    x

  • 8/11/2019 Full Text 22

    13/108

    xi

  • 8/11/2019 Full Text 22

    14/108

  • 8/11/2019 Full Text 22

    15/108

    AUV

    AutonomousUn

    derwaterVehicle

    GNC

    Gu

    idance

    ,Navig

    ationan

    dControl

    HIL

    Har

    dware

    inthe

    Loop

    LOS

    Lineof

    Sight

    LP

    Low

    Pass

    NED

    North-E

    ast-

    Dow

    n

    NED

    North-E

    ast-

    Dow

    n

    PID

    Proportional-In

    tegral-D

    erivat

    ive

    SMC

    Sliding

    Mo

    deControl

    UAS

    Unmanned

    Aerial

    System

    UAV

    Unmanned

    Aerial

    Veh

    icle

    UDP

    User

    Datagram

    Protoco

    l

    xiii

  • 8/11/2019 Full Text 22

    16/108

  • 8/11/2019 Full Text 22

    17/108

    1.1

    UAVusedinsearchand

    rescueoperations....

    ......

    3

    2.1

    Illustrationof

    LineofS

    ightprincip

    le.......

    ......

    8

    2.2

    Lineofsight

    ......

    ...............

    ......

    9

    2.3

    Nav

    igat

    ion

    inxyp

    lanew

    ithcirc

    leofacceptance.

    ......

    10

    2.4

    Block

    diagramofa

    kine

    mat

    iccontrol

    ler

    .....

    ......

    11

    3.1

    Controlsurfaceson

    Ces

    sna

    172SP

    ........

    ......

    14

    3.2

    Block

    diagramofa

    PID

    control

    ler

    ........

    ......

    15

    3.3

    Sta

    bilityof

    PIDcontro

    ller

    .............

    ......

    16

    3.4

    Slidingmo

    dew

    ithchattering

    dueto

    delay

    incont

    roller

    ...

    20

    3.5

    Phaseportraitofsl

    idingmo

    dew

    ithboun

    dary

    layer

    .....

    21

    4.1

    Block

    diagramofa

    Guid

    ance

    ,Nav

    igat

    ionan

    dControl

    System

    23

    4.2

    Block

    diagramofgu

    idancesystem

    ........

    ......

    24

    4.3

    Block

    diagramof

    imple

    mente

    dkinemat

    iccontro

    lforalti-

    tudegu

    idance

    .....

    ...............

    ......

    26

    4.4

    1stor

    der

    LPfilter

    ..

    ...............

    ......

    27

    4.5

    Block

    diagramofcontrolsystem

    .........

    ......

    28

    4.6

    Block

    diagramofa

    PID

    control

    lerw

    ithanti-w

    indup

    ....

    29

    4.7

    Illustrationof

    Cessna1

    72sp

    inX

    -plane

    ......

    ......

    34

    5.1

    Simu

    lin

    kX

    -Planecomm

    un

    icat

    ion

    .........

    ......

    35

    5.2

    Simu

    lin

    kblock

    diagram

    ,commun

    icat

    ionw

    ithX-P

    lane

    ...

    36

    5.3

    North-E

    astan

    dAltitude

    [m]w

    ithout

    disturbances

    andpay-

    load

    ,measure

    dvs

    desired

    .............

    ......

    38

    5.4

    Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ithout

    disturba

    ncesan

    d

    pay

    load

    ,measure

    dvsdes

    ired

    ...........

    ......

    39

    xv

  • 8/11/2019 Full Text 22

    18/108

    5.10Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ithpay

    loa

    d,

    measure

    dvs

    des

    ired

    ....................

    ..........

    45

    5.11Angleofattackan

    dsi

    desl

    ipangle

    [deg

    ]wit

    hpay

    load

    ,mea-

    sure

    dvs

    des

    ired

    ...............

    ..........

    46

    5.12North-E

    astan

    dAltitu

    de

    [m]w

    ithpay

    load

    ,measure

    dvs

    des

    ired

    ....................

    ..........

    47

    5.13Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ithpay

    loa

    d,

    measure

    dvs

    des

    ired

    ....................

    ..........

    48

    5.14Angleofattackan

    dsi

    desl

    ipangle

    [deg

    ]wit

    hpay

    load

    ,mea-

    sure

    dvs

    des

    ired

    ...............

    ..........

    49

    5.15North-E

    astan

    dA

    ltitu

    de

    [m]w

    ith20ktwin

    d,

    measure

    dvs

    des

    ired

    ....................

    ..........

    50

    5.16Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ith20ktw

    ind

    ,measure

    d

    vs

    des

    ired

    ..................

    ..........

    51

    5.17Angleofattackan

    dsi

    desl

    ipangle

    [deg

    ]w

    ith20ktw

    ind

    ,

    measure

    dvs

    desire

    d

    ............

    ..........

    52

    5.18North-E

    astan

    dA

    ltitu

    de

    [m]w

    ith20ktwin

    d,

    measure

    dvs

    des

    ired

    ....................

    ..........

    53

    5.19Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ith20ktw

    ind

    ,measure

    d

    vs

    des

    ired

    ..................

    ..........

    54

    5.20Angleofattackan

    dsi

    desl

    ipangle

    [deg

    ]w

    ith20ktw

    ind

    ,

    measure

    dvs

    desire

    d

    ............

    ..........

    55

    5.21North-E

    astan

    dA

    ltitu

    de

    [m]w

    ith40ktwin

    d,

    measure

    dvs

    des

    ired

    ....................

    ..........

    56

    5.22Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ith40ktw

    ind

    ,measure

    d

    vs

    des

    ired

    ..................

    ..........

    57

    xv

    i

  • 8/11/2019 Full Text 22

    19/108

    xv

    ii

  • 8/11/2019 Full Text 22

    20/108

  • 8/11/2019 Full Text 22

    21/108

    3.1

    Controlsurfaces....

    ...............

    ......

    13

    4.1

    Cessna

    Spec

    ifications.

    ...............

    ......

    33

    xix

  • 8/11/2019 Full Text 22

    22/108

  • 8/11/2019 Full Text 22

    23/108

    There

    aremany

    differentop

    inio

    nsofw

    hatan

    Unmanned

    Aer

    ialVeh

    icle

    (UAV)is

    .Many

    bel

    ievethattheyareon

    lyuse

    dform

    ilitarypurposes

    ,but

    thisis

    far

    fromthetruth

    .Themost

    important

    defin

    itionof

    anUAVisthat

    itisanaerialve

    hiclew

    ithouta

    pilot

    ,Unmanned

    Aer

    ialVeh

    icleSystem

    Assoc

    iation

    [2012b].Withoutp

    ilotmeansthattheaerial

    vehicledonot

    havea

    piloton

    boardnorap

    ilotonthegroun

    dinacontrolcenter,re

    ferred

    toasa

    Groun

    dControl

    Station.

    Iftheaerialve

    hiclehasap

    ilotongroun

    d

    whocancommun

    icatew

    iththe

    vehicle

    ,itisre

    ferredtoa

    sa

    Unmanned

    Aer

    ial

    System

    (UAS).The

    UAV

    ispreprogrammedan

    dissupposedto

    do

    theop

    erat

    ionan

    dcome

    bac

    kw

    ithout

    human

    interference,

    whilean

    UAS

    isremotelyoperated

    duringthe

    operat

    ion

    .

    In

    the

    dictionary,the

    UAV

    isdefinedas:Apowered,aerialvehicle

    thatdoesnotcarryahumanoperator,usesaerodynamic

    forcestopro-

    videvehiclelift,canflyautonom

    ously,canbeexpendable

    orrecoverable,

    andca

    ncarryalethalornonleth

    alpayload.Ballisticorsem

    iballisticvehi-

    cles,cruisemissiles,andartilleryprojectilesarenotconsideredunmanned

    aerial

    vehicles.

    AlsocalledUAV.

    The

    Free

    Dictionary

    [201

    2].

    1.1

    Motivation

    Origin

    allythe

    UAVswere

    designed

    form

    ilitaryoperat

    ion

    s,but

    inthese

    dayst

    heyare

    des

    igned

    fornonm

    ilitaryoperat

    ionsaswell.

    The

    UAVscan

    beuse

    dto:

    1

  • 8/11/2019 Full Text 22

    24/108

    andmanyothertaskwhereitisnotnecessarytohaveahumanpiloton

    board

    .The

    UAVtechnologyw

    illnot

    beapp

    lied

    in

    commercialav

    iation

    industry

    inthenear

    futureas

    itishar

    dforpeopletotrustamac

    hine

    without

    hav

    inga

    humanp

    ilotobserv

    ing.

    Therearemanyadvan

    tages

    forusing

    UAVs,suc

    has:

    1.Lowcost

    2.Nonee

    dforqual

    ified

    pilots

    3.Savetime

    byusingt

    woormore

    UAVsatthes

    ametime

    4.Nonee

    dtomake

    humanconsi

    derat

    ionsw

    hen

    des

    ign

    ingthe

    UAV

    5.Canoperate

    inareasthatare

    dangerous

    forhumans

    6.Have

    longoperat

    ion

    timew

    ithout

    loos

    ingprec

    ision

    Unmanned

    Aer

    ialVeh

    icle

    System

    Associat

    ion

    [2012a].Figure

    1.1illus-

    trates

    howan

    UAVcanb

    euse

    dforsearch

    ing

    inopenseas

    .By

    hav

    ing

    many

    UAVswor

    kingtogether

    inasearchoperat

    ions

    ,livescan

    besave

    d.

  • 8/11/2019 Full Text 22

    25/108

    Figur

    e1.1:UAVuse

    dinsearch

    andrescueoperat

    ions,

    Jo

    hansen

    [2011]

    Ho

    wever

    ,therearesome

    disadvantagesofusingan

    UA

    Van

    dsomeof

    themare

    listed

    here:

    1.Limitat

    iononpay

    load

    2.Can

    loosecontactw

    ithgro

    un

    dbase,must

    havea

    backupp

    lan

    3.Easytocras

    h

    4.Ap

    ilotcanmon

    itorw

    ider

    areas

    1.2

    Whathasbeend

    one

    UAV

    Inwart

    ime,manygreat

    inventionsaremade.

    Thishol

    ds

    for

    UAVstoo.

    Duringthe

    Amer

    ican

    CivilWar

    (1861-1865)the

    firstun

    mannedaerial

    vehiclewastested

    .Itwasa

    balloonthatcarr

    iedexp

    losive

    san

    ddropped

  • 8/11/2019 Full Text 22

    26/108

    morethan900civilianswhileinjuring35

    .000

    .

    Inthe

    1960san

    d70s

    ,the

    Un

    ited

    State

    developed

    UAVsthatwere

    launched

    fromap

    lanean

    dremotelycontrol

    ledby

    operatorsw

    ithinthe

    plane.

    Later

    inthe

    1970s

    and80s,

    Israel

    developed

    smal

    ler

    UAVs.

    They

    cou

    ldtransm

    itlivev

    ideo

    witha

    360-degreev

    iew

    .S

    incetheyweresmal

    l

    theywere

    inexpensivetopro

    ducean

    ddifficu

    lttoshoot

    down

    .

    Althoughthe

    UAVtec

    hnology

    hasgonethrough

    ahuge

    development

    throughoutthe

    20thcentury

    ,itwas

    first

    inthe

    199

    0sthat

    itgot

    its

    big

    breakthroughw

    iththe

    Pre

    datormade

    bythe

    U.S.D

    epartmentof

    Defense

    ,

    How

    Stu

    ffWor

    ks

    [2012].

    The

    first

    UAVswaspro

    grammedto

    fly

    inastra

    ight

    lineoracirc

    leunti

    l

    itranoutof

    fuelan

    dfelldown

    ,muchthesameasd

    rones

    donow

    .Later

    theygotra

    diocommunicationsan

    dcou

    ldremotelyo

    peratethe

    UAV

    ,an

    d

    nowtheyarepreprogramm

    edw

    ithon

    boardcontrola

    ndgu

    idancesystems.

    Thegoal

    istocreate

    UAV

    sthatmake

    dec

    isionsby

    themse

    lves

    ,w

    ithout

    human

    interference

    .

    Autopilot

    Tomakean

    UAVfly

    itnee

    dsanautop

    ilot

    .Thereare

    different

    kindsof

    autop

    ilots,

    fromthesimpleonesuse

    dinsmal

    lprivat

    eboatstomorecom-

    plexsystemsuse

    don

    for

    instancesu

    bmar

    ines

    ,oilt

    ankersan

    daircra

    fts.

    The

    firstattemptatanautop

    ilotwasash

    ipan

    d

    airp

    lanestab

    ilizer

    in

    1914by

    Elmer

    Sperry

    (1860-1930)an

    dhiscompany

    Sperry

    Gyroscope

    Company,

    IEEEGlobal

    History

    Networ

    k[2012]

    .Spe

    rryan

    dNicholas

    Mi-

    norsky

    (1855-1970),w

    hoformu

    late

    dtheProportional-Integral-Derivative

    control

    Bennett

    [1984]

    ,wo

    rkedtogetheran

    dmai

    da

    hugecontr

    ibutionto

  • 8/11/2019 Full Text 22

    27/108

    Theguidancesystemisanimpor

    tantpartofanautopilot.Itdeterminesa

    pat

    ht

    ofollow

    basedoncomman

    dedsignals,suchaswayp

    oints

    ,altitu

    de,

    speed,

    etc.given

    byanoperator

    .

    Gu

    idancemet

    ho

    dsmade

    formar

    inecraftscaneasily

    beapp

    liedto

    aerial

    vehicles,espec

    iallymetho

    ds

    forautonomousunderwaterve

    hicles

    (AUV)sincetheytoooperatein

    6degreesof

    free

    dom

    .Brhaugan

    dPet-

    tersen

    [2005]uses

    Lineof

    Sightmet

    ho

    dforcross-trac

    kcon

    trol

    forun

    der-

    actuat

    edautonomousve

    hiclesonan

    AUVan

    dBre

    ivikan

    dFossen

    [2008]

    proposes

    differentgu

    idance

    laws

    for

    AUVs.

    Theseareal

    lba

    sedonstra

    ight

    linesb

    etweencomman

    dedwayp

    oints

    .Ifthepat

    hdoesnotconsistsof

    stra

    ight

    lines

    ,thepat

    hhasto

    be

    parametrize

    dan

    da

    kinem

    aticcontrol

    ler

    canbe

    app

    lied

    ,Skjetneetal

    .[2003]an

    dFossen

    [2011b].

    Control

    When

    itcomestothecontrolsystem

    inanautop

    ilot

    ,thechoicesaremany.

    Them

    ostuse

    dcontrol

    ler

    inthe

    industry

    isthethreeterm

    ,proportional-

    integral-derivative(PID),contro

    llerw

    hichdates

    bac

    kto1

    890s

    ,w

    iththe

    firstpract

    icalexamp

    lefrom

    1911by

    Elmer

    Sperry

    ,Bennett

    [1984]

    .This

    isalin

    earcontrol

    ler,

    but

    ithasb

    eenuse

    donnon

    linearsystemssuchas

    for

    anUA

    Vquadrotor

    Sal

    ihetal

    .[2

    010]an

    dforan

    UAVfixed

    -wing

    Albaker

    andR

    ahim[2011]

    .Thesearedecoup

    ledcontrol

    lers

    ,givin

    gyou

    different

    contro

    llers

    forspee

    d,

    altitu

    de,p

    itchangle,

    hea

    dingangle,

    etc.

    For

    bas

    ic

    PIDt

    heorysee

    Balchenetal.

    [2003]an

    dPIDcontrol

    fo

    rmar

    inecraft

    Fossen

    [2011b].

    Slidingmo

    decontrol

    isanot

    herw

    idelyuse

    dcontrol

    ler.

    Thisisanon-

    linear

    met

    ho

    dan

    dw

    illbemorero

    bustw

    henuse

    donanon

    linearsystem

    .

  • 8/11/2019 Full Text 22

    28/108

    y

    g

    other

    .The

    firstw

    illbea

    PIDcontrol

    ler,muchused

    inthe

    industry

    ,an

    d

    thesecon

    da

    Sliding

    Mode

    control

    ler.

    The

    lastcontr

    oller

    isnon

    linearan

    d

    morero

    bustthanthe

    PID

    .Slidingmo

    dew

    illon

    lybeuse

    dforp

    itchan

    d

    altitu

    decontrol

    ,w

    hile

    PID

    willta

    kecareofturn

    ingcontrolaswel

    lasp

    itch

    andaltitu

    de.

    Theautopilot

    des

    ign

    issupposedto

    ha

    ndlevary

    ingpay

    load

    andw

    ind

    .Thegu

    idancesystemw

    illmakeapat

    hforthe

    UAVto

    follow

    basedonwaypointsgiven

    .

    Theoutl

    ineofth

    isthes

    isisas

    follows:

    InChapter2t

    hetheoryofgu

    idance

    systemw

    illbepresente

    d.

    Controlsystemtheory

    is

    descr

    ibed

    inChapter

    3,w

    hilethe

    des

    ignoftheoveral

    lgu

    idance

    ,nav

    igatio

    nan

    dcontrolsystem

    isinChapter

    4.Simu

    lati

    onstu

    diesan

    dresu

    lts

    for

    the

    differentcontrol

    strategiesproposed

    ispresente

    dinChapter

    5.Conclusionan

    dsuggestions

    for

    furt

    herwor

    kw

    illbefo

    un

    dinChapter

    6.

  • 8/11/2019 Full Text 22

    29/108

    Toma

    keanautop

    ilot

    ,itisnec

    essaryto

    haveagoo

    dgu

    idancesystem

    .

    There

    aremanywaysofmak

    ing

    agu

    idancesysteman

    ditisadvantageous

    tokno

    ww

    hattheautop

    ilot

    issu

    pposedto

    dow

    henchoosingthemet

    ho

    d.

    Isitsupposedtotrac

    katargetw

    hichismov

    ingorstan

    dingst

    ill,isit

    suppose

    dtotrac

    katra

    jectoryor

    just

    followapre

    definedpat

    h?Track

    ing

    atargetoratra

    jectory

    isoften

    time

    depen

    dent,w

    hilea

    pat

    h-f

    ollow

    ing

    met

    ho

    distime

    indepen

    dent.

    In

    thisthes

    isapat

    h-f

    ollowingmet

    ho

    dforgu

    idance

    isu

    sed

    .Sinceth

    is

    autop

    ilot

    issupposedtocontro

    lanaerialve

    hicle

    ,thegu

    idancesystem

    musth

    andlethree

    dimensions.Itw

    illbe

    decoup

    ledintotwoparts

    ,one

    for

    hor

    izonta

    lmot

    ion

    ,North-E

    ast,a

    ndone

    forvert

    icalmot

    ion

    ,altitu

    de.

    The

    guidan

    cesystemw

    illbemak

    ing

    thepat

    hto

    follow

    inNort

    h-E

    ast-

    Down

    (NED)coor

    dinates

    ,w

    herex

    isN

    orth

    ,y

    isEastan

    dz

    isdown

    ,oraltitu

    de,

    see

    Ap

    pen

    dixA

    .Thealtitu

    deis

    thenegat

    iveof

    down

    .

    2.1

    LineofSightGuidanceforPath-Following

    Lineo

    fSight

    (LOS)isagu

    idanc

    emet

    ho

    dw

    hichcan

    beus

    edforal

    lthree

    scenar

    iospresente

    dabove;targe

    ttrac

    king,tra

    jectorytrac

    kingan

    dpat

    h-

    follow

    ing.

    Themet

    ho

    disclassifiedasathree-pointgu

    idanc

    esc

    heme,since

    itinvo

    lvesare

    ferencepoint,the

    UAVan

    datargetoraset

    pointthe

    UAV

    issupposedtogoto

    .ThenameLineofSightcomes

    from

    theprincip

    le

    ofthe

    met

    ho

    d.

    Itissupposedto

    followthe

    line,

    LOSve

    ctor

    ,fromthe

    7

  • 8/11/2019 Full Text 22

    30/108

    Figure

    2.1:Illustrationof

    Lineof

    Sight

    princip

    le

    Consi

    dertwowaypointspk=

    [xk

    yk]R2and

    pk+1

    =[x

    k+1

    yk+1

    ]

    R2

    andastra

    ight

    line

    betweenthem

    ,Figure

    2.2.Th

    eaim

    istomakethe

    aircra

    ftfollowth

    isstra

    ight

    line,

    bymak

    ingthecross-t

    rackerroreassmal

    l

    aspossi

    ble:

    lim

    t

    e(t)=

    0

    (2.1

    )

    wherethecross-trac

    kerror

    isgiven

    by:

    e(t)=

    [x(t)xk

    ]sin(

    k)+

    [y(t)yk

    ]cos

    (k

    )

    (2.2

    )

    andtheangle,k,

    use

    dto

    rotatethenorth-a

    xisinthepat

    h-fi

    xedre

    ference

    framew

    ithor

    igininp

    n k,in

    Figure

    2.2,iswrittenas:

    k:=atan

    2(y

    k+1yk,x

    k+1xk

    )S:

    =[,

    ]

    (2.3

    )

    Toensurethatthecross-t

    rackerrore(t)

    0forbot

    hcases,enclosure-

    basedor

    lookahea

    d-b

    ased

    steeringgu

    idanceprinciplescan

    beuse

    d.

    Since

    the

    lookahea

    d-b

    asedmeth

    odhassevera

    ladvantagesovertheenclosure-

    basedmet

    ho

    d,

    Fossen

    [2011b],a

    lookahea

    d-b

    aseda

    pproac

    hw

    illbeuse

    d

    inth

    isthes

    is.

  • 8/11/2019 Full Text 22

    31/108

    Figure

    2.2:Line

    ofsight,

    Johansen

    [2011]

    Lookahead-Basedsteering

    Thelo

    okahea

    d-b

    asedapproac

    hisaverysimp

    leapproac

    han

    discom

    bined

    oftwo

    parts:

    d(e)

    =

    p+

    r(e)

    (2.4

    )

    where

    p

    =k

    (2.5

    )

    from

    E

    quat

    ion

    2.3.

    pisca

    lled

    thepath-tangentialangle

    .Thevelocity-

    pathrelativeangle

    r(e)can

    beimp

    lemente

    dasacontrol

    laww

    ithon

    lya

    proport

    ionalaction:

    r

    (e)=

    arctan

    (K

    pe)

    (2.6

    )

    where

    Kp

    (t)

    =1/(t)>

    0an

    d

    isthe

    lookahea

    ddistance.

    Th

    econtrolsystemuses

    head

    ingangleinstea

    dofcourseanglean

    d

    itisnecessarytotransformthecourseangletothe

    hea

    dingangle

    by:

    d=d

    (2.7

    )

    where

    isthesi

    des

    lipanglewh

    ichcan

    becompute

    dby:

    =

    arcs

    in(v U

    )

    (2.8

    )

  • 8/11/2019 Full Text 22

    32/108

    andyiseastfromtheNEDcoordinateframe.When

    theaircraftisinside

    acirc

    lew

    ithra

    diusR

    k+1

    R1

    aroun

    dthepoint[xk+1

    yk+1

    ],

    ascan

    beseen

    inFigure

    2.3,th

    egu

    idancesystemshould

    changetothenext

    waypoint.

    Thepos

    itionof

    theaircra

    fthastosatisfy:

    [xk+1x

    (t)]2+

    [yk+1y(t)

    ]2R

    2 k+1

    (2.1

    0)

    attimettochangethewaypoint,

    Fossen

    [2011b].

    Figure

    2.3showsthec

    ircleofacceptanceprincip

    leinthetwo

    dimen-

    sionalp

    lane.

    Figure

    2.3:Nav

    igat

    ion

    inxyp

    lanew

    ithcircle

    ofacceptance

  • 8/11/2019 Full Text 22

    33/108

    tan

    (w/u)andismeasuredpitchangle.=()iscalledtheflight

    path.Donotm

    ixth

    isw

    iththek

    inLOS

    .

    Transform

    ing

    Equat

    ion

    2.11

    tothe

    des

    iredsignalweget:

    hd=U(

    d

    )

    (2.1

    2)

    Thea

    imistore

    ducethe

    difference

    betweenmeasure

    daltitu

    dean

    ddes

    ired

    altitud

    etozero

    ,h

    dh

    0,by

    feed

    ingthecontrolsystem

    des

    iredp

    itch

    angle.

    Thiscan

    be

    done

    byapp

    lyinga

    P-c

    ontrol

    ler,more

    descr

    ibed

    in

    Chapt

    er3,suchthatthe

    des

    ired

    pitchangledbecomes:

    d=

    1 U(hd

    K

    p(h

    dh))+

    (2.1

    3)

    Kp

    isacontrol

    lerga

    inan

    disgiven

    byK

    p=

    2>

    0w

    her

    eisa

    des

    ign

    param

    eter

    .h

    disgiven

    bythere

    ferencemo

    delw

    hichw

    illbe

    descr

    ibed

    in

    Section

    4.1.Thiskinemat

    iccont

    roller

    isillustrate

    dinablock

    diagram

    in

    Figure

    2.4.

    Figure

    2.4:Block

    diagramofa

    kinemat

    iccontrol

    ler

  • 8/11/2019 Full Text 22

    34/108

  • 8/11/2019 Full Text 22

    35/108

    Co

    ntrolTheo

    ry

    An

    UA

    Visto

    becontrol

    ledinthismasterthes

    is.

    Inthesimu

    lationpro-

    gram

    beinguse

    d,

    X-P

    lane,thereareno

    UAVflightmode

    l.Instea

    d,

    the

    contro

    lsystem

    des

    igned

    inth

    ism

    asterthes

    isisapp

    liedona

    Cessna

    172SP

    .

    ThisC

    essna

    has

    fourcontrolsur

    facesgiven

    inTab

    le3.1an

    dillustrate

    din

    Figure

    3.1.

    Tab

    le3.1:

    Controlsurfaces

    ControlSurface

    Action

    Motor

    Spee

    dforwar

    d

    Surge

    Aileron

    Ban

    kedTurn

    Roll

    Elevator

    Takeo

    ff

    Pitch

    Ru

    dder

    Turn

    ing

    Yaw

    13

  • 8/11/2019 Full Text 22

    36/108

    Figure

    3.1:Controlsurfaceson

    Cessna

    172SP

    3.1

    PIDControl

    Inth

    isthes

    istwo

    different

    controlsystemsw

    illbedes

    igned

    .The

    firstone

    isaregu

    lar

    Proportional-

    Integral-D

    erivat

    ive

    (PID)control

    ler.

    PIDisthe

    mostw

    idelyuse

    dcontrol

    ler

    inthe

    industry

    because

    itiseasyto

    imp

    le-

    mentan

    dmainta

    in.

    Thecontrol

    ler

    islinearan

    dishe

    reapp

    liedtoa

    highly

    non

    linearsystem

    ,but

    itw

    illwor

    knonet

    heless.

    Theaimofa

    PIDcontrol

    ler

    istomaketheerrorof

    thesignal

    ,the

    differ-

    ence

    betweenwante

    dsignalan

    dactualsignal

    ,assmal

    laspossi

    ble

    ,i.e.go

    tozero

    ,bymak

    ingcontro

    lsignalstotheprocess:

    lim

    t

    e=

    lim

    t

    xdx

    0

    (3.1

    )

    Thiscan

    be

    donew

    ith

    four

    differentcontrol

    lers,

    Balchenetal

    .[2003]

    Proportional

    (P)con

    trol

    ler

    Proportional-I

    ntegra

    l(PI)control

    ler

    Proportional-D

    erivative

    (PD)control

    ler

    Proportional-I

    ntegra

    l-Der

    ivat

    ive

    (PID)contro

    ller

  • 8/11/2019 Full Text 22

    37/108

    contro

    ller.isthecontrolsignal,whichwillbesenttotheprocess.

    Th

    eproportionaltermgivesa

    noutputthat

    isproportionaltotheerror.

    Tooh

    ighproportionalga

    inK

    pc

    angiveanunstab

    leprocess

    .

    Th

    eintegralterm

    isproportionalto

    bot

    hthe

    durationo

    ftheerroran

    d

    themagn

    itu

    deof

    it.

    The

    integr

    alterm

    dea

    lsw

    ithstea

    dy

    -stateerror

    by

    acceleratethemovementoftheprocesstowar

    dssetpoint.

    It

    cancontr

    ibute

    toan

    overshoot

    because

    itrespon

    dstoaccumu

    late

    derror

    fromthepast

    which

    can

    beso

    lved

    byad

    dingt

    he

    der

    ivat

    iveterm

    .

    Th

    eder

    ivat

    ivetermslowsdo

    wntherateofchangeofthecontrolsig-

    nalan

    dmakestheovershootsm

    aller.

    Thecom

    binedcontrol

    ler-process

    stab

    ility

    isimprove

    dbythe

    deriv

    ativeterm

    ,but

    itcou

    ldmaketheprocess

    unstab

    lebecause

    itissensitivetonoise

    intheerrorsignal

    ,Wikiped

    ia-

    PID[2

    012]

    .

    Figure

    3.2showsa

    block

    diagramofaregu

    lar

    PIDcontrol

    ler.

    F

    igure

    3.2:Block

    diagramofa

    PIDcontrol

    ler,

    Johansen

    [2011]

  • 8/11/2019 Full Text 22

    38/108

    Figure

    3.3:Stabi

    lityof

    PIDcontrol

    ler,

    Johansen

    [2011]

    3.2

    SlidingModeControl

    Sliding

    Mo

    de

    Control

    (SM

    C)isaro

    bust-n

    onlinearcontrol

    ler,muchuse

    d

    onmar

    ineve

    hicles,

    Fossen

    [2011b].Sincemar

    ineveh

    iclesan

    daerialve

    hi-

  • 8/11/2019 Full Text 22

    39/108

    +

    +(,)

    (3)

    where

    xisse

    lectedstates

    from

    =N

    E

    D

    TR6

    and=

    UV

    W

    P

    Q

    RTR6,

    depen

    dentofw

    hatshould

    becontrol

    led

    .

    uisth

    econtrolsignal;motor

    ,ru

    dder

    ,ai

    leronorelevator,

    M,

    R,

    Aor

    Eres

    pective

    ly.

    f(x,t

    )isanon

    linear

    funct

    ion

    descr

    ibing

    the

    dev

    iation

    from

    linearity

    intermsof

    distur

    bancesan

    dunmo

    deled

    dynam

    ics,

    Fossen

    [2011b

    ].

    Letx=

    Q

    ZT

    ,w

    here

    Z=D

    ,an

    du=E

    ,A

    andBmatrix

    becom

    es:

    A=

    a11

    a12

    0

    10

    0

    0U

    0

    0 ,

    B=

    b10 0

    (3.5

    )

    forpit

    chan

    daltitu

    decontrol

    .T

    he

    feed

    bac

    kcontrol

    lawis

    writtenas:

    u=

    kx+u0

    (3.6

    )

    where

    kR3

    isfeed

    bac

    kga

    invector

    ,compute

    dbypolep

    lacement.

    By

    substituting

    Equat

    ion

    3.6intoE

    quat

    ion

    3.4weget:

    x=Ax+B(

    kx+u0

    )+f(x,t

    )

    =(ABk

    )x+Bu0

    +f(x,t

    )

    =A

    cx+Bu

    0+f(x,t

    )

    (3.7

    )

    To

    fin

    dagoo

    dcontrol

    law,d

    efinethesl

    idingsurface

    s=hx

    (3.8

    )

  • 8/11/2019 Full Text 22

    40/108

    AssuminghB=0,ch

    oosethenonlinearcontrollawas:

    u0

    =hB1[hxd

    hf(x,t

    )sgn

    (s)]>

    0

    (3.1

    0)

    wheref(x,t

    )istheestimateoff(x,t

    )an

    dsgn

    (s)isthesignum

    funct

    ion:

    s

    gn(s)

    = 1,

    s>

    0

    0,

    s=

    0

    1,s||h||||f(x,t

    )||

    (3.2

    0)

    where

    ||X||detonatesthenormofX

    .

  • 8/11/2019 Full Text 22

    42/108

    whenthetrajectoryreach

    estheslidingsurface,s=

    0,ina,itbeginsto

    dri

    ftaway

    dueto

    delay

    be

    tweenthetimesigntossw

    itchesan

    dthecon-

    trol

    lersw

    itches

    .Whenthe

    control

    lersw

    itches

    ,thetr

    ajectoryreversesan

    d

    themot

    ionw

    illbetowardsthesl

    idingsurfaceagain

    .

    Figure

    3.4:Slidingmo

    dew

    ithchattering

    dueto

    delay

    incontrol

    ler,

    Khal

    il

    [2002] C

    hatteringcan

    leadto

    lowcontrolaccuracy

    ,wear

    andtearofactuators

    andinworstcase

    leadto

    unmo

    deled

    high

    -frequent

    dynam

    icsw

    hichcan

    leadtoworseper

    formance

    ofthesysteman

    dunstab

    ility.

  • 8/11/2019 Full Text 22

    43/108

    Figure

    3.5:Phaseportraitofsl

    idingmo

    dew

    ithboun

    dary

    layer,

    Fossen

    [2011b

    ]

    Thea

    imofsl

    idingmo

    decontro

    listomakeacontrol

    lawthatensures

    s

    0

    infin

    itetime,

    Figure

    3.5.

    Th

    isisthe

    bas

    icidea

    beh

    indsl

    idingmo

    de,

    butthereare

    many

    different

    approac

    hesofusingsl

    idingmode

    inautop

    ilot

    des

    ign

    foran

    UAV

    .

  • 8/11/2019 Full Text 22

    44/108

  • 8/11/2019 Full Text 22

    45/108

    De

    signofGN

    CSystem

    Anau

    top

    ilotconsistofa

    Gu

    idance

    ,Nav

    igat

    ionan

    dContr

    ol(GNC)Sys-

    tem

    ,F

    igure

    4.1.Gu

    idancetake

    scareof

    inputtothesys

    tem

    ,inputsas

    waypo

    intsan

    ddes

    iredspee

    d,a

    nddeterm

    inethe

    des

    ired

    pat

    hfromthe

    curren

    tlocationoftheaircra

    fttothe

    des

    iredwaypoint.

    Gu

    idance

    isof-

    ten

    de

    coup

    ledontore

    ferencemo

    delan

    dgu

    idancesystemw

    herere

    ference

    mo

    del

    dea

    lsw

    ithcomman

    dedsignalsan

    dgu

    idance

    determ

    inesthepat

    h.

    Nav

    igat

    ion

    determ

    inesthe

    locat

    ionan

    daltitu

    deoftheaircraftatagiven

    time.

    Thecontrolsystemensuresthattheaircra

    ftfollowst

    he

    des

    iredpat

    h

    andaltitu

    de

    byman

    ipu

    latingthecontrolsurfaces

    .

    Figure

    4.1:Block

    diagramofaG

    uidance

    ,Nav

    igat

    ionan

    dC

    ontrol

    System

    23

  • 8/11/2019 Full Text 22

    46/108

    Figure

    4.2:B

    lock

    diagramofgu

    idance

    system

    4.1.1

    ReferenceModel

    Signals

    intothegu

    idance

    systemaregiven

    bythere

    ferencemo

    del

    .An

    operator

    determ

    inesw

    herethe

    UAVissupposedto

    goinNorth-E

    astco-

    ordinates

    ,atw

    hichaltitu

    dean

    dspee

    d.

    Theseare

    comman

    dedsignals.

    Sincethegu

    idancesystem

    foraltitu

    de

    doesnottoleratestepsas

    inputs

    ,

    there

    ferencemo

    del

    hastosmoot

    houtth

    issignals.

    Thiscan

    be

    done

    by

    ath

    ird

    -order

    Low

    Pass(L

    P)filterw

    iththestructure:

    xd r(s)

    =h

    lp(s)

    (4.1

    )

    wherexd

    isdes

    iredstate,

    risthere

    ferencesignalg

    iven

    byoperatoran

    d

    hlp

    isthe

    LPfilter

    .Thechoiceof

    filtershou

    ldbebasedonthep

    hysics

    ofthesystem

    itissuppos

    edtowor

    kon

    ,an

    dformar

    inecraftan

    daerial

  • 8/11/2019 Full Text 22

    47/108

    hdh

    c=

    3

    s3+

    (2+1)s2+

    (2+

    1)2s+3

    (4.3

    )

    where

    hc

    isthecomman

    dedaltitu

    dean

    dh

    disthe

    des

    ired

    altitu

    dew

    hich

    willbesenttothegu

    idancesyst

    em.

    Th

    ecomman

    dedaltitu

    deis

    set

    byanoperator

    bysp

    ecify

    ingw

    hich

    altitud

    eisdes

    iredatw

    hattime.

    Un

    derta

    keoff

    ,thealtitud

    eissettozero

    thefir

    stsecon

    ds,

    becausetheU

    AVnee

    dsacertainspeed

    tobeab

    leto

    liftoff.

    Pos

    ition

    inNEDan

    dspee

    disse

    nt

    directlytogu

    idance

    ,pos

    itionasway-

    points

    andspee

    dasavectorwithcomman

    dedspee

    dwante

    dat

    different

    times,

    asforaltitu

    de.

    4.1.2

    GuidanceSystem

    LOSforNorth-East

    TheL

    OSgu

    idancemet

    ho

    dfor

    calcu

    lating

    des

    ired

    hea

    din

    ganglewas

    des

    ign

    edas

    descr

    ibed

    inSection

    2.1.Waypointsarecom

    man

    ded

    byan

    operat

    orinthesimu

    lation

    fileSim

    GNCSystem

    .m,

    Append

    ixB

    .Tosw

    itch

    fromo

    newaypointtothenext,

    acirc

    leofacceptancetest

    isapp

    lied

    ,as

    mentioned

    .To

    fin

    dthera

    diuso

    fthecirc

    le,

    testsimu

    lation

    has

    been

    done

    andth

    eva

    lue

    has

    beensettoR

    k+1

    =1600inEquat

    ion

    2.10

    .

    Whenca

    lcu

    latingthe

    des

    ired

    hea

    dingangle

    fromcourse

    anglegiven

    by

    theLO

    Smet

    ho

    d,

    Equat

    ion

    2.7,

    itisnecessaryto

    havethesi

    des

    lipangle

    .

    Th

    isangle

    isper

    fect

    lymeasure

    dinthesimu

    lationpro

    gram

    X-p

    lane,

    anddoesnotnee

    dto

    beca

    lculate

    d.

  • 8/11/2019 Full Text 22

    48/108

    Figure

    4.3:Block

    diagram

    ofimp

    lemente

    dkinemat

    iccontrol

    foraltitu

    de

    guidance

    LPforSpeed

    Sincethe

    UAVdoesnot

    han

    dlestepsas

    input,th

    ecomman

    dedspee

    d

    nee

    dsto

    besmoot

    hed

    .Th

    iscan

    be

    done

    byapp

    lyinga

    1stor

    der

    lowpass

    filter:

    UdU

    c=

    11+Ts

    (4.4

    )

    whereU

    disthe

    des

    iredspeedgiventothecontrolsystem

    ,Uc

    iscomman

    ded

    spee

    dan

    dTisthetimeconstantgiven

    byT

    =1/

    >0.The

    LPfilter

    is

    imp

    lemente

    das

    inFigure

    4.4.

  • 8/11/2019 Full Text 22

    49/108

    flectionsaresaturatedduetophysicallimitationsontheaircraft.Aileron,

    elevatoran

    dru

    dder

    doesnot

    have

    360degreesofoperat

    ionarea

    .

    4.2

    Navigation

    Thenav

    igat

    ionsystem

    issupposedto

    determ

    inethepos

    itionoftheaircra

    ft

    atag

    iventimet.

    Itcou

    ldinclu

    dea

    filter

    for

    filter

    ingmeasurementnoise

    caused

    by

    for

    instancenoise

    inse

    nsors

    ,waves

    ,currentan

    dw

    ind

    .Oftenan

    observ

    erisuse

    dfor

    filter

    ingan

    ds

    tateestimat

    iontoestimateunmeasura

    ble

    signalsorestimatestates

    ifthesignals

    dropsout.

    In

    thisthes

    ishowever

    ,perfectmeasure

    dsignalsareassumedan

    dan

    observ

    eristhereforenotnecessary

    .

    4.3

    ControlSystem

    Thecontrolsystem

    has

    beenim

    plemente

    dfirstw

    ithPID

    control

    ,then

    withslidingmo

    decontrol

    .ThePIDcontrol

    ler

    has

    beenm

    adetotestthe

    guidan

    cesystem

    ,han

    dleturn

    ing

    operat

    ionsan

    dtocompareper

    formance

    withslidingmo

    decontrol

    .

    4.3.1

    PID

    Thefirstcontrol

    ler,

    PIDcontrol

    ler,was

    imp

    lemente

    das

    de

    scri

    bed

    inSec-

    tion

    3.1.

    Thecontrolsystemw

    asdecoup

    ledinto

    fourparts

    ,forspee

    d

    contro

    l,ro

    llcontrol

    ,p

    itchcontro

    lan

    dyawcontrol

    ,see

    Fig

    ure

    4.5.

  • 8/11/2019 Full Text 22

    50/108

    Figure

    4.5:

    Block

    diagramofcontrolsystem

    WithM,

    A,

    E,

    Ras

    inputtomotor

    ,ai

    leron,

    elevatoran

    dru

    dder

    ,

    respective

    ly,

    andU,,,

    asspee

    d,

    roll

    ,p

    itchan

    d

    yaw

    ,respective

    ly,

    we

    getthecontrolequat

    ions

    inEquat

    ion

    4.5:

    M=K

    p(U

    dU(t))+

    Ki

    t 0(U

    dU())d+Kd

    d dt(

    UdU(t))

    A=K

    p(

    d(t))+K

    i

    t 0(

    d())d+K

    dd d

    t(

    d(t))

    E=K

    p(

    d(t)

    )+K

    i

    t 0(

    d(

    ))d+K

    dd dt

    (d(t)

    )

    R=K

    p(

    d(t))+

    Ki

    t 0(

    d())d+Kd

    d dt(

    d(t))

    (4.5

    )

    Since

    X-p

    lane

    haveperfectmeasurementsthere

    isnonee

    dtota

    kethe

    der

    ivat

    iveofthesignalsan

    dthe

    lastterm

    intheequat

    ionsabove,except

  • 8/11/2019 Full Text 22

    51/108

    givea

    satisfyingbehavior.

    Anti-Windup

    Anti-w

    indup

    isan

    importantpartof

    PIDcontrol

    .Itisapp

    liedto

    PID

    contro

    ltopreventovershootma

    de

    bythe

    integralterm

    in

    Equat

    ion

    4.5.

    Theovers

    hootmayoccurw

    hen

    alargechange

    insetpointoccurs

    ,then

    theintegraltermaccumu

    latesa

    sign

    ificanterror

    duringth

    erise

    ,alsore-

    ferred

    toasw

    indup

    .Itthenovershootsan

    dcontinuesto

    increaseasthe

    accum

    ulate

    derror

    isoffset

    byer

    rors

    intheot

    her

    direction

    .

    To

    preventth

    isasaturation

    isapp

    liedonthecontrolle

    rsoutputan

    d

    thisis

    subtracte

    dfromthe

    integra

    lactionsignal

    .Withanti-w

    indup

    ,Fig

    -

    ure

    3.2ischange

    dto

    Figure

    4.6.

    Figure

    4.6:Block

    diagramofa

    PIDcontrol

    lerw

    ithan

    ti-w

    indup

  • 8/11/2019 Full Text 22

    52/108

    Q Z

    = a11

    a12

    0

    1

    0

    0

    0

    U0

    0 Q

    Z

    + b10 0

    E+

    0 0

    U0

    (4.7

    )

    from

    Equat

    ion

    3.4.

    Theequat

    ion

    for

    Qisfoun

    dbysystem

    identi

    fica

    tionan

    d

    =

    Pcos()

    sin()Q

    (4.8

    )

    h

    =

    U0

    sin

    ()+V

    cos()sin

    ()+Wcos()cos()U0W(4

    .9)

    whereh

    =Z

    ,V

    =P

    =0,

    W

    =U0an

    dassumedsmal

    lva

    luesofan

    d

    .Slidingsurface

    from

    Equat

    ion

    3.8becomes:

    s=h1

    (Q

    Qd

    )+h2

    (d)+h3

    (ZZ

    d)

    (4.1

    0)

    From

    Equat

    ion

    3.6:

    u=kx+u0

    =kx+hB1

    [hxd

    hf(x,t

    )

    sgn

    (s)]

    (4.1

    1)

    thep

    itchan

    daltitu

    decon

    trol

    law

    foru=Eis:

    E=k1Qk3Z+

    1h1b

    1[h1

    Qd+h2

    d+h3

    Zdh3f3sat

    (s)](4

    .12)

    sgn

    (s)isrep

    lace

    dbysat(s)toavoi

    dchattering,as

    exp

    lained

    inSection

    3.2.

  • 8/11/2019 Full Text 22

    53/108

    used,whereunknownparametersinalinearregressionmodelwillbees-

    timate

    d.

    Consi

    derkindepen

    den

    tvariab

    lesx1,x2,,x

    kan

    dnobserva-

    tionsy1,y2,,yn,

    thisgivethe

    mu

    ltiplelinearregression

    mo

    del:

    yi=

    1x1i

    +2x2i

    +

    +

    kxki

    +ii

    =1,2,,n

    and

    n>k

    (4.1

    3)

    which

    willbeestimatedas:

    yi=b1x1i

    +b

    2x2i

    +

    +bkxki

    +ei

    (4.1

    4)

    where

    ian

    dei

    isran

    domerroran

    dresi

    dualassociatedwiththeresponse

    ofyiw

    hilebiistheestimateo

    f

    iestimated

    fromsamp

    lesof

    data

    by

    applying

    leastsquaresystem

    ide

    nti

    ficationmet

    ho

    d.

    Th

    emu

    ltiplelinearregression

    mo

    delcan

    bewrittenona

    morecompact

    forma

    s:

    y

    =X

    +

    (4.1

    5)

    where: y

    = y1y2 . . . yn

    ,

    X=

    x11

    x21

    xk1

    x12

    x22

    xk2

    . . .

    . . .

    . . .

    x1n

    x2n

    xkn

    ,

    = 1

    2 . . .

    k

    (4.1

    6)

    No

    wwewantto

    fin

    dabtha

    tm

    inimizes:

    SSE

    =

    n i=1

    e2i=

    n i=1

    (y1

    b1x1ib2x2ibkxki

    )2

    =(yX

    b)(yXb)

    (4.1

    7)

  • 8/11/2019 Full Text 22

    54/108

    (

    )

    y

    (

    )

    Walpole

    [2002]

    .

    Forp

    itchtheun

    known

    parametersarea11,a12a

    ndb1inEquat

    ion

    4.7

    suchthatthemu

    ltiplelinearregressionmo

    del

    becom

    es:

    Q

    =Q

    E

    a11

    a12b1

    (4.2

    1)

    whereQisp

    itchrate

    ,isp

    itchange

    lan

    dEisdesiredelevator

    deflection

    .

    4.4

    X-Plane

    Tosimu

    latetheautop

    ilot,

    itisnecessaryto

    havea

    goo

    daircra

    ftmo

    del

    .

    Therearemanymet

    hods

    togetanaircra

    ftmo

    del;

    makeone

    inMat

    lab

    Simu

    lin

    kbasedonequationsofmot

    ion

    foranaircraft

    ,useamo

    delalready

    made

    inMat

    labSimu

    link

    or,

    ashas

    been

    done

    inth

    isthes

    is,

    usea

    flight

    simu

    latorprogram

    .X

    -Plane,

    X-P

    lane

    9[2012]

    ,isas

    imu

    latormain

    lyuse

    d

    tosimu

    late

    flights

    byuseo

    fpedalsan

    djoyst

    ickan

    dto

    des

    ignnewaircra

    fts.

    Inth

    isthes

    isthe

    flightsim

    ulator

    has

    beenuse

    dtosimu

    latethe

    des

    igned

    autop

    ilot

    .To

    dothat

    ,the

    data

    fromtheautop

    ilotsystem

    has

    beensentto

    the

    flightsimu

    lator

    byaU

    ser

    Datagram

    Protoco

    l(U

    DP)block

    inMat

    lab

    Simu

    lin

    kan

    dan

    inbu

    iltp

    lugin

    inX

    -Plane.

    Beforethesignalsenters

    X-P

    lane,

    ithas

    beenal

    locate

    dto

    fit

    input

    for

    X-P

    lane.

    X-P

    lane

    isse

    ttingmotor

    ,leftan

    dright

    aileron

    ,elevatoran

    d

    rudder

    ,butthecontrolsy

    stem

    isgivenoutmotor,

    aileron

    ,elevatoran

    d

    rudder

    .Thismeansthatthesignals

    forai

    leron

    deflection

    hasto

    besp

    lit

  • 8/11/2019 Full Text 22

    55/108

    where

    LAmeans

    leftai

    leronan

    dR

    Ameansrightai

    leron

    .Theru

    dder

    is

    weighte

    dlowsince

    itisdes

    irable

    touseai

    lerons

    forturn

    .

    4.4.1

    Cessna172SP

    Asme

    ntioned

    inChapter

    3,aC

    essna,notan

    UAV

    ,hasb

    eenuse

    dasan

    aircra

    ftmo

    del

    inth

    isthes

    is,

    Fig

    ure

    4.7.The

    Cessna

    hast

    hecontrolsur-

    faces:

    aileronon

    bot

    hw

    ingsand

    rudderonthetrai

    lingedgeofthevert

    ical

    stab

    ilizeruse

    dfor

    hea

    ding,eleva

    tor

    bac

    konthetrai

    linge

    dgeofthe

    hor-

    izonta

    lstab

    ilizer

    forp

    itchan

    da

    motor

    for

    forwar

    dthrust

    .Spec

    ifications

    for

    Cessnacan

    beseen

    inTab

    le

    4.1.

    Ta

    ble4.1:Cessna

    Spec

    ificat

    ions,

    Cessna

    Aircraft

    Company

    [2012]

    Length

    8.28m

    Heig

    ht

    2.72m

    Wingspan

    11.0

    0m

    WingA

    rea

    16.2

    0m2

    Weig

    ht

    779kg

    Max

    Takeoff

    Pay

    load

    378kg

    Max

    Takeo

    ffWeight

    1157kg

  • 8/11/2019 Full Text 22

    56/108

    Figure

    4.7:Illu

    strationof

    Cessna

    172spi

    nX

    -plane

  • 8/11/2019 Full Text 22

    57/108

    Ca

    seStudy

    Toma

    kehar

    dware

    inthe

    loop(

    HIL)test

    ingofthegu

    idan

    cean

    dcontrol

    system

    des

    ign

    ,the

    flightsimulat

    orX

    -Plane

    has

    beenused

    .Thegu

    idance

    andco

    ntrolsystem

    isdes

    igned

    in

    Mat

    labSimu

    lin

    kan

    dthe

    controlsignals

    aresentto

    X-P

    lane

    bya

    X-P

    lanep

    lugin

    ,see

    Figure

    5.1for

    X-P

    lanecom-

    municat

    ion

    .From

    X-P

    lane,themeasure

    dsignalsaresenttothegu

    idance

    andco

    ntrolsystem

    inSimu

    link,

    see

    Figure

    5.2.

    Figure

    5.1:Simulin

    kX

    -Planecommun

    icat

    ion

    35

  • 8/11/2019 Full Text 22

    58/108

    Figure

    5.2:Simu

    lin

    kb

    lock

    diagram

    ,commun

    icationw

    ithX

    -Plane

    Thecommun

    icat

    ionw

    ithX

    -Plane

    isdone

    byUDPprotoco

    l.This

    makes

    itpossi

    bletorunt

    hesimu

    lation

    inMat

    labS

    imu

    lin

    kan

    dX

    -Plane

    ontwo

    differentcompute

    rs.

    However

    ,inth

    isthesis

    ital

    lrunonone

    computer.

    Thesimu

    lation

    isdone

    inrealtime.

    Since

    X-P

    lane

    isused

    asflightmo

    delan

    dv

    isua

    lization

    ,thegu

    idance

    andcontrolsystem

    has

    be

    entestedona

    Cessna

    insteadofan

    UAV

    .The

    X-P

    lanesimu

    lator

    hadno

    UAVavai

    lable

    .As

    longasthecontrolsignals

    sent

    fromthecontrolsystemget

    distr

    ibute

    dtotherightcontrolsurfaces

    ontheaircra

    ft,

    thegu

    ida

    ncean

    dcontrolsystem

    des

    igned

    inth

    isthes

    is

    cou

    ldbeapp

    liedtoanyU

    AVsorregu

    laraircra

    fts.

    Thepurposeofth

    isthe

    siswasto

    invest

    igate

    differ

    entmet

    ho

    ds

    for

    UAV

    autop

    ilot

    des

    ign

    .Twod

    ifferentcontrol

    lerswerec

    hosen

    ,Proportional-

    Integral-D

    erivat

    ive

    (PID)

    ,an

    dsl

    idingmo

    decontrol

    .The

    PIDisa

    linear

    control

    lerw

    hichdoesnot

    nee

    dmo

    delparameter

    in

    format

    ion

    ,w

    hilethe

    slidingmo

    de

    isbasedont

    heaircra

    ftmo

    del

    .Sliding

    mo

    de

    ismorero

    bust

    thatcan

    han

    dlemo

    deluncertainties

    .PIDisw

    idely

    use

    dan

    disaper

    fect

    control

    lertocompare

    behav

    iorw

    ithot

    hercontrolm

    etho

    ds.

    An

    UAVmust

    beable

    tooperateun

    dercertaincon

    ditions.

    Itw

    illbe

    exposedtow

    indan

    dithasto

    beab

    letocarryapay

    loadsuchascameras

    ,

    sensorsan

    dcommun

    icatio

    nequ

    ipment.

    Thegu

    idancean

    dcontrolsystem

    willbetested

    inthe

    follow

    ing

    fourscenar

    ios:

    Without

    disturbancesan

    dpay

    load

    ,Section

    5.1

    Withpay

    load

    ,now

    ind

    ,Section

    5.2

  • 8/11/2019 Full Text 22

    59/108

    g

    ,

    g

    10000

    metersnorthan

    d8000meterseast

    .Whenreac

    hing

    analtitu

    deof

    500meters,

    itw

    illremainat

    500

    .

    Whenthesimu

    lationstarts,

    des

    iredaltitu

    de

    issettozerosothatthe

    speedcontrol

    ler

    istheon

    lyonew

    orking.

    Thisisbecausetheaircra

    ftnee

    ds

    acertainspee

    dto

    beab

    letota

    keoffan

    dth

    iscon

    ditionho

    lds

    foral

    lfour

    testca

    ses.

    Theta

    keoffspee

    dissetto

    80knots,w

    hilethe

    cru

    isespee

    dis

    setto

    48knots.

    Foreachcase

    ,sevenp

    lotswi

    llbepresente

    d;pos

    itionin

    North-E

    astvs

    des

    ired

    ,measure

    daltitu

    devsdesire

    d,

    measure

    dro

    ll,

    pitch

    andyawangle

    vs

    desired

    ,angleofattackan

    dthesi

    des

    lipangle.

    Th

    esimu

    lations

    have

    beend

    one

    inMat

    labSimu

    lin

    ka

    ndX

    -plane

    in-

    terface

    has

    beenuse

    dasaircraft

    mo

    delan

    dv

    isual

    izat

    ion,

    Section

    4.4.To

    simula

    te,

    usethem

    inigu

    ide

    inAppen

    dixCforcomputer

    setupan

    drun

    theattachedSimGNCSystem.m

    file

    .Makesureal

    lfiles

    listed

    inAppen

    dix

    Bare

    inthesame

    folder

    .

    5.1

    Withoutdisturbancesandpayload

    Firstthe

    UAVhas

    beensimulate

    dw

    ithper

    fectcon

    dition

    s,w

    ithout

    dis

    -

    turban

    cessuchasw

    indan

    dwith

    outpay

    load

    .

  • 8/11/2019 Full Text 22

    60/108

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    20000

    2000

    4000

    6000

    East

    North

    Measured

    Desired

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4

    1.6

    1.8

    2x104

    0

    100

    200

    300

    400

    500

    Time

    Altitude

    Figure

    5.3:North-E

    astan

    dAltitu

    de

    [m]w

    ithoutd

    isturbancesan

    dpay-

    load

    ,measure

    dvs

    des

    ired

    Ascan

    beseen

    from

    Figure

    5.3,thecontrolsyste

    mfollowsthe

    des

    ired

    pat

    h.

    There

    isasmal

    lover

    shootof

    40meters

    inheadi

    ngatthe

    firstturn

    ing

    maneuver,

    butth

    isisconsi

    deredsmal

    lenoughtono

    tcausepro

    blems.

  • 8/11/2019 Full Text 22

    61/108

    0

    0.5

    1

    1.5

    2

    x104

    200

    20

    Time

    PitchAngle

    0.2

    0.4

    0.6

    0.

    8

    1

    1.2

    1.4

    1.6

    1.8

    2

    x104

    1000

    100

    Time

    YawAngle

    Measured

    Desired

    Figure

    5.4:Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ithout

    dis

    turbancesan

    d

    pay

    loa

    d,

    measure

    dvs

    des

    ired

    From

    Figure

    5.4itisseenth

    atthecontrolsystem

    follo

    wsthe

    des

    ired

    roll,pitchan

    dyawanglessatisfa

    ctor

    ily.

    Allanglesare

    ind

    egrees

    .

  • 8/11/2019 Full Text 22

    62/108

    x104

    Time

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4

    1.6

    1.8

    2

    x104

    2

    1.5

    1

    0.50

    0.51

    Time

    SideslipAngle

    Figure

    5.5:Angleofattac

    kan

    dsi

    desl

    ipangle

    [deg]

    without

    disturbances

    andpay

    load

    ,measure

    dvs

    des

    ired

    Figure

    5.5showsthe

    angleofattackan

    dsidesl

    ipangle.

    Angleof

    attack

    istheangle

    betweenare

    ference

    lineonthea

    ircraftan

    dthevector

    representingthere

    lativem

    otion

    .Sides

    lipangle

    istheangle

    betweenthe

    aircra

    ftcenterl

    inean

    dthe

    vectorofre

    lativew

    indwo

    rkingontheaircra

    ft.

    Theangleofattack

    isvary

    ingastheaircra

    ftascend

    san

    dthenstab

    ilizes

    whentheaircra

    ftisatthecru

    isingaltitu

    de.

    Thes

    ides

    lipangle

    issmal

    l,

    closetozerow

    ithsmal

    lsp

    ikes

    .

  • 8/11/2019 Full Text 22

    63/108

    0

    5000

    10000

    15

    000

    20000

    2000

    4000

    East

    Nort

    Measured

    Desired

    0.5

    1

    1.5

    2

    x104

    0

    100

    200

    300

    400

    500

    Time

    Altitude

    Figure

    5.6:North-E

    astan

    dAltitu

    de

    [m]w

    ithout

    disturbancesan

    dpay-

    load

    ,measure

    dvs

    des

    ired

    In

    Figure

    5.6,thep

    lot

    foraltitu

    de

    istheon

    lyoneof

    in

    terest

    ,because

    thePI

    Dcontrol

    lersta

    kescareof

    hea

    ding.

    Thesl

    idingmod

    econtrol

    ler

    for

    pitchan

    daltitu

    decontrol

    isableto

    follow

    des

    iredaltitudesatisfactori

    ly.

    Atthe

    endofthesimu

    lation

    ,sm

    allosci

    llat

    ionsoccur.

  • 8/11/2019 Full Text 22

    64/108

    0

    0.5

    1

    1.5

    2 x104

    200

    20

    Time

    PitchAngle

    0

    0.5

    1

    1.5

    2

    x104

    1000

    100

    Time

    YawAngle

    Measured

    Desired

    Figure

    5.7:Rol

    l,p

    itcha

    ndyawangle

    [deg

    ]w

    itho

    ut

    disturbancesan

    d

    pay

    load

    ,measure

    dvs

    desire

    d

    Rol

    lan

    dyawangleon

    Figure

    5.7arecontrol

    led

    by

    PIDcontrolan

    d

    havethesame

    behav

    ioras

    inFigure

    5.4.Thep

    itchangle

    hasare

    lative

    ly

    bigerror

    inthe

    beg

    inning

    andattheen

    d.

    Inthem

    iddlethemeasure

    d

    pitch

    followsthe

    des

    iredp

    itch

    .

  • 8/11/2019 Full Text 22

    65/108

    0

    05

    5

    x104

    Time

    0.2

    0.4

    0.6

    0

    .8

    1

    1.2

    1.4

    1.6

    1.8

    2

    x104

    2

    1.5

    1

    0.50

    0.51

    Time

    SideslipAngle

    Figure

    5.8:Angleofattackan

    dsi

    desl

    ipangle

    [deg

    ]w

    ithout

    disturbances

    andpay

    load

    ,measure

    dvs

    des

    ire

    d

    Th

    eangleofattack

    inFigu

    re5.8varies

    inthe

    begin

    ningan

    dthen

    stab

    ilizesw

    hentheaircra

    ftreac

    hes

    itscru

    isingspee

    d,as

    itdidfor

    PID

    contro

    l.Thesi

    des

    lipangle

    isthesameas

    for

    PIDcontrol.

    5.2

    Withpayload

    TheG

    NCsystem

    has

    beenteste

    dw

    ithacertainamountofpay

    load

    .An

    UAVmaycarrysensors

    ,cameras

    ,commun

    icat

    ionequipm

    ent,etc.an

    d

  • 8/11/2019 Full Text 22

    66/108

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    20000

    2000

    4000

    East

    North

    Desired

    0

    0.5

    1

    1.5

    2

    x104

    1000

    100

    200

    300

    400

    500

    600

    Time

    Altitude

    Figure

    5.9:North-E

    astan

    dAltitu

    de

    [m]w

    ithpayloa

    d,

    measure

    dvs

    de-

    sire

    d

    The

    PIDcontrol

    hasnop

    roblems

    follow

    ingthe

    des

    iredpat

    horaltitu

    de

    whenpay

    load

    isad

    ded

    ,as

    can

    beseen

    from

    Figure

    5.9.Theovershootat

    the

    firstturn

    ingmaneuver

    isthesameas

    forsimulat

    ionw

    ithout

    distur-

    bancesan

    dpay

    load

    .

  • 8/11/2019 Full Text 22

    67/108

    0.2

    0.4

    0.6

    0.

    8

    1

    1.2

    1.4

    1.6

    1

    .8

    2

    x104

    200

    20

    Time

    PitchAngle

    0

    0.5

    1

    1.5

    2

    x104

    500

    50

    100

    Time

    YawAngle

    Measured

    Desired

    Figure

    5.10:

    Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ithpay

    load

    ,measure

    dvs

    des

    ired

    From

    Figure

    5.10itcan

    be

    seenthatthep

    itchcontro

    ldonee

    dto

    wor

    kh

    arderw

    hensimu

    latingwithpay

    loadthanw

    ithout.T

    hisy

    ieldson

    ly

    during

    takeoff

    ,an

    dnotafterthe

    aircra

    fthasascen

    dedtoc

    ruisealtitu

    de.

  • 8/11/2019 Full Text 22

    68/108

    0

    0.5

    1

    1.5

    2

    x104

    Time

    0

    0.5

    1

    1.5

    2

    x104

    2

    1.5

    1

    0.50

    0.51

    Time

    SideslipAngle

    Figure

    5.11:

    Angleofatta

    ckan

    dsi

    desl

    ipangle

    [deg

    ]w

    ithpay

    load

    ,mea-

    sure

    dvs

    des

    ired

    Therearemoresp

    ikes

    inbot

    hangleofattacka

    ndthesi

    des

    lipangle

    whenpay

    load

    isad

    dedto

    theaircra

    ft.

    Duringta

    keo

    ffan

    dascen

    ding,the

    angleofattackvariesmore

    andhavegreatersp

    ikesth

    antheprev

    iouscase

    .

  • 8/11/2019 Full Text 22

    69/108

    2000

    4000

    600

    0

    8000

    10000

    12000

    14000

    16000

    20000

    East

    0.5

    1

    1.5

    2

    x104

    0

    100

    200

    300

    400

    500

    Time

    Altitude

    Figure

    5.12:

    North-E

    astan

    dAltitu

    de

    [m]w

    ithpay

    load

    ,m

    easure

    dvs

    de-

    sire

    d

    When

    app

    liedpay

    load

    ,theslid

    ingmo

    decontrolosci

    llatesa

    bitmore,

    Figure

    5.12

    .Itfollowsthe

    desiredaltitu

    deas

    itascen

    ds,butshowsome

    oscillationsw

    henreac

    hingthea

    ltitu

    deof

    500meters.

    Whenturn

    ing,at

    time1

    000to

    1500

    ,itisab

    leto

    ke

    epthealtitu

    dew

    ithoutosci

    llat

    ions.

    This

    isthesecon

    dturn

    ,w

    hilethe

    first

    isduringthecl

    imbing.

    After

    fin

    ishturn

    ,

    ithas

    someosci

    llat

    ionsthat

    decreases

    .

  • 8/11/2019 Full Text 22

    70/108

    0

    0.5

    1

    1.5

    2

    x104

    200

    20

    Time

    PitchAngle

    0.2

    0.4

    0

    .6

    0.8

    1

    1.2

    1.4

    1

    .6

    1.8

    2x104

    1000

    100

    Time

    YawAngle

    Measured

    Desired

    Figure

    5.13:

    Rol

    l,p

    itcha

    ndyawangle

    [deg

    ]w

    ithp

    ayload

    ,measure

    dvs

    des

    ired A

    gain

    ,thesl

    idingmod

    econtrol

    isnotab

    leto

    followthe

    des

    iredp

    itch

    angle

    inthe

    beg

    inn

    ingan

    dattheen

    d,

    Figure

    5.13

    .Rol

    lan

    dyawangle

    follow

    des

    iredangle.

  • 8/11/2019 Full Text 22

    71/108

    0

    05

    5

    x104

    Time

    0.5

    1

    1.5

    2

    x104

    2

    1.5

    1

    0.50

    0.51

    Time

    SideslipAngle

    Figure

    5.14:

    Angleofattackand

    sidesl

    ipangle

    [deg

    ]w

    ith

    pay

    load

    ,mea-

    suredvs

    des

    ired

    Th

    eangleofattackan

    dthe

    sides

    lipangle,

    Figure

    5.14

    ,arethesame

    asfor

    simu

    lationw

    ithoutpay

    loa

    dan

    dw

    inddisturbances.

    5.3

    WithWind20knots

    Un

    der

    takeoff

    itispre

    fera

    bleto

    have

    direct

    hea

    dw

    ind.Therunway

    is

    directed

    West-

    Eastsotheappl

    iedw

    indiscom

    ing

    from

    theeast

    .The

    applie

    dw

    indissetafter

    discuss

    ionw

    ithanexper

    ience

    dCessnap

    ilot

    .

  • 8/11/2019 Full Text 22

    72/108

    2000

    4000

    6000

    8000

    10000

    120

    00

    14000

    0

    5000

    10000

    East

    North

    Measured

    Desired

    0

    2000

    400

    0

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    0

    100

    200

    300

    400

    500

    Time

    Altitude

    Figure

    5.15:

    North-E

    astan

    dAltitu

    de

    [m]w

    ith20k

    tw

    ind

    ,measure

    dvs

    des

    ired

  • 8/11/2019 Full Text 22

    73/108

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    200

    Time

    RollAngle

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    160

    00

    200

    20

    Time

    PitchAngle

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    1000

    100

    Time

    YawAngle

    Measured

    Desired

    Figure

    5.16:

    Rol

    l,p

    itchan

    dyaw

    angle

    [deg

    ]w

    ith20ktwin

    d,

    measure

    dvs

    des

    ired

  • 8/11/2019 Full Text 22

    74/108

    2000

    4000

    6000

    8000

    10000

    12000

    14

    000

    16000

    2024

    Time

    AngleofAttack

    2000

    4000

    6000

    8000

    10000

    12000

    1400

    0

    16000

    2

    1.5

    1

    0.50

    0.51

    Time

    SideslipAngle

    Figure

    5.17:

    Angleofattackan

    dsi

    desl

    ipangle[

    deg

    ]w

    ith20ktw

    ind

    ,

    measure

    dvs

    des

    ired

    Therearea

    lotofsp

    ikes

    inthemeasure

    dangleofa

    ttac

    kan

    dthesi

    des

    lip

    angle

    inFigure

    5.17

    .The

    angleofattackstartspo

    sitive

    duringta

    keoff

    ,

  • 8/11/2019 Full Text 22

    75/108

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    0

    5000

    10000

    East

    North

    Measur

    ed

    Desired

    0

    0.5

    1

    1.5

    2

    x104

    0

    100

    200

    300

    400

    500

    Time

    Altitude

    Figure

    5.18:

    North-E

    astan

    dAltitu

    de

    [m]w

    ith20ktwin

    d,

    measure

    dvs

    des

    ired

  • 8/11/2019 Full Text 22

    76/108

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4

    1.6

    1.8

    2x104

    200

    Time

    RollAngle

    0

    0.5

    1

    1.5

    2 x104

    60

    40

    200

    20

    Time

    PitchAngle

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4

    1.6

    1.8

    2x104

    1000

    100

    Time

    YawAngle

    Measured

    Desired

    Figure

    5.19:

    Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ith20ktw

    ind

    ,measure

    dvs

    des

    ired T

    heon

    lyp

    lotof

    interest

    isherethep

    itchanglein

    Figure

    5.19

    .Itnever

  • 8/11/2019 Full Text 22

    77/108

    0.2

    0.4

    0.6

    0

    .8

    1

    1.2

    1.4

    1.6

    1.8

    x104

    202

    Time

    AngleofAttac

    0.2

    0.4

    0.6

    0.8

    1

    1.2

    1.4

    1.6

    1.8

    2

    x104

    2

    1.5

    1

    0.50

    0.51

    Time

    SideslipAngle

    Figure

    5.20:

    Angleofattackan

    dsi

    desl

    ipangle

    [deg

    ]w

    ith20ktw

    ind

    ,

    measu

    redvs

    des

    ired

    An

    gleofattackan

    dthesi

    deslipangle,

    inFigure

    5.20

    ,be

    havesthesame

    asfor

    pure

    PIDcontrol

    inFigure

    5.17

    .On

    lychange

    isaso

    mew

    hat

    bigger

    angleofattack

    inthe

    beg

    inn

    ing,

    atthestartofta

    keoff

    .T

    he

    hugesp

    ikes

  • 8/11/2019 Full Text 22

    78/108

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    0

    5000

    10000

    East

    North

    Measured

    Desired

    2000

    4000

    6000

    8000

    10000

    12000

    1400

    0

    16000

    18000

    0

    100

    200

    300

    400

    500

    Time

    Altitude

    Figure

    5.21:

    North-E

    astan

    dAltitu

    de

    [m]w

    ith40k

    tw

    ind

    ,measure

    dvs

    des

    ired

  • 8/11/2019 Full Text 22

    79/108

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    200

    20

    Time

    RollAngle

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    200

    20

    Time

    PitchAngle

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    1000

    100

    Time

    YawAngle

    Measured

    Desired

    Figure

    5.22:

    Rol

    l,p

    itchan

    dyaw

    angle

    [deg

    ]w

    ith40ktwin

    d,

    measure

    dvs

    des

    ired

  • 8/11/2019 Full Text 22

    80/108

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    2024

    Time

    AngleofAttack

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    2

    1.5

    1

    0.50

    0.51

    Time

    SideslipAngle

    Figure

    5.23:

    Angleofattackan

    dsi

    desl

    ipangle[

    deg

    ]w

    ith40ktw

    ind

    ,

    measure

    dvs

    des

    ired

    Theangleofattack

    is

    slightly

    biggeratthesta

    rtoftheta

    keoffan

    d

    decreasesastheaircra

    fthits

    itcru

    isingspee

    d,

    Figure

    5.23

    ,w

    henthew

    ind

  • 8/11/2019 Full Text 22

    81/108

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    20000

    2000

    4000

    6000

    8000

    East

    North

    Measur

    ed

    Desired

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    1000

    100

    200

    300

    400

    500

    Time

    Altitude

    Figure

    5.24:

    North-E

    astan

    dAltitu

    de

    [m]w

    ith40ktwin

    d,

    measure

    dvs

    des

    ired

    Thes

    lidingmo

    decontrol

    ler

    for

    pitchan

    daltitu

    de

    doesn

    otseemtoget

    affecte

    dbythe

    increase

    dw

    ind.From

    Figure

    5.24itcanbeseenthatthe

  • 8/11/2019 Full Text 22

    82/108

    0

    0.5

    1

    1.5

    2

    x104

    20

    Time

    2000

    4000

    6000

    8000

    10000

    12000

    140001

    6000

    18000

    60

    40

    200

    20

    Time

    PitchAngle

    0

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    1000

    100

    Time

    YawAngle

    Measured

    Desired

    Figure

    5.25:

    Rol

    l,p

    itchan

    dyawangle

    [deg

    ]w

    ith40ktw

    ind

    ,measure

    dvs

    des

    ired W

    ith40knotsw

    inddis

    turbancethep

    itchcontroller

    ,Figure

    5.25

    ,seems

    tobehave

    betterthan

    itd

    idw

    ith20knotsw

    ind

    ,Figure

    5.19

    .Itst

    illhas

    anhugeerror

    inthe

    beginn

    ing

    duetothep

    hysica

    llim

    itat

    ions,

    butthe

    error

    issmal

    ler

    intheend

    .However

    ,itisst

    illnotsu

    ccessfu

    l.

  • 8/11/2019 Full Text 22

    83/108

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    2

    Time

    A

    2000

    4000

    6000

    8000

    10000

    12000

    14000

    16000

    18000

    2

    1.5

    1

    0.50

    0.51

    Time

    SideslipAngle

    Figure

    5.26:

    Angleofattackan

    dsi

    desl

    ipangle

    [deg

    ]w

    ith40ktw

    ind

    ,

    measu

    redvs

    des

    ired

    Forsl

    idingmo

    decontrol

    ,the

    angleofattack

    ,Figure

    5.26

    ,issmoot

    her

    during

    takeoffthan

    itisfor

    PID

    control

    ,Figure

    5.23

    .It

    startspos

    itive

    andco

    nvergestoaconstantnega

    tiveangleastheaircra

    ftr

    eachescru

    ising

    altitud

    e,w

    ithon

    lysmal

    lsp

    ikesoccurr

    ing.

    Th

    esi

    des

    lipangle

    isthesam

    eas

    for

    PIDcontrol

    .Sides

    lipanglecor-

    respon

    dsto

    hea

    dingcontrolwhich

    ispre

    formed

    by

    PIDco

    ntrol

    .

  • 8/11/2019 Full Text 22

    84/108

    wind,hasanerrorof100and200meterswhen

    flyingstraightnorth

    atw

    indspee

    dof

    20

    knotsan

    d40knots

    fromt

    heeast

    ,respective

    ly.

    Inhea

    ding,

    PIDcontrol

    ,there

    isanovershootof

    40meters

    from

    des

    iredpat

    hw

    hentu

    rning

    insimu

    lationsw

    itho

    utw

    inddisturbances

    andpay

    loadan

    dwit

    hpay

    load

    .Theovershoot

    increasestoapprox

    i-

    mately

    400metersw

    ith20knotsw

    indspee

    dan

    d700w

    ith40knots

    windspee

    d.

    PIDhassmal

    loscillat

    ions

    inaltitu

    dew

    henexposedtow

    ind

    .

    PIDhassl

    ightly

    largerosci

    llat

    ionsw

    henexpose

    dto

    40knotsw

    ind

    spee

    dthanw

    henexposedto

    20knotsw

    indspee

    d.

    Slidingmo

    de

    hasosci

    llat

    ions

    inaltitu

    de

    inall

    fourcases.

    Slidingmo

    de

    hassam

    eper

    formancesw

    henexposedto

    20knotsw

    ind

    spee

    dasw

    henexpose

    dto

    40knotsw

    indspeed

    .

    PIDfollows

    des

    iredro

    ll,

    pitchan

    dyawanglen

    icely

    inal

    lfourcases,

    worstper

    formanceinyaww

    henexposedtowin

    d.

    Slidingmo

    dealmostneverreac

    hes

    des

    iredpitc

    h.

    Des

    iredp

    itch

    is

    somet

    imesgreatert

    hanthep

    hysica

    llim

    itat

    ion

    sontheaircra

    ft.

    Angleofattack

    issm

    oot

    her

    forsl

    idingmo

    dethan

    for

    PIDcontrol

    .

    Sides

    lipangle

    isclos

    etozero

    duringal

    lfourcases

    ,but

    ithassome

    spikes

    .

  • 8/11/2019 Full Text 22

    85/108

  • 8/11/2019 Full Text 22

    86/108

  • 8/11/2019 Full Text 22

    87/108

    Work

    6.1

    Conclusions

    Inthis

    thes

    isa

    Gu

    idance

    ,Nav

    iga

    tionan

    dControl

    (GNC)sy

    stem

    foruse

    in

    autop

    ilot

    des

    ign

    foran

    UAVhas

    been

    des

    ignedan

    dtested

    witha

    Cessna

    172SP

    flightmo

    del

    inX

    -Planefl

    ightsimu

    lator.

    The

    flightmo

    del

    has

    been

    un

    known

    ,a

    black

    box

    ,duringthesimu

    lati