DEVELOPMENT OF PROCEDURES FOR FLIGHT...

DEVELOPMENT OF PROCEDURES FOR FLIGHT TESTING UAVS USING THE

ARDUPILOT SYSTEM

A dissertation submitted to The University of Manchester for the degree

of Master of Science in the Faculty of Engineering and Physical Sciences

2011

KHRISTOPHER KABBABE

SCHOOL OF MECHANICAL, AEROSPACE AND CIVIL ENGINEERING

DEVELOPMENT OF PROCEDURES FOR FLIGHT TESTING UAVS USING THE ARDUPILOT SYSTEM [2011]

2

Contents

Table of Figures ...................................................................................................................6

Abstract ..............................................................................................................................8

Declaration.........................................................................................................................9

Intellectual Property Statement.........................................................................................10

Acknowledgments.............................................................................................................12

Nomenclature ...................................................................................................................13

Acronyms..........................................................................................................................14

Chapter 1: Introduction....................................................................................................15

1.1. Objectives...........................................................................................................16

1.2. The report...........................................................................................................16

Chapter 2: Literature Survey ............................................................................................18

2.1. Scope..................................................................................................................18

2.2. UAV....................................................................................................................18

2.3. Autopilots...........................................................................................................20

2.4. Simulation of Autopilots.......................................................................................20

2.5. PID Based Autopilots ...........................................................................................22

2.6. Fuzzy Based Autopilots........................................................................................23

2.7. Robust H Based Autopilot...................................................................................24

2.8. Available Autopilots.............................................................................................25

2.9. Research and Teaching Value of the Autopilot.......................................................31

2.10. Conclusion ..........................................................................................................37

Chapter 3: Background Theory .........................................................................................38

3.1. Inertial Measurement Unit (IMU) .........................................................................38

3.2. Navigation Algorithm ..........................................................................................42

Chapter 4: MATLAB Flight Simulation Software ................................................................45

4.1. Flight Simulator Version 10.7 ...............................................................................45

4.2. Flight simulator Autopilot Modes .........................................................................47


3

4.3. Flight Simulator m-files........................................................................................48

4.4. Autopilot Simulink blocks and functions................................................................49

Chapter 5: ArduPilot: Apparatus and Setup ......................................................................53

5.1. Components of the ArduPilot System ....................................................................53

5.2. Experimental Setup and Aircraft Integration .........................................................57

5.3. ArduPilot Source Code Summary..........................................................................60

5.4. ArduIMU Source Code Summary...........................................................................63

Chapter 6: Autopilot Testing Procedures ..........................................................................65

6.1. Simulation ..........................................................................................................65

6.2. Aircraft Ground Testing........................................................................................68

6.3. PRE-Flight Testing................................................................................................72

6.4. Flight Test Procedures..........................................................................................75

Chapter 7: Analysis of Flight Tests ....................................................................................78

7.1. Stabilize Mode Flight Tests...................................................................................78

7.2. FBW A Mode.......................................................................................................82

7.3. Failsafe (RTL).......................................................................................................83

7.4. FBW B Mode .......................................................................................................84

7.5. Autonomous Mode..............................................................................................84

7.6. Differences between Simulation and Physical FLIGHT Testing.................................85

Chapter 8: Proposed Flight Test Procedures for the 4th Year Aerospace Design Module ....87

8.1. Preparing The ArduPIlot.......................................................................................87

8.2. Requirements for installing the ardupilot ..............................................................92

8.3. Before taking-off.................................................................................................92

8.4. Testing Manual Mode..........................................................................................92

8.5. Testing in Stabilize Mode .....................................................................................93

8.6. TEsting in FBW Mode...........................................................................................93

8.7. Testing in Autonomous Mode...............................................................................95

8.8. POST PROCESSING DATA......................................................................................96

Chapter 9: Recommendations For Future Work ................................................................97

9.1. About the MFS ....................................................................................................97

9.2. About the ardupilot .............................................................................................98

9.3. About flight testing..............................................................................................99


4

9.4. About Post processing data..................................................................................99

Chapter 10: Challenges and Limitations........................................................................100

10.1. Arduino Programming .......................................................................................100

10.2. Flying the Aircraft..............................................................................................100

10.3. Weather Limitations..........................................................................................101

Chapter 11: Conclusion................................................................................................102

11.1. The Project........................................................................................................102

11.2. Technology assessment .....................................................................................102

11.3. The ArduPilot System.........................................................................................102

11.4. Simulation ........................................................................................................102

11.5. Testing..............................................................................................................103

11.6. Instruction Guide...............................................................................................103

Bibliography....................................................................................................................104

Chapter 12: Appendices...............................................................................................108

Appendix A: Modified MATLAB Flight Simulator User Guide [36]......................................109

Appendix B: Proposed MATLAB Autopilot User Guide ......................................................119

Appendix C: MA TLAB Fli ght Simul ator Source Code ......................................................126

C.1. MASTER.M ............................................................................................................126

C.2. AIRFRAME_C ONFIGURATION.M ............................................................................128

C.3. CLOUDSFLY_AIR FRAM E.M ..........................................................................................130

C.4. Airframe_CALC.M ...................................................................................................132

C.5. AutoPilotConfig.M..................................................................................................134

C.6. ESC _Settings.M......................................................................................................135

C.7. Waypoints_CONFIG.M .............................................................................................135

C.8. STARTUP_Config.M .................................................................................................136

C.9. SimMechanics_COORDINATE_SYSTEM _SETUP.M..........................................................138

C.10. MODE_settings.M.................................................................................................140

C.11. PID_SETTING S.M ...................................................................................................141

C.12. PID_TUNER.M .......................................................................................................142

C.13. AP_PID_TUNER ....................................................................................................144

C.14. AP_Dataextract...................................................................................................145

C.15. NAV FUNC TION .....................................................................................................148


5

C.16. DCM FUNCTION .....................................................................................................149

Appendix D: ArduPilot Modi fied Source Code ...............................................................150

D.1. Modifications to ArduPilot_2_7 ..............................................................................150

D.2. AP _CloudsFly_CONFIG.H.......................................................................................150

D.3. AP _CloudsFly_PID_SETTINGS ................................................................................155

Appendix E: ArduIMU Modified Source Code ................................................................158

E.1. ARDUIMU.H .........................................................................................................158

Appendix F: Wind Tunnel Te sting Re sults......................................................................160

F.1. COMP ARATIVE RESUL TS .......................................................................................160

F.2. Results with installed static tube.............................................................................161

Appendix G: ArduPilot Housing CA D Drawing................................................................162

G.1. MAIN ARDUPIL OT HOUSING ......................................................................................163

G.2. ARDUIMU AND GPS MODUL E ...................................................................................164

G.3. PITOT ADAP TER .......................................................................................................165

G.4. EX TEN SION MOUNT...................................................................................................166

G.5. CAMERA EXTENSION MOUN T .....................................................................................167

Appendix H: Proposed Pre-Flight Check List ..................................................................168

Appendix I : Proposed Flight Test Log.............................................................................171

Appendix J: M apped GE KML Data For Stabili ze Test .....................................................174

Appendix K: Content of Attached CD ............................................................................175

WORD COUNT 22,176


6

Table of Figures

Figure 2.1 Ground Control Station examples.......................................................................19

Figure 2.2 Test Platform Block Diagram (As proposed in [11] ) ............................................21

Figure 2.3 Proposed Test Platform Block Diagram...............................................................22

Figure 2.4 PID Control examples [4], [11]............................................................................23

Figure 2.5 Architecture of the self-adaptive fuzzy PID controller [14]...................................24

Figure 2.6 Comparison curves of step response [14]............................................................24

Figure 2.7 Configuration and Layout of Components [17]....................................................26

Figure 2.8 Proposed Block Diagram for the LEGO NXT Based Autopilot................................27

Figure 1.9 Paparazzi Components Layout [7] ......................................................................28

Figure 2.10 Paparazzi Autopilot Global Overview [18] ........................................................28

Figure 2.11 Paparazzi Board with GPS ................................................................................28

Figure 2.12 ArduPilot Component Assembled .....................................................................30

Figure 2.13 ArduPilot Schematics [19] ................................................................................30

Figure 2.14: System Architecture and Autopilot Block Diagram [20]....................................32

Figure 2.15 Hardware System Configuration Schematic [8] .................................................34

Figure 2.16 Block Diagram and Assembled Autopilot Hardware [8].....................................35

Figure 3.1 MEMS 2-DOF Accelerometer ..............................................................................39

Figure 3.2 Difference between 2 and 3 DOF Accelerometer Based IMUs [31], [32]................39

Figure 3.3 MEMS Gyroscope...............................................................................................40

Figure 3.4 Geometry Definition for Waypoint Following......................................................42

Figure 3.5 Flight Path of UAV with Wind Speed - Distance vs. Line Transition ......................43

Figure 3.6 Waypoint Smoothening Function Effect ..............................................................44

Figure 4.1 Screenshot of Flight Simulator V10.7 Graphic Interface.......................................45

Figure 4.2 SIMULINK Autopilot Blocks ................................................................................50

Figure 4.3 RC Aircraft System Analogy................................................................................50

Figure 4.4 Receiver Simulink Block......................................................................................51

Figure 4.5 Horizontal and Vertical Position of the Aircraft Over Time ..................................52

Figure 5.1 ArduPilot Complete Setup ..................................................................................53

Figure 5.2 ArduPilot Board.................................................................................................54

Figure 5.3 ArduShield V2 Flat Board ...................................................................................54

Figure 5.4 ArduIMU V2 Flat Board......................................................................................55

Figure 5.5 U-Blox 5 GPS......................................................................................................55


7

Figure 5.6 XBee Pro S1 System ...........................................................................................56

Figure 5.7 Initial Installation of the ArduPilot .....................................................................58

Figure 5.8 Wiring Problem inside the Airframe ...................................................................58

Figure 5.9 Proposed ArduPilot System Housing V1..............................................................59

Figure 5.10 Proposed ArduPilot System Housing V2 ............................................................59

Figure 5.11 Complete System Housing ................................................................................60

Figure 5.12 Clouds Fly Airframe With The ArduPilot System Installed ..................................60

Figure 5.13 Fly by Wire vs. Normal Transmitter Demand .....................................................63

Figure 6.1 Flight Path for Autonomous Flight Simulation ....................................................67

Figure 6.2 Impact Test at 0.5 m ..........................................................................................68



Figure 6.5 Damage to the Fuselage at Drop Test No. 3 ........................................................69

Figure 6.6 Wind Tunnel Configuration ................................................................................70

Figure 6.7 Broughton Park FC.............................................................................................73

Figure 6.8 Field Setup.........................................................................................................73

Figure 6.9 ArduPilot Flight Test Waypoints.........................................................................77

Figure 7.1 Pitch and Roll Settling Time During 1st Flight Test ..............................................79

Figure 7.2 Pitch and Roll Settling Time During 2nd Flight Test...............................................79

Figure 7.3 Pitch and Roll Settling Time during 3rd Flight Test ..............................................80

Figure 7.4 Mapped Google Earth KML Data from the Stabilize Test .....................................80

Figure 7.5 Attitude Beyond the 180 Degrees Bank Angle.....................................................81

Figure 7.6 Time to Maximum Pitch and Bank Angle During 5th Flight Test ............................82

Figure 7.7 Time to Maximum Bank and Pitch Angle During 6th Flight Test...........................83

Figure 7.8 Waypoint for Autonomous Test 1 .......................................................................85

Figure 7.9 Autonomous Simulation at 4m/s wind speed ......................................................86

Figure 8.1 FTDI Connection To The ArduPilot ......................................................................88

Figure 8.2 ArduPilot Configuration Tool v1.4.14..................................................................90

Figure 8.3 Broughton Park F.C............................................................................................91

Figure 8.4 Installing the IMU..............................................................................................92

Figure 9.1 Proposed GUI Framework ..................................................................................98


8

Abstract

Unmanned aerial vehicles are complex systems created to substitute the human in

missions deemed as Dirty, Dull or Dangerous [1]. With the current maturity of the

technology and the high level of demand, academics are now interested in research

opportunities to develop innovative and unique UAV solutions.

At the University of Manchester, the 4th year Aerospace Engineering students have the

task to develop a design process that will ultimately result in the production of an

autonomous UAV, which is to be flight tested using the Arduino based autopilot.

During the development of this project, an autopilot simulation module and several

post-processing tools were created and are fully operational for them to be deployed

to the students. Within this report all these tools are presented and discussed.

A guide was written for student to setup and deploy the ArduPilot into their UAVs. This

manual shall minimize the time it will take to students to use the ArduPilot.


9

Declaration

No portion of the work referred to in the dissertation has been submitted in support of

an application for another degree or qualification of this or any other university or

other institute of learning;


10

Intellectual Property Statement

I. The author of this dissertation (including any appendices and/or schedules to

this dissertation) owns certain copyright or relatedrights in it (the Copyright)

and s/he has given The University of Manchester certain rights to use such

Copyright, including for administrative purposes.

II. Copies of this dissertation, either in full or in extracts and whether in hard or

electronic copy, may be made only in accordance with the Copyright, Designs

and Patents Act 1988 (as amended) and regulations issued under it or, where

appropriate, in accordance with licensing agreements which the University has

entered into. This page must form part of any such copies made.

III. The ownership of certain Copyright, patents, designs, trademarks and other

intellectual property (the Intellectual Property) and any reproductions of

copyright works in the dissertation, for example graphs and tables

(Reproductions), which may be described in this dissertation, may not be

owned by the author and may be owned by third parties. Such Intellectual

Property and Reproductions cannot and must not be made available for use

without the prior written permission of the owner(s) of the relevant

Intellectual Property and/or Reproductions.

IV. Further information on the conditions under which disclosure, publication and

commercialisation of this dissertation, the Copyright and any Intellectual

Property and/or Reproductions described in it may take place is available in

the University IP Policy (see:

http://documents.manchester.ac.uk/display.aspx?DocID=487), in any relevant

Dissertation restriction declarations deposited in the University Library, The

University Librarys regulations (see:

http://www.manchester.ac.uk/library/aboutus/regulations) and in The

Universitys Guidance for the Presentation of Dissertations.


11

This dissertation is dedicated to my parents, grandparents, siblings, my nephew, and

my girlfriend. I would also like to dedicate my work to General de Divisin Eutimio

Fuguett Borregales, Aviator of the Venezuelan Air Force, who mentored me during my

first aviation related project.


12

Acknowledgments

First, I would like to thank Gareth Roberts, whose guidance and patience made the

process of learning about the Arduino system a great experience. Also, David

Langkamp who arranged for the wind tunnel testing, his input on the report was also

greatly appreciated.

To Alberto, Ivan, Christophe, Ian, Ashley, and Chaun, thanks for creating such a great

environment to work in.

I would also like to thanks Dr. Crowther; whose interest, enthusiasm and guidance

throughout the project kept me motivated and engaged.


13

Nomenclature

B Bearing

D Derivative Gain

d Euclidean Distance

e error

h Height

I Integral Gain

K Arbitrary Constant

P Proportional Gain

Q Given Vector

R Rotation Matrix

V Velocity Vector

X X-Axis

x Arbitrary Position along X axis

Y Y-Axis

y Arbitrary Position along X axis

Z Z-Axis

ADC Analogue to Digital Convertor

Rx Accelerometer Reading

Vref Reference Voltage

VzeroG Zero G-Force Voltage

Greek Letter

Angle of Attack

Pitch Angle

Yaw Angle

Roll Angle


14

Acronyms

AC Aerodynamic Centre

CAD Computer Aid Design

CG Centre of Gravity

DCM Direction Cosine Matrix

DOF Degree of Freedom

FBW Fly by Wire

FTDI Future Technology Devices International

GCS Ground Control Station

GPS Global Positioning System

HIL Hardware-in-Loop

IDE Integrated Development Environment

IMU Inertial Measurement Unit

KML Keyhole Mark-up Language

MEMS MicroElectroMechanical Systems

MFS MATLAB Flight Simulator

PID Proportional, Integral, Derivative

PWM Pulse Width Modulation

RC Remote Control

RPM Revolutions per Minute

SUAV Small UAV

UAS Unmanned Aerial System

UAV Unmanned Aerial Vehicle


15

Chapter 1:

Introduction

Never fly the 'A' model of anything.

Ed Thompson.

Unmanned aerial vehicles are complex systems created to substitute the human in

missions deemed as Dirty, Dull or Dangerous [1]. With the current maturity of the

technology and the high level of demand, UAVs are finding their way into civilian

applications: law enforcement, search and rescue, etc. With this tendency due to

increase, academics are now interested in research opportunities to develop

innovative and unique UAV solutions.

UAVs rely on autopilots to performautonomously. Although commercial autopilot are

available theirhigh cost, and license protected source code, makes them useless for

research purposes. Open source autopilots offer a cheap and highly customizable

platform to research on, the ArduPilot is one of them [2].

At the University of Manchester, the 4 th year Aerospace Engineering students have the

task to develop a design process that will ultimately result in the production of an

autonomous UAV, which is to be flight tested using the Arduino based autopilot

ArduPilot.

The ArduPilot system, as an open source solution, is design to fit a wide range of

airframes [2]. Also, it requires the user to program the code to fit a particular aircraft

and tune the PID gains to achieve the desired performance. This involves a steep

learning curve and requires time to adjust the gains. Within the time constraints

inherent to term projects it is necessary to streamline the process and develop a set of

instructions aimed to minimize the challenge involved.

This project aims to produce a set of instructions that would be used by students to

flight test their UAVs using the ArduPilot. It also aims to create a simulation tool for

autonomous flying using an existing flight simulator created by Dr. W. Crowther [3].

To accomplish these goals a general objective was settled.


16

1.1. OBJECTIVES

The main objective of this project is to develop a set of tools and instructions that

student would use to flight test their UAVs.

To accomplish the main objective it will be necessary to:

Investigate how autopilot works, what systems are available and how other academic

institutions are using autopilots as learning tools.

Gain understanding on the ArduPilot system.

Develop an Autopilot module to be integrated into the MATLAB Flight Simulator code.

Install the ArduPilot system in a commercially available airframe.

Flight test the system to gain knowledge on its features and behaviour.

Write an instruction manual based on the experience gained from field testing the

ArduPilot.

1.2. THE REPORT

This report is the compilation of the accomplishments and knowledge obtained

throughout this project.

The literature survey aims to introduce the reader to the current state of related UAV

technology: revising its definition, giving a current view on the systems available,

summarising the research on autopilot being made for teaching purposes and, by

describing their theory of operation.

Then, the theory behind Inertial Measurement Unit (IMU), and how MicroElectro-

Mechanical devices (MEMS) are used to tell the autopilot its current orientation, will

be explained. It will also introduce key concepts on UAV navigation; the same

principles are used in the ArduPilot and in the autopilot module created for MATLAB.

The MATLAB Flight Simulator (MFS) code will then be introduced. Highlighting the

changes made by the author to improve the user interface and to add the autopilot

module. In this section, the blocks and scripts that constitute the autopilot will be

explained in detail. It will be shownthat the simulated autopilot was design as a simile

to the ArduPilot.

The next section will containa detailed description on the ArduPilot as a system, rather

than as a stand-alone entity. It will thenbe shown how the system was ground tested

and integrated to the airframe for flight testing.


17

Flight testing procedures used to assess the performance of the ArduPilot+CloudsFly

integration will be introduced. It will also be shown how the functionality of the

MATLAB autopilot system was tested.

Next, the data obtained via telemetry will be presented and discussed. The

performance of the ArduPilot across its main flight modes, and the sensitivity of the

PID settings, will be the main topics of discussion.

Additionally, using the experience and know-how obtained during this project, a

manual with step-by-step instructions for the will be proposed. The manual includes

guidance on how to use the ArduPilot systemand how to flight test their vehicles using

the tools designed for that purposed.

Limitations and challenges experienced withinthe project will thenbe listed. Solutions

to the problems encountered will be presented, and recommendations on how to

overcome some of the limitations encountered will also be given.

Then, recommendation will be given on future work that should be carried out. High

altitude testing (FABLAB), Hardware-In-Loop simulation (HIL) and furtherdevelopment

on the MFS for pilot training and robustness are some of the topics discussed.

Lastly, the conclusions of the project will be listed. These will be based on the

accomplishment of the objectives raised, and the overall experience using the

ArduPilot system.


18

Chapter 2:

Literature Survey

Where am I?.

Charles A. Lindbergh, upon arrival in Paris

This literature survey aims to cover the research made so far towards achieving the

objectives of the project.First the objectives and scope of the literature survey will be

explained. The literature will be then reviewed in a structured fashion, by addressing

each point in separate.

First, based on work undertaken by the author [1], and other researchers [4], the main

requirements for the constructiona UAV will be revised. Thensimulation and theories

for the construction of an autopilot will be discussed with a special emphasis on PID

controllers [4], [5].

The off-the-shelve and open-source autopilots will be reviewed and compared, based

on the work in [4], [6], [2] and [7].

Lastly the work being undertakenat the Virginia Commonwealth University (VCU) and

Georgia Institute of Technology (GT) will be reviewed; special attention will be given to

the studies that are in progress at the University of Manchester.

2.1. SCOPE

Theory: Special attention will be given to simple PID controllers. Most open-source

autopilots will be using PID control loop to construct the autopilot.

Systems: Commercial autopilots will be mentioned to raise awareness of their

availability and capabilities. A deeper insight will be given to open-source autopilots,

especially to the Paparazzi and ArduPilot systems, as they can be used in student

related projects.

2.2. UAV

An Unmanned Aerial Vehicles (UAV) is an aircraft is described as a powered aircraft

capable of performing its mission without a human operator on board. Since its


19

introduction in the 50s [4] UAVs have been used by the military for missions

considered to be dull, dirty and/or dangerous [1].

In previous work undertaken by the author, it is discussed as series of mission

sceneries were UAV are preferred to humans. Examples of these are dangerous

mission, for example, flying near weather disaster zones or reconnaissance missions

over hostile territory [1]. Advantages and disadvantages of UAVs were also discussed

in [1] as well as reasons why UAVs are systems worth studying. In his report, the

author also attempts to categorize UAV depending on theirrole and size/performance.

However only two types of UAVs can be said to fall within the scope of this project:

1.1.1. Remotely Piloted Vehicles (RPV)

RPVs are defined as those which are remotely flown by a pilot. They rely on two main

components to function [8].

1) The Ground Control Station (GCS)

RPVs are flown using a distant cockpit, oftenreferred as GroundControl Station (GCS).

Shown in Figure 2.1, the GCS is defined in [9] as a unit where an operator (pilot) can

send and receive instructionto one or more airvehicles that have been deployed. For

his thesis Anderson [9] designs and implements a ground control station to be used

with Linkping Universitys existing UAV float.

Figure 2.1 Ground Control Station examples

His research however is limited to the GCS; no consideration is given to the

autonomous potential of suchsystems. However his researchdoes allow visualizing of

the data that can be obtained with the right combination of resources.

In [10]and [8] the GCS is revisited from a status-feed point of view. The GCS is used a

tool to supervise the flight of autonomous systems.


20

2) Aircraft Systems

[9] And [8] presents a list of some of the components that are required to guide,

navigate and communicate with the UAV. A GPS sensor, engine RPM sensor,

accelerometers, angular rate sensors, absolute and differential pressure sensors to

calculate speed andbattery voltage sensor are some of the systems mention in their

papers. A micro controller is also essential as it provides communication with the GCS

[8].

1.1.2. Autonomous UAVs

UAVs are defined as those vehicles which can be used to perform a mission without

any kind of human involvement. This capability is only possible with the use of an

autopilot.

2.3. AUTOPILOTS

As previously discussed, autonomous UAV are those which are able to carry out the

complete mission by its own means. They are able to take-off, climb, loiter and land

(and taxi) without any human intervention [4], [8], [1], [10].

In [11] it is recognized that autopilots were first introduced as an aid to stabilize the

aircraft after a disturbance. They are today complex systems which are used to

navigate the aircraft, manage its sub-systems and stability augmentation.

This inheritedcomplexity demand for simulation of the system to be carried out before

the device is fitted to an aircraft. In [11] it is suggested that MATLAB / Simulink and X-

Plane can be used as a test platform for an autopilot system.

2.4. SIMULATION OF AUTOPILOTS

2.4.1. MATLAB / Simulink and X-Plane Approach

In their paper Ribeiroand Oliveira [11] propose that MATLAB / Simulink can be used to

model the autopilot module, X-Plane is used to model the aircraft dynamics and then a

separate microcontroller to visualize the changes on the control surfaces. This

configuration is shown in Figure 2.2.

This configurationseems convenient as MATLAB / Simulink allows for accurate control

design to be carried out with maximum customization. Also, X-plane can be used to

simulate very accurate flight dynamics, as it relies on blade element theoryto calculate


21

the forces acting on the entire aircraft. This simplifies the task as no flight dynamics

simulation needs to be design on MATLAB [11].

Figure 2.2 Test Platform Block Diagram (As proposed in [11] )

However there is an obvious cost penalty for this configuration. X-Plane is not open-

source software. A license is required to use the simulator; additionally a special USB

key is needed to unlock some features that are not available to the standard

installation of X-plane [12].

A microcontrollercard is also required in order to link MATLAB with the physical model

of the aircraft. A User Datagram Protocol (UDP) needs to be setup between the two

computers being used to simulate the equipment [11].

2.4.2. ArduPilot and X-Plane Approach

As in [11], Bin and Justice [13], propose that the direct link between the Arduino

Integrated development Environment (IDE) can be exploitedfor research purposes. In

their paper, it is proposed that between the Arduino IDE and X-plane all the data

required for a flight test can be collected.

As in the previous approach, this solution toorelies on a licence dependent software,

which, aside from the aggregated cost, cannot be modified therefore one is limited to

the capabilities of X-Plane and the Arduino IDE.

2.4.3. MATLAB / Simulink and ArduPilot Approach

Although the approach in [11] is proved to be stable and accurate, it is deemed as an

expensive configuration andit is proposedby the author that a similar simulation can

be carried out without the use of X-Plane and therefore using only one computer.

The approach proposed in [13] does not rely on MATLAB but on an open source IDE.

This may be a good method to implement the final product (compiling and installing on


22

the microcomputer). However during the design andsimulation stage, MATLAB seems

to be the most appropriate tool to simulate the autopilot module.

The proposed diagram is shown in Figure 2.3.

As with any simulation, there are many ways of simulating autopilots, there are many

levels of complexity that can be added, and many different theories that can be

applied. As indicated in Figure 2.3 a PID based controller will be used in our model.

However, in [4] there is a section dedicated entirely to the different types of autopilot

based on different control theories. Some of whichwill be summarizedbelow. This will

assist on understanding the advantages and compromises taken when using PID

controllers.

Figure 2.3 Proposed Test Platform Block Diagram

2.5. PID BASED AUTOPILOTS

PID is the most widely used controller for autopilots [2], [4], [8], [7]. Its fundamental

equation can be expressed as

() = () ()+ + ()] ) )] ( 2.1 )

Where represents the output value of the controller; e(k) would be the current

error to be corrected as read on the kth sample [5], [14]. Represented as:

= () () ( 2.2 )

Kp, Ki and Kd are the proportional, integral and derivative coefficients respectively. Each

of these is represented by the general functions:

= () ( 2.3 )


23

= ()

( 2.4 )

=

() ( 2.5 )

PID controllers are widely usedas they are simple and extremely effective. They dont

require exhaustive computation to yield a solution. They only required for the correct

gains (P I D coefficients) to be correctly tuned for it to work. In [11] it is suggested that

the system gains can be accurately obtained using the root locus method.

Figure 2.4, shows a simple traditional PID control loop on the left and a more complex

multi-level PID controller used to correct roll attitude by an autopilot.

Figure 2.4 PID Control examples [4], [11]

The configuration shown on the right of Figure 2.4 is widely used and can be

implemented as a cascade of PID controllers to improve robustness [13], [8], [14],

[15], [11]. In most cases they will be arranged in levels were the outer levels will be

used to compute the x and y coordinate of the aircraft (direction) to determine the

correct heading. The middle layers are used to determine is used to stabilize the

heading and altitude. Lastly, the inner layer is used to stabilize attitude (pitch and roll)

[4].

One drawback is that the system will be optimal for only one point of the flight

envelope. As the control gains are not inherently variable, fuzzy controls are needed to

add this feature.

2.6. FUZZY BASED AUTOPILOTS

Shengyi et al [14], explains how the gains of a PID controller can be setup to be self-

adaptive with the use of Fuzzy control methods. In specific:

= + Where : 55

= + Where : 0.02 0.02

= + Where : 8 8


24

The structure of the control circuit, Figure 2.5, shows the double loop that would be

required, one for the control response, and an inner one for the fuzzy control to

decide on the gains. However there is an evident trade-off to be made. The extra

loop would increase the computational time, although this is not a problem for

personal computers, it is something worth considering when using microprocessors.

Figure 2.5 Architecture of the self-adaptive fuzzy PID controller [14]

In their paper they show how, for a longitudinal control, the settling time of

disturbances is reduced and the overshoot controlled. Figure 2.6 shows the settling

time after a step input.

Figure 2.6 Comparison curves of step response [14]

2.7. ROBUST H BASED AUTOPILOT

Aranda and Sanchez [16], claim that by building an autopilot based on non-classical

control theory (for example H) some of the uncertainties and need-for-experience

required for classical theory based autopilots could be avoided; leading to a more

robust system.

IT is based on the assumption that the stable poles close to the transmission zeros can

be cancelled. They identified that the cancelation is not desired when there are poles

closed to the imaginary axis. As an example: The longitudinal poles:

------------ Short Period Mode

----------------Phugoid Mode


25

The above results show that there is a pole very close to the imaginary axis, the

Phugoid mode. This mode is by definition poorly damped [16].

In their approach they have developed the lateral and longitudinal controls in separate

modules. They separated each in an inner and outer loop, the inner loop to stabilize

the aircraft and the outer to follow the navigation instructions. A detailed list of the

results and time response of the system using robust control theory can be found in

[16].

While they manage to successfully build a flight control system which works for

longitudinal disturbance, it seems to fail in lateral mode, where it failed to recover

from engine problems [16].

2.8. AVAILABLE AUTOPILOTS

Reference [4] offers a very complete survey of available autopilot systems, some of the

products described are summarized below. Although open-source autopilot were not

discussed in [4], their research value as cost-effective, fully configurable solutions

makes them essential in this research, for these reasons some of them are summarized

as well.

2.8.1. Commercial Off-the-shelf Autopilots

Some of the commercially available autopilot solutions are discussed in Table 2.1.

Table 2.1 Commercial off-the-shelf Autopilots (3)


26

Off-the-shelf commercial autopilots are good when it is needed for a UAV to perform a

desired task in a very specific way.However most of these systems cancost thousands

of pounds and are closed sources. Therefore their customization is very limited to what

the company has left for the user to configure. This renders them useless for the

applications intended in this research.

2.8.2. Open-source Autopilots

2.8.2.1.LEGO NXT Based Autopilot

Overview:

One of the simplest open source autopilots available is based on the LEGO Mindstorm.

The autopilot uses a proportional control loop that calculates and corrects the attitude

of the aircraft [17].

Requirements and Configuration:

The NXT autopilot requires:

- LEGO NXT Computer.

- 2 specially design gyros.

- A 3-axis accelerometer.

- A specially design servo multiplex.

- A GPS sensor with Bluetooth capabilities.

- Knowledge in RobotC programming.

- Complete system around 600 USD.

Figure 2.7 shows the layout of the components around the autopilot and on the

aircraft.

Figure 2.7 Configuration and Layout of Components [17]

Theory of Operation:


27

To engage the autopilot the toggle switch on the transmitter is used. Then a servo will

press a LEGO touch sensor which will then literally move the control in place. This

assembly is shown in the picture on the far right of Figure 2.7.

The block diagram, Figure 2.8 shows how the NXT computer is feed with the data from

the gyros and accelerometer then, by using a proportional control it calculates the drift

to get back on course and keep the aircraft flying straight.

Figure 2.8 Proposed Block Diagram for the LEGO NXT Based Autopilot

One of the problems with this configuration is uncovered in [17]. The gyros used are

not open source as of the time of this research.They had been heavily modified as the

standard gyros will give deflectionas rate andnot the actual degrees. This modification

can be used in software withthe use of Kalmanfilters.However the NXT is not capable

of performing the computations needed for the conversion, requiring hardware that

had the function built into it [6].

Additionally the NXT does not offer support for transmitting Pulse Width Modulation

(PWM) signals. Therefore, a custom design servo multiplex is required to communicate

with the servos.

Conclusion:

Another problem is the weight of the system. The NXT computer uses 6 AA sized

batteries which lead to the conclusion that this configuration wouldnot be feasible for

small UAVs (SUAVs).

2.8.2.1.Paparazzi Autopilot

Overview:

By relying in complete PID loops and Kalman filters to stabilize and navigate the

aircraft. The Paparazzi is probably one of the most capable open-source autopilot

available.



28

The Paparazzi autopilot requires the following components, which may be arranged as

shown in Figure 2.9:

- Autopilot Board. (A)

- Battery. (B)

- Datalink Modem and Antenna. (D)

- GPS Receiver. (G)

- IR Sensor Board. (I)

- Motor and Controller. (M)

- RC Receiver. (R)

- Servos. (S)

- Camera (for video telemetry). (P)

- Knowledge on Linux or Mac OSX Programming.

Theory of Operation

The Paparazzi uses PID loops to calculate and correct error. For some application it

simplifies the loop by setting the ID components to 0. For example, the pitch control

loop uses only a proportional loop as pitch is usually well damped. This simplifies

development as there is no need to tweak extra gains in the controllers [15].

A global overview of the systemis shown in Figure 2.10. The cascade of PID controllers

are used to stabilize and control in pitch roll andyaw angles, as well as, climb, altitude,

course and navigation loops.

Figure 2.10 Paparazzi Autopilot Global Overview [18]

Figure 2.11 Paparazzi Board with GPS

Figure 2.9 Paparazzi

Components Layout [7]


29

One of the mayor issues with this solution is that intensive programming knowledge is

required. Additionally, the autopilot cannot be programmed using windows OS. Linux

or Mac OSX is required. For an Aerospace Engineering student this represents a steep

learning curve.

Conclusion

With the use of Kalman filters to improve accuracy the accuracy of their hardware,

Paparazzi is one of the most capable open-source autopilots. It is also one of the most

compact designs, Figure 2.11.

However the programming involves requires great experience in Linux or Mac OSX

programming, which is simply not available within the targeted users.

2.8.2.2.Arduino Based Autopilot

Overview:

The ArduPilot is an autopilot system based on the open-source Arduino software and

hardware. It comprises of a main board, an IMU shield or thermopile sensors, the GPS

module and a pressure sensor to measure speed.

As the Paparazzi autopilot, the ArduPilot also relies on PID controllers to function.


The ArduPilot is a complete autopilot solution when all of the following components

are used together. The layout of the components is shown in Figure 2.12, where the

autopilot module is shown on the left, the IMU+ in the centre and the GPS module on

the right.

- ArduPilot board.

- Shield expansion kit with airspeed sensor.

- GPS module.

- XY and Z sensors or ArduIMU+.

- FTDI cable for programming.

- Xbee module for telemetry.

- Arduino IDE software.

- ArduPilot ConfigTool (software) for waypoint uploading.


30

Figure 2.12 ArduPilot Component Assembled

Theory of Operation:

The ArduPilot is the least expensive solution currently available. At around 200 GBP it

is possible to have a complete autopilot system [2].

As the Paparazzi autopilot the ArduPilot also uses a cascade of PID controllers to

stabilize and navigate the aircraft.KalmanFilters are not used on the PID controllers as

the processor currently on the board is not as capable as the one on the Paparazzi (the

ArduPilot uses a ATMega328 embedded processor). Therefore, it is slightly less

accurate than the Paparazzi although, with the right PID settings, the difference in

performance will not be notable for most applications.

The ArduPilot schematics are shown in Figure 2.13. It shows how the ArduPilot takes

input from all the available sensors: Airspeed sensor, IMU and GPS. One of the main

features of this autopilot is that it does not require an external MUX for servo output.

Instead the boardtranslates the required instruction to PWM signals. This results in a

weight and space requirement reduction (very important for SUAV applications).

Figure 2.13 ArduPilot Schematics [19]


31

The ArduIMU+ shield, shown in the centre of Figure 2.12, comprises of 3 axis

accelerometers and gyroscopes which are used to obtain the current attitude of the

aircraft. This information is then process by the autopilot module. Where the current

attitude and GPS location is compared and corrected accordingly [13].

To upload GPS waypoints a configurationtool (ConfigTool) is used. This program works

using Google Earth to map the position of the desired waypoints. The waypoints are

then uploaded to the board. In order to store the waypoints, modification were

needed in the short memory access of the board (EEPROM) to allow for the waypoints

to be uploaded and retained even after the board is turned off. However, all the

modifications and most of the programming needed to have an operative autopilot

come standard in a downloadable folder. The user is only required to adjust the gains

and some minor configuration on the code [2].

One of the drawbacks of the autopilot is the refresh rate of the GPS. The ArduPilot calls

for position at a rate of 1Hz, which implies that the accuracy of the position will be

inversely proportional to the speedof the UAV. It would be beneficial to install a more

capable processor on the board to add filters to provide the autopilot with the

capability of predicting position in between GPS readings to improve accuracy at

medium and high speeds.

Another drawback is the fact that the autopilot cannot receive data via telemetry while

the aircraft is flying (without any modification).

Conclusion:

The ArduPilot is the most user-friendly open-source autopilot currently available. It

relies on PID controllers to work and is very compact and light weight, ideal for SUAV

applications. It is also the cheapest, which makes it a good candidate for student

projects.

2.9. RESEARCH AND TEACHING VALUE OF THE AUTOPILOT

2.9.1. Autopilot System Design as a Capstone in Computer Design Projects at Virginia

Commonwealth University (VCU)

Objective

To develop a flight control system for an autonomous UAV, using existing hardware

based on the Armel FPSLIC device [20].


32

Configuration

An existing UAV platform was used so that students were able to concentrate on the

autopilot system [20]. The following components were used:

- 2 FQM-117B converted army target drones with the following modifications:

o Addition of landing gear (for standard take-off and landing capabilities).

o Updated radio equipment.

o Installation of on-board GPS module.

o New upgraded engines (specifications not available).

- GPS System.

- An Atmel FPSLIC device containing an 8-bit microcontroller and a 40k gate-

equivalent field programmable gate array for students to implement hardware

and software to the board.

- A student design GCS based on Visual Basic.

- Custom code for linking the GCS to the autopilot, coded using C++.

Procedure

The main objective of the Design project was to introduce students to real applications

for embedded system and software design. The task was to produce a fully functional

Flight Control System (FCS). The architecture of the system used is similar to that

offered in the open source solutions previously described in the sense that

microcontroller will take inputs from the GPS and then calculate the control surface

deflection required to steer the aircraft to the selected waypoint [20]. VCUs proposed

System and computer architecture is shown in Figure 2.14.

Figure 2.14: System Architecture and Autopilot Block Diagram [20]

After designing and installing theirsolutions, the student would then test their system

on the air. The model would be flown by the lecturer in manual mode and then

triggered to function autonomously in the air. The students were responsible of


33

monitoring the GCS to verify the position of the UAV relative to the waypoints.

Students were also responsible of modifying targeted waypoints and study the

response of the system to the change in command from the GCS [20].

Project Results

In one semester students were able to successfully implement a FCU to a UAV using

standard hardware. The interests of students inthe project lead them to continue the

development after the semester was officially ended, adding video telemetry

capabilities to their system. The end product was entered in the Association for UAS

international (AUVSI) competition, were they receivedfirst place in overall competition

and in the flight portion of it [20].

This main requirement of the project was to develop a fully functional UAS for video

surveillance. However, no evidence is shown on how well the systems performed

during the flight tests, norany resulting dynamic behaviourcharacteristics. VCUs case

study helps to demonstrate the genuine interest that Autonomous UAVs receives from

academia.

2.9.2. Undergraduate Teaching at Georgia Tech

Objective

The main purpose of Georgia Tech project is to develop a low-cost UAV test-bed as a

teaching resource for undergraduate and graduate Aerospace Engineering students

[8].

Configuration

This paper acknowledges that some of the current off-the-shelf autopilot offer limited

access to the configuration of their system. This makes then difficult to modify,

therefore, rendering them useless for teaching and research activities.

They address the problem by creating their own software and hardware. To balance

the cost they have decided to use a commercially available aircraft. The components

used by GT include:

- Goldsberg Decathlon ARF R/C aircraft.

- Autopilot. Containing:


34

o Sensor board. Equipped with: 3 rate gyros, 3 2-axis accelerometers, a 3-axis

magnetometer, 2 pressure sensors, GPS receiver interface and power

regulating circuitry.

o Inertial sensors. Equipped with: 3 ADXS150 angular rate sensors, 3 ADXL202

dual-axis accelerometers, a 3-axis magnetometer module (HMC2003) and a

Motorola GPS receiver.

- Other sensors include: RPM measurement hardware, a differential pressure

sensors with custom made pitot tubes to measure altitude and speed; also, a

potentiometer is used to accurately obtain the voltage outputted by the PWM

motors. This helps to accurately obtain the deflection of the control surfaces

during autonomous operations.

- Telemetry modem.

- In-house developed GCS.

The experimental schematic is shown in Figure 2.15. This schematic map shows the

configuration of the system as used in flight.

Figure 2.15 Hardware System Configuration Schematic [8]

Procedure

After the design printing and assembly of the PCB board for the autopilot, shown in

Figure 2.16. GT students were involved withthe calibration of the inertial sensors, the

magnetometer and the control surface deflection. This step was vital for the correct

installation of their custom made autopilot system.

Static test performed on the inertial sensor were used to determine noise level on

each sensor. The magnetometer was tested to account for disturbance from nearby

ferrous materials.Control surfaces were tested by measuring the output voltage from

the servo motors. The results from these tests are discussed in detailed in [8].


35

Figure 2.16 Block Diagram and Assembled Autopilot Hardware [8]

Completed the tests, attentionwas given towards identifying the dynamic behaviour

of the aircraft. For this, a linear model for longitudinal and lateral dynamics was

derived from data obtained from a flight test. Theoretical linear model was formulated

for a given flight condition using the stability derivatives. In their paper it is proposed

that these stability derivatives can be obtained by extensive wind tunnel testing or by

using the aircraft geometric and mass characteristics. Such characteristics were

obtained by GT by constructing a detailed CAD model to obtain mass and inertial

properties of the aircraft and then computing the derivatives using MATLAB.

Extended Kalman filter are used to complement the signals from the sensors and the

GPS and to reduce noise.

Findings

This paper summarized the effort undertaken by the Aerospace Engineering

Department at GT. The main configuration of the autopilot was explained in great

detailed. It was also explain the calibration criteria and methodology for the different

sensors.

The next step in their project is to design the control loops to be usedby the autopilot

to control the aircraft. From [8] it is not clear however what the involvement of the

students (especially for the graduate students) will be. The plan for inclusion of this

tool to the current teaching modules was now not presented.


36

2.9.3. Teaching at The University of Manchester

Objective

The intentions of the University of Manchester are three-fold [3]:

- To produce autonomous flight test procedures for student to implement on their

4th year design project.

- To substitute the flight testing exercise currently takenplace on board an aircraft

property of Cranfield University, to decrease cost and enhance student

experience.

- To create an autopilot module in MATLAB / SIMULINK environment to add to the

flight simulator writtenby University staff. This would also be used by the 4th year

design students.

Proposed Configuration

Flight test procedures will be design following existing protocols for flight testing [21],

[22], taking special considerations to the requirement applicable to UAVs [23] and

utilizing a cost-effective, flexible and open source autopilot, such as the ArduPilot.

The main purpose of the procedures to be designed is to evaluate the performance of

the UAVs being design by the 4th year design students by developing their flight

envelope maps.

As a complement, an autopilot module will be added to the MATLAB / SIMULINK flight

simulator, replicating the ArduPilot controllers. The module will provide the students

with a tool to test the autonomous capabilities of their UAVs in a synthetic

environment, reducing the risk of damaging the aircraft. It will also provide the

capability of adjusting the gains on the PID controllers to optimize the autopilot to

each aircraft.

The students task will be to post process the results obtained from the flight tests.

Assembling their flight test envelopes andassessing the performance of their models.

The only drawback perceived so far is that none of the aircraft built flies sufficiently

stable as to add an autopilot module. This makes impossible to perform any kind of

flight test.

However flight test can be performed using a commercially available UAV, following

the approach from GT and VCU. This would prove valuable during the second year


37

flight dynamic laboratory [21]. The same flight test procedures could be applied to an

existing airframe to teach 2nd year Aerospace Students about flight testing, aircraft

performance andaircraft dynamics; effectively substituting the need to hire external

equipment from Cranfield University.

2.10.CONCLUSION

UAV or UAS are complex system which comprises of 3 main components to work: a

Ground Control Station or GCS to control and / or monitor the current state of the

aircraft, the aircraft to be flown and an autopilot module which stabilizes and

navigates the aircraft through the desired path.

It was understood that, autopilots canbe complexsystems and are often constructed

using cascades of PID control loops. It relies on inputs from an IMU or thermopiles and

GPS to read the aircraft attitude and location to then compare it with the flight path

described inthe missionplan. The autopilot then corrects as needed by deflecting the

control surfaces.The amount of the deflection and the settling time will dependon the

gains used on the PID controllers. However these gains will be optimum at one point

in the flight envelope.

To obtain and validate these gains, it was shown different methods of simulating

autopilots. It was concludedthat a MATLAB / SIMULINK simulation would be the most

convenient and cost-effective solution as the simulator already exist.

It was also studied how fuzzy controls can be introduced to create self-adaptive PID

gains. However the increase in performance is linked to an increase in operational

requirements, which might be out of reach for less-expensive autopilots.

Different off-the-shelve commercial autopilots were visited, concluding that a close-

source autopilot, while very useful for commercial applications, its useless for

research applications andstudent projects. These systems were deemed too expensive

as well.

Open-source autopilots were investigated in further detailed. It was concluded that

the most convenient solution is to implement the Arduinobased autopilot ArduPilot.

Lastly different case-studies were introduced on how Universities across the word uses

autopilot and UAV systems for academic purposes.


38

Chapter 3:

Background Theory

It is possible to fly without motors, but not without knowledge and skill.

Wilbur Wright.

One of the main challenges when using autopilot is describing which way is up. The

open source autopilots described in the literature survey relay mainly in two

technologies: The thermopile, which uses thermal reading to seek for warmest point,

and assumes it is the ground. The other is a more precise tool, widely used in

commercial autopilots as well, an Inertial Measurement Unit, or IMU.

This chapter describes the basic principles of the Inertial Measurement Unit (IMU). It

will describe the functionality of MicroEletroMechanical system (MEMS)

accelerometers and gyroscopes. Then, it will explain how the direction cosine matrix

(DCM) algorithm is used for inertial navigation. Lastly, it will show the navigation

model used within the autopilot and ArduPilot. These concepts are necessary to truly

understand the basic theory behind todays open source autopilot technology.

3.1. INERTIAL MEASUREMENT UNIT (IMU)

Inertial Measurements Units (IMUs) are electronic devices, composed of

accelerometers and gyroscopes, which are used to track a crafts position and

orientation with respect to an arbitrary coordinate system [24] [25] [26].

IMUs are preferred overthermopile sensors for UAV applications, as they tend to be

more precise and are not weather or element dependent [27]. They are also more

precise than magnetic based navigation. One disadvantage of IMUs is the drift error.

Drift error results from the addition over time of smaller positioning errors of the

device. However, external speed sensors and GPS sensors can be installed to correct

drift errors [25].


39

3.1.1.Accelerometers

Accelerometers are electromechanical devices used tomeasure acceleration. They can

be static, like the force of gravity, or dynamic, like the ones experience from a change

in velocity. Nowadays they are fairly common in electronics, they can be found in

smartphones, tablets, gaming consoles and other devices [28].

A common of accelerometers are those made using microscopic piezoelectric crystals.

As acceleration changes, the crystal gets stressed.This stress creates a voltage change

which is then translated into acceleration. [29]

Other types of accelerometers are those made using Microelectromechanical Systems

(MEMS) techniques. Where a torsional bar attached to a mass on top of a pedestal

measures the acceleration of a device in movement. Figure 3.1 shows a CAD

representation of this arrangement. The flat design of this configuration makes it

possible to fit these accelerometers inside a chip [30].

Figure 3.1 MEMS 2-DOF Accelerometer

An IMU uses, either a 3-axis MEMS accelerometeror two 2-axis MEMS accelerometers

placed normal to each other, to measure the acceleration on all axes while the aircraft

is in motion. Both configurations are shown in Figure 3.2.

Figure 3.2 Difference between 2 and 3 DOF Accelerometer Based IMUs [31], [32]


40

Figure 3.2 above shows the components of the forces acting inside the accelerometer.

To compute the G force associated to each component a transformation must be made

from the ADC reading to the g force. Equation ( 3.1 ) shows the transformation [25].

=

( 3.1 )

Where, Rx is the ADC reading from the accelerometer. Vref is the voltage supplied. The

Divider depends on the bit module, for a 12-bit module the divider would be (2 12-1 =

4095). VzeroG is ADC reading when no force is exerted. Lastly, everything is divided by

the sensitivity of the device.

Accelerometers can be used to measure acceleration while the aircraft experiences a

change in state. For this reasona gyroscope is also neededto measure the orientation

of the aircraft.

3.1.2. Gyroscopes

Gyroscopes are mechanical devices used to measure orientation. They are used for

applications where inertial measurements are required. A basic gyroscope consists on

a mass which is free to rotate around all 3 axes.

An IMU uses a MEMS gyroscope to sense the orientation of an aircraft. The principle of

operation is similar to the mechanical one. A proof mass, point 1 in Figure 3.3 is

allowed to displace on all 3 axes. The movement generates a change in capacitance.

The capacitance is translated into a digital signal, which is then converted into an angle

by the processor [33]. Figure 3.3 shows a CAD model of a MEMS gyroscope.

Figure 3.3 MEMS Gyroscope


41

3.1.3.Direction Cosine Matrix (DCM)

As described above, accelerometers andgyroscopes are used to obtain the attitude of

the aircraft. However, this information is given with respect to the devices they are on,

the IMU in this instance, and therefore with respect to the aircrafts [26].

An algorithm is requiredto convert these intoEuler angles. First, the acceleration and

translationvalues obtained from accelerometers and gyroscopes must be transformed

by multiplying by a 3 x 3 rotational matrix. The resulting matrix will be accelerations

and translation with respect to the ground. Lastly, these values are converted into

Euler angles by performing a filter based on the DCM [34].

For example:

Given a vector Q (representing velocity, position, acceleration or translations) such

that:

=

( 3.2 )

A 3 x 3 rotational matrix R can be used to rotate Q into another reference frame.

Where:

=

( 3.3 )

The translation from the aircraft to earth axis is carried out by multiplying the matrix R

and Q:

= ( 3.4 )

Where the relation between R and the Euler angles is the direction cosine matrix:

= + +

( 3.5 )

Where c is the cosine and s is the sine. From equations ( 3.3 ), ( 3.4 ) and ( 3.5 ). The

Euler angles, pitch, roll and yaw are given by equation ( 3.6 ):

===

()

(,)

(,)( 3.6 )

Pitch is limitedto -90 to 90 degrees. The function atan2 is used to obtain the values of

roll and yaw for all 4 quadrants [34].


42

Open source IMUs, such as the ArduIMU and the IMU shield of the ArduPilot Mega,

relies on DCM algorithms to obtain the aircrafts orientationwith respect to the ground

[32].

3.2. NAVIGATION ALGORITHM

3.2.1.Distance to waypoint

To calculate the distance to the next waypoint, the easiest approach is to compute the

3D Euclidean distance between those two points, equation ( 3.7 ). This approximation

is valid as the distance between the waypoints is relatively small for SUAVs Figure 3.4

below shows a representation of the problem.

= ( +) +)) ) ) ( 3.7 )

In full size UAV, the distance between waypoints maybe sufficient as to create the

need to implement a distance functionbasedon great circle navigation. Howeversuch

applications are not covered here.

3.2.2.Bearing to waypoint

The bearing is computed simply by calculating the angle that the aircraft velocity

makes with the waypoint. The current velocity vector is computed using the DCM

algorithm previously discussed. Figure 3.4, below, shows the required bearing angle

[35].

Figure 3.4 Geometry Definition for Waypoint Following

From the problem described in Figure 3.4, geometry yields

=

( 3.8 )

However, a better practice is to use the function atan2 as it would also give the

quadrant at which the waypoint is located from the aircraft axis [35].


43

=

( 3.9 )

It is important to clarify that the solution is given in radians, from to . Further

manipulation might be necessary depending on the axis being used. An example on

how this manipulation is carried out is given in the Navigation function within the

autopilot block of the Simulink model, appendix C.15.

3.2.3.Waypoint Transition

Waypoint transition refers to the instruction, givento the autopilot, on how and when

the aircraft is considered to have reached the current designatedwaypoint and it to be

directed to the next.

[35] Explains two possible scenarios on whenthe aircraft is considered to have reached

the destination.

For the first method, the aircraft is consideredto have reached the waypoint when the

distance betweenthan is lesser or equal to a certain threshold. Shown in Figure 3.5,

this is the method used within the ArduPilot code and the Simulink simulation. Care

must be exercise when defining the threshold as flight conditions (including wind

speed) might impede the aircraft from crossing it.

Another method is to define that the waypoint will be reached when a line between

the waypoints has been crossed. Figure 3.5 below shows how the line transition

approach can result in a faster flight when high wind speeds are present.

Figure 3.5 Flight Path of UAV with Wind Speed - Distance vs. Line Transition

How the aircraft behaves just before, and right after, is also important. Transition can

be carried out with an immediate effect on the heading and attitude of the aircraft.

This is the method used in the simulation.


44

Another method, also described by [35] involves a smoothening algorithm to be

developed. This method uses a radius from the waypoint to turn the aircraft in the

direction of the next waypoint. Both methods of transitionare represented Figure 3.6.

Figure 3.6 Waypoint Smoothening Function Effect


45

Chapter 4:

MATLAB Flight Simulation Software

"The pilot is still the pilot, whether he is at a remote console or on the flight deck.With the

potential for thousands of these unmanned aircraft in use years from now, the standards for

pilot training need to be set high to ensure that those on the ground and other users of the

airspace are not put in jeopardy."

Mark Rosenker, NTSB Chairman, 2007.

This chapter describes the changes that were made to the MATLAB Flight Simulator

code createdby Dr. Crowther. It aims to highlight the changes and the added functions

that were made during the development of the autopilot module. A modified version

of the Flight Simulator User Guide is included in appendix A.1. The latest version of the

code is also included as appendix C.

4.1. FLIGHT SIMULATOR VERSION 10.7

The simulator was developed by Dr. Crowther [36] in the Simulink environment, to be

used as a teaching andtraining tool for the students of the aerospace design module.

It is mainly used to flight test the UAVs designed by students and to train the pilots ina

risk free environment.Figure 4.1 below shows a screenshot of the graphic interface of

the simulation while running.

Figure 4.1 Screenshot of Flight Simulator V10.7 Graphic Interface

Figure 4.1 shows the main features of the simulation environment: aircraft, synthetic

word, and force representation lines on the aerodynamic surfaces. Basic user


46

controlled parameters include: camera selection, simulation controls and arrow keys

to change current view.

The simulation works by dividing the model into small sections and then performs

blade element theory on these areas. This allows accurate prediction of the forces and

moments acting on the aircraft. Further information can be obtained in the Flight

Simulator user guide [36].

4.1.1. System Requirements:

- A computer running MATLAB 2007b or higher (version 2010 is not recommended due

to bugs within the visualization toolbox)

- Real Time block (Slows down simulation to run close to real time)

- Aerospace toolbox

- Transmitter with USB connection or any other kind of controller device

4.1.2. Included Airframes:

- Default fixed winged aircraft

- Clouds Fly model

4.1.3. Features:

- Flight dynamics model based on blade element theory

- Collision detection

- Graphics based on VRML

- SolidWorks model compatible

- Stability augmentation (within Autopilot)

- Autopilot capability

- Wind tunnel mode

- Fully configurable models

- Ability to visualize settling time to adjust PID gains

4.1.4. Limitations:

- Autopilot does not control rudder. This was made to imitate the current configuration

used in the ArduPilot.

- The simulation, in any mode, requires the presence of a control device.

- The simulation does not model the physics of the electronic speed controller (ESC) of

an electric powered aircraft.


47

- Atmospheric changes are not simulated.

- The navigation function within the autopilot does not check for cross track errors.

4.2. FLIGHT SIMULATOR AUTOPILOT MODES

4.2.1. Mode 1 Manual:

Fly as a normal RC aircraft. For this mode the Simulink model bypasses the autopilot

block to fly the aircraft as a standard radio controlled model.

4.2.2. Mode 2 Stabilize:

Stabilize is a stability augmentation code inwhich the aircraft goes back to level when

the transmitter stick is released. This mode utilizes a PID control loop to return the

aircraft back to level flight.

This is represented on the Simulink model by stating that, for any signal within the

threshold, the demanded attitude is level flight. Is the example below the threshold is

set 5% transmitter signal.

4.2.3. Mode 3 Autonomous:

In automatic mode, all channels will be controlled by the autopilot block to follow a

series of waypoints previously defined by the user. The algorithm used for waypoint

following is the same as explained in section 3.2.

4.2.4. Mode 4 Wind Tunnel:

Wind Tunnel mode fixes the CG of the aircraft in a specific location of the world. By

having a fixed model, the aerodynamic properties of the aircraft canbe determined at

any fixed attitude. For this mode to work users must set the degree of freedom to

weld and the wind speedmust be correctly set. Further instructions are provided in

the Flight Simulator User Guide.

4.2.5. Mode 5 Fly by wire:

This semi-automatic mode translates the current position of the transmitterand flights

the aircraft to the pointed direction. This mode reads the signals from the transmitter

as an angle demand, instead of the regular rate demand. The maximum bank andpitch

angle are designated by the user.


48

When this mode is selected, assuming a maximum banking angle of 60 degrees, a full

right movement of the aileron channel on the transmitter will be translated as a 60

degree positive bank angle demand. When the control is release the aircraft will go

back to level flight.

4.3. FLIGHT SIMULATOR M-FILES

4.3.1. Master.m

[Appendix C.1]. Master is the simulation initialization file. Within this file the most

common changes can be made to the simulation. Flight mode, wind speed, initial

position, among other parameters can be change before running the simulation. This

file also calls all necessary scripts. Therefore, there is no need to manually command

any other script to be run.

4.3.2. Airframe_Configuration.m

[Appendix C.2]. This script houses all the information regarding the configuration,

geometry and characteristics of the airframe to be simulated.

4.3.3. Airframe_Calc.m

[Appendix C.4]. This script runs when Airframe_Configuration terminates. It contains

all the parameters that do not require user input.

4.3.4. AutoPilotConfig.m

[Appendix C.5]. This script runs Mode_Settings.m and PID_Settings.m. It provides the

simulation with the initial demanded Euler angles, the speed control mode and the

threshold of the autonomy the autopilot will have over the transmitter.

4.3.5. Mode_Settings.m

[Appendix C.10]. This m-file turns on and off the signals incoming from the controller

and the autopilot depending on which mode is selected.

4.3.6. PID_Settings.m

[Appendix C.11]. This script contains the inner and outer PID gains. They are to be

modified every time a new airframe is installed or the design is changed. The file

PID_Tuner.m, explained below, can be used to set the gains.


49

4.3.7. ESC_Settings.m

[Appendix C.6]. Depending on the autonomy mode of the speed controller, this script

turns on and off the signals from the autopilot to take control of the speed loop.

Although no ESC is simulated, the Simulink block and configuration script has been

named that way for the function they perform.

4.3.8. Waypoints_Config.m

DEVELOPMENT OF PROCEDURES FOR FLIGHT...

Documents

Transcript of DEVELOPMENT OF PROCEDURES FOR FLIGHT...