Modelling, Flight Dynamics and Inner-Loop

Control of an Urban Air Mobility Vehicle Subject

to Empirically-Developed Urban Airflow
Disturbances

by

Richard Grant McKercher

A Thesis submitted to
the Faculty of Graduate Studies and Research
in partial fulfilment of
the requirements for the degree of
Master of Applied Science

Ottawa-Carleton Institute for

Mechanical and Aerospace Engineering

Department of Mechanical and Aerospace Engineering

Carleton University
Ottawa, Ontario, Canada
November 2021
 Copyright ©
2021 - Richard Grant McKercher

ii
 The undersigned recommend to
the Faculty of Graduate Studies and Research
acceptance of the Thesis

Modelling, Flight Dynamics and Inner-Loop Control of an

Urban Air Mobility Vehicle Subject to
Empirically-Developed Urban Airflow Disturbances

Submitted by Richard Grant McKercher

in partial fulfilment of the requirements for the degree of
Master of Applied Science

F. Khouli, Supervisor

A. S. Wall, Supervisor

G. L. Larose, Supervisor

R. Miller, Department Chair

Carleton University
2021

iii
Abstract

Advanced Air Mobility in general, and Urban Air Mobility in particular, are expected
to play a significant role in improving transportation of people and goods in increas-
ingly growing urban centres while contributing to reduction in emissions and reaching
the goal of sustainable urban growth. The goal of this work was to investigate the per-
formance of a novel controller as an autopilot for a generic aircraft operating in urban
airflow. In this investigation, the extraction of geometric, inertial and aerodynamic
properties of a generic urban air taxi, similar in design to the Bell Nexus 4EX, was
carried out. Subsequently the properties were used as a basis for the development of
a linearized flight dynamics numerical model of the aircraft. Flights about a trimmed
condition at cruising speed were simulated including cases where the urban air taxi
was immersed in empirically-developed urban airflow disturbances. The open-loop re-
sponse of the urban air taxi was assessed and classical proportional-integral-derivative
and Active Disturbance Rejection Control inner-loop controllers were developed. The
performance of the controllers in presence of the aforementioned urban airflow distur-
bance were then examined and compared. Active Disturbance Rejection Control was
shown to perform similarly to proportional-integral-derivative control in the flight of
the aircraft immersed in a representative urban airflow environment. Future investi-
gations were recommended to increase the model fidelity to understand how unsteady
aerodynamics would affect the aircraft motion and that real world considerations be
included to further test Active Disturbance Rejection Control.

iv
Acknowledgments

I would like to acknowledge the financial support for this project from the Natural
Sciences and Engineering Research Council of Canada (NSERC), the National Re-
search Council of Canada (NRC), and Carleton University. Their support allowed for
this work to be conducted.
I would also like to acknowledge Maryam Al Labbad for providing support and
assistance when working with the data that she collected and inviting me to observe
the wind tunnel testing. The data was a crucial part of the project and observing
the wind tunnel tests was invaluable in gaining knowledge about the urban airflow
environment.
My final acknowledgement is to my supervisors. I would like to thank Dr. Fidel
Khouli, Dr. Alanna Wall, and Dr. Guy L. Larose for their continuous support and
assistance while I conducted this work. Their support made it possible for me to
successfully complete this project.

v
Table of Contents

Abstract iv

Acknowledgments v

Table of Contents vi

List of Tables viii

List of Figures x

Nomenclature xv

1 Introduction 1
1.1 Scope and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . 2

2 Literature Review 4
2.1 Experimental and Computational Urban AirFlow Research . . . . . . 4
2.2 Flight Dynamics and Control of an Urban Air Taxi/UAM Vehicle . . 6
2.3 Active Disturbance Rejection Control . . . . . . . . . . . . . . . . . . 8

3 Flight Dynamics Modelling of a Generic Urban Air Taxi 11

3.1 Generic Urban Air Taxi Design Choice . . . . . . . . . . . . . . . . . 12
3.2 Geometric and Inertial Parameter Estimation . . . . . . . . . . . . . 16
3.2.1 Geometric Parameter Estimation . . . . . . . . . . . . . . . . 16

vi
 3.2.2 Inertial Parameter Estimation . . . . . . . . . . . . . . . . . . 21
3.3 Aerodynamic Parameter Estimation . . . . . . . . . . . . . . . . . . . 26
3.3.1 Lift-Curve Slope Coefficient Estimation . . . . . . . . . . . . . 26
3.3.2 Longitudinal Static Analysis . . . . . . . . . . . . . . . . . . . 28
3.3.3 Estimation of Stability Derivatives . . . . . . . . . . . . . . . 35
3.4 State-Space Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.5 Urban Air Flow Disturbances . . . . . . . . . . . . . . . . . . . . . . 68

4 Inner-Loop Controller Design 74

4.1 Autopilot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.2 PID Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.3 ADRC Controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77

5 Results and Discussion 85

5.1 Tuning Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.2 Disturbance Response Comparison . . . . . . . . . . . . . . . . . . . 97

6 Conclusion and Future Work 104

List of References 107

Appendix A Simulink Model Implementation 112

Appendix B Lateral autopilot tuning results 115

Appendix C Longitudinal and lateral A and B matrix estimations 126

vii
List of Tables

3.1 Wing section span and chord estimates for the generic UAT . . . . . 17
3.2 Wing geometric parameters . . . . . . . . . . . . . . . . . . . . . . . 18
3.3 Canard geometric parameters . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Vertical tail geometric parameters . . . . . . . . . . . . . . . . . . . . 20
3.5 Control surface geometric parameters . . . . . . . . . . . . . . . . . . 24
3.6 Mass moments of inertia about the principal axes . . . . . . . . . . . 26
3.7 Lift-curve slope coefficients . . . . . . . . . . . . . . . . . . . . . . . . 27
3.8 Control effectiveness coefficients . . . . . . . . . . . . . . . . . . . . . 28
3.9 Horizontal flight trim conditions . . . . . . . . . . . . . . . . . . . . . 35
3.10 Longitudinal non-dimensional stability derivatives . . . . . . . . . . . 36
3.11 Lateral non-dimensional stability derivatives . . . . . . . . . . . . . . 36
3.12 Alpha derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.13 Forward speed derivatives . . . . . . . . . . . . . . . . . . . . . . . . 38
3.14 Pitch rate derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.15 Side slip derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.16 Roll rate derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.17 Yaw rate derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.18 Control derivatives . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.19 Longitudinal dimensional stability derivatives . . . . . . . . . . . . . 49
3.20 Lateral dimensional stability derivatives . . . . . . . . . . . . . . . . 50

viii
3.21 Longitudinal dimensional control derivatives . . . . . . . . . . . . . . 51
3.22 Lateral dimensional control derivatives . . . . . . . . . . . . . . . . . 51
3.23 Longitudinal A matrix eigenvalues, period, and damping . . . . . . . 63
3.24 Lateral A matrix eigenvalues, period, and damping . . . . . . . . . . 63
3.25 New lateral A matrix eigenvalues, period, and damping . . . . . . . . 68
4.1 PID controller gain values . . . . . . . . . . . . . . . . . . . . . . . . 76
4.2 ADRC gain values . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

ix
List of Figures

1.1 Expected flight phases for single UAT trip . . . . . . . . . . . . . . . 3

3.1 Bell Nexus 4EX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3.2 Bell Boeing V-22 Osprey . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.3 Bell X-22 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
3.4 Top View of Bell Nexus 4EX . . . . . . . . . . . . . . . . . . . . . . . 15
3.5 Side view of Bell Nexus 4EX . . . . . . . . . . . . . . . . . . . . . . . 15
3.6 Top view of generic UAT layout . . . . . . . . . . . . . . . . . . . . . 17
3.7 Side view of Bell Nexus 4EX . . . . . . . . . . . . . . . . . . . . . . . 19
3.8 Front view of Bell Nexus 4EX . . . . . . . . . . . . . . . . . . . . . . 20
3.9 Flap chord factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.10 Flap span factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.11 Box body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.12 Wing lift geometric parameters . . . . . . . . . . . . . . . . . . . . . 27
3.13 Generic UAT layout side view . . . . . . . . . . . . . . . . . . . . . . 28
3.14 Effect of a fuselage or nacelle on neutral-point position . . . . . . . . 33
3.15 Roll damping part 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.16 Roll damping part 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.17 Roll damping part 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.18 Wing yawing derivative parameter . . . . . . . . . . . . . . . . . . . . 46

x
3.19 Notation for body axes, angular and linear displacements, rates, forces
and moments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.20 (a) Definition of α, (b) Definition of β . . . . . . . . . . . . . . . . . 55
3.21 Characteristic transients. (a) Long period mode, (b) Short period mode 57
3.22 (a) Vector diagram of long period mode, (b) Vector diagram of short
period mode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.23 Motion of longitudinal dynamic modes . . . . . . . . . . . . . . . . . 59
3.24 Motion of spiral mode . . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.25 Vector diagram of Dutch roll mode . . . . . . . . . . . . . . . . . . . 61
3.26 Motion of Dutch roll mode . . . . . . . . . . . . . . . . . . . . . . . . 62
3.27 Change in lateral eigenvalues due to change in Lr . . . . . . . . . . . 64
3.28 Change in lateral eigenvalues due to change in Yv . . . . . . . . . . . 66
3.29 Change in lateral eigenvalues due to change in Lv . . . . . . . . . . . 67
3.30 Top down view of city model . . . . . . . . . . . . . . . . . . . . . . . 70
3.31 Side view of city model . . . . . . . . . . . . . . . . . . . . . . . . . . 71
3.32 PSD of u direction with modal values . . . . . . . . . . . . . . . . . . 72
3.33 PSD of v direction with modal values . . . . . . . . . . . . . . . . . . 72
3.34 PSD of w direction with modal values . . . . . . . . . . . . . . . . . . 73
3.35 Variations in wind speed in each coordinate direction . . . . . . . . . 73
4.1 PID controller subsystem . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.2 Uncontrolled longitudinal response to unit step disturbance in w . . . 77
4.3 Uncontrolled lateral response to unit step disturbance in v . . . . . . 77
4.4 Controlled longitudinal response to unit step disturbance - PID . . . 78
4.5 Controlled lateral response to unit step disturbance - PID . . . . . . 78
4.6 ADRC topology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.7 ADRC autopilot subsystem . . . . . . . . . . . . . . . . . . . . . . . 82
4.8 ADRC elevator channel subsystem . . . . . . . . . . . . . . . . . . . 83

xi
4.9 ESO subsystem block . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
5.1 PID tuning responses - longitudinal . . . . . . . . . . . . . . . . . . . 86
5.2 PID tuning responses zoomed - longitudinal . . . . . . . . . . . . . . 87
5.3 Rudder PID tuning responses . . . . . . . . . . . . . . . . . . . . . . 87
5.4 Aileron PID tuning response . . . . . . . . . . . . . . . . . . . . . . . 88
5.5 Rudder PID tuning responses with tuned aileron channel . . . . . . . 88
5.6 Effect of ESO gains on state estimation error . . . . . . . . . . . . . . 89
5.7 Effect of ESO gains on state estimation error - zoomed view . . . . . 90
5.8 Vertical airspeed response due to ESO tuning . . . . . . . . . . . . . 90
5.9 Vertical airspeed response due to ESO tuning - zoomed view . . . . . 91
5.10 Effect of tuning parameter gains on estimation error - β2 = 8 . . . . . 92
5.11 Effect of tuning parameter gains on estimation error - β2 = 12 . . . . 92
5.12 Vertical airspeed response due to parameter tuning - β2 = 8 . . . . . 93
5.13 Vertical airspeed response due to parameter tuning - β2 = 12 . . . . . 93
5.14 Vertical airspeed response due to parameter tuning - detailed view of
select parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.15 Vertical airspeed response of tuned PID and ADRC autopilots - step
disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.16 Longitudinal airspeed response of tuned PID and ADRC autopilots -
step disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
5.17 Lateral airspeed response of tuned PID and ADRC autopilots - step
disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.18 Variations in bank angle for tuned PID and ADRC autopilots - step
disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.19 Variations in α angle for tuned PID and ADRC autopilots - step dis-
turbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97

xii
5.20 Variations in β angle for tuned PID and ADRC autopilots - step dis-
turbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.21 Variations in elevator angle for tuned PID and ADRC autopilots - step
disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
5.22 Variations in rudder angle for tuned PID and ADRC autopilots - step
disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.23 Variations in aileron angle for tuned PID and ADRC autopilots - step
disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.24 Longitudinal airspeed response of tuned PID and ADRC autopilots -
urban airflow disturbance . . . . . . . . . . . . . . . . . . . . . . . . 100
5.25 Lateral airspeed response of tuned PID and ADRC autopilots - urban
airflow disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.26 Variations in bank angle for tuned PID and ADRC autopilots - urban
airflow disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.27 Variations in α angle for tuned PID and ADRC autopilots - urban
airflow disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.28 Variations in β angle for tuned PID and ADRC autopilots - urban
airflow disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.29 Variations in elevator angle for tuned PID and ADRC autopilots -
urban airflow disturbance . . . . . . . . . . . . . . . . . . . . . . . . 102
5.30 Variations in rudder angle for tuned PID and ADRC autopilots - urban
airflow disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
5.31 Variations in aileron angle for tuned PID and ADRC autopilots - urban
airflow disturbance . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103
A.1 Top level simulink block diagram . . . . . . . . . . . . . . . . . . . . 112
A.2 State-space model arrangement . . . . . . . . . . . . . . . . . . . . . 113
A.3 Wind disturbance subsystem . . . . . . . . . . . . . . . . . . . . . . . 113

xiii
A.4 Wind disturbance state-space model . . . . . . . . . . . . . . . . . . . 114
A.5 Controller/Autopilot subsystem . . . . . . . . . . . . . . . . . . . . . 114
B.1 Effect of tuning parameter gains on estimation error - rudder channel β1 115
B.2 Yaw rate response due to ADRC tuning - rudder channel β1 . . . . . 116
B.3 Effect of tuning parameter gains on estimation error - rudder channel β2 116
B.4 Yaw rate response due to ADRC tuning - rudder channel β2 . . . . . 117
B.5 Effect of tuning parameter gains on estimation error - rudder channel β3 117
B.6 Yaw rate response due to ADRC tuning - rudder channel β3 . . . . . 118
B.7 Effect of tuning parameter gains on estimation error - rudder channel r0 118
B.8 Yaw rate response due to ADRC tuning - rudder channel r0 . . . . . 119
B.9 Effect of tuning parameter gains on estimation error - rudder channel h0 119
B.10 Yaw rate response due to ADRC tuning - rudder channel h0 . . . . . 120
B.11 Effect of tuning parameter gains on estimation error - aileron channel β1 120
B.12 Bank angle response due to ADRC tuning - aileron channel β1 . . . . 121
B.13 Effect of tuning parameter gains on estimation error - aileron channel β2 121
B.14 Bank angle response due to ADRC tuning - aileron channel β2 . . . . 122
B.15 Effect of tuning parameter gains on estimation error - aileron channel β3 122
B.16 Bank angle response due to ADRC tuning - aileron channel β3 . . . . 123
B.17 Effect of tuning parameter gains on estimation error - aileron channel r0 123
B.18 Bank angle response due to ADRC tuning - aileron channel r0 . . . . 124
B.19 Effect of tuning parameter gains on estimation error - aileron channel h0 124
B.20 Bank angle response due to ADRC tuning - aileron channel h0 . . . . 125

xiv
Nomenclature

Aircraft Model Parameters

A State matrix
Aa,j Magnitude of wind speed variation at frequency speci-
fied by index j at location a
AR Aspect ratio
ARF Vertical tail aspect ratio
B Control vector/matrix
CD Drag coefficient
CD0 Drag coefficient at the reference condition
CL Overall aircraft total lift coefficient
CLc Total lift coefficient of canard
CLtrim Overall aircraft lift coefficient at trim condition
CLwb Total lift coefficient of wing-body
CL0 Lift coefficient at reference condition
(CL )0 Zero αwb angle lift coefficient
CLα 3D lift-curve slope coefficient
Cl α 2D lift-curve slope coefficient
CLδ Change in 3D lift coefficient due to trailing edge surface
deflection

xv
CLδe Change in overall aircraft lift coefficient due to elevator
deflection
Cl δ Change in 2D lift coefficient due to trailing edge surface
deflection
Cm Overall aircraft pitching moment coefficient
Cmacwb Pitching moment coefficient about wing-body aerody-
namic centre
Cm p Pitching moment coefficient due to propulsive effects
Cm 0 Pitching moment coefficient at zero α angle
Cm α Pitching moment coefficient due to change in α
Cmδe Change in overall aircraft pitching moment coefficient
due to elevator deflection
CT Thrust coefficient
CW 0 Weight force coefficient at the reference condition
CX Non-dimensional force coefficient in x-direction
CZ Non-dimensional force coefficient in z-direction
Cm Non-dimensional pitching moment coefficient
Cl Non-dimensional rolling moment coefficient
Cn Non-dimensional yawing moment coefficient
Cy Non-dimensional force coefficient in y-direction
G(K) Theodorsen function
H Aircraft body height
Iy Mass moment of inertia about vehicle y-axis
Ix′ , Iz′ , Izx
′
Mass moments and product of inertia about stability
axes
Ix , Iy , Iz Mass moments of inertia about reference axis
Ixp , Iyp , Izp Mass moments of inertia about principal axes

xvi
K1 Deflection surface chord factor
K2 Deflection surface span factor
Kn Static margin
L Aircraft body length
Lv , Lp , Lr Roll axis moment with respect to lateral airspeed, roll
rate, and yaw rate
M Mach number
M0 Mach number at reference condition
Mu , Mw , Mq , Mẇ Pitching moment with respect to forward speed, vertical
speed, pitch rate and rate of change of vertical speed
Nv , Np , Nr Yaw axis moment with respect to lateral airspeed, roll
rate, and yaw rate
N Total number of frequencies
S1 First section of main wing, main fixed wing
S2 Second section of main wing, ducted fan wing
S3 Third section of main wing, main wing tip
S Wing-body planform area
Saa,j The magnitude of the auto-spectrum for frequency com-
ponent j
Sc Canard planform area
SF Vertical tail planform area
Sx Area of a particular wing section
T Period
VH Horizontal tail (canard) volume ratio relative to centre
of gravity
V̄H Horizontal tail (canard) volume ratio relative to aerody-
namic centres

xvii
VV Vertical tail volume ratio
V Aircraft speed
W Aircraft weight
W Aircraft body width
Xu , Zu , Mu X, Z-direction force, and pitching moment with respect
to forward airspeed
Xw , Zw , Mw X, Z-direction force, and pitching moment with respect
to vertical airspeed
Yv , Yp , Yr Y-direction force with respect to lateral airspeed, roll
rate, and yaw rate
Zẇ , Mẇ Z-direction force and pitching moment with respect to
vertical change in airspeed
Zq , Mq Z-direction force and pitching moment with respect to
pitch rate
∆Xc , ∆Zc , ∆Mc Change in X, Z-direction force and pitching moment due
to control input
∆Yc , ∆Lc , ∆Nc Change in Y-direction force, roll, and yaw axis moments
due to control input
a Overall aircraft lift-curve slope coefficient
aa Change in lift coefficient due to aileron deflection
ac Lift-curve slope coefficient of the canard
ae Change in lift coefficient due to elevator deflection
aF Vertical tail lift-curve slope coefficient
af Change in wing-body lift due to flap deflection
ar Change in vertical tail lift-curve slope coefficient due to
rudder deflection
awb Lift-curve slope coefficient of wing-body

xviii
a∞ 2D aircraft lift coefficient
b Wing span
b1 S1 section span
b2 S2 section span
b3 S3 section span
b∗ Span of a particular wing section
c Wing chord at specific location along span
croot Chord at wing root
ctip Chord at wing tip
c̄ Mean aerodynamic chord
c̄F Mean aerodynamic chord of vertical tail
cf /c Deflection surface chord ratio
e Oswald efficiency factor
g Gravitational acceleration constant
h Overall aircraft neutral point
hn Overall aircraft neutral point location as chord percent-
age from leading edge
hnwb Wing-body aerodynamic centre location as chord per-
centage from leading edge
hnw Wing neutral point
j Frequency index value
lc Distance between canard aerodynamic centre and the
centre of gravity
¯lc Distance between wing-body and canard aerodynamic
centres
m aircraft mass
n Real part of complex value

xix
ni , nf Span factor parameters
p Change in aircraft roll rate
pd Dynamic pressure
q Change in aircraft pitch rate
r Change in aircraft yaw rate
r Point mass lever arm for moment of inertia calculation
t Time
thalf Time for the amplitude of oscillation to dampen to half
the initial amplitude
t/c Airfoil thickness ratio
uo Initial vehicle airspeed
u Change in aircraft forward airspeed
v Change in aircraft lateral airspeed
va Time-series representation of wind variation
w Change in aircraft vertical airspeed
x Forward direction in body coordinates
x state variable vector
ẋ rate of change of state variables
x̄ Location of aerodynamic centre in chord direction start-
ing from leading edge
ȳ Location of aerodynamic centre along span of wing/tail
y Lateral direction in body coordinates
z vertical direction in body coordinates
zF Vertical distance between the centre of gravity and fin
aerodynamic centre
ΛLE Leading edge sweep angle
ΛLEF Vertical tail leading edge sweep angle

xx
Λc/4 Sweep back angle at quarter chord location
Λβ Compressible sweep parameter
αwb Angle of attack of wing-body, also used for canard
α Angle of attack relative to overall aircraft zero-lift line
αtrim Angle of attack at trim condition
α̇ Change in rate of change of angle of attack
β Aircraft side slip angle
β̂ Prandtl-Glauert compressibility factor
δe Elevator deflection angle
δf Flap deflection angle
δei Initial elevator defection angle
δetrim Elevator angle at trim condition
ϵ Wing-body angle of attack at the reference condition
θo Initial pitch angle
θ Change in aircraft inclination angle
θtwist Wing twist angle
κ Aerodynamics estimation parameter
λ Taper ratio
λF Vertical tail taper ratio
π Pi
ρ Air density
ψa,j Phase angle of variation specified by index j at location
a
ωa,j Frequency of variation specified by index j at location a
ω Imaginary part of complex value
Controller Parameters
F (t) Total disturbance

xxi
G(t) Rate of change of total disturbance
b Estimation of control matrix
b0 Estimation of control matrix
e PID controller set-point error
e1 , e2 Active Disturbance Rejection Control(ADRC) set-point
error and error derivative
f han, f al ADRC Non-linear functions
f e, f e1 Non-linear function of error and error derivative
a, a0 , a1 , a2 , d, sa , sy , y f han function parameters
k0 , k1 , k2 PID controller gain values
c, r0 , h0 , b0 , α1 , α2 , δ, β1 , β2 , β3 ADRC parameters
u Control input
v, v1 , v2 Set-point signals
z1 , z2 , z3 ADRC state estimates

xxii
Chapter 1

Introduction

There is increased interest in aircraft operations in unserved and underserved local,

regional, intraregional and urban areas with the goal of improved and sustainable
passenger and cargo transportation. This new area in aviation is being referred to as
Advanced Air Mobility (AAM) and Urban Air Mobility (UAM) [1, 2]. It is predicted
that 60% of the world’s population will live in urban areas by 2030 [3] and UAM
operations are already being conducted today, where in October 2021, a fully au-
tonomous drone performed the first lung transport in a dense urban setting [4]. This
increase in urban population and access to new aerospace technology is contributing
to increasing aircraft operations over and within cities.
The challenges for UAM include, but are not limited to, GPS reception issues
near buildings, security risks with data transmission, developing new air-traffic con-
trol technologies, and noise and safety requirements. One aspect of the urban envi-
ronment that affects aircraft performance and safety are turbulent wind conditions.
These conditions occur in the urban environment and form from the general wind
interaction with the surface roughness of the surrounding area as well as the spe-
cific flow interaction with tall buildings. Turbulence can severely affect the flight
performance of aircraft and will play a significant role in how aircraft operations are
regulated and conducted in the urban environment.

1
 2

It is expected that UAM vehicles will be either entirely autonomously controlled

or be piloted by a human through a stability augmentation system (SAS). To real-
ize a successful, safe and sustainable future with UAM, novel control concepts must
be analysed under conditions that correspond to those which UAM vehicles will en-
counter. This thesis presents the development of a generic Urban Air Taxi (UAT)
dynamical model operating in the horizontal cruise phase of flight for the purposes of
evaluating the suitability of different control systems for UAM. An inner-loop flight
controller was designed using classical Proportional-Integral-Derivative (PID) control
and Active Disturbance Rejection Control and the performance of each controller was
compared while the aircraft model was perturbed. The turbulent conditions corre-
sponding to urban airflow were created using experimental data collected during an
investigation of urban airflow [5]. The experimental urban airflow data were com-
bined with the UAT dynamic model and the controller, and was simulated using
MATLAB’s Simulink to determine if ADRC would be a suitable controller for UAM
operations.
This thesis is organized into; Chapter 1 which introduces the work presented in this
thesis; Chapter 2 which presents the current literature related to the topics discussed;
Chapter 3 which presents the development of the generic UAT flight model; Chapter
4 which presents the two control architectures and implementation in the simulation;
Chapter 5 presents the results of the analysis; and Chapter 6 concludes the thesis.

1.1 Scope and Limitations

This work aimed to develop a generic Urban Air Taxi flight dynamics model that
included a novel controller as an inner-loop autopilot so as to to test the performance
of the UAT and controller while being subjected to urban airflow disturbances which
were based off empirical data. The scope of the research is limited to investigating
 3

the controller performance as an inner-loop autopilot for a generic UAT operating

during the horizontal cruise phase of flight. Effects associated with the ducts, duct
actuation, differential thrust and unsteady aerodynamic effects are excluded. Figure
1.1 shows the expected phases of flight for a single trip where the cruise phase is the
focus of this research.

Figure 1.1: Expected flight phases for single UAT trip

Chapter 2

Literature Review

This chapter presents a review of the current literature covering experimental and
computational urban airflow research, flight dynamics and control of UAT/UAM
vehicles and Active Disturbance Rejection Control. Each topic is discussed in its
corresponding section.

2.1 Experimental and Computational Urban Air-

Flow Research

When winds are present in an urban environment it is common to experience gusty

conditions and high wind speeds that change throughout the area. The investigation
by Adkins [6] demonstrated that high levels of turbulence are common in the urban
environment, which produces operational challenges to autonomous flight control and
stability. The investigation concluded with the following summary of certain urban
airflow characteristics that are of relevance to flight dynamics of UAT:

ˆ At low altitudes, the turbulence intensity in suburban areas can be as much as

twice what is found in the rural environment.

ˆ The turbulence intensity in high rise city centres can be three times higher as

4
 5

compared to the rural environment.

ˆ Disturbed flow can exist at a distance from a building up to three times the
height of that building.

The last observation implies that a UAT will experience gusts and disturbances
throughout the urban environment.
As part of the development of the ACELAB simulator for electric VTOL (eV-
TOL) by NASA, scientists have measured wind speeds and directions in key locations
throughout downtown San Francisco [7]. This collected data were used to create more
accurate wind models for use in the ACELAB simulator. One of the conclusions of
this program was that it could be very difficult to manually operate a UAM vehi-
cle in certain areas of the city or near tall buildings. They noted that in the past
for aircraft development, accurate wind modelling was not required for analysing the
performance of aircraft and only basic wind features were acceptable. However, UAM
vehicles will be operating in near constantly changing turbulent wind conditions when
in the urban environment and therefore requires more accurate wind models.
Within the research group where the work presented in this thesis was conducted,
a colleague completing a Masters thesis conducted an experimental investigation into
Urban Airflow considering UAM [5]. This involved performing wind tunnel tests with
a scale model of a urban centre, Toronto was used. The flow approaching the city was
adjusted to match the velocity and turbulence profiles that are associated with wind
flowing past areas with different surface roughness. This was done to match the wind
directions on the model with the correct wind profile due to the upwind building and
terrain roughness. Data from these experiments were used to synthesize the urban
airflow disturbance used in the current research.
 6

2.2 Flight Dynamics and Control of an Urban Air

Taxi/UAM Vehicle

Guidance and navigation of aerial systems in urban environments is one of the sev-
eral challenges that is being addressed by researchers to achieve safe and sustainable
UAM. Yang and Wei proposed a computational guidance algorithm with collision
avoidance capabilities for use in UAM applications [8]. The problem was formu-
lated using a Markov Decision Process and solved using Monte Carlo Tree Search
algorithm. Through numerical experiments, it was demonstrated that this method
shows promising performance in path planning while avoiding collisions in a rela-
tively dense airspace. Similarly, Cotton and Wing examined applying a decentralized
air traffic management system for Unmanned Aircraft Systems [9] operating in the
urban environment. The future of UAM may involve thousands of different sized air-
craft operating in the urban environment, this presents an unprecedented challenge
to current air-traffic management techniques. In their research, they proposed using
Airborne Trajectory Management (ABTM) as a solution to address the dense oper-
ations encountered in UAM applications by combining minimal rules-based airspace
management with on-board surveillance systems using conflict detection and reso-
lution algorithms to maintain safe separation distances. They concluded that safe
and high density UAM operations are possible if this type of air traffic management
methodology is developed along with other UAM infrastructure and recommend that
NASA and commercial Unmanned Aerial System (UAS) developers consider sim-
ulation and further analytical development of ABTM to evaluate its performance.
Developing models that capture the dynamics of the various aircraft types that will
perform UAM operations will aid in creating simulations to test aircraft management
methodologies for UAM operations.
 7

To certify and operate these vehicles safely, a thorough understanding of the ve-
hicle dynamics and control requirements are required. Most of the research in the
flight dynamics and control of aerial vehicles in urban environments has focused on
autonomous small systems [10] [11] [12] [13]. Galway et al. presented a methodol-
ogy that predicted flight performance of a UAV operating in an urban gust environ-
ment [14] where Computational Fluid Dynamics (CFD) was used to create a flow field
of an urban environment with several buildings represented as rectangular bodies. A
flight dynamics model of the Aerosonde UAV [15] was formulated and immersed in
the simulated flow field to assess the dynamics response and the flight performance
of the UAV. The investigation provided insight into the flight performance of UAVs
in the airflow fields of buildings and concluded that the characteristics of the flow
field obtained from CFD simulations are representative of the actual flow; however,
additional future validations are needed. The study recommended that future inves-
tigations should focus on investigating flight controller design to investigate control-
lability and control surface deflection in addition to establishing the wind conditions
that cause the aircraft to become uncontrollable. The work done by Raza focused
on the autonomous position control of a quadrotor platform in the presence of urban
gusts [16]. Large Eddy Simulation (LES) was used to predict the flow field around
a single building. The transient wind velocities were used to create a realistic flow
environment in which a custom-built quadrotor prototype, named TARA, was im-
mersed. This created a simulation environment to test the position-hold performance
of different control algorithms. Simulation results were verified by performing phys-
ical flight tests in the wake of a building that is similar to the one in the simulation
study. A novel position-hold technique was proposed and it was demonstrated to be
capable of maintaining a position-hold within a single body length in the presence of
urban gusts.
The ideas explored, and the conclusions reached, in these important studies have
 8

laid the groundwork for future investigations in UAM in general and the flight dy-
namics and control of UAT in particular. Su et al. has developed a flight dynamics
model of a tiltrotor aircraft with controller architectures to enable a smooth transi-
tion between vertical and horizontal flight [17]. A linear time-invariant state-space
model was created for different vehicle trim conditions related to the rotor tilt posi-
tion. These state-space models for the different trim conditions were used to create
a linear parameter-varying model. Model predictive control methodology was used
to design a flight controller to enable the transition of the aircraft between vertical
and horizontal flight modes. It was demonstrated that the controller was capable
of performing a smooth transition and it was stated that future work would include
unsteady effects in the aerodynamic loads. This work has pioneered the development
of a controller for the transition phase of Vertical Take Off and Landing (VTOL) air-
craft, which characterises most UAT designs. However, the design of this controller
made extensive use of a system model which is not always available and may not
capture all dynamic effects when subjected to complex wind conditions.

2.3 Active Disturbance Rejection Control

In the investigation by Reza [16], PID control was used for the position-hold con-
troller because PID control dominates commercial [16], and hobby, UAV systems.
PID control still remains a popular control architecture because it is fundamentally
error-driven rather than model-driven and does not require detailed models of the
system which are difficult and expensive to generate. This is confirmed by Gao who
reported that over 90% of industrial control uses PID architecture [18]. It is antici-
pated that UAM vehicles will require robust controllers that will need to achieve good
performance in the turbulent urban air environment. PID control is widely used in
real world applications but UAM will involve new flight operations with new vehicle
 9

designs in complex new environments. PID control might be reaching its useful limit,
if not passed already, with modern control engineering challenges. An evolution of
PID control that addresses its limitations and pulls from modern control develop-
ments could provide a system that is truly adoptable by industry and suitable for
control of UAT and other UAM vehicles.
Active Disturbance Rejection Control (ADRC) has been developed as the suc-
cessor to PID control [19]. It addresses several issues in classical PID control and
aims to create a controller that is better suited to handle complex and non-linear
systems [19], which are commonly found in aerospace engineering. These issues are:

ˆ The error computation;

ˆ Noise degradation in the derivative control;

ˆ Oversimplification and the loss of performance in the control law in the form of
a linear weighted sum; and

ˆ Integral control complications.

ADRC solves these issues using four distinct measures:

ˆ A simple differential equation for transient trajectory generation;

ˆ A noise-tolerant tracking differentiator;

ˆ Non-linear control laws; and

ˆ The concept and method of total disturbance estimation and rejection.

Good control performance can be achieved with this controller without a highly ac-
curate system model, also a popular feature of PID control, though if the system
model is available it can be used to improve controller performance [20]. ADRC has
 10

been shown to outperform the classical PID controller in applications spanning motor
control, attitude tracking, industrial control and control of chemical processes [21].
ADRC has been demonstrated in aerospace applications such as missile attitude con-
trol [22], morphing wing structure control [23], and control of a small unmanned
helicopter [24]. This makes ADRC a suitable candidate for future aerospace control
challenges.
In this investigation an inner-loop controller of a generic UAT, that was immersed
in an empirically-developed urban airflow model, was developed using ADRC and its
performance was compared to classical PID control.
Chapter 3

Flight Dynamics Modelling of a Generic

Urban Air Taxi

To develop the flight dynamics model of a generic UAT operating in the horizontal
cruise flight phase, one must obtain, or propose, a generic conceptual UAT design;
however, these designs are not readily available since most UAT are still in the de-
velopment phase and their designs are considered proprietary. In this investigation,
a generic UAT configuration was identified and an in-development prototype by a
world-renowned aerospace company was selected as basis for the development of a
simplified flight dynamics model. The development began with estimating the geom-
etry of the platform based on publicly available data and pictures. The inertial and
the aerodynamic properties were then estimated using known aircraft design guide-
lines and thin airfoil aerodynamics theory respectively. The aerodynamic parameters
associated with the aircraft geometry were estimated and used to generate the state-
space model for trimmed level flight. The estimated model parameters were then
used in conjunction with classical aircraft flight dynamics theory in trimmed cruise
condition to develop a representative state-space model that was used to assess the
dynamics response of the UAT when flying in an urban airflow environment and to
develop appropriate vehicle attitude stabilization control laws.

11
 12

3.1 Generic Urban Air Taxi Design Choice

The Bell Nexus 4EX is a proposed UAT being developed by Bell Textron. A concep-
tual depiction of this platform is shown in Figure 3.1. This platform was chosen as
the basis for developing the flight dynamics model in this investigation since it has
many features common to other proposed UAT platforms and it is being developed
by an aerospace company that has an established footprint in the VTOL and the
rotorcraft market.

Figure 3.1: Bell Nexus 4EX [25]

An in-service aircraft that was developed by Bell Textron and shares many VTOL
characteristics with their proposed Bell Nexus 4EX UAT is the V-22 shown in Figure
3.2 and its experimental predecessor the X-22 shown in Figure 3.3. The V-22 is a
military tilt-rotor vehicle developed in partnership by Bell and Boeing and is used for
special operations and air assault as well as passenger and cargo transportation [26].
While the X-22 is a V/STOL experimental aircraft developed by Bell for the United
States Navy during the 1960’s and into the 1980’s [27]. The success of the V-22
demonstrates the expertise of Bell Textron in VTOL and further justifies the selection
of their UAT platform, the Nexus 4EX, as the platform adopted in this investigation.
 13

Figure 3.2: Bell Boeing V-22 Osprey [26]

Figure 3.3: Bell X-22; VSTOL test platform for US Navy [27]
 14

The Nexus 4EX shows a very similar design layout to the X-22, which demonstrated
both vertical and horizontal flight modes during its development. It is assumed that
the experience gathered in developing the X-22 was used to inform the design and
the development of the Nexus 4EX. This reinforces the belief that the UAT design
proposed by Bell Textron will potentially see real world applications in the near future.
Though other designs exist, such as the UATs in development by Joby Aviation [28],
Volocopter [29] and Airbus [30], the fact that Bell has proven commercial VTOL
experience in market success, naturally renders their design a sensible choice to use
as a basis for a generic UAT design.
The Nexus 4EX configuration consists of four ducted fans that can articulate to
produce lift for vertical take-off or landing and thrust for horizontal flight. It uses
fixed lifting surfaces in the form of the main wing and the vertical stabilizer. Within
the ducts, a horizontal surface produces lift in the horizontal flight phase. Extending
outboard from the aft ducted fans are additional wing tip sections for added lift.
Longitudinal stabilization is accomplished with a canard configuration where the
forward horizontal surface is contained within the forward ducted fans. There are
several movable trailing edge surfaces that can be used for control and stabilization
in horizontal flight. Flaps on the main fixed wing, a rudder on the vertical stabilizer
and adjustable trailing edge surfaces in the ducted fans. The movable trailing edge
surfaces in the aft ducted fans function as ailerons while the surfaces in the forward
ducted fans act as an elevator. The trailing edge surfaces on the main fixed wing
can be seen in Figure 3.4 while the trailing edge surfaces in the ducted fans and
on the rudder can be seen in Figure 3.5. For control of the 4EX, thrust vectoring
through duct rotation and differential thrust might be used but in this investigation
only the elevator, rudder and ailerons are used for control. In addition, the ducts are
expected to add to the total lift produced but these additional lift and drag effects
were neglected in this investigation.
 15

Figure 3.4: Top view of Bell Nexus 4EX [25]

Figure 3.5: Side view of Bell Nexus 4EX [31]

 16

3.2 Geometric and Inertial Parameter Estimation

The geometric parameters, which were essential for later estimating the aerodynamics
parameters of the platform, were estimated using available images of the Bell Nexus
4EX. A simple mass distribution model was used to estimate the inertial parameters
where the fuselage was considered as a uniform mass distribution and point masses
were used at the ducted fan locations to account for the mass in the motors, wing
and canard. Available group weight data for the X-22 presented in [32] was used with
this simple mass distribution model to estimate the inertial parameters.

3.2.1 Geometric Parameter Estimation

A simplified generic UAT design that was suitable for developing the linearized flight
dynamics model was created based of the estimated geometric and inertial properties
of the Bell Nexus 4EX. To facilitate this process, a simple layout was produced to
identify geometric values, which is shown in Figure 3.6. The pictures shown in Figures
3.4 to 3.8 were used to extract an estimate of the Bell Nexus 4EX geometry given a
reported duct diameter of 8 feet [33] (2.4 m), which was used as a scaling parameter.
As shown in Figure 3.6, the wing is composed of three distinct segments, labelled
as S1, S2 and S3 respectively. S1 is the fixed portion of the wing, S2 is the portion of
the wing within the duct and S3 is the wing tip section that extends outboard from
the duct. The span, root chord and tip chord were estimated for each segment. These
estimates are listed in Table 3.1.
Using the wing span and chord estimates the mean aerodynamic chord was cal-
culated using Equation 3.1 [34]

Z b/2
2
c̄ = c2 dy (3.1)
S 0

where c̄ is the mean aerodynamic chord, S is the wing planform area, b is the total
 17

Table 3.1: Wing section span and chord estimates for the generic UAT

Wing Section Section Span [m] Root Chord [m] Tip Chord [m]
S1 3.117 1.33 1.163
S2 2.438 0.451 0.451
S3 0.6096 0.451 0.281

Figure 3.6: Top view of generic UAT layout

 18

wing span and c is the chord length at a distance y from the aircraft centre line.
Using the formula for the area of a right trapezoid in Equation 3.2, the area of each
of the wing segments, S1, S2 and S3 were calculated and are presented in Table 3.2.

(croot + ctip )b∗

Sx = (3.2)
2

In Equation 3.2, croot and ctip are the wing root and tip chord lengths, b∗ is the span
of an individual segment and Sx is the area of the particular section. The planform
area for segment S2 was calculated based the area of a rectangle using Equation 3.3.

Sx = cb∗ (3.3)

The geometric parameters for the main wing are presented in Table 3.2.

Table 3.2: Wing geometric parameters

S1 3.88 m2
S2 1.01 m2
S3 0.22 m2
b 12.33 m
c̄ 1.08 m
S 10.23 m2
(AR)wb 11.42
Thickness Ratio (t/c) 0.12

The canard consists of the two forward ducted fan sections; therefore, the canard
geometric parameters were taken to be similar to those of wing segment S2. The
canard geometric parameters are presented in Table 3.3. For both the main wing
and the canard, the width of the aircraft fuselage where the wing/canard attaches to
it was included in the span estimations. This was done to account for the effect of
 19

wing-body interaction on the lift coefficient [35].

Table 3.3: Canard geometric parameters

bc 6.44 m
c̄c 0.45 m
Sc 2.90 m2
(AR)c 14.27
Thickness Ratio (t/c) 0.12

Figure 3.7: Side view of Bell Nexus 4EX [36]

The vertical tail geometric parameters were estimated using Figure 3.7. The loca-
tion of the aerodynamic centre of the vertical tail was required to calculate the rolling
moment. This location was measured from the leading-edge point where the vertical
tail joins the fuselage. In the x-direction (stream-wise direction) the aerodynamic cen-
tre is taken at the quarter chord point location. In the negative z-direction (vertical
tail span-wise direction), the aerodynamic centre, ȳ, was estimated using [34]:
 20

Figure 3.8: Front view of Bell Nexus 4EX [36]

1 + 2λ
ȳ = b (3.4)
3(1 + λ)

where λ is the taper ratio. The vertical tail geometric parameters are listed in Table
3.4.

Table 3.4: Vertical tail geometric parameters

Span (bF ) 1.76 m

Mean Aerodynamic Chord (c̄F ) 1.12 m
Planform Area (SF ) 1.99 m2
Aspect Ratio (ARF ) 1.47
Thickness Ratio (t/c) 0.12
Taper Ratio (λF ) 0.43
Leading Edge Sweep Angle (ΛLEF ) 47.6°
x̄ 0.30 m
ȳ 0.76 m

Certain geometric parameters for the elevator, rudder and ailerons were required
to estimate the control effectiveness of these surfaces. The control effectiveness for a
given control surface deflection was calculated using estimation methods provided in
Appendix B of Reference [34]. This appendix provides data to estimate the stability
 21

and control derivatives for rigid subsonic flight. The information is derived from data
in the USAF Datcom and data sheets of the Royal Aeronautical Society of Great
Britain. Using Equation 3.5 in conjunction with the graphs shown in Figures 3.9 and
3.10, the control surface effectiveness, CLδ , for each movable surface was estimated.

 
CLα
CLδ = Clδ K1 K2 (3.5)
Clα

In Equation 3.5, Clδ is the change in two dimensional lift coefficient due to control
surface deflection, CLα is the three dimensional lift-curve slope of the entire aircraft
and Clα is the two dimensional lift-curve slope of the entire aircraft. The parameter
K1 , presented in Figure 3.9, represents the effect on control effectiveness due to the
percent of the wing chord taken up by the trailing edge surface. The parameter
K2 , presented in Figure 3.10, represents the effect on control effectiveness due to the
percent of the wing span taken up by the trailing edge surface. The flap chord ratio,
cf /c, is the ratio of the moveable surface chord to the chord of the entire surface
and was used for estimating K1 . The parameters ηi and ηo are the distances to the
inboard and outboard edges of the movable surface, as shown in image of the wing
and on the plot in the top left and top right of Figure 3.10 respectively. These control
surface geometric parameters for the flaps, elevator, rudder and ailerons are listed in
Table 3.5.

3.2.2 Inertial Parameter Estimation

To estimate the inertial properties for the generic UAT, a simple rectangular prism was
used to represent the aircraft fuselage and point masses at the locations of the ducted
fans were used to account for the weight of the wing, motors and other components.
The height, width, and length were based off the body of the Nexus 4EX. This
simple representation is shown in Figure 3.11. The mass moment of inertia about the
 22

Figure 3.9: Flap chord factor. Reproduced from Reference [34]

principle axes of this rectangular prism were calculated using equations 3.6, 3.7 and
3.8.

1
Ix = m(H2 + W2 ) (3.6)
12

1
Iy = m(H2 + L2 ) (3.7)
12

1
Iz = m(W2 + L2 ) (3.8)
12

The height, H, width, W, and length, L, estimated using Figures 3.7 and 3.8, were
 23

Figure 3.10: Flap span factor. Reproduced from Reference [34]

found to be 1.868 m, 1.564 m and 8.431 m respectively. The mass fraction method
presented by Roskam [32] was used to estimate the weight of certain distributed
components based on their weight as a fraction of the total aircraft weight.
In Reference [32], the group weight data of several NASA X-Planes are provided.
The group weight data take components within certain systems and groups them to
provide a mass fraction for that group of components. The group weight data in
Table A13.4b in Reference [32] provide mass fractions of several major component
groups for a version of the X-22, the X-22a. The X-22a uses a turbo-shaft system
to power the ducted fans while the Nexus 4EX uses electric motors and distributed
 24

Table 3.5: Control surface geometric parameters

Flap Parameters
cf /c 0.273
ni 0.136
no 0.532
Elevator Parameters
cf /c 0.48
ni 0.243
no 1
Rudder Parameters
cf /c 0.305
ni 0.09
no 0.874
Aileron Parameters
cf /c 0.48
ni 0.506
no 0.901

batteries. Although the mass of the components and their distribution are likely dif-
ferent for each of these vehicles, their close similarities in configuration render the data
provided a suitable estimation for the inertial parameter estimation. Errors in this
estimation would ultimately affect the values of the eigenvalues which describe the
dynamic modes of the aircraft. The estimated dynamic modes were assessed against
the stability requirements for certified aircraft so that any errors in the estimation
procedure could be accounted for by slightly altering the dynamic modes eigenvalues
to ensure a characteristic response from the model. For the moment of inertia esti-
mations, the mass fractions for the power plant group and wing group were used to
estimate point mass values, which were placed at the centre of each of the ducted
 25

Figure 3.11: Box body

fans. These fractions are 0.35 and 0.054 respectively. The moment of inertia, I, of a
point mass at a perpendicular distance from the axis of rotation was calculated using

I = mr2 (3.9)

where r is the shortest distance to the axis of interest. The mass associated with
the point masses was removed from the mass used in the rectangular prism body.
The moment of inertia about each axis due to the point masses were added to the
principal axis moments of inertia for the rectangular prism. The total moments of
inertia about the principle axes are presented in Table 3.6.
 26

Table 3.6: Mass moments of inertia about the principal axes

Ixp 12.33 kg · m2
Iyp 1.08 kg · m2
Izp 10.23 kg · m2

3.3 Aerodynamic Parameter Estimation

In this section, the vehicle aerodynamic parameters and coefficients which were needed
to form the linearized flight dynamics model are estimated based on potential flow
and thin airfoil aerodynamics theories. The lift-curve slope and control effectiveness
coefficients were estimated for the main lifting and control surfaces. A longitudinal
static analysis was conducted to determine certain aerodynamic reference points of
the aircraft in a trimmed condition and to estimate the static longitudinal stability
and control derivatives. The longitudinal and the lateral stability derivatives that
are not calculated using potential flow and thin airfoil aerodynamics theories were
estimated based on the data provided in Appendix B of Reference [34]. These non-
dimensional quantities were converted to dimensional quantities for the state-space
model.

3.3.1 Lift-Curve Slope Coefficient Estimation

Based on Reference [37] and the geometric parameters shown in Figure 3.12, the two-
dimensional lift-curve slope, a∞ , of an infinite wing with a finite thickness of t, chord
length of c and leading edge sweep angle ΛLE is given by Equation 3.10 such as,

 
t
a∞ = 1.8π 1 + 0.8 cos ΛLE (3.10)
c

For a finite lifting/control surface with an aspect ratio of AR, the lift-curve slope, a,
 27

Figure 3.12: Wing lift geometric parameters

is modified as in Equation 3.11 such as,

a∞
a= a∞

(3.11)
1 + πAR

Applying Equations 3.10 and 3.11 to the wing, canard and the vertical tail yields the
2D and 3D lift curve slopes presented in Table 3.7. The 3D lift-curve slope values for
the wing-body, the canard, and the vertical tail are represented with awb , ac and aF
respectively.

Table 3.7: Lift-curve slope coefficients

Wing-Body
(a∞ )wb 6.198 1/rad
awb 5.285 1/rad
Canard
(a∞ )c 6.198 1/rad
ac 5.445 1/rad
Vertical Tail
(a∞ )F 4.167 1/rad
aF 2.192 1/rad
 28

The control effectiveness derivatives were calculated using Equation 3.5 and the co-
efficients obtained through Figures 3.9 and 3.10 in combination with the estimated
control surface geometric parameters. The control effectiveness coefficients are listed
in Table 3.8 where af , ae , ar and aa are the control effectiveness coefficients for the
flaps, elevator, rudder and ailerons respectively.

Table 3.8: Control effectiveness coefficients

af 2.105 1/rad
ae 3.970 1/rad
ar 4.253 1/rad
aa 1.930 1/rad

3.3.2 Longitudinal Static Analysis

Figure 3.13: Generic UAT layout side view

A static longitudinal analysis was conducted to determine the static longitudinal

stability derivatives and aircraft reference trim conditions. This also provided the
required reference configuration for the elevator and flaps that achieved the steady
 29

state, horizontal cruise condition for the aircraft. The analysis followed the procedure
presented in Reference [34].
A side view of the generic UAT layout is shown in Fig. 3.13. The lift coefficient
of the wing-body, CLwb , consists of the part due to the wing-body angle of attack and
the part due to the flap control surface deflection as shown in Equation 3.12,

CLwb = awb αwb + af δf (3.12)

where αwb is the wing-body angle of attack and δf is the flap deflection angle. The lift
coefficient of the canard, CLc , consists of the part due to the canard angle of attack
and the part due to elevator control surface deflection as shown in Equation 3.13,

CLc = ac αwb + ae δe (3.13)

where δe is the elevator deflection angle. Although the canard lies within the forward
ducted fan, which complicates the local airflow surrounding it and part of the airflow
approaching the main wing, the simplifying assumption of the canard local angle of
attack being similar to the wing-body combination was employed.
The total lift coefficient for the aircraft, CL , was calculated using Equation 3.14,
where the canard lift coefficient was modified by the canard wing-body area ratio to
account for the different non-dimensionalization of these parameters.

Sc
CL = CLwb + CL (3.14)
S c

The overall aircraft lift coefficient can then be written in terms of the control surface
deflections and the wing-body angle of attack as shown in Equation 3.15,

Sc Sc
CL = af δf + ae δe + (awb + ac )αwb (3.15)
S S
 30

Equation 3.15 can be rewritten as,

CL = (CL )0 + aαwb (3.16)

where (CL )0 is the lift coefficient when αwb is zero and a is the lift-curve slope for the
entire aircraft, written as

 
ac Sc
a = awb 1 + (3.17)
awb S

A new angle of attack is defined, α, which is measured from the zero lift line for the
entire aircraft. The aircraft lift coefficient is redefined as

CL = aα (3.18)

By equating Equations 3.16 and 3.18, α and αwb can be related using

af ae Sc
α − αwb = δf + δe (3.19)
a aS i

Equation 3.19 introduces δei as the initial elevator deflection. The initial value of
δei was taken as zero and the elevator angle needed to maintain the aircraft trim
condition was then determined.
The overall pitching moment coefficient, Cm , about the aircraft centre of gravity
was calculated using

Cm = Cmacwb + Cmp + CL (h − hnwb ) + V̄H CLc (3.20)

where Cmacwb is the portion of the pitching moment coefficient associated with the
pitching moment about the wing-body aerodynamic centre, Cmp is the pitching mo-
ment coefficient due to the thrust and V̄H is the canard horizontal tail volume ratio,
defined as,
 31

¯lc Sc
V̄H = (3.21)
c̄S

where ¯lc is the distance between the canard and wing-body aerodynamic centres.
The propulsion pitching moment was considered constant with respect to changes in
angle of attack and is therefore neglected. The constant pitching moment was much
smaller than the other aerodynamic coefficients and was also neglected. The pitching
moment coefficient was written in terms of control surface deflections and aircraft
angle of attack as shown in Equation 3.22,

 
ac Sc ac
Cm = Cmacwb +Cmp + V̄H ae 1 − δei + V̄H af δf +a(h−hnwb )α+ V̄H ac α (3.22)
aS a

Equation 3.22 can be rewritten in the form,

Cm = Cm 0 + Cm α α (3.23)

where Cm0 is given by,

 
ac Sc ac
Cm0 = Cmacwb + Cmp + V̄H ae 1− δei − V̄H af δf (3.24)
aS a

and Cmα is written as,

Cmα = a(h − hn ) (3.25)

where hn is the location of the neutral point relative to the wing leading edge as
shown in Figure 3.13. The location of this point relative to the centre of gravity
indicates the static stability in the longitudinal direction of the aircraft. The aircraft
is statically stable in the longitudinal direction when the centre of gravity is ahead
 32

of the neutral point. This yields a negative value for Cmα and the aircraft is said to
have positive pitch stiffness. The parameter h is the location of the centre of gravity
measured from the leading edge of the main fixed wing, which is positive moving
toward the tail, and reported as a percentage of the mean aerodynamic chord.
The longitudinal location of the neutral point relative to the wing leading edge
was calculated using,

ac
hn = hnwb − V̄H (3.26)
a

where hnwb is the location of the wing-body neutral point relative to the wing leading
edge. The neutral point for a wing is located 25% of the wing chord back from the
wing leading edge [34]. When the wing is combined with a fuselage to create a wing-
body, the neutral point shifts forward producing the wing-body neutral point. The
shift in the wing neutral point due to the wing-body combination is determined using
Figure 3.14. The static margin, Kn , can be used to gauge the pitch stiffness of the
aircraft or impose limits on the location of the centre of gravity [34], it is defined with
Equation 3.27.

Kn = hn − h (3.27)

Generally, the centre of gravity location of an aircraft is fixed. Given a fixed

static margin, the angle of attack, or aircraft trim condition, can be changed through
deflection of the elevator. To determine the control surface deflections and the trim
condition angle of attack for level flight, the change in the overall aircraft lift force and
moment due to elevator deflection was required. The total lift coefficient including
the change in lift due deflection of the elevator was calculated using Equation 3.28

CL = CLα α + CLδe δe (3.28)

 33

Figure 3.14: Effect of a fuselage or nacelle on neutral-point position. Reproduced

from Reference [34]

The total pitching moment coefficient including the change in pitching moment due
to deflection of the elevator was calculated using Equation 3.29

Cm = Cm0 + Cmα α + Cmδe δe (3.29)

where Cmδe and CLδe are the change in pitching moment and lift coefficients for the
aircraft with respect to elevator deflection. CLδe and Cmδe were calculated using
Equations 3.30 and 3.31 such as,
 34

Sc
CLδe = ae (3.30)
S

Cmδe = ae V̄H − CLδe (h − hn ) (3.31)

The reference conditions were determined by recognizing that when the aircraft
is in constant level flight, the lift force is equal to the aircraft weight force and
the pitching moment about the aircraft centre of gravity is zero. The trimmed lift
coefficient, CLtrim , for stable horizontal level flight was calculated using equation
3.32 [34]

W
CLtrim = 1 (3.32)
ρ V 2S
2 0

where W is the aircraft weight force. Rewriting Equations 3.28 and 3.29 in a matrix
form as shown in Equation 3.33, then solving for the trim angle of attack and elevator
deflection resulted in Equations 3.34 and 3.35,

    
 CLα CLδe  αtrim  CLtrim 
= (3.33)
    
  
    
Cmα Cmδe δetrim −Cm0

CLtrim Cmδe + CLδe Cm0

αtrim = (3.34)
det

−CLα Cm0 + CLtrim Cmα

δetrim = (3.35)
det

where det is given by

det = CLα Cmδe − CLδe Cm0 (3.36)

 35

The calculated reference conditions are presented in Table 3.9. The air density cor-
responds to the air density at approximately 1000 feet in altitude at 20°C.

Table 3.9: Horizontal flight trim conditions

Airspeed (V ) 130 KTS (150 mph [33])

Air Density (ρ) 1.19 kg/m3
Aircraft Mass (m) 3175 kg [33]
Elevator Angle (δe ) 5.77°
Flap Deflection Angle (δf ) 5.00°
α 8.70°
αwb 7.16°
h -1.435
hn -1.285
Kn 0.15

3.3.3 Estimation of Stability Derivatives

The coefficients that describe how the aerodynamic forces and moments on the aircraft
change with respect to the state of the aircraft are referred to as the aerodynamic
derivatives or stability derivatives [34]. These coefficients are specific to the aircraft
design and configuration and are difficult to determine. Advanced methods such
as CFD and wind tunnel testing are required to obtain accurate estimates of these
derivatives. In Reference [34], semi-empirical methods to estimate several of the
stability derivative are provided, these are based on USAF Datcom and data sheets
from the Royal Aeronautical Society of Great Britain. Those methods along with
analytic calculations were used to estimate the stability derivatives.
The stability derivatives can be separated into those which affect longitudinal
and those which affect lateral motion. The longitudinal and the lateral stability
 36

derivatives of relevance are shown in Tables 3.10 and 3.11 respectively. The forces
and the moment, in the x and z directions and about the aircraft y axis respectively,
that affect the aircraft motion in the longitudinal direction are described by the non-
dimensional coefficients Cx , Cz , and Cm respectively. Each of these coefficients are
affected by the longitudinal motion of the vehicle, which include the forward speed
of the vehicle u, the angle of attack α, the pitch rate q, and the rate of change of the
angle of attack α̇. The lateral direction force and moments, in the y direction and
about the x and z axes respectively, are described by the non-dimensional coefficients
Cy , Cl and Cn . Each of these coefficients are affected by the lateral motion of the
vehicle, which include, the side slip angle β, roll rate p, and yaw rate r.
Table 3.10: Longitudinal non-dimensional stability derivatives

Cx Cz Cm
u C xu C zu Cmu
α C xα C zα Cmα
q Cxq C zq Cm q
α̇ Cxα̇ Czα̇ Cmα̇

Table 3.11: Lateral non-dimensional stability derivatives

Cy Cl Cn
β C yβ Cl β Cnβ
p C yp Clp Cnp
r C yr Cl r Cnr

Non-Dimensional Stability Derivatives

This section provides the equations and tables that were used to estimate the non-
dimensional stability derivatives for the generic urban air taxi in a horizontal cruise
 37

configuration.
The alpha derivatives, Cmα , Cxα , and Czα , which represent the change in pitching
moment and the forward and vertical force coefficients, were calculated using Equation
3.25, from the longitudinal analysis, and Equations 3.37 and 3.38 [34] such as,

2CL0
Cxα = CL0 − CL (3.37)
πARe α

Czα = −CLα (3.38)

where CL0 is the lift at the reference condition and is equal to CLtrim while e is the
Oswald efficiency factor, taken as 0.75 [38]. The estimates of the angle of attack, α,
derivatives are shown in Table 3.12.

Table 3.12: Alpha derivatives

CXα 0.5665
CZα -6.8307
Cmα -1.0246

The force in the forward direction and vertical direction and the pitching with
respect to the forward speed, u, derivatives, CXu , CZu and Cmu , were calculated using
Equations 3.39, 3.40 and 3.41 [34] such as,

     
∂CT ∂CD ∂CD ∂CD
CXu = M0 − + ρu20 + CT u 1− (3.39)
∂M ∂M 0 ∂pd 0 ∂CT 0

     
∂CL ∂CL ∂CL
CZu = −M0 − ρu20 − CTu (3.40)
∂M 0 ∂pd 0 ∂CT 0
 38

     
∂Cm ∂Cm ∂Cm
Cmu = M0 + ρu20 + CTu (3.41)
∂M 0 ∂pd 0 ∂CT 0

where M0 is the Mach number at the reference condition, ρ is the air density, and u0
is the reference forward speed. ∂C − ∂C

∂CL

, ∂M 0 and ∂C


∂M
T D
∂M 0
m
∂M 0
represent compress-
     
ibility effects, ∂C D
, ∂CdL and ∂C m
represent the effects due to aeroelasticity,
   ∂pd 0 ∂p  0 ∂pd
0
1 − ∂C D
∂CT
∂CL
, ∂C T
and ∂C m
∂CT
represent the change in drag, lift and moment coef-
0 0 0
ficients with respect to the thrust coefficient. The aircraft was considered rigid so the
aeroelastic effects are neglected and the compressibility effects were also neglected.
The change in drag, lift, and moment coefficients due to change in thrust coefficient
were neglected as thrust control is not considered in this model. The thrust coeffi-
cient with respect to forward speed u, CTu , for constant-power propulsion systems,
was estimated with Equations 3.42 and 3.43 [34] such as,

CTu = −3CT0 (3.42)

CT0 = CD0 + CW0 sin θ0 (3.43)

where CD0 is the drag coefficient at the reference condition, θ0 is pitch angle at the
trim condition and CW0 is the weight force coefficient at the reference condition and
is equal to the lift coefficient at the trim condition, calculated with Equation 3.32.
The estimates of the forward speed, u, derivatives are shown in Table 3.13.

Table 3.13: Forward speed derivatives

CXu -0.2978
CZu 0
Cmu 0
 39

The longitudinal stability derivatives with respect to pitch rate, q, for the wing,


CZq wing and (Cmq )wing , were calculated using Equations 3.44 and 3.45 [34] while
the pitch rate stability derivatives for the canard, (CZq )canard and (Cmq )canard , were
calculated using Equations 3.46 and 3.47 [34] such as,

CZq wing
= −CLq = 2CLα (h − h0 ) (3.44)

(Cmq )wing = C̄mq − 2CLα (h − h̄)2 (3.45)

(CZq )canard = 2ac VH (3.46)

lc
(Cmq )canard = −2ac VH (3.47)
c̄

where h0 , h̄ and C̄mq for subsonic flight are typically 0.75, 0.5 and 0 respectively [34];
VH is the horizontal tail volume ratio with respect to the centre of gravity location,
which was calculated using Equation 3.48 and can be related to V̄H through Equation
3.49; lc is the distance between the centre of gravity and the canard aerodynamic
centre and it is related to l¯c through Equation 3.50 as all presented below,

lc Sc
VH = (3.48)
c̄S

Sc
VH = V̄H + (h − hnwb ) (3.49)
S

lc = l¯c + c̄(h − hnwb ) (3.50)

The estimates of pitching moment and vertical force with respect to the pitch rate, q,
 40

derivatives are shown in Table 3.14 where the forward force with respect to the pitch
rate derivative is considered zero.

Table 3.14: Pitch rate derivatives

CZq -25.7177
Cm q -56.7287

The pitching moment, forward and vertical force with respect to α̇ derivatives
represent the effect of unsteady flow on the aircraft stability derivatives. The The-
ordorsen function, G(k) [34], is used to determine the unsteady effect on lift in the
context of thin airfoil theory [34]. The change in aircraft lift and moment due to
unsteady flow, CLα̇ and Cmα̇ , can be calculated using Equations 3.51 and 3.52 [34]
such as,

G(k)
CLα̇ = π + 2π (3.51)
k

   
1 G(k) 1
Cmα̇ =π h− + 2π h− (3.52)
2 k 4

Due the high speed of the aircraft relative to the wind speed fluctuations, these
unsteady effects were neglected in this investigation.
The β, or the side slip angle, stability derivatives, Cyβ , Clβ and Cnβ , represent
the change in lateral force, rolling moment and yawing moment coefficients during
a side slip condition. The β stability derivatives for side force, rolling moment, and
yawing moment coefficients due to the vertical tail were calculated using Equations
3.53, 3.54, and 3.55 [34],

 

∂σ SF
C yβ tail
= −aF 1− (3.53)
∂β S
 41

 

∂σ SF zF
Clβ tail
= −aF 1 − (3.54)
∂β Sb

 

∂σ
Cn β tail
= Vv aF 1− (3.55)
∂β

where zF is the vertical distance between the centre of gravity and fin aerodynamic
centre and Vv is the vertical tail volume ratio and is given by Equation 3.56 [34],
where lF is the horizontal distance between the centre of gravity and the vertical tail
∂σ
aerodynamic centre; side wash effects, represented by the term ∂β
, are neglected. The
estimates of the side slip, β, derivatives are shown in Table 3.15.

SF lF
Vv = (3.56)
Sb

Table 3.15: Side slip derivatives

Cyβ -0.4264
Clβ -0.0587
Cn β 0.1623

The roll rate derivatives for the fin, Cyp tail
and Cnp tail
, which affect the side
force and yawing moment coefficients, were calculated using Equations 3.57 and 3.58
[34] respectively. The effect of roll rate on the rolling and yawing moment coefficients
due to the main wing, Clp and Cnp , were calculated using Equations 3.59 and 3.60 [34]
such as,

SF
(Cyp )tail = −aF (3.57)
S

 

zF ∂σ
Cnp tail
= aF V v 2 − (3.58)
b ∂β
 42

Cl p Γ
   
βClp κ

Clp =
+ ∆Clp Drag (3.59)
κ CL =0 β Clp Γ=0

 
Cnp
Cnp = CL (3.60)
CL CL =0
M

where κ and β̂ were determined from Equations 3.61 and 3.62 such as,

β̂Clα
κ= (3.61)
2π

√
β̂ = 1 − M2 (3.62)

C
where M is the Mach number, Clα is the 2D lift-curve slope coefficient and ( CnLp )CL =0
M
in equation 3.60 was calculated using Equation 3.63 [34] such as,

  
Cn AR + 4 cos Λc/4 ARB + 0 Cn p
( p )CL =0 = ( )C =0 (3.63)
CL M ARB + 4 cos Λc/4 AR + 0 CL ML=0

where Λc/4 is the sweep back angle at the quarter chord location, B was calculated
C
with Equation 3.64 and ( CnLp )CL =0 was calculated using Equation 3.65 such as
M =0

q
2
B = 1 − M 2 cos Λc/4 (3.64)

Cn p 1 AR + 0
( )CL =0 = − (3.65)
CL M =0 6 AR + 4 cos Λc/4
(Clp )Γ
The value of (Clp )Γ=0
in Equation 3.59 was determined using the plots shown in Figures
3.15, 3.16 and 3.17 where each plot corresponds to a different taper ratio, λ, of 0.25,
0.5 and 1 respectively. The estimates of the coefficients for the roll rate, p, derivatives
are given in Table 3.16.
 43

Table 3.16: Roll rate derivatives

C yp -0.1173
Cl p -1.1936
Cnp -0.3203

Figure 3.15: Roll damping part 1. Reproduced from Reference [34]

The yaw rate, r, derivatives were estimated as follows; the side force, rolling
moment, and yawing moment coefficients due to the vertical tail, (Cyr )tail , (Clr )tail
and (Cnr )tail , were calculated using Equations 3.66, 3.67 and 3.68 [34] such as,

 
SF lF ∂σ
(Cyr )tail = aF 2 + (3.66)
S b ∂r̂

 
SF zF lF ∂σ
(Clr )tail = aF 2 + (3.67)
S b b ∂r̂
 44

Figure 3.16: Roll damping part 2. Reproduced from Reference [34]

 
lF ∂σ
(Cnr )tail = −aF Vv 2 + (3.68)
b ∂r̂
∂σ
where ∂ r̂
represents the effect on side wash due to non-dimensional yaw rate, r̂, this
effect was neglected. The wing is also capable of producing a rolling moment due to
the yaw rate. The wing and canard rolling moment coefficients due to yaw rate, Clr ,
were calculated using equation 3.69 [34]

     
Cl r ∆Clr ∆Clr
Cl r = CL + Γ+ θtwist (3.69)
CL CL =0 Γ θtwist
M

where Γ is the dihedral angle and θtwist is the wing twist angle, both are taken as
 45

Figure 3.17: Roll damping part 3. Reproduced from Reference [34]

 
Cl r
zero. CL CL =0
was calculated from Equation 3.70 [34],
M

A(1−B 2 )
1+
   
Clr 2B(AB+2) Cl r
= (3.70)
CL CL =0 1+0 CL CL =0
M M =0
 
Clr
and CL CL =0
was taken from the plot shown in Figure 3.18. Figure 3.18 uses
M =0
the aspect ratio, taper ratio and sweep back angle at the quarter chord to estimate
 
Cl r
CL C =0
. To use the figure, start from the aspect ratio and follow a vertical line
L
M =0
until the correct taper ratio curve is reached, then move horizontally left until the
corresponding sweep back angle line is reached, follow a vertical line downward to the
 46

x axis to yield the estimated value. The yaw rate derivatives are given in Table 3.17.

Table 3.17: Yaw rate derivatives

Cyr 0.3247
Clr 0.6183
Cnr -0.1236

Figure 3.18: Wing yawing derivative parameter. Reproduced from Reference [34]

Non-Dimensional Control Derivatives

The control derivatives were used to determine the effect of control inputs on the
vehicle motion. The longitudinal control derivatives for the change in aircraft lift,
 47

CLδe , and moment, Cmδe , coefficients due to elevator deflection were presented in
Equations 3.30 and 3.31. The change in the vertical, z direction, force coefficient,
CZδe , due to elevator deflection was calculated using Equation 3.71 [34],

CZδe = −CLδe (3.71)

the lateral control derivative for the aileron input, Clδa , is presented in Equation 3.72,

ya
Clδa = aa (3.72)
b

where ya is the distance from the centre of gravity to the aileron centre.
The lateral control derivatives for the rudder input, Cyδr , Clδr and Cnδr , are pre-
sented in Equations 3.73, 3.74, and 3.75 [34] such as,

 
∂σ SF
Cyδr = −ar 1− (3.73)
∂β S

 
∂σ SF zF
Clδr = −ar 1− (3.74)
∂β Sb

 
∂σ
Cnδr = Vv ar 1− (3.75)
∂β

The estimates of the control derivatives are shown in Table 3.18.

Table 3.18: Control derivatives

CZδe -1.127
Clδa 0.6788
Cyδr -0.8273
Clδr -0.1138
Cnδr 0.3150
 48

Dimensional Derivatives

The estimated non-dimensional stability derivatives are dimensionalized according to

the generic UAT dimensions to formulate the state-space model. The dimensional
parameters were calculated using the Equations presented in Tables 3.19, 3.20, 3.21
and 3.22 [34].
 49

Table 3.19: Longitudinal dimensional stability derivatives

X Z M

ρu0 SCW0 sin θ0 −ρu0 SCW0 cos θ0 1

u (3.76) (3.77) ρu0 c̄SCmu (3.78)
1 1 2
+ ρu0 SCXu + ρu0 SCZu
2 2

1 1 1
w ρu0 SCXα (3.79) ρu0 SCZα (3.80) ρu0 c̄SCmα (3.81)
2 2 2

1 1 1
q ρu0 c̄SCXq (3.82) ρu0 c̄SCZq (3.83) ρu0 c̄2 SCmq (3.84)
4 4 4

1 1 1 2
ẇ ρc̄SCXalpha
˙
(3.85) ρc̄SCZα̇ (3.86) ρc̄ SCmα̇ (3.87)
4 4 4
 50

Table 3.20: Lateral dimensional stability derivatives

Y L N

1 1 1
v ρu0 SCyβ (3.88) ρu0 bSClβ (3.89) ρu0 bSCnβ (3.90)
2 2 2

1 1 1
p ρu0 bSCyp (3.91) ρu0 b2 SClp (3.92) ρu0 b2 SCnp (3.93)
4 4 4

1 1 1
r ρu0 bSCyr (3.94) ρu0 b2 SClr (3.95) ρu0 b2 SCnr (3.96)
4 4 4
 51

Table 3.21: Longitudinal dimensional control derivatives

X Z M

1 2 1 2 1 2
δe ρu SCXδe (3.97) ρu SCZβ (3.98) ρu Sc̄Cmβ (3.99)
2 0 2 0 2 0

Table 3.22: Lateral dimensional control derivatives

Y L N

1 2 1 2 1 2
δa ρu SCYδa (3.100) ρu SbClδa (3.101) ρu SbCnδa (3.102)
2 0 2 0 2 0

1 2 1 2 1 2
δr ρu SCYδr (3.103) ρu SbClδr (3.104) ρu SbCnδr (3.105)
2 0 2 0 2 0
 52

3.4 State-Space Model

A state-space model was used to represent the generic UAT dynamic response in the
time domain. A typical state-space model has the form,

ẋ = Ax + Bu (3.106)

where x is a vector which represents the current state variables while ẋ is a vector
which represents the rate of change with respect to time of the state variables. The
A matrix describes how the state variables evolve in time given the current vehicle
state. The B matrix represents how the state variables change given a certain control
input while u is a vector which represents the control input(s). The forms of the A
and B matrices are provided in Reference [34] based on applying Newton’s second law
to a rigid aircraft to yield a set of nonlinear equations that are then linearized about
a trimmed condition. The aircraft symmetry allows for the linearized equations to
be split into longitudinal and lateral dynamics. The aircraft longitudinal and lateral
dynamics are given in Equations 3.107, 3.108 and 3.109 [34] such as,

     
∆u̇ ∆u  ∆Xc 
     m 
     
     
 ẇ  w  ∆Zc 
     m−Zẇ 
  = ALong   +   (3.107)
     
     ∆Mc Mẇ ∆Zc 
 q̇   q   I + I (m−Z ) 
     y y ẇ 
     
     
∆θ̇ ∆θ 0
 53

 
Xu Xw
 m m
0 −g cos θo 
 
 
 Z +mu

 Zu Zw q o −mg sin θo 
 m−Zẇ m−Zẇ m−Zẇ m−Zẇ 
ALong =
  (3.108)

1 Mẇ Zu 1 Mẇ Zw 1 Mẇ (Zq +muo ) Mẇ mg sin θo 
 I [Mu + m−Z ] Iy
[Mw + m−Zẇ
] Iy
[Mq + m−Zẇ
] − Iy (m−Zẇ ) 

 y ẇ
 
 
0 0 1 0

      
Yv Yp Yr

∆Yc
 v̇   m m m
− uo g cos θo  v   m

       
            
   Lp
    
 ṗ   Lv + I ′ N + I ′
N Lr
+ I ′
N 0   p   ∆Lc + I ′ N 
   Ix′ zx v Ix′ zx p Ix′ zx r    Ix′ zx c 
(3.109)

 =   + 
            
Lp + NI ′p
   ′ ′ ′    ′
 ṙ   Izx Lv + NI ′v Lr + NI ′r  r  Izx ∆Lc + ∆N
 
Izx Izx 0  ′
c
Iz 
   z z z    
       
       
ϕ̇ 0 1 tan θo 0 ϕ 0

As previously mentioned, the equations of motion are simplified based on small-

perturbations from a reference condition, hence the state variables represent a change
from the reference condition. In Equations 3.107 and 3.109, the ∆ symbol is included
in front of certain state variables to show that value may have a non-zero reference
condition.
The inertia-like terms, Ix′ , Iz′ and Izx
′
, in Equation 3.109 are given by Equations
3.110, 3.111 and 3.112 [34] such as,

Ix′ = Ix Iz − Izx
2


/Iz (3.110)

Iz′ = Ix Iz − Izx
2


/Ix (3.111)

′ 2


Izx = Izx / Ix Iz − Izx (3.112)
 54

where Ix , Iz , and Izx are the moments and the product of inertia about the aircraft
stability axes, which were calculated using [34],

Ix = Ixp cos2 ϵ + Izp sin2 ϵ (3.113)

Iz = Ixp sin2 ϵ + Izp cos2 ϵ (3.114)

1

Izx = Izp − Ixp sin 2ϵ (3.115)
2

where Ixp and Izp are the moments of inertia about the principal axes and ϵ is taken
as the wing-body angle of attack at trim condition. The final estimation of the
longitudinal and lateral A and B matrices are shown in Appendix C.
Figures 3.19 and 3.20 show the positive direction of the aircraft variables. Figure
3.19 shows the positive directions for the forces, moments, airspeed and rotation rates.
Figure 3.20 shows the positive directions for the angle of attack, α, and side slip angle,
β.

Figure 3.19: Notation for body axes, angular and linear displacements, rates, forces
and moments. Reproduced from reference [34]
 55

Figure 3.20: (a) Definition of α, (b) Definition of β. Reproduced from Reference [34]

Stability Analysis

The stability of the aircraft was analysed through the dynamic modes of the aircraft.
The dynamic modes of the system describe characteristic motions in the longitudinal
and lateral directions. These motions are characterized by how each state variable
changes in relation to the other state variables. For example in the longitudinal
motion, the long period, or Phugoid mode, is characterized by large changes in the
pitch angle, θ, and the forward speed, ∆u, with small changes (basically negligible)
in the vertical speed, w, also represented with α, and the pitch rate, q. The dynamic
modes are captured in the mathematics as the eigenvalues of the system.
The eigenvalues of the longitudinal and the lateral state matrices were determined
to analyse the control-fixed dynamic stability of the aircraft and to ensure that the
estimated aircraft dynamics yield a response typical of an aircraft. The characteristic
equation is determined using Equation 3.116 [34] and the eigenvalues taken from the
roots of the characteristic polynomial such as,

det(A − λI) = 0 (3.116)

The eigenvalues can be real or complex valued. The imaginary part of a complex
 56

eigenvalue represents the oscillatory nature of the response while the real part rep-
resents its decay or growth in time. The general form of a complex eigenvalue is
represented by,

λ = n ± ωi (3.117)

√
where n is the real part, ω is the complex part and i = −1. The period of the
response is defined as [34],

2π
T = (3.118)
ω

and the time for the amplitude of oscillation to dampen to half the initial amplitude
is defined as [34],

0.693
thalf = (3.119)
|n|

The dynamic stability of the aircraft is assessed against the requirements in the
Transport Canada airworthiness manual Chapter 523, which covers normal, utility,
aerobatic, and commuter category aeroplanes [39]. Section 523.2005 states that the
commuter category covers multi engine aircraft with seating for 19 or less, excluding
pilots, that have a maximum certified take-off weight of 8618 kg (19,000 lbs) or less
[40]. Section 523.2145 contains the following stability requirements that are relevant
in this study; have static longitudinal, lateral, and directional stability in normal
operations; have dynamic short period and Dutch roll stability in normal operations;
and no aeroplane may exhibit any divergent longitudinal stability characteristic so
unstable as to increase the pilot’s workload or otherwise endanger the aeroplane
and its occupants [40]. Equivalent requirements are stated in the Federal Aviation
Administration (FAA) part 23 and European Union Aviation Safety Agency (EASA)
 57

CS-23 regulations.
The dynamic modes describe specific motion profiles that an aircraft may follow
when perturbed. In the longitudinal direction, an aircraft has a long period mode and
short period mode. Figure 3.21 shows the characteristic perturbation in the aircraft
motion for the longitudinal modes.

Figure 3.21: Characteristic transients. (a) Long period mode, (b) Short period
mode. Reproduced from Reference [34]

The long period mode is generally characterized by low damping while the short period
mode has high damping. This is shown in parts (a) and (b) of Figure 3.21 respectively.
Figure 3.22 shows the vector diagram for the long period and short period modes
in part (a) and (b) respectively. These diagrams show how the longitudinal state
variables evolve in time with respect to each other for each dynamic mode. The
 58

values presented in the figures are from an example using the Boeing 747 taken from
Reference [34].

Figure 3.22: (a) Vector diagram of long period mode, (b) Vector diagram of short
period mode. Reproduced from Reference [34]

The motion of the longitudinal dynamic modes are illustrated in Figure 3.23. In the
figure the long period mode, or phugoid mode, is shown to experience a larger change
 59

in the pitch angle. which is shown in the top plot of the figure, the path of the aircraft
is also shown just above the top plot. The path shows large variation from the initial
horizontal path. The bottom two plots show the short period mode for stable and
unstable motion. The plots show that the angle of attack changes the most while the
image of the path of the aircraft shows that it remains relatively horizontal.

Figure 3.23: Motion of longitudinal dynamic modes. Reproduced from Reference

[41]
 60

In the lateral direction, an aircraft has three dynamic modes, two of which are non-
oscillatory and the other is oscillatory. The two non-oscillatory modes are referred
to as the spiral mode and the rolling convergence mode. The oscillatory mode is
referred to as the Dutch roll mode. Using the same example from Reference [34] with
the Boeing 747, the spiral mode is characterised by exponential decay, for a stable
mode, which follows the aircraft angle ratios for side slip angle, β, bank angle, ϕ,
and yaw angle, ψ, of β : ϕ : ψ = −0.00119 : −0.177 : 1. In this mode the yaw
angle changes at a certain rate with the bank angle changing at a rate that is an
order of magnitude lower with the side slip angle changing at a rate an additional
two orders of magnitude lower. The order and magnitude of this ratio characterises
the spiral mode. This motion is illustrated using Figure 3.24. The figure shows an
example of the unstable spiral mode motion where the bank angle steadily increases
with increasing yaw angle. In the case of a stable spiral mode, the bank angle will
decrease with increasing yaw.

Figure 3.24: Motion of spiral mode. Reproduced from Reference [42]

The rolling convergence mode is characterised by the ratio of aircraft angle variables
of β : ϕ : ψ = −0.0198 : 1 : −0.0562. This mode describes motion that is almost
 61

pure rotation about the aircraft x-axis. The Dutch roll mode, which is also referred
to as the lateral oscillation, is characterised by the magnitude and the phase differ-
ence in the lateral state variables described in the vector diagram shown in Figure
3.25. The diagram shows the side slip angle and yaw angle changing with similar
magnitude approximately 180° out of phase. The non-dimensional roll rate, p̂, and
non-dimensional yaw rate, r̂, are changing approximately 150° out of phase with the
roll rate magnitude approximately and order of magnitude larger than the yaw rate.
The motion of the Dutch roll mode is illustrated in Figure 3.26. The figure shows
how the the bank angle and yaw angle change with respect to each other.

Figure 3.25: Vector diagram of Dutch roll mode. Reproduced from Reference [34]

The eigenvalues for the longitudinal long period and short period modes are pre-
sented in Table 3.23. The short period eigenvalue was calculated to have a negative
real part indicating exponential decay in the amplitude of the motion, this satisfies the
short period longitudinal stability requirements in Section 523.2145 of the Transport
Canada airworthiness manual. This motion is also highly damped which is indicated
 62

Figure 3.26: Motion of Dutch roll mode. Reproduced from Reference [42]

by the thalf value of 0.8 seconds. The long period eigenvalue was also calculated to
have a negative real part indicating dynamic stability, which exceeds the requirement
in Section 523.2145, of the same airworthiness manual, where divergent motion is sta-
bilized though pilot input. The static longitudinal stability requirement is satisfied
through the negative Cmα coefficient. The satisfaction of these requirements shows
that the estimated longitudinal state matrix predicts characteristic motion typical of
a certified aircraft.
The eigenvalues for the lateral Dutch roll mode, rolling mode and spiral mode are
presented in Table 3.24. The Dutch roll mode and rolling mode both show dynamic
stability through the negative real part of the eigenvalues, however, the spiral mode
 63

Table 3.23: Longitudinal A matrix eigenvalues, period, and damping

Mode Type Eigenvalue Period [s] thalf [s]

Long Period Mode

−0.0119 ± 0.1772i 35.5 58.2
(Phugoid)

Short Period Mode

−0.867 ± 1.3561i 4.6 0.8
(Pecking)

Table 3.24: Lateral A matrix eigenvalues, period, and damping

Mode Type Eigenvalue Period [s] thalf [s]

Dutch Roll Mode −0.1126 ± 1.4442i 4.35 6.15

Rolling Mode −2.3215 ± 0i N/A 0.3

Spiral Mode 0.0613 ± 0i N/A 11.3

shows a positive real eigenvalue, indicating an unstable mode. For this study, the
lateral coefficients are adjusted to stabilize the spiral mode while maintaining the
stability of the Dutch roll mode and spiral mode.
A parametric study was performed to identify the parameters that most influence
the spiral mode eigenvalue. This was accomplished by scaling each dimensional aero-
dynamic derivative individually from 0% to 250% of the original value and observing
the effect on the lateral eigenvalues. It was found that changing the rolling moment
coefficient with respect to yaw rate, Lr , had the largest effect on the stability of the
spiral mode. The effect of changing Lr on the eigenvalues is shown in Figure 3.27
where the numbers above and to the right of the points correspond to the scaling
value to produce that eigenvalue. Figure 3.27 shows that when the Lr coefficient
value was decreased to below approximately 8% of its original value, the spiral mode
became stable. To stabilize the spiral mode the Lr coefficient value was adjusted to
 64

Figure 3.27: Change in lateral eigenvalues due to change in Lr

 65

5% of its original value. Adjusting Lr to 5% of its original value also effects the rolling
and Dutch roll modes. The damping of the rolling mode value is increased while the
damping in the Dutch roll mode decreases. The increase in rolling mode damping
is acceptable but the decrease in the Dutch roll mode damping is not ideal because
it increased the time for the combined lateral-directional motion to dampen out. In
Canada, prior to September 2021, in the airworthiness Chapter 523, the requirements
for the combined lateral-direction oscillations were to reduce in amplitude to 10% of
the original amplitude within 7 cycles while below 18,000 feet [43]. This section was
removed in the recent revision which follows revisions made by the Federal Aviation
Administration (FAA). This standard was used in this study to guide the adjustment
of parameters to ensure the combined lateral-direction motion is typical for a certified
aircraft. It was found that increasing the side force with respect to lateral velocity
coefficient, Yv , increased the damping of the Dutch roll mode while also increasing
the damping in the rolling and spiral modes but by a much smaller amount. This
effect is shown in Figure 3.28. It was also found that increasing the rolling moment
with respect to lateral velocity coefficient, Lv , increased the damping of the rolling
mode, decreased the damping in the Dutch roll mode, and decreased the positive real
value of the spiral eigenvalue, providing a stabilizing effect. The effect of changing
Lv on the lateral eigenvalues is shown in Figure 3.29.
 66

Figure 3.28: Change in lateral eigenvalues due to change in Yv

 67

Figure 3.29: Change in lateral eigenvalues due to change in Lv

 68

Based on the analysis it was chosen to scale the original values for Lr , Yv , and
Lv by 0.05, 2.1, and 1.2 respectively. This provided the desired effect of stabilizing
the spiral mode, maintaining the stability of the other modes, and ensured that the
Dutch roll mode demonstrated the motion described earlier. The eigenvalues of the
lateral state matrix were recalculated using the new adjusted coefficient values and
are shown in Table 3.25.

Table 3.25: New lateral A matrix eigenvalues, period, and damping

Mode Type Eigenvalue Period [s] thalf [s]

Dutch Roll Mode −0.0727 ± 1.393i 4.5 9.5

Rolling Mode −2.468 ± 0i N/A 0.28

Spiral Mode −0.0025 ± 0i N/A 277.2

3.5 Urban Air Flow Disturbances

The urban airflow disturbance used in the simulation was generated from the empirical
data that were developed by a colleague researching urban airflow within the UAM
research group. The empirical airflow data were collected for a scaled urban setting
extracted from downtown Toronto. The scale model was subjected to various wind
directions with wind velocity and turbulence profiles that are similar to the ones
encountered in an urban environment [5]. The data are presented as a Power-Spectral
Density (PSD) plot in the u, v, and w directions for a specific location in the city. The
original data are at model scale 1:400, this requires that frequency be scaled down
by 400 and the PSD magnitude scaled up by 400. Equation 3.120 [44] is used to
convert the Power-Spectral Density data into time-domain variations of wind speed
in each coordinate direction. This equation uses the frequency components of the
 69

data to generate sine waves which represent a time-domain wind speed variation.
The superposition of several of these individual frequency components results in the
more complex wave form that represents the wind speed variation in time.

N
X
va = (Aa,j cos (ωa,j t + Ψa,j )) (3.120)
j=1

The subscripts a and j are index values corresponding to a spatial location and a
frequency value respectively; N is the index value for the maximum frequency; ωa,j
is the frequency value; Ψa,j is the phase of the frequency; t is time; and Aa,j is the
amplitude of the specific frequency component. Aa,j is calculated using [44],

p
Aa,j = 2Saa,j ∆f (3.121)

where ∆f is the constant spacing between frequencies and Saa,j is the magnitude of
the auto-spectrum for frequency component j, which was taken from the PSD plots.
Figure 3.30 presents a top down view of the model city layout with measurement
points identified while Figure 3.31 shows a side view of the city model with the
elevation of the measurements identified. The data for point 32 was chosen to generate
the disturbance. Although the measurement point is below the building canopy it
was chosen to provide a potential worst case scenario wind variation. This point
shows higher turbulence intensity in the v direction, mid turbulence levels in the w
direction and lower turbulence levels in the u direction, as indicated by the colour
coded legend in Figure 3.30. This would have an increased effect on the lateral motion
of the aircraft, which is a potential issue for directional control and navigation of
UAT and this point provides a wider variability in the wind speeds in each direction
as compared to the other points.
To generate the time-domain airflow disturbance a cut-off frequency and number
of constant width frequency windows are required. The PSD plot for the u direction,
 70

Figure 3.30: Top down view of city model. Reproduced from Reference [5]

in Figure 3.32, shows that the energy contained within the turbulence decreases by
approximately 2 orders of magnitude for approximately every 1 Hz. For this reason,
1 Hz was chosen as the cut-off frequency for all directions and it was divided into
three equally spaced windows. Figures 3.32, 3.33 and 3.34 show the PSD plots for
the u, v, and w directions. Labelled on each plot is the cut-off frequency for each
section and the frequency and time-domain amplitude of each mode. Using the modal
information obtained for each direction and Equation 3.120, the total wind speed
variation in each direction was calculated and is shown in Figure 3.35. Although
different locations within the city of Toronto have different turbulence levels, the
 71

Figure 3.31: Side view of city model. Reproduced from Reference [5]

spectra shapes were similar. As a result, the modal behaviour for different cases was
similar, and as a result only the case given in Figures 3.32 to 3.35 are shown. A single
urban airflow disturbance case is used because the PSD plot at each location showed
similar trends. Future studies will examine controller response to other features,
beyond the turbulence spectrum.
 72

Figure 3.32: PSD of u direction with modal values

Figure 3.33: PSD of v direction with modal values

 73

Figure 3.34: PSD of w direction with modal values

Figure 3.35: Variations in wind speed in each coordinate direction

Chapter 4

Inner-Loop Controller Design

This chapter presents the PID and ADRC architectures and the methodology for
tuning the controller gains.

4.1 Autopilot

An autopilot is used to control the motion of an aircraft without direct human in-
fluence. The autopilot functions using feedback control where a measured value that
represents some vehicle state is subtracted from a set-point value. The set-point value
represents the desired value of the state and the difference produced is called the error
which is used to drive the input signal which controls the system. When a perturba-
tion to the system or set-point change occurs, error is produced which causes the input
signal to drive the system to reduce the error, resulting in the system maintaining or
following the set-point.
The autopilot was designed to maintain the aircraft trim, this is referred to as the
inner-loop controller as only the aircraft trim is controlled and not the path of the
aircraft. In the longitudinal direction the angle of attack was controlled through the
elevator by monitoring the aircraft pitch angle. In the lateral direction the side slip
angle was maintained by controlling the yaw and bank angles through the yaw rate

74
 75

and bank angle respectively. PID and ADRC architectures were used to create the
inner-loop controller.

4.2 PID Controller

The PID law for the control input is shown in Equation 4.1 [19],

Z t
de
u = k0 e dτ + k1 e + k2 (4.1)
0 dt

where e is the feedback error, t is time, k0 is the integral gain, k1 is the proportional
gain, and k2 is the derivative gain. The PID controller was implemented in the
Controller/Autopilot subsystem block shown in Figure A.5. Within Figure A.5, two
subsystem blocks are shown, one for PID and one for ADRC. The PID controller
was assembled within the PID subsystem block and its arrangement is shown in
Figure 4.1. Figure 4.1 shows how the longitudinal and lateral set-points and feedback
signals were combined to produce the error which was used to create the control
signal following the PID control law. The longitudinal control signal was directly
outputted from this subsystem while the lateral control signals were combined as
a vector and then outputted from the subsystem. Using the simulation, different
PID combinations were tested and it was chosen to use PID control for the elevator
channel, proportional only (P) control for the rudder channel, and PID control for
the aileron channel. The rudder channel controller used the yaw rate as the feedback
signal and this would cause issues when the integral (I) and derivative (D) terms were
included. The derivative term would become large in response to relatively large yaw
rates and result in errors in the simulation and the integral term would degrade the
rudder controller performance.
The controller gains were tuned using a step input to disturb the longitudinal
and lateral motion. In the longitudinal motion the vertical airspeed, w variable, was
 76

Figure 4.1: PID controller subsystem

disturbed, while in the lateral motion the lateral airspeed, v variable, was disturbed.
The uncontrolled response of the aircraft to this unit step disturbance for the longi-
tudinal and lateral motion are shown in Figures 4.2 and 4.3. The simulation was used
to iteratively alter the gains for each PID controller to generally improve the system
response as compared to the uncontrolled system response. The gains were chosen to
minimize the settling time, overshoot, and rebound in the system response and make
the response follow a path which demonstrates more critically damped behaviour.
The gain values are shown in Table 4.1. The controlled response of the aircraft to a
unit step perturbation for the longitudinal and lateral motion are shown in Figures
4.4 and 4.5.

Table 4.1: PID controller gain values

Parameter Elevator Rudder Aileron

k1 50 0.5 2
k2 30 0 1
k0 20 0 1
 77

4.3 ADRC Controller

The ADRC topology, shown in Figure 4.6, consists of the transient profile generator,
the non-linear weighted sum, and the extended state observer. The contribution of
each part is discussed.

Figure 4.2: Uncontrolled longitudinal response to unit step disturbance in w

Figure 4.3: Uncontrolled lateral response to unit step disturbance in v

 78

The transient profile generator is used to create a signal that the system, also
termed the plant, is able to track because if a set-point is changed the resulting step
change to the input is difficult for the plant to track. The transient profile generator
alters the abrupt step change in the input signal into a smooth signal which is fed
into the controller. The transient profile generator is not included in this controller

Figure 4.4: Controlled longitudinal response to unit step disturbance - PID

Figure 4.5: Controlled lateral response to unit step disturbance - PID

 79

Figure 4.6: ADRC topology [19]

design because the set-point remains constant.

ADRC uses a non-linear weighted sum function, as compared to the linear
weighted sum used in PID control. The non-linear function is an improvement over
the linear weighted sum because with the linear weighted sum, the error can approach
zero in infinite time but with the non-linear function, the error can reach zero more
quickly in finite time [19]. The non-linear function was generated through numerical
simulation [19] and is as follows,

u0 = f han(v1 − v, v2 , r0 , h0 ) (4.2)

where h0 and r0 are controller parameters, v, v1 , and v2 are the set-point signals, and
f han(v1 − v, v2 , r0 , h0 ) is

a 
f han = −r0 − sign(a) sa − r0 sign(a) (4.3)
d

where the parameters in Equation 4.3 are calculated using the following list of equa-
tions:
 80

d = h0 r02

a0 = h0 v2

y = v1 + a0
p
a1 = d(d + 8|y|)

a2 = a0 + sign(y)(a1 − d)/2

sy = (sign(y + d) − sign(y − d))/2

a = (a0 + y − a2 )sy + a2

sa = (sign(a + d) − sign(a − d))/2

The extended state observer (ESO) is a key idea developed by Han [19]. It is
introduced using the following example, described by Han in Reference [19], of a
second-order single-input single-output (SISO) system:

ẋ1 = x2

ẋ2 = f (x1 , x2 , w(t), t) + bu (4.4)

y = x1

where y is the measured output to be controlled, u is the input, and f (x1 , x2 , w(t), t)
is a multi-variable function of the system states, the external disturbance, and time.
Unlike other modern controllers, ADRC does not require that the multi-variable
function be explicitly known. The multi-variable function can be considered as
F (t) = f (x1 , x2 , w(t), t) and is simply something to be overcome by the control sig-
nal. F (t) is denoted as the “total disturbance” and is treated as an additional state,
x3 = F (t), with Ḟ (t) = G(t), where G(t) is unknown. Using these definitions, the
 81

system described in Equation 4.4 can be written as

ẋ1 = x2

ẋ2 = x3 + bu
(4.5)
x˙3 = G(t)

y = x1

Pulling from modern control theory, a state observer is constructed and is denoted as
the extended state observer (ESO). It follows the continuous form

e = z1 − y

f e = f al(e, 0.5, δ)

f e1 = f al(e, 0.25, δ)
(4.6)
z˙1 = z2 − β1 e

z˙2 = z3 − β2 f e

z˙3 = −β3 f e1

where z1 , z2 , and z3 are the state estimates of x1 , x2 , and x3 respectively; e is the

error determined by the difference between the estimated state z1 and the measured
output y; β1 , β2 , and β3 are the observer gains; b, termed b0 in Figure 4.6, is an
estimate of the system B matrix; and f al is a non-linear function described by



e
|x| ≤ δ


 δ 1−α
,
f al(e, α, δ) = (4.7)

 |e|α sign(e), |x| ≥ δ



where α and δ are tuning parameters. The calculations for f e and f e1 use α values of
0.5 and 0.25 respectively, where these values were taken from [19]. The ESO estimates
the value of F (t) = f (x1 , x2 , w(t), t) which allows the control law (u0 − F (t))/b to
 82

reduce the plant shown in Equation 4.4 to

ẋ1 = x2

ẋ2 = u0 (4.8)

y = x1

This example shows the power of ADRC in its ability to reject any unknown system
disturbance and force the system to behave ideally while following the control signal.
Figure A.5 shows the ADRC subsystem block in Simulink, which contains the
ADRC implementation. The ADRC channel layout is shown in Figure 4.7 where
the channels for the elevator, rudder, and ailerons were separated into their own
subsystems.

Figure 4.7: ADRC autopilot subsystem

 83

Each channel subsystem contains the same ADRC layout. The elevator controller
subsystem is shown in Figure 4.8. Here the combination of the ESO and the non-
linear feedback yields the ADRC architecture for the elevator channel, noting that
the tracking differentiator is not included, as previously mentioned. The non-linear
feedback was calculated using the f han function described in Equation 4.2 and was
implemented in Simulink using a Matlab function block. The ESO described by
Equation 4.6 is contained within the ESO subsystem block shown in Figure 4.8. The
ESO layout is shown in Figure 4.9 where the f al function was implemented using a
Matlab function block.

Figure 4.8: ADRC elevator channel subsystem

Figure 4.9: ESO subsystem block

 84

The gains were tuned using a step input to disturb the longitudinal and lateral
motion. In the longitudinal motion the w variable was perturbed, while in the lateral
motion the v variable was perturbed. Similar to the PID controllers, the simulation
was used to iteratively alter the gains for each ADRC to generally improve the system
response as compared to the uncontrolled system response and make it more critically
damped. The gain values are shown in Table 4.2. The controlled response of the
aircraft to a unit step perturbation for the longitudinal and lateral motion are shown
in Chapter 5.

Table 4.2: ADRC gain values

Parameter Elevator Rudder Aileron

h0 5 0.01 0.05
r0 5 20 0.05
b0 5 10 25
c 1 1 1
β1 100 4 200
β2 12 150 10
β3 15 0.1 10
α 0.5 0.5 0.5
δ 0 0 0
α1 0.25 0.25 0.25
δ1 0 0 0
Chapter 5

Results and Discussion

This chapter presents the results and the observations from the controller tuning and
response of the system when subjected to an urban airflow disturbance.

5.1 Tuning Observations

The following observations were made during the tuning of each controller.

PID Controller Tuning Observations

The PID tuning process for the longitudinal and the lateral trim controllers was
accomplished by testing different combinations of the P (proportional), I (integral)
and D (derivative) terms and their respective gain values. Figure 5.1 shows the
response of the vertical airspeed, w, using different PID controller configurations
while a disturbance in the form of a unit step in the vertical wind speed was applied.
The uncontrolled response is shown using the blue line as compared to the controlled
responses, where unity gains are used for the P, PD, and PID configurations. Figure
5.1 shows that a pure proportional controller with unity gain significantly improves
over the uncontrolled vehicle response, where the controlled responses in Figure 5.1
follow a profile that is close to being critically damped. The P only controller response

85
 86

Figure 5.1: PID tuning responses - longitudinal

does show a small amplitude oscillation which is eliminated when the derivative term
is introduced in the PD configuration. While the PID controller with unity gains
improves the response, further tuning was expectedly found to yield a better one.
The PD, PID, and tuned PID controllers are represented in Figure 5.1 by the yellow,
purple, and green lines respectively but are close in magnitude such that the graph
shows them essentially as being identical. Figure 5.2 shows only the first fifteen
seconds of the response with the uncontrolled one excluded to better visualize the
difference in performance between the PD, PID, and tuned PID controllers. The PID
configuration with tuned gains was used for the longitudinal control of the aircraft to
compare against the response when using the ADRC autopilot.
Lateral control used the rudder and the ailerons and by the nature of the vehicle
dynamics, rudder and aileron inputs each individually affect both the yaw and the
roll motions. Changes in the tuning of one controller will therefore affect the motion
in both yaw and roll. The lateral trim that was to be maintained is the side slip angle,
which is represented by the lateral velocity v. The lateral velocity is most sensitive to
rudder input, so this controller was tuned first. The aileron controller was then tuned
 87

Figure 5.2: PID tuning responses zoomed - longitudinal

to maintain a wing level condition. Figure 5.3 shows the rudder tuning responses with
no aileron control input while Figure 5.4 shows the aileron tuning responses with the
rudder channel tuned. Figure 5.5 shows the rudder tuning responses with the aileron
channel tuned.

Figure 5.3: Rudder PID tuning responses

 88

Figure 5.4: Aileron PID tuning responses

Figure 5.5: Rudder PID tuning responses with tuned aileron channel

The trial and error method used to tune the PID controller was used as a bench-
mark to test the tuning of the ADRC.
 89

ADRC Tuning Observations

The ADRC was tuned in a similar manner to the PID controller but with the addition
of the tuning of the ESO gains. The tuning methodology was to set the controller
gains to unity then tune the ESO gains until the estimation of the feedback state was
reasonably close to the actual one. Then the controller parameter gains were tuned to
adjust the system response to minimize overshoot and settling time and to generally
improve the system response as compared to the uncontrolled one.
The longitudinal autopilot with the elevator channel was tuned first starting with
the ESO, where each gain parameter was initially set to unity. Starting with β1 , the
parameters were individually adjusted in a iterative manner using the simulation to
reduce the error between the estimated state and the actual feedback state. With the
tuning parameters, h0 and r0 , set to unity, the β1 , β2 , and β3 , values were adjusted
where the effect of adjusting these parameters is shown in Figure 5.6.

Figure 5.6: Effect of ESO gains on state estimation error

Figure 5.7 shows the same plot as the one shown in Figure 5.6 but is zoomed in to
highlight finer details. The ESO gains that resulted in the smaller amplitude yellow
 90

Figure 5.7: Effect of ESO gains on state estimation error - zoomed view

Figure 5.8: Vertical airspeed response due to ESO tuning

and green lines were chosen as the potential gains. The gains represented by the
green line were chosen given their reduced overshoot while the gains represented by
the yellow line were chosen because of the smaller oscillation amplitude at steady
state. The vertical airspeed response is shown in Figure 5.8 while Figure 5.9 shows
the same information as in Figure 5.8 but scaled to highlight finer details.
 91

Figure 5.9: Vertical airspeed response due to ESO tuning - zoomed view

The tuning parameters, h0 and r0 , were then adjusted to improve the system
response and to select which specific gains to use for the ESO. Figure 5.10 and Figure
5.11 show the results of the ESO error due to adjusting the tuning parameters for a
β2 value of 8 and 12 respectively. Figures 5.10 and 5.11 show that changing only the
tuning parameters has an effect on the ESO performance. The difference between
the two figures is that in 5.11 the initial overshoot is reduced but the steady state
oscillation amplitude is larger, as compared to Figure 5.10. The final value for the
gains were chosen by comparing the vertical airspeed responses of the system. The
system responses due to varying the tuning parameters are shown in Figures 5.12
and 5.13 for a β2 value of 8 and 12 respectively. A selection of the responses are
shown in Figure 5.14 from both the β2 equal to 8 and 12 cases. Figure 5.14 shows
that the response represented by the orange line provides the smallest overshoot and
minimal settling time as compared to the other responses. The gains represented by
the orange line response are chosen for the longitudinal ADRC autopilot.
The vertical airspeed responses for both the tuned PID and ADRC autopilots are
shown in Figure 5.15. Figure 5.16 shows the same response as in Figure 5.15 but
 92

Figure 5.10: Effect of tuning parameter gains on estimation error - β2 = 8

Figure 5.11: Effect of tuning parameter gains on estimation error - β2 = 12

with the step disturbance included in the figure. Figure 5.15 shows that the ADRC
autopilot has a marginally better response to a step disturbance as compared to the
PID autopilot.
The lateral autopilots were tuned in a similar manner to the longitudinal autopilot
 93

Figure 5.12: Vertical airspeed response due to parameter tuning - β2 = 8

Figure 5.13: Vertical airspeed response due to parameter tuning - β2 = 12

where the controller parameters were initially set to unity and the ESO gain param-
eters were adjusted to minimize the state estimation error. During this process, the
system response was monitored to inform the ESO tuning process. After reaching
a suitable error and system response, the tuning parameters were adjusted to fur-
ther improve the system response. After adjusting the tuning parameters, the ESO
 94

Figure 5.14: Vertical airspeed response due to parameter tuning - detailed view of
select parameters

Figure 5.15: Vertical airspeed response of tuned PID and ADRC autopilots - step
disturbance

gains were further adjusted. This process was repeated until a suitable response was
achieved. The same process was completed for the aileron channel and then fine ad-
justments were made between the rudder and the aileron channels to further improve
the response.
 95

Figure 5.16: Longitudinal airspeed response of tuned PID and ADRC autopilots -
step disturbance

Results from the entire tuning process are not presented here but rather the state
estimation error and the system response are shown for the rudder and the aileron
channels, where the final tuned gain values are adjusted by several scaling values from
10% to 200%. These results show that the chosen gain values provide a potential
optimal solution for this specific case. These results are shown in Figures B.1 to B.20
in Appendix B.
Figure 5.17 shows the response of the lateral airspeed for the tuned lateral PID and
ADRC autopilots and Figure 5.18 shows the bank angle response. The lateral velocity
response, which is controlled through the rudder, shows similar performance between
the ADRC and PID controllers, as was observed in the longitudinal control. The bank
angle response, which is controlled through the ailerons, shows the ADRC performing
better than the PID controller. The PID controller response could be improved with
further tuning, however, the magnitude of the bank angle change is significantly small
for both controllers that changes in the bank angle are essentially negligible and the
rudder control is dominant for this aircraft in this type of disturbance.
 96

Figure 5.17: Lateral airspeed response of tuned PID and ADRC autopilots - step
disturbance

Figure 5.18: Variations in bank angle for tuned PID and ADRC autopilots - step
disturbance

The change in the angle of attack, α, and the side slip angle, β, are shown in
Figures 5.19 and 5.20. The variations presented in these figures below show similar
performance between the two controllers. Figures 5.21, 5.22 and 5.23 show the control
inputs to the system that were generated by the ADRC and the PID controllers. It
 97

was observed that the ADRC exhibits an oscillation in the control signal. From
the tuning process, it was observed that when the ESO was tuned to improve the
controller performance, an oscillation was present. It is assumed that this plays a
role in the oscillation observed in the control signal.

Figure 5.19: Variations in α angle for tuned PID and ADRC autopilots - step
disturbance

5.2 Disturbance Response Comparison

The responses of the vertical airspeed, lateral airspeed and bank angle of the system
to the urban airflow disturbance are shown in Figures 5.24, 5.25 and 5.26 respectively,
while the change in α and β angles are shown in Figures 5.27 and 5.28 respectively.
For reference, the urban airflow disturbance were extracted from the disturbance
presented in Figure 3.35 of Chapter 3.
Both controllers exhibited similar performance in their ability to reduce the mag-
nitude of the longitudinal and the lateral motions when subjected to an urban airflow
disturbance. Both controllers also demonstrated similar smoothing of the aircraft
 98

Figure 5.20: Variations in β angle for tuned PID and ADRC autopilots - step
disturbance

Figure 5.21: Variations in elevator angle for tuned PID and ADRC autopilots - step
disturbance

motion as compared to the air motion in the vertical and lateral directions. The
bank angle was better maintained by the ADRC, but as mentioned in the tuning
observations this difference in performance may be reduced by further PID tuning.
Figures 5.29, 5.30 and 5.31 show the elevator, rudder and aileron control angles
 99

Figure 5.22: Variations in rudder angle for tuned PID and ADRC autopilots - step
disturbance

Figure 5.23: Variations in aileron angle for tuned PID and ADRC autopilots - step
disturbance

respectively. The magnitude of each of the control inputs were suitably small such
that the linear methods used to determine the control effectiveness values were valid.
The fastest rate of change of a control surface is demonstrated by the elevator surface
where the elevator traverses approximately 0.8 degrees in 0.8 seconds giving a 1 degree
 100

Figure 5.24: Longitudinal airspeed response of tuned PID and ADRC autopilots -
urban airflow disturbance

Figure 5.25: Lateral airspeed response of tuned PID and ADRC autopilots - urban
airflow disturbance

per second change. This rate is assumed to be well achievable by a modern elevator
control system.
 101

Figure 5.26: Variations in bank angle for tuned PID and ADRC autopilots - urban
airflow disturbance

Figure 5.27: Variations in α angle for tuned PID and ADRC autopilots - urban
airflow disturbance
 102

Figure 5.28: Variations in β angle for tuned PID and ADRC autopilots - urban
airflow disturbance

Figure 5.29: Variations in elevator angle for tuned PID and ADRC autopilots -
urban airflow disturbance
 103

Figure 5.30: Variations in rudder angle for tuned PID and ADRC autopilots - urban
airflow disturbance

Figure 5.31: Variations in aileron angle for tuned PID and ADRC autopilots - urban
airflow disturbance
Chapter 6

Conclusion and Future Work

A classical flight dynamics model of a generic urban air taxi was developed in this
research project. The generic urban air taxi was based on the Bell Nexus quadro-
tor 4EX UAM vehicle. This unique flight dynamics model was used to compare the
performance of ADRC to classical PID control in its ability to maintain the trim con-
dition of the generic urban air taxi when subjected to a disturbance. The disturbance
was generated using empirical urban airflow data to create wind disturbances that
were representative of winds in an urban environment [5].
It was found that the controllers performed similarly for comparable tuning efforts.
More time was required to tune the ADRC controller but there were more parameters
to tune. Also the I (integral) and D (derivative) terms were not included in the rudder
control which made the controller simpler and tuning faster. However, a similar (trial
and error) tuning process can effectively be used for both controllers. This shows
that ADRC may be suitable for wider adoption in the industry given that tuning
would be similar to the tuning of PID control, which is widely used, and does not
require any precise knowledge of the flight dynamics model for initial tuning of the
gain parameters. It should also be noted that the simpler PID implementation would
not be expected to work well in a physical system implementation, and those issues
which would effect the PID controller may be better handled by ADRC. Those issues

104
 105

include the linear combination of terms which is not ideal for the control of non-linear
systems and issues related to the I (integral) and D (derivative) terms, all of which
are addressed by ADRC.
The literature on ADRC presents stability analyses that mathematically prove
that positive real gain values exist that will yield a stable system, with all gains or
just the ESO gains related through a certain parameter. The publications present gain
values for the system but do not elaborate on how the gains were selected and simply
state that numerical simulations were used to converge to these values. The ESO
gain values in the literature often follow the form β1 < β2 < β3 . In this investigation,
it was observed that a large β3 term could lead to steady state offset. It is also
worth mentioning that the systems which ADRC has been applied to are often small
UAV’s, whereas in this work ADRC is applied to a much larger vehicle. In addition,
the system model used linearised, small-disturbance equations of motion. This could
mean that the advantages of ADRC were not being fully tested in the context of
the current study. Furthermore, the observed similar performance of the ADRC to a
PID controller may indicate that suitable heuristic methods of tuning ADRC may be
possible to develop.
The generic urban air taxi model that was developed for this analysis applies only
to horizontal flight at cruise speeds with no duct actuation and does not consider
unsteady aerodynamic effects. As the aircraft speed decreases, the unsteady aerody-
namic effects would have a larger influence on the aerodynamic forces and moments.
This would occur when the urban air taxi is approaching the point where it would re-
configure from horizontal flight to vertical flight near or in an urban setting. Also, the
originally estimated lateral matrix exhibited instabilities that were stabilized through
the stability analysis. The existence of these instabilities may be true or a result of
the assumptions and simplifications used in the model development. Increased model
fidelity could be used to investigate the stability characteristics of these aircraft.
 106

ADRC is shown to perform similarly to PID in the control of a UAM vehicle

that was immersed in a representative urban airflow environment which captures
the turbulent wind speeds. However, further investigation is recommended where
unsteady aerodynamic effects, increased model fidelity and more realistic effects are
considered. Based on these conclusions, the following future work is recommended:

ˆ The ADRC controller be implemented in hardware and be made to control a

non-linear physical system to investigate its performance in a physical labora-
tory setting;

ˆ The generic UAT flight dynamics model should be extended to include non-
linear dynamics and that the controllers tuned using the linear model be applied
to the non-linear model and the performance assessed;

ˆ Unsteady aerodynamic effects should be considered in the dynamic model to

extend the model to lower cruise speeds;

ˆ More advanced modelling techniques and tools should be used to develop higher
fidelity models to identify stability characteristics for UAT vehicles;

ˆ The ADRC should be applied to existing models that capture the dynamics
of the transition from vertical to horizontal flight to test ADRC performance
against other modern controllers; and

ˆ A more detailed mass distribution model for the generic urban air taxi should
be created to improve model fidelity.
List of References

[1] NASA, “Advanced air mobility mission overview.” [Online]. Available: https:
//www.nasa.gov/aam/overview/ [Accessed: Jul. 5, 2021], Jan 2021.

[2] Airbus, “Urban air mobility.” [Online]. Available: https://www.airbus.

com/innovation/zero-emission/urban-air-mobility.html [Accessed: Jul.
5, 2021], 2021.

[3] PRB, “Urban population to become the new majority world-

wide.” [Online]. Available: https://www.prb.org/resources/
urban-population-to-become-the-new-majority-worldwide/ [Accessed:
Jun. 22, 2021], Jul 2007.

[4] The Canadian Press, “Drone delivers lungs for transplant to toronto hospital
in world 1st, health network says.” [Online]. Available: https://www.cbc.ca/
news/canada/toronto/first-lung-transplant-drone-1.6208057 [Accessed:
Oct. 25, 2021], Oct. 2021.

[5] M. A. Labbad, “The effects of urban settings on airflow characterisitcs for urban
air mobility applications,” Master’s thesis, Carleton University, 2021.

[6] K. A. Adkins, “Urban flow and small unmanned aerial system operations in
the built environment,” International Journal of Aviation, Aeronautics, and
Aerospace, vol. 6, no. 1, p. 10, 2019.

[7] J. L. Archdeacon and N. Iwai, Aerospace Cognitive Engineering Laboratory

(ACELAB) Simulator for Urban Air Mobility (UAM) Research and Develop-
ment, ch. 1, pp. 1–2. AIAA AVIATION 2020 FORUM, 2020.

[8] X. Yang and P. Wei, “Autonomous on-demand free flight operations in urban air
mobility using monte carlo tree search,” in International Conference on Research
in Air Transportation (ICRAT), Barcelona, Spain, 2018.

107
 108

[9] W. B. Cotton and D. J. Wing, “Airborne trajectory management for urban air
mobility,” in 2018 Aviation Technology, Integration, and Operations Conference,
p. 3674, 2018.

[10] G. Cai, J. Dias, and L. Seneviratne, “A survey of small-scale unmanned aerial

vehicles: Recent advances and future development trends,” Unmanned Systems,
vol. 2, no. 02, pp. 175–199, 2014.

[11] E. N. Johnson and M. A. Turbe, “Modeling, control, and flight testing of a small-
ducted fan aircraft,” Journal of Guidance, Control, and Dynamics, vol. 29, no. 4,
pp. 769–779, 2006.

[12] M. Xue and M. Wei, “Small unmanned aerial vehicle flight planning in urban
environments,” Journal of Aerospace Information Systems, pp. 1–9, 2021.

[13] R. Niemiec, F. Gandhi, M. Lopez, and M. Tischler, “System identification and

handling qualities predictions of an evtol urban air mobility aircraft using modern
flight control methods,” in 76th Annual Forum of the Vertical Flight Society,
Montreal, Canada, 2020.

[14] D. Galway, J. Etele, and G. Fusina, “Modeling of the urban gust environment
with application to autonomous flight,” in AIAA Atmospheric Flight Mechanics
Conference and Exhibit, p. 6565, 2008.

[15] Textron systems, “Aerosonde: Small unmanned aircraft system.” [On-

line]. Available: https://www.textronsystems.com/products/aerosonde [Ac-
cessed: Jul. 26, 2021], 2021.

[16] S. A. Raza, Autonomous UAV control for low-altitude flight in an urban gust
environment. PhD thesis, Carleton University, 2015.

[17] W. Su, S. Qu, G. G. Zhu, S. S.-M. Swei, M. Hashimoto, and T. Zeng, “A

control-oriented dynamic model of tiltrotor aircraft for urban air mobility,” in
AIAA Scitech 2021 Forum, p. 0091, 2021.

[18] Z. Gao, “Active disturbance rejection control: a paradigm shift in feedback con-
trol system design,” in 2006 American Control Conference, pp. 7–pp, IEEE,
2006.

[19] J. Han, “From pid to active disturbance rejection control,” IEEE transactions
on Industrial Electronics, vol. 56, no. 3, pp. 900–906, 2009.
 109

[20] Y. Zhang, Z. Chen, X. Zhang, Q. Sun, and M. Sun, “A novel control scheme
for quadrotor UAV based upon active disturbance rejection control,” Aerospace
Science and Technology, vol. 79, pp. 601–609, 2018.

[21] Y. Huang and W. Xue, “Active disturbance rejection control: methodology and
theoretical analysis,” ISA transactions, vol. 53, no. 4, pp. 963–976, 2014.

[22] L. Hui, F. Yu, L. Fengyi, S. Chuang, and Y. Zhide, “Missile attitude control
in a large flight envelope based on the linear active disturbance rejection con-
trol approach,” in Proceedings of 2014 IEEE Chinese Guidance, Navigation and
Control Conference, pp. 1219–1223, IEEE, 2014.

[23] R. Shi, J. Song, and W. Wan, “Active disturbance rejection control system design
for a morphing wing structure,” Asian Journal of Control, vol. 17, no. 3, pp. 832–
841, 2015.

[24] D. Zhang, H. Duan, and Y. Yang, “Active disturbance rejection control for small
unmanned helicopters via levy flight-based pigeon-inspired optimization,” Air-
craft Engineering and Aerospace Technology, pp. 946–952, 2017.

[25] Bell Textron inc., “Bell nexus.” [Online]. Available: https://www.bellflight.

com/products/bell-nexus [Accessed: Jul. 5, 2021], 2021.

[26] Bell Textron inc., “Bell boeing v-22 osprey.” [Online]. Available: https://www.
bellflight.com/products/bell-boeing-v-22 [Accessed: Jul. 5, 2021], 2021.

[27] Bell Textron inc., “Vintage bell: The bell x-22 aircraft.”
[Online]. Available: https://news.bellflight.com/en-US/
188762-vintage-bell-the-bell-x-22-aircraft [Accessed: Jul. 5, 2021],
2020.

[28] Joby Aviation, “All electric air mobility.” [Online]. Available: https://www.
jobyaviation.com/ [Accessed: Jul. 6, 2021]., 2021.

[29] Volocopter, “We bring urban air mobility to your life.” [Online]. Available:
https://www.volocopter.com/ [Accessed: Jul. 6, 2021], 2021.

[30] Airbus, “City airbus.” [Online]. Available: https://www.airbus.com/

innovation/zero-emission/urban-air-mobility/cityairbus.html [Ac-
cessed: Jul. 6, 2021].
 110

[31] Read Sector, “CES attendees wait in huge lines to sit in bell’s re-
designed air taxi.” [Online]. Available: https://readsector.com/
ces-attendees-wait-in-huge-lines-to-sit-in-bells-redesigned-air-taxi/
[Accessed: Jul. 6, 2021], 2020.

[32] J. Roskam, Airplane Design - Part V: Component Weight Estiomation.

Lawrence: Design, Analysis and Research Corporation (DARcorporation), 2018.

[33] O. Johnson, “Bell unveils electric four-ducted nexus 4ex at

ces 2020.” [Online]. Available: https://verticalmag.com/news/
bell-unveils-electric-four-ducted-nexus-4ex-at-ces-2020/ [Accessed:
Jul. 6, 2021, 2020.

[34] B. Etkin and L. D. Reid, Dynamics of Flight: stability and control. New York:
Wiley, 1996.

[35] J. John D Anderson, Aircraft Performance and Design. New York: Tata
MaGraw-Hill, 2010.

[36] Aicrovision, “3d bell nexus 4ex flying car - 2020 model.”
[Online]. Available: https://www.turbosquid.com/3d-models/
3d-bell-nexus-4ex-2020-model-1523481?referral=aicrovision,%20Mar.
%202020 [Accessed: Jul. 6, 2021], 2020.

[37] M. V. Cook, Flight dynamics principles [a linear systems approach to aircraft

stability and control] (2nd ed.). Oxford: Butterworth-Heinemann/Elsevier, 2007.

[38] J. McIver, “Cessna 172 performance assesment.” [Online]. Available: http://

www.temporal.com.au/aero.htm [Accessed: Oct. 14, 2021], May 2003.

[39] Transport Canada, “Airworthiness manual chapter 523 preamble - canadian

aviation regulations (cars).” [Online]. Available: https://tc.canada.
ca/en/corporate-services/acts-regulations/list-regulations/
canadian-aviation-regulations-sor-96-433/standards/
airworthiness-manual-chapter-523-normal-category-aeroplanes/
airworthiness-manual-chapter-523-preamble-canadian-aviation-regulations-cars
[Accessed: Oct. 21, 2021], Sept. 2021.

[40] Transport Canada, “Airworthiness manual chapter 523 subchap-

ter a - general, subchapter b - flight - canadian aviation reg-
ulations (cars).” [Online]. Available: https://tc.canada.ca/
en/corporate-services/acts-regulations/list-regulations/
 111

canadian-aviation-regulations-sor-96-433/standards/
airworthiness-manual-chapter-523-normal-category-aeroplanes/
airworthiness-manual-chapter-523-subchapter-general-subchapter-b-flight-canadi
523-2005 [Accessed: Oct. 21, 2021], Sept. 2021.

[41] H. H. Hurt, Aerodynamics for naval aviators. Office of the Chief of Naval Oper-
ations, Aviation Training Division, 1965.

[42] M. A. U. Amir, A. Maimun, S. Mat, M. Saad, and M. Zarim, “Wing in ground

effect craft: a review of the state of current stability knowledge,” in International
Conference on Ocean Mechanical and Aerospace for Scientists and Engineer,
2016.

[43] FAA, “Code of federal regulations.” [Online]. Available: https://www.

ecfr.gov/on/2017-01-03/title-14/chapter-I/subchapter-C/part-23/
subpart-B/subject-group-ECFR0f5b2c830cfe3db/section-23.181 [Ac-
cessed: Nov. 3, 2021], Jan. 2017.

[44] A. Wall, S. Zan, R. Langlois, and F. F. Afagh, “Correlated turbulence mod-

elling: An advancing fourier series method,” Journal of Wind Engineering and
Industrial Aerodynamics, vol. 123, pp. 155–162, 2013.
Appendix A

Simulink Model Implementation

The time-domain simulation was created using Simulink, where a state-space model
was formed and integrated in time to obtain the system response. The top level
simulink block diagram set up is shown in Figure A.1.

Figure A.1: Top level simulink block diagram

The State-Space Model subsystem contains the A and B matrices for the longi-
tudinal and lateral motion. The arrangement of the State-Space Model subsystem
is shown in Figure A.2. The MATLAB Function block was used to implement the
state-space model mathematics described in Equations 3.107 and 3.109. The wind
disturbance was implemented using the Wind Disturbance subsystem shown in Figure
A.1. The arrangement of this subsystem is shown in Figure A.3. Within the Wind
Disturbance subsystem, the same state matrices were used to calculate the rate of
change of the state variables due to a wind disturbance. This approach was selected

112
 113

given that the wind disturbance affects the u, v and w states of the aircraft which are
inputs to the state matrices. The arrangement of this subsystem is shown in Figure
A.4.

Figure A.2: State-space model subsystem

Figure A.3: Wind disturbance subsystem

The Controller/Autopilot subsystem block contains the control architecture for

PID and ADRC. The arrangement of this subsystem is shown in Figure A.5. The
details of the PID and ADRC subsystem blocks are discussed in Chapter 4. It should
be noted that when running the simulation one of either the ADRC or PID subsystem
block was commented out so that only one controller is running in a given simulation.
 114

Figure A.4: Wind disturbance state-space model

Figure A.5: Controller/Autopilot subsystem

Lastly, the Analysis subsystem block, shown in Figure A.1, was used to quickly visu-
alize the system response when tuning the controllers and to save data for analysis.
Appendix B

Lateral autopilot tuning results

Figure B.1: Effect of tuning parameter gains on estimation error - rudder channel
β1

115
 116

Figure B.2: Yaw rate response due to ADRC tuning - rudder channel β1

Figure B.3: Effect of tuning parameter gains on estimation error - rudder channel
β2
 117

Figure B.4: Yaw rate response due to ADRC tuning - rudder channel β2

Figure B.5: Effect of tuning parameter gains on estimation error - rudder channel
β3
 118

Figure B.6: Yaw rate response due to ADRC tuning - rudder channel β3

Figure B.7: Effect of tuning parameter gains on estimation error - rudder channel
r0
 119

Figure B.8: Yaw rate response due to ADRC tuning - rudder channel r0

Figure B.9: Effect of tuning parameter gains on estimation error - rudder channel
h0
 120

Figure B.10: Yaw rate response due to ADRC tuning - rudder channel h0

Figure B.11: Effect of tuning parameter gains on estimation error - aileron channel
β1
 121

Figure B.12: Bank angle response due to ADRC tuning - aileron channel β1

Figure B.13: Effect of tuning parameter gains on estimation error - aileron channel
β2
 122

Figure B.14: Bank angle response due to ADRC tuning - aileron channel β2

Figure B.15: Effect of tuning parameter gains on estimation error - aileron channel
β3
 123

Figure B.16: Bank angle response due to ADRC tuning - aileron channel β3

Figure B.17: Effect of tuning parameter gains on estimation error - aileron channel
r0
 124

Figure B.18: Bank angle response due to ADRC tuning - aileron channel r0

Figure B.19: Effect of tuning parameter gains on estimation error - aileron channel
h0
 125

Figure B.20: Bank angle response due to ADRC tuning - aileron channel h0
Appendix C

Longitudinal and lateral A and B matrix

estimations

 
−0.0381 0.0724 0 −9.81
 
 
 
−0.2943 −0.8731 64.895 0 
 
ALong =


 (C.1)
−0.0283 −0.8466
 
 0 0 
 
 
 
0 0 1 0
 
 0 
 
 
 
−9.6033
 
BLong =


 (C.2)
 
 4.3085 
 
 
 
0

126
 127

 
−0.1145 −0.0924 −66.4142 9.81
 
 
 
−0.0193 −2.3771 0.0461 0 
 
ALat =


 (C.3)
 0.0258 −0.3963 −0.1239
 
0 
 
 
 
0 1 0 0
 
−7.0499 0 
 
 
 
−2.0144 14.4390
 
BLat =


 (C.4)
 
 3.3694 0.8762 
 
 
 
0 0

Initial estimate for lateral state, ALat , matrix with unstable spiral mode.

 
−0.0545 −0.0924 −66.4142 9.81
 
 
 
−0.0156 −2.3771 1.2014 0 
 
ALat =


 (C.5)
 0.0260 −0.3963 −0.0538
 
0 
 
 
 
0 1 0 0