See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/308877050

Nonlinear Dynamic Inversion of a Flexible

Aircraft

Article · December 2016

DOI: 10.1016/j.ifacol.2016.09.058

CITATION READS

1 101

4 authors, including:

Ryan Caverly Anouck Girard

University of Michigan University of Michigan
20 PUBLICATIONS 52 CITATIONS 172 PUBLICATIONS 1,316 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

The Large Gain Theorem View project

Saturated Control of Flexible-Joint Manipulators View project

All content following this page was uploaded by Ryan Caverly on 25 September 2017.

The user has requested enhancement of the downloaded file.

20th IFAC Symposium on Automatic Control in Aerospace
20th IFAC Symposium on Automatic Control in Aerospace
August 21-25,
20th IFAC
IFAC 2016. Sherbrooke,
Symposium Quebec,
on Automatic
Automatic Canada
Control in Aerospace
Aerospace
20th
August Symposium
21-25, on
2016. Sherbrooke, Control
Quebec, in
Canada
August 21-25, 2016. Sherbrooke, Available
August 21-25, 2016. Sherbrooke, Quebec, Canada
Quebec, online at www.sciencedirect.com
Canada

ScienceDirect
IFAC-PapersOnLine 49-17 (2016) 338–342
Nonlinear
Nonlinear Dynamic
Dynamic Inversion
Inversion of
of a
a Flexible
Flexible
Nonlinear Dynamic Inversion
Aircraft of a Flexible
Aircraft
Aircraft
Ryan James Caverly ∗ Anouck R. Girard ∗∗
∗
Ryan James Caverly ∗ ∗ Anouck R. Girard ∗ ∗∗
Ilya V. Kolmanovsky
Ryan
Ryan James
James Caverly ∗ James
Caverly Richard
∗ Anouck
Anouck R. Forbes
R. Girard
Girard ∗
Ilya V. Kolmanovsky ∗ James Richard Forbes ∗∗
Ilya V. Kolmanovsky ∗ James Richard Forbes ∗∗
∗
Ilya V. Kolmanovsky James Richard Forbes ∗∗
∗ Department of Aerospace Engineering, University of Michigan,
∗ Department of Aerospace Engineering, University of Michigan,
1320 Beal of
∗ Department
Department Avenue, AnnEngineering,
Aerospace Arbor, MI 48109, USA of
University (emails:
Michigan,
1320 Beal of Aerospace
Avenue, AnnEngineering, University
Arbor, MI 48109, USA of Michigan,
(emails:
caverly@umich.edu,
1320
1320 Beal
Beal Avenue,
Avenue, anouck@umich.edu,
Ann
Ann Arbor,
Arbor, MI
MI ilya@umich.edu.)
48109,
48109, USA
USA (emails:
(emails:
∗∗caverly@umich.edu, anouck@umich.edu, ilya@umich.edu.)
Department of Mechanical
∗∗caverly@umich.edu,
caverly@umich.edu, Engineering, ilya@umich.edu.)
anouck@umich.edu,
anouck@umich.edu, McGill University,
∗∗ Department of Mechanical Engineering, ilya@umich.edu.)
McGill University,
817
∗∗ Sherbrooke
Department Street
of
DepartmentStreet West, Montreal,
Mechanical
of Mechanical Quebec,
Engineering,
Engineering, Canada
McGill
McGill H3A 0C3
University,
University,
817 Sherbrooke West, Montreal, Quebec, Canada H3A 0C3
817 (email:
Sherbrooke
817 Sherbrooke Streetjames.richard.forbes@mcgill.ca)
West,
West, Montreal, Quebec, Canada H3A
Montreal, Quebec, Canada
Streetjames.richard.forbes@mcgill.ca) H3A 0C3
0C3
(email:
(email: james.richard.forbes@mcgill.ca)
(email: james.richard.forbes@mcgill.ca)
Abstract: This paper investigates the use of dynamic inversion for the attitude control of a
Abstract: This paper investigates the use of dynamic inversion for the attitude control of a
flexible aircraft.
Abstract:
Abstract: This Input-output
ThisInput-output
paper investigates
paper feedback
investigates linearization-based
thelinearization-based
the use of
use of dynamic
dynamic inversiondynamicfor
inversion inversion is usedcontrol
the attitude
attitude to linearizeof aa
flexible aircraft. feedback dynamicfor the
inversion is usedcontrol
to linearizeof
the angular
flexible
flexible aircraft.
aircraft.velocity and
Input-output
Input-output forward translational
feedback
feedback velocity
linearization-based
linearization-based equations
dynamic of motion.
inversion
dynamic inversion is A
used
is used proportional-
to
to linearize
linearize
the angular velocity and forward translational velocity equations of motion. A proportional-
integral-derivative
the angular
angular velocity (PID)and controller
forward based on thevelocity
forward translational directionequations
cosine matrix (DCM) that describes
the
integral-derivative velocity (PID)and controller based on thevelocity
translational directionequations
cosine matrix of
of motion.
motion.
(DCM) A proportional-
A that
proportional-
describes
the attitude of the(PID)
integral-derivative
integral-derivative aircraft
(PID) is used,based
controller
controller in addition
based on
on the
the to a proportional-integral
direction
direction cosine
cosine matrix
matrix (DCM)
(DCM) (PI)that controller
that describes
describes to
the attitude of the aircraft is used, in addition to a proportional-integral (PI) controller to
maintain
the
the attitude
attitudea desired
of the airspeed.
aircraft
of theairspeed.
aircraft is Numerical
is used, in simulations
addition
used, in simulations
addition toareto are
a performed using
proportional-integral
a proportional-integral the proposed
(PI) dynamic
controller
(PI) controller to
to
maintain a desired Numerical performed using the proposed dynamic
inversion acontroller,
maintain as well as more practical implementations ofusing
the controller thatdynamic
include
inversion a
maintain desired
desired airspeed.
controller, airspeed.
as well as Numerical
Numerical simulations
simulations
more practical are
are performed
performedof
implementations using the
the proposed
the controllerproposed thatdynamic
include
saturated controller,
inversion control inputs andasnomore knowledge of the aircraft’s flexible states. In simulation, the
saturated controller,
inversion control inputs as
as wellandas
well nomore practical
practical
knowledge implementations
implementations
of the aircraft’s flexible of
of the
the controller
controller
states. that
that include
In simulation, include
the
practical controllers
saturated
saturated control
control inputs
inputssuccessfully
and
and no
no stabilize a of
knowledge
knowledge offlexible
the
the aircraftflexible
aircraft’s
aircraft’s with reasonable
flexible states.
states. In
In control surface
simulation,
simulation, the
the
practical controllers successfully stabilize a flexible aircraft with reasonable control surface
deflections.
practical
practical controllers
controllers successfully
successfully stabilize
stabilize aa flexible
flexible aircraft
aircraft with with reasonable
reasonable control control surface
surface
deflections.
deflections.
deflections.
© 2016, IFAC
Keywords: (International
Dynamic Federation
inversion, of Automatic
feedback Control)aircraft
linearization, Hostingcontrol,
by Elsevier Ltd. All rights
saturation control,reserved.
Keywords: Dynamic inversion, feedback linearization, aircraft control, saturation control,
PID control,
Keywords:
Keywords: flexible inversion,
Dynamic
Dynamic wings. feedback linearization,
linearization, aircraft
aircraft control,
control, saturation
saturation control,
control,
PID control, flexible inversion,
wings. feedback
PID control,
PID control, flexible
flexible wings.
wings.
1. INTRODUCTION
1. INTRODUCTION
1.
1. INTRODUCTION
INTRODUCTION
Dynamic inversion has been shown to be an efficient con-
Dynamic inversion has been shown to be an efficient con-
trol method
Dynamic
Dynamic for aircraft,
inversion
inversion has due to
has been
been its ability
shown
shown to
to be toan
beto antransform
efficient
efficient non-non-
con-
con-
trol method for aircraft, due to its ability transform
linear
trol
trol equations
method
method for
for of motion
aircraft,
aircraft, due
due into
to
to linear
its
its ability
abilityequations
to
to over
transform
transform large
non-
non-
Fb
linear equations of motion into linear equations over large Fb
operating
linear
linear regimes
equations
equations of (Enns
motion
of (Enns et
motionetinto al.,
into 1994).
linear Dynamic
equations inversion
over large Fb
operating regimes al., linear
1994).equations
Dynamicover large
inversion F
Fa
a F b
is a nonlinear
operating
operating control
regimes
regimes technique
(Enns
(Enns et al., that
1994).
et al., that
1994).is equivalent
Dynamic to input-
inversion
Dynamic inversion
is a nonlinear control technique is equivalent to input- F
Faa
output
is
is feedback
aa nonlinear
nonlinear linearization
control
control technique (Sastry
technique that
that is and Isidori,
is equivalent
equivalent to 1989;
to input-
input-
output feedback linearization (Sastry and Isidori, 1989;
Hovakimyan
output
output feedback
feedbacket al., 2005).
linearization
linearization As the
(Sastryperformance
and Isidori,limits
1989;of
Hovakimyan et al., 2005). As (Sastry and Isidori,
the performance 1989;
limits of
aircraft
Hovakimyan
Hovakimyan continueet to
al.,
et al., expand,
2005).
2005). As lighter
As the and
the and consequently
performance
performance limits
limitsmoreof
of
aircraft continue to expand, lighter consequently more
flexible continue
aircraft aircraft are produced. This flexibility has morebeen
flexible continue
aircraft aircraft are to
to expand,
expand,
produced. lighter
lighterThisand consequently
andflexibility
consequently has morebeen Fig. 1. Visualization of a flexible aircraft with body frame
modeledaircraft
flexible (Tuzcu et areal.,produced.
2007; PatilThis et al., 1999; Caverly
flexibility has and Fig.
been Fig.
1. Visualization of a flexible aircraft with body frame
F
1. and inertial frame Fa . aircraft
flexible aircraft are produced. This flexibility
modeled (Tuzcu et al., 2007; Patil et al., 1999; Caverly and Fig. F has been 1.b Visualization
Visualization
and inertial frame of
of aa flexible
Fa . aircraft with
flexible with body
body frame
frame
Forbes, 2015)
modeled
modeled (Tuzcu
(Tuzcu and et incorporated
al.,
al., 2007;
2007; Patil
etincorporated into
Patil et the
al.,
et the dynamic
1999;
1999; Caverly
al., dynamic inversion
Caverly and
and Fbbb and
and inertial
inertial frame
frame F Faa ..
Forbes, 2015) and into inversion F
control framework
Forbes,
Forbes, 2015)
2015) and (Gregory, 2001;
and incorporated
incorporated into Dillsaver
the
the dynamic
intoDillsaver dynamic et al., 2013). of the aircraft’s forward velocity in the dynamic inver-
inversion
inversion
control framework (Gregory, 2001; et al., 2013). of the aircraft’s forward velocity in the dynamic inver-
However,
control
control the work (Gregory,
framework of Gregory2001; (2001) and Dillsaver
Dillsaver et al., et al. sion
2013). of theprocess,
aircraft’s which
forwardallows for the
velocity in linearization
dynamic of
the dynamic the
inver-
However, the work of Gregory (2001) and Dillsaver2013).
framework (Gregory, 2001; Dillsaver et al., et al. of sionthe aircraft’s
process, forward
which allows velocity
for the in the
linearization ofinver-
the
(2013) only
However,
However, consider
the
the work
work the
of
of longitudinal
Gregory
Gregory (2001)
(2001) dynamics
and
and of a flexible
Dillsaver
Dillsaver et
et al.
al. aircraft’s
sion
sion process,
process,angular which
whichvelocity
allows
allows and for
for forward
the
the velocity equations
linearization
linearization of
of the
the
(2013) only consider the longitudinal dynamics of a flexible aircraft’s angular velocity and forward velocity equations
aircraft.only
(2013) There existstheliterature on dynamics
the use of of aadynamic
flexible of motion.angularA proportional-integral-derivative attitude con-
aircraft.only
(2013) consider
consider
There existsthe longitudinal
longitudinal
literature on dynamics
the use of of flexible
dynamic aircraft’s
of motion.angular
aircraft’s velocity
velocity and and forward
forward velocity
A proportional-integral-derivative velocity equations
equations
attitude con-
inversion There
aircraft. for attitude
exists control of flexible the spacecraft (Tafa- trol law based on the direction cosine matrix (DCM) is
inversion There
aircraft. exists
for attitude literature
literature
control on
on the
of flexible use
use ofof dynamic
spacecraft (Tafa- of
dynamic trolmotion.
of motion.
law based A
A proportional-integral-derivative
proportional-integral-derivative
on the direction cosine matrix attitude
attitude
(DCM) con-
con-
is
zoli and Khorasani,
inversion
inversion for
for attitude
attitude 2004; Malekzadeh
control
control of
of flexible
flexible et spacecraft
al., 2010), (Tafa-
spacecraft which considered
(Tafa- trol
trol law
law alongon
based
based on with
the
the adirection
proportional-integral
direction cosine
cosine matrix
matrix (PI) control
(DCM)
(DCM) is
is
zoli and Khorasani, 2004; Malekzadeh et al., 2010), which considered along with a proportional-integral (PI) control
contains
zoli
zoli and
and techniques that
Khorasani,
Khorasani, 2004;
2004; can be used aset
Malekzadeh
Malekzadeh inspiration
et al.,
al., 2010),
2010), for the law
which
which for airspeed
considered
considered along
along control.
with
with a
a The proposed dynamic
proportional-integral
proportional-integral (PI)
(PI)inversion
control
control
contains techniques that can be used as inspiration for the law for airspeed control. The proposed dynamic inversion
dynamic techniques
contains inversion ofthat a flexible beaircraft. the controller is implemented in simulation on a flexible high-
dynamic techniques
contains inversion ofthat can
can be
a flexible used
used as
aircraft. as inspiration
inspiration for for the law
law for
for airspeed
controller airspeed control.
control. The
is implemented The proposed
proposed dynamic
in simulation dynamic
on a flexible inversion
inversion
high-
dynamic
dynamic inversion of a flexible aircraft. altitude
controller long-endurance
is implemented (HALE)
in aircraft.
simulation on More
a practical
flexible high-
The novelinversion
contribution of a offlexible
this workaircraft.
includes the design and controller is implemented(HALE)
altitude long-endurance in simulationaircraft.on More
a flexible high-
practical
The novel contribution of this work includes the design and forms altitude
altitude of long-endurance
the proposed controller
long-endurance (HALE)
(HALE) are alsoMore
aircraft.
aircraft. implemented,
More practical
practical
analysis
The novel of a controller
contribution
The novelofcontribution of that
this employs
work includesdynamicthe inversion
design and forms of the proposed controller are also implemented,
analysis a controllerofthat this work
employs includesdynamicthe design
inversion and which include limits on controller
the allowable are control surface de-
for attitude
analysis
analysis of and
of aaand airspeed
controller
controller that
thatcontrol
employs
employs of flexible
dynamic
dynamic aircraft.
inversion
inversion In forms forms of
of the
which include the proposed
proposed
limits on controller
the allowable are also
also implemented,
control implemented,
surface de-
for attitude airspeed control of flexible aircraft. In flections,
which
which no
include
include knowledge
limits on
limits on of of
the the
thethe flexible
allowable
allowable coordinates
control surfacein the
de-
particular,
for
for attitude
attitude theand
and use of dynamic
airspeed
airspeed control
control inversion
of
of flexible
flexible with practical
aircraft.
aircraft. In
In flections, no knowledge flexible control
coordinates surface de-
in the
particular, the use of dynamic inversion with practical flections, dynamic
flections, inversion
no knowledge
no knowledge process,
of and
the
of the both
flexible
flexible control
coordinates
coordinates surfacein de-
the
in the
constraints
particular,
particular, on
the the
use number
of of
dynamic measured
inversion stateswithand on
practicalthe dynamic inversion process, and both control surface de-
constraints the on the usenumber
of dynamic inversion
of measured stateswithand practical
on the flection
dynamiclimits
dynamic inversion
inversionand no knowledge
process,
process, and both
and of flexible
both control
controlcoordinates.
surface de-
surface de-
magnitude of
constraints
constraints
magnitude of on the
onthe the control
number
thecontrol
number inputs
of
inputs
is considered.
of measured
measured
is considered.states
statesAs As
and
andproposed
on the flection
on the
proposed flection
limits
limits
and no
and no
knowledge
knowledge
of flexible
of flexible
coordinates.
coordinates.
in Tafazoli ofand
magnitude Khorasani
the Khorasani
control inputs (2004),
inputs dynamic As
is considered.
considered. inversion
proposed flection limits and no knowledge
of The remainder of this paper proceeds as follows. A presen-of flexible coordinates.
magnitude
in Tafazoli of andthe control is
(2004), dynamic As proposed
inversion of The remainder of this paper proceeds as follows. A presen-
theTafazoli
in angular andvelocity equation(2004),
Khorasani of motion dynamicis performed.
inversion Theof tation
The of the equations
remainder of this
this paperof motion
paper proceeds of as
a flexible
follows. aircraft is
in Tafazoli and Khorasani (2004), dynamic
the angular velocity equation of motion is performed. The The inversion of tationremainder of
of the equations proceeds
of motion of as follows.
a flexible A presen-
A presen-
aircraft is
dynamic
the
the angular
angular inversion
velocity presented
equation
velocity presented ofin
equation ofinmotion this
motion paper
is differs
performed.
is performed. from
The
The tation included
tation of in
of inthe Section
equations
theSection
equations 2. Section
of motion
of motion 3 includes
of a the
flexible
of a flexible dynamic
aircraft
aircraftin-in-
is
is
dynamic inversion this paper differs from included 2. Section 3 includes the dynamic
that
dynamic
dynamic in Tafazoli
inversion
inversion and Khorasani
presented
presented in (2004)
this
in (2004) by
paper
this paper the inclusion
differs
differs from version
included
from included framework
in Section
in Sectionused used
2. to control
Section
2. Section 3 the
includes
3 includes attitude
the of a
dynamic
the dynamic flexible
in-
in-
that in Tafazoli and Khorasani by the inclusion version framework to control the attitude of a flexible
that
that in in Tafazoli
Tafazoli and and Khorasani
Khorasani (2004) (2004) by by the
the inclusion
inclusion version version framework
framework used used toto control
control the the attitude
attitude of of a
a flexible
flexible
Copyright © 2016 IFAC 338
Copyright©©2016,
2405-8963 2016IFAC
IFAC (International Federation of Automatic Control)
338Hosting by Elsevier Ltd. All rights reserved.
Copyright
Copyright
Peer © 2016
review ©under IFAC
2016responsibility
IFAC 338
338Control.
of International Federation of Automatic
10.1016/j.ifacol.2016.09.058
 IFAC ACA 2016
August 21-25, 2016. Quebec, Canada Ryan James Caverly et al. / IFAC-PapersOnLine 49-17 (2016) 338–342 339

aircraft. Numerical simulations of the proposed controller T  ba ba ba 

for a = [a1 a2 a3 ] . Defining CT ba = c1 c2 c3 and
and variations of the proposed controller are presented in baT
c
T T T
Section 4, followed by concluding remarks in Section 5. = c1 c2 c3
ba ba ba leads to
 
2. DYNAMIC MODEL 0 −cba 3 cba
2
Γbab =
 cba
3 0 −cba 3
.
Consider the flexible aircraft shown in Fig. 3 whose equa- −c2 c1
ba ba
0
tions of motion are   The matrices in (1) are further expressed as
τ  
Mν̇ + Dq̇ + Kq + fnon (q, ν) = B̂ , (1) Mrr Mrω Mre
T
  M = MT rω Mωω Mωe ,

where qT = rcw
T T
a cba qT e are the generalized coordi- Mre Mωe Mee
T T
   
nates, ν T = ṙcw
T T
ωbba q̇T are the augmented general- fnon,r
a e
fnon = fnon,ω ,
ized velocities, ra is the position of the center of mass of
cw
fnon,e
the aircraft relative to an unforced particle w expressed in
the inertial frame Fa , cba is a column matrix that contains D = diag{0, 0, Dee },
the nine entries of the DCM that describes the attitude of K = diag{0, 0, Kee },
the aircraft relative to Fa , qe are the elastic coordinates ee ≥ 0 and Kee = Kee ≥ 0.
where Dee = DT T
used to describe the elastic deformation of aircraft, ωbba is
the angular velocity of the aircraft relative to Fa expressed 3. DYNAMIC INVERSION
in the body frame Fb . The matrix M = MT > 0 is the
mass matrix, D is the damping matrix, K is the stiffness The output of the system is chosen to be
matrix, fnon (q, ν) is a column matrix containing nonlinear  
ωbba
terms including aerodynamic forces, B̂ is the matrix that y= T , (4)
12 Cba ṙcw
a
distributes the inputs to the dynamic equations, τ is the
body torque input to the aircraft, and T is the thrust where 1T 2 Cba ṙa
cw
is the forward velocity of the aircraft,
input to the aircraft that acts along the longitudinal axis 12 = [0 1 0], Cba is the DCM that describes the attitude
T

of the aircraft. Although aircraft do not have actuators of Fb relative to Fa , and ṙcw a is the velocity of the aircraft’s
that can exert an arbitrary body torque, a body torque center of mass relative to an unforced particle expressed
is considered as a input in this paper for simplicity and in Fa . Taking the time derivative of (4) yields
to allow for the use of a DCM-based attitude control  
law. The pitch, roll, and yaw moments can be mapped  ω̇bba 
to the appropriate control surface deflections by solving ẏ = T × ,
12 −ωbba Cba + r̈cw a
a control allocation problem (Durham, 1994; Härkegård,  
2004; Bodson, 2002). The equation of motion in (1) is  fω (x) + Gωτ (x)τ + GωT (x)T 
rewritten in first-order state-space form as = T × ,
  12 −ωbba Cba + fr (x) + Grτ (x)τ + GrT (x)T
τ  
ẋ = f(x) + G(x) , (2) τ
T = f̄(x) + Ḡ(x) . (5)
  T
where xT = qT ν T ,
  where
Γν  
 f (x)   fω (x) 
f(x) =  r  , f̄(x) = T × ,
fω (x) 12 −ωbba Cba + fr (x)
fe (x)  
  G (x) GωT (x)
0 0 Ḡ(x) = T ωτ .
12 Grτ (x) 1T
2 GrT (x)
 G (x) GrT (x) 
G(x) =  rτ , (3)
Gωτ (x) GωT (x) Based on the structure of (5) the control input is chosen
Geτ (x) GeT (x) to be  
  τ  
fr (x) = Ḡ−1 (x) v̄ − f̄(x) , (6)
T
fω (x) = −M−1 (Dq̇ + Kq + fnon (q, ν)) ,  
fe (x) which gives ẏ = v̄, where v̄T = vT ω vr . Based on the
T
  chosen aircraft inputs and outputs, the matrix Ḡ(x) will
Grτ (x) GrT (x)
Gωτ (x) GωT (x) = M−1 B̂, always be non singular. This may not always be true
Geτ (x) GeT (x) when using dynamic inversion with a different set of inputs
and/or outputs. Substituting (6) into (5) gives
Γ = diag{1, Γbab , 1},
and ċ = Γb ωb . As shown in de Ruiter and Forbes
ba ba ba
ω̇bba = vω , (7)
(2014), the relationship ċba = Γba ba  
b ωb stems from Poisson’s ×
ba× 1T −ωbba Cba + r̈cw = vr . (8)
equation, Ċba = −ωb Cba , where 2 a
 
0 −a3 a2 Equation (7) represents the attitude dynamics, while (8)
a× = a3 0 −a1 , represents the forward velocity dynamics. A PID control
−a2 a1 0 law that incorporates the DCM directly is considered

339
 IFAC ACA 2016
August 21-25, 2016. Quebec, Canada Ryan James Caverly et al. / IFAC-PapersOnLine 49-17 (2016) 338–342
340

for the attitude dynamics (Goodarzi et al., 2013). By

using the DCM directly, the attitude control law is well-
defined globally and avoids common issues, such as the
kinematic singularities associated with Euler angles and
the unwinding of quaternions (Chaturvedi et al., 2011).
The control law is given by
 v
vω = Kp Cbd − CT bd − Kd ωbba
 t 
 v
+ Ki Cbd − CT
bd − kωbba dτ, (9)
0
where Kp = diag{kp,1 , kp,2 , kp,3 }, Kd = diag{kd,1 , kd,2 , kd,3 },
Ki = diag{ki,1 , ki,2 , ki,3 }, 0 < kp,j < ∞, j = 1, 2, 3,
0 < kd,j < ∞, j = 1, 2, 3, 0 < ki,j < ∞, j = 1, 2, 3,
0 < k < ∞, Cbd is the DCM representing the relative
attitude between Fb and a desired reference frame, Fd ,
Cbd = Cba CT v
da is the attitude error, and (·) : R → Rn
n×n

is the uncross operator defined as

 
0 −a3 a2
Fig. 3. Animation of the HALE aircraft used in the
A = −AT = a3 0 −a1 ,
numerical simulations of Section 4.
−a2 a1 0
T Table 1. Main specifications of HALE aircraft
where Av = [a1 a2 a3 ] . The PI airspeed controller used
used in the simulations of Section 4.
is given by
 t Property Value
vr = −kp,r (V − Vd ) − ki,r (V − Vd ) dτ, (10) aircraft wingspan 16.5 m
0 aircraft length 4m
where V is the airspeed of the aircraft relative to the wind, aircraft mass 10 kg
Vd is the desired airspeed, 0 < kp,r < ∞ is the proportional wing airfoil NACA 0012
control gain, and 0 < ki,r < ∞ is the integral control wing chord length 0.5 m
gain. PI control is used, rather than PID control, since a wing aspect ratio 33
wing thickness 0.03 m
measurement of V̇ is not readily available.
wing elastic modulus 40 GPa (carbon fiber)
The control inputs τ and T depend on the flexible coor- horizontal stabilizer airfoil NACA 0012
dinates, qe , and the flexible coordinate rates, q̇e , which in horizontal stabilizer chord length 0.5 m
practice are not measured. Without knowledge of these horizontal stabilizer span 4.5 m
vertical stabilizer airfoil NACA 0012
coordinates the dynamic inversion will not completely vertical stabilizer chord length 0.5 m
cancel out the system’s nonlinear dynamics, which could vertical stabilizer span 1m
have a significant impact on the closed-loop system’s per-
formance and stability. The simulations presented in Sec- the pitch axis of the aircraft, and a deflection of the rudder
tion 4 include a comparison of the closed-loop performance provides a body torque about the yaw axis of the aircraft.
with knowledge of the flexible coordinates and flexible For the simulated aircraft, control surface deflections lead
coordinate rates to the closed-loop performance without to a change in airfoil camber, which causes a change in the
knowledge of these states. lift and drag produced by the airfoil.

4. NUMERICAL EXAMPLE The first simulation is performed using the dynamic inver-
sion control law of (6) with perfect knowledge of the model,
including the flexible coordinates and flexible coordinate
In this section a numerical example is presented to illus-
rates (denoted “dyn inv” in Fig. 2). This represents the
trate the benefits of using dynamic inversion for the control
ideal case, where all necessary model information and
of a flexible aircraft. The aircraft used in simulation is a
measurements are available. The chosen control gains are
HALE vehicle characterized by its high aspect ratio and
low structural weight. Examples of HALE aircraft include 2
X-HALE (Cesnik et al., 2012), Helios (AeroVironment, kp,j = 5 (rad/s ), j = 1, 2, 3,
Inc., 2015), and Phantom Eye (Boeing, 2015). The HALE kd,j = 5 (1/s), j = 1, 2, 3,
aircraft used in simulation is based on the dynamic model 3
of a kiteplane with flexible wings developed by Caverly and ki,j = 1 (rad/s ), j = 1, 2, 3,
Forbes (2015) with the specifications listed in Table 1. −3
k = 1 × 10 (s/rad),
Although a body torque input is calculated in (6), this kp,r = 5 (1/s),
is converted to approximately equivalent control surface 2
ki,r = 0.5 (1/s ).
deflections (to make for a simpler interpretation of the
results) by making some simplifying assumptions. It is The control gains are tuned to obtain short rise and set-
assumed that a deflection of the ailerons provides a pure tling times with little overshoot in the Euler angle and
body torque about the longitudinal axis of the aircraft, forward velocity responses. Different response character-
a deflection of the elevators provides a body torque about istics can be obtained by different tuning of the control

340
 IFAC ACA 2016
August 21-25, 2016. Quebec, Canada
Ryan James Caverly et al. / IFAC-PapersOnLine 49-17 (2016) 338–342 341

8 0
0.5
reference
6 -2 dyn inv
0 dyn inv sat
-4 dyn inv no flex sat
4
θ (deg)

ψ (deg)
φ (deg)
dyn inv no flex
-0.5 -6
2 reference reference
dyn inv dyn inv -8
dyn inv sat -1 dyn inv sat
0 dyn inv no flex sat dyn inv no flex sat -10
dyn inv no flex dyn inv no flex
-2 -1.5 -12
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Time (s) Time (s) Time (s)
(a) (b) (c)
150 80 50
dyn inv dyn inv
dyn inv sat 60 dyn inv sat
0
100 dyn inv no flex sat dyn inv no flex sat
dyn inv no flex δa (deg) 40 dyn inv no flex
δe (deg)

-50

δr (deg)
50 20
-100
0 dyn inv
0 dyn inv sat
-20 -150 dyn inv no flex sat
dyn inv no flex
-50 -40 -200
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Time (s) Time (s) Time (s)
(d) (e) (f)
15.05 4000 0.6 0.6

2000 0.4 0.4

15
V (m/s)

wr (m)

wl (m)
T (N)

0 0.2 0.2
reference
14.95 dyn inv dyn inv dyn inv dyn inv
dyn inv sat -2000 dyn inv sat 0 dyn inv sat 0 dyn inv sat
dyn inv no flex sat dyn inv no flex sat dyn inv no flex sat dyn inv no flex sat
dyn inv no flex dyn inv no flex dyn inv no flex dyn inv no flex
14.9 -4000 -0.2 -0.2
0 5 10 15 20 0 5 10 15 20 0 5 10 15 20 0 5 10 15 20
Time (s) Time (s) Time (s) Time (s)

(g) (h) (i) (j)

Fig. 2. Closed-loop system response plots of (a) pitch angle θ, (b) roll angle φ, (c) yaw angle ψ, (d) elevator deflection
δe , (e) aileron deflection δa , (f) rudder deflection δr , (g) airspeed V , (h) thrust T , (i) right leading-edge wingtip
deflection wr , and (j) left leading-edge wingtip deflection wl . The dotted lines in (d), (e), (f), and (h) represent the
control limits used when actuator saturation is considered.
gains. The following simulations use the same control gains The results of the four simulations are presented in Fig. 2.
without any retuning, in order to make a fair comparison In particular the closed-loop responses of the pitch angle
to the results of the first simulation. θ, roll angle φ, yaw angle ψ, elevator deflection δe , aileron
deflection δa , rudder deflection δr , airspeed V , thrust T ,
The second simulation is performed using the same con-
right leading-edge wingtip deflection wr , and left leading-
trol law with knowledge of the flexible coordinates and
edge wingtip deflection wl are included.
flexible coordinate rates, but with saturation limits on
the allowable control surface deflections set to ±45 deg As expected, the simulation with perfect dynamic inver-
and minimum thrust of 0 N (denoted “dyn inv sat” in sion exhibits excellent performance, as the desired aircraft
Fig. 2). The chosen limits do not reflect the maximum attitude and airspeed are attained fairly quickly without
control surface deflections of any particular aircraft, but any steady-state error. Unfortunately, the control surface
are simply chosen to demonstrate the effect of control deflections used to achieve this performance are unrealis-
constraints. tic. The controller without knowledge of the flexible states
and no saturation (“dyn inv no flex”) performs similarly
The third simulation is performed with the control law
to the ideal controller with a slightly slower response. Both
of (6) assuming knowledge of the rigid aircraft dynamics,
controllers that include saturation induce some vibrations,
but no knowledge of the flexible coordinates and flexible
but ultimately stabilize the aircraft. Even with the con-
coordinate rates and the same saturation limits from the
trollers that include saturation, the controller that has
previous simulation (denoted “dyn inv no flex sat” in
knowledge of the flexible states outperforms the one that
Fig. 2). This means that any terms involving qe or q̇e
does not.
in (6) are omitted in the dynamic inversion process.
The final simulation is also performed without knowledge 5. CONCLUSIONS
of the flexible states, but no saturation limits (denoted
“dyn inv no flex” in Fig. 2). A dynamic inversion control algorithm for the attitude and
airspeed control of a flexible aircraft was presented in this

341
 IFAC ACA 2016
August 21-25, 2016. Quebec, Canada Ryan James Caverly et al. / IFAC-PapersOnLine 49-17 (2016) 338–342
342

paper. The angular velocity and forward velocity equations control design. International Journal of Control, 59(1),
of motion are inverted and a synthetic input is chosen using 71–91.
a DCM-based PID attitude control law and a PI airspeed Goodarzi, F., Lee, D., and Lee, T. (2013). Geometric
control law. The effectiveness of the proposed dynamic nonlinear PID control of a quadrotor UAV on SE(3). In
inversion controller is demonstrated in simulation, and Proceedings of the European Control Conference, 3845–
is compared to slightly more realistic variations of the 3850.
same controller. As anticipated, the idealistic proposed Gregory, I.M. (2001). Stability result for dynamic inversion
dynamic inversion controller outperforms the other more devised to control large flexible aircraft. In AIAA
realistic controllers, but the other controllers still produce Guidance, Navigation, and Control Conference, 4282.
reasonable results. Härkegård, O. (2004). Dynamic control allocation using
constrained quadratic programming. Journal of Guid-
Future work will investigate the use of an extended
ance, Control, and Dynamics, 27(6), 1028–1034.
Kalman filter (EKF) to estimate the flexible states of
Hovakimyan, N., Lavretsky, E., and Sasane, A.J. (2005).
the aircraft for use in the dynamic inversion control law.
Dynamic inversion for nonaffine-in-control systems via
This would allow for the practical implementation of the
time-scale separation: Part i. In Proceedings of
dynamic inversion control law of (6) with the inclusion of
the American Control Conference, 3542–3547. Port-
flexible states, which could potentially lead to improved
land, OR.
performance, as seen in the closed-loop response of the
Malekzadeh, M., Naghash, A., and Talebi, H.A. (2010).
ideal controller in Section 4.
Control of flexible spacecraft using dynamic inversion
Additional future work will consider control surface deflec- and µ-synthesis. Journal of Vibration and Control,
tions as the aircraft’s inputs. This will likely be a fly-by- 17(13), 1938–1951.
wire flexible aircraft, where each aileron and each elevator Patil, M.J., Hodges, D.H., and Cesnik, C.E.S. (1999). Non-
can be operated independently. This would increase the linear aeroelasticity and flight dynamics of high-altitude
number of system inputs and could possibly lead to im- long-endurance aircraft. In Structures, Structural Dyan-
proved closed-loop performance and stability properties. mics, and Materials Conference.
The addition of wing-based actuators (e.g., piezo-electric Sastry, S.S. and Isidori, A. (1989). Adaptive control of
materials, control moment gyros, etc.) may be investigated linearizable systems. IEEE Transactions on Automatic
to allow for more control authority over the closed-loop Control, 34(11), 1123–1131.
response of the flexible coordinates. Tafazoli, S. and Khorasani, K. (2004). Nonlinear control
and stability analysis of spacecraft attitude recovery.
IEEE Transactions on Aerospace and Electronic Sys-
REFERENCES tems, 42(3), 825–845.
Tuzcu, I., Marzocca, P., Cestino, E., Romeo, G., and
AeroVironment, Inc. (2015). Helios (UAV): Frulla, G. (2007). Stability and control of a high-
UAS Advanced Development Center. URL altitude, long endurance UAV. Journal of Guidance,
https://www.avinc.com/uas/adc/helios/. Control, and Dynamics, 30(3), 713–721.
Bodson, M. (2002). Evaluation of optimization methods
for control allocation. Journal of Guidance, Control,
and Dynamics, 25(4), 703–711.
Boeing (2015). Boeing: Phantom Eye. URL
http://www.boeing.com/defense/phantom-eye/.
Caverly, R.J. and Forbes, J.R. (2015). Dynamic model-
ing, trajectory optimization, and control of a flexible
kiteplane. IEEE Transactions on Control Systems Tech-
nology. Under Review.
Cesnik, C.E.S., Senatore, P.J., Su, W., Atkins, E.M.,
and Shearer, C.M. (2012). X-HALE: A very flexible
unmanned aerial vehicle for nonlinear aeroelastic tests.
AIAA Journal, 50(12), 2820–2833.
Chaturvedi, N.A., Sanyal, A.K., and McClamroch, N.H.
(2011). Rigid-body attitude control. IEEE Control
Systems Magazine, 31(3), 30–51.
de Ruiter, A.H.J. and Forbes, J.R. (2014). General iden-
tities for parameterizations of SO(3) with applications.
J. Appl. Mech., 81(7), 071007.
Dillsaver, M.J., Cesnik, C.E.S., and Kolmanovsky, I.V.
(2013). Trajectory control of very flexible aircraft
with gust disturbance. In AIAA Atmospheric Flight
Mechanics (AFM) Conference. Boston, MA.
Durham, W.C. (1994). Constrained control allocation:
Three-moment problem. Journal of Guidance, Control,
and Dynamics, 17(2), 330–336.
Enns, D., Bugajski, D., Hendrick, R., and Stein, G. (1994).
Dynamic inversion: An evolving methodology for flight

342
View publication stats