Aerodynamic modeling and State-Space model extraction of a UAV using

DATCOM and Simulink

Ansar Rauf, Muhammad Aamir Zafar & Zeeshan Humza Akhtar

Ashraf RF Planning Officer, Telenor Pakistan,
Department of Electrical Engineering, Paris Plaza, F11 Markaz,
College of E & M.E, National University of Sciences Islamabad, 44000, Pakistan
and Technology, Rawalpindi, Punjab, Pakistan [email protected]
[email protected] & {zeeshan-ashraf &
ansarrauf}@hotmail.com

Abstract—Unmanned aerial vehicle (UAV) is a remotely [1]. Input range is defined for linear model by solving force,
controlled and auto-piloted vehicle capable to sustaining moment and kinematic equations we get dimensionless
lift due to aerodynamic forces. It finds its application in derivatives. Next step in model building is finding out the
military for reconnaissance and airstrikes, carrying out numerical values of these aerodynamic derivatives [3].
search and rescue operations, terrain mapping and so on. Techniques for finding out the derivatives include wind
To develop a robust and adaptive control system of a tunnel testing, Computational fluid dynamic analysis and
UAV a good mathematical model based on aerodynamic equational dynamic modeling softwares like DATCOM
analysis is required. This paper presents aerodynamic which was originally developed by USAF for estimating
modeling and subsequently state-space model extraction of a aerodynamic control and stability derivatives and Advanced
high wing design UAV. The technique demonstrated in the Aircraft Analysis by DAAR Corporation [4]. Wind tunnel
paper serves as a procedure to investigate flight dynamic testing is a promising and traditional method of finding out
principles, model validation and basics for Autopilot the derivatives, but it requires bigger setup and the cost of
implementation. A fairly accurate plant model is an important
doing wind tunnel testing is also high [7, 5]. Computational
first step in Autopilot design and pilot training. The test
benches used extract model’s key parameters are off-the-shelf fluid analysis softwares are also getting popular these days
custom build. Aerodynamic derivates are found out using for modeling purposes and aircrafts of more complex
DATCOM. Aerosim toolbox of Simulink is used to linearize structure are modeled in them. More robust and
and trim the model for a given set of inputs and state-space computationally extensive technique of system
model is found out from the trimming results. identification using modern control theory is also
implemented [6]. The method used in this paper is by using
Index Terms DATCOM; Aerosim; aerodynamics; state- digital DATCOM. It provides reliable results for standard
space; aircraft configurations. UAV used for modeling purpose is a
standard high wing design. After finding derivatives
I. INTRODUCTION numerical values, aircraft model is trimmed using Aerosim
Model building is the fundamental process in designing toolbox of Simulink to get state-space model of the UAV.
a control mechanism for any system. It is necessary to The paper is organized in the following way. In section
incorporate the physical existence of any system in to its II nonlinear aircraft model is described using the basic force,
mathematical equivalent model so that mathematical laws moment and kinematic equations. In section III linear model
can be applied to analyze the behavior and responses of that and its conditions are discussed. Section IV comprises of
system. Modeling of aircraft includes the modeling of its trimming techniques used and in section V results from
control surfaces, their response to different flight conditions DATCOM and Aerosim are shown. Conclusion and future
and their behavior to various changes happening to aircraft work is discussed in section VI.
[2]. Mathematical modeling of aircraft is also necessary due
II. AERODYNAMIC MODELING
to fact that the testing of real aircraft is very costly and long
process requiring specialized locations and apparatus. So The standard 6-DOF (six degree of freedom) equations
development of model goes far beyond just the control are used to manipulate the non-linear model of aircraft. For
system design. aerodynamic modeling we first need to know our aircraft’s
Model building is the step-wise formulation of system linear model, which is extracted from non-linear 6-DOF
matrices. It starts form non-linear aircraft model which model. Our state vector is:
consists of force, moment and kinematic equations. Linear XT = [U V W    P Q R] (1)
model is obtained from non-linear model using small Body axis 6-DOF equations can be given as
perturbation theory and considering steady-state conditions
___________________________________
978-1-61284-840-2/11/$26.00 ©2011 IEEE


• Force Equations
UT = [thl el ail rdr ] (17)
RV − QW − g 0 sin θ + FX
U = (2) Where the elements of the vector are throttle setting,
m
elevator deflection, ailerons deflection and rudder deflection.
− RU − PW − g 0 sin ϕ cos θ + F Y III. LINEAR MODEL
V = (3) In designing a control mechanism it is necessary to
m convert the non-linear model into the linear model. The
stability derivatives for linear systems can be estimated
QU − PV − g 0 cos ϕ cos θ + F Z
W = (4) quickly and accurately. Now in aircraft it is relatively easy
m to model steady state flight with linear model as compare to
climb and landing flight conditions. Steady state flight
• Kinematic Equations conditions are as follow
ϕ = P + tan θ (Q sin ϕ + R cos ϕ ) (5) • Steady state flight:
P`, Q, R`, VT, `, `=0 (18)
θ = Q cos ϕ − R sin ϕ (6)

Q sin ϕ + R cos φ This flight condition can be further elaborated as,

ψ = (7)
cos θ • Steady wing level flight:
• Moment Equations ' ' '
φ , φ , θ ,ψ = 0 , P , Q , R = 0 (19)
P = (c1 R + c 2 P )Q + c3 L + c 4 N (8)
• Steady turning flight

Q = c5 RP − c 6 ( P
2 2
− R ) + c7 M (9) φ , θ = 0 ,=turn rate (20)

• Steady pull up
R = ( c 2 P − c 2 R )Q + c 4 L + c 9 N (10)
' '
φ , φ ,ψ = 0 , =pull – up rate (21)
where
2
{( J Y − J Z ) J Z − J XZ }
• Steady roll
c1 = (11) ' '
θ ,ψ = 0 , φ = roll rate
'
r (22)

In steady state flight conditions all the angular rates

{( J X − J Y + J Z ) J XZ } must be zero or constant, in addition airspeed, angle of
c = (12) attack and
(12 side-slip angle must also be zero or constant. This
2 r
leads to the fact that aerodynamic and thrust moments must
J J (J − J X ) be zero or constant.
c = Z , c = XZ , c5 = Z (13) By the process of linearization we can decouple
3 r 4 r JY aircraft’s state-space equations into longitudinal and lateral
dynamics. The entire modeling process is based on the
J XZ 1 assumption that the motion is constrained to small
c6 = , c7 = (14) disturbances about an equilibrium trim state.
JY JY
IV. TRIMMING IN AEROSIM TOOLBOX
2
{( J X − J Y ) J X + J XZ } Simulink Aerosim is used to calculate Trim
c8 = (15) Conditions. It includes a Simulink model of the aircraft and
r
a Matlab script which performs the trim and the
J linearization of the aircraft dynamics.
2
c = X , Γ = J X J Z − J XZ (16) The Simulink model is shown in Fig.8. The attitude
9 Γ equations in the aircraft model are implemented using the
The forces and moment components (Fx, Fy, Fz, L, M, Euler angle representation which provides more intuitive
N) must broken down into aerodynamic and thrust trim results. Several aircraft control inputs - the ap, the
contributions. Here we will define our controlling vector U mixture and the ignition are assumed constant through-out
as the trim procedure. The values for these constants, as well


as the aircraft model block parameters are read from a
Matlab structure created by the trim program.

Pitching moment coefficient (Cn)

Angle of attack (deg)
Figure 2 Basic pitch moment coefficient

• Longitudinal Control derivatives:

Figure 1: Simulink Model for Trim and Linearization

Lift coefficient (CL)

Elevator deflection (deg)

Figure 4 Lift coefficient due to elevator deflection
Pitching moment coefficient (Cn)

Figure 2: Simulink Trim Parameters

V. RESULTS
A. DATCOM results:

Stability and control derivatives were found out in

DATCOM and resultant graphs were imported into Matlab Angle of attack (deg)
for curve fitting and slope finding. In the paper graphs of Figure 5 Pitch moment coefficient due to elevator deflection
some of the important derivatives are given and numerical
values of all derivatives used are tabulated in Table 1. In the
selected graphs shown below vertical axis is dimensionless
• Lateral Stability derivatives:
quantity.
• Longitudinal Stability derivatives:
Roll Moment Coefficient due to
Lift coefficient (CL)

beta(Clb)

Angle of attack (deg)

Angle of attack (deg)
Figure 6 Roll moment coefficient due to beta per degree
Figure 3 Lift due to alpha


 C. Aerosim results for Trim Condition calculation &
Side force due to roll rate (Cyp)

Linearization

INPUTS:
Elevator = -0.1580, Aileron = -0.0070, Rudder = 0.0016
Throttle = 0.3290
STATES:
u =19.95 m/s, v = 0.01 m/s, w = 1.45 m/s, p = 0.00 deg/s
q = -0.00 deg/s, r = -0.00 deg/s, phi = -0.02 deg,
theta = 4.14 deg, psi = 0.20 deg, Alt = 210.00 m,
Angle of attack (deg) Fuel = 2.00 kg, Engine rpm = 4244 rot/min
Figure 7 Side force coefficient due to roll rate per degree
OUTPUTS:
• Lateral Control derivative: Airspeed =20.00 m/s, Sideslip =0.02 deg, AOA = 4.14 deg
Bank = -0.02 deg, Pitch = 4.14 deg, Heading = 0.20 deg
Roll Moment Coefficient (Cl)

Altitude = 210.00 m

• Longitudinal Dynamics:
The longitudinal state equation is given as:

(23)
Aileron deflection (deg)
Figure 8 Roll moment coefficient due to aileron deflection

B. Numerical Values The four response transfer functions obtained in the solution
Numerical values of the derivatives were tabulated using of equation (23) may conveniently be written:
linear ranges of the plots and then curve fitting in Matlab.
They are tabulated below (all values are in per radian): (24)
Derivaitive Numerical Value
CL0 0.293
CLa 4.2781 (25)
CLde 0.2554
CLalphadot 1.059
CLq 7.6
CDmin 0.03267
CDde 0.0049
(26)
CDda 0.01
CYbeta -0.0571
CYda 0.0
CYp -0.04321 (27)
CYr 0.08954673
Cm0 0.00464
Cma -2.3465987
Cmde -0.6535653 Open-Loop Longitudinal Linearization Results:
Cmalphadot -2.7100
Cmq -11.3297 State vector: x = [u w q h ]
Clbeta -0.05324 Input vector: u = [elevator throttle]
Clda -0.21567 Output vector: y = [Va  q h]
Clp -0.45199
Clr 0.042
Cnbeta 0.034845 State matrix: A =
Cnda -0.01
Cndr -0.0519 -0.1982 0.593 1.245 -9.779 -0.0001 0.0101
Cnp -0.0395 -0.7239 -3.9848 18.7028 -0.6286 0.0009 0
Cnr -0.0447773
0.3537 -5.5023 -5.4722 0 0 0
0 0 1.00 0 0 0
TABLE I AERODYNAMIC DERIVATIVES
0.0641 -0.9979 0 19.9997 0 0
27.838 1.7894 0 0 -0.0086 -2.1436


Control matrix: B = Observation matrix: C =

0.2281 0 0.0500 0 0 0 0
-4.6830 0 0 1.0000 0 0 0
-36.1341 0 0 0 1.0000 0 0
0 0 0 0 0 1.0000 0
0 0 0 0 0 0 1.0000
0 375.0287
Eigenvalue: -11.6226
Observation matrix: C = Time constant = 0.0860 s
Eigenvalue: -0.2262 +/- 3.0218 i
0.9979 0.0641 0 0 0 0 Damping = 0.0746, natural frequency = 3.0303 rad/s,
-0.0032 0.0499 0 0 0 0 period= 2.0793 s
0 0 1.0 0 0 0 Eigenvalue: 0.0084
0 0 0 1.0 0 0 Time constant = -119.4471 s
0 0 0 0 1.0 0

Eigenvalue: -4.6216 +/- 10.0909 i VI. CONCLUSION

Damping = 0.4164, natural frequency = 11.0989 rad/s, The progressive model building approach followed in
period = 0.6227 s the paper is robust and effective. State-space model of UAV
Eigenvalue: -2.3230 from basic aerodynamic equations and by using DATCOM
Time constant = 0.4305 s and Simulink is derived. The model obtained is thus reliable
Eigenvalue: -0.0391 +/- 0.6316 i enough to implement more robust control system design like
Damping = 0.0618, natural frequency = 0.6328 rad/s, Autopilots and Auto navigation systems. Modeling
period= 9.9484 s technique used in this paper can be made more adaptive to
Eigenvalue: -0.0003 accommodate for airplane designs other than high wing.
Time constant = 3829.8894 s Using the current plant model a robust Autopilot system is
to be implemented in future.
• Lateral Dynamics:
REFERENCES
Open loop results: [1] Brian Stevens and Frank Lewis, “Aircraft Control and Simulation”,
John Wiley and sons publications U.S.A, 1992.
State vector: x = [v p r  ]
[2] Reed Siefert Christiansen, “Design of an autopilot for small
Input vector: u = [aileron rudder] Unmanned aerial vehicles”. Department of Electrical and Computer
Output vector: y = [ p r  ] Engineering Brigham Young University, August 2004.
[3] Robert Nelson, Flight Stability and Automatic Control,
State matrix: A = WCB/McGraw-Hill, Ohio, 1998.
[4] Abzug, M.J., and Larrabee, E.E.: ‘Airplane stability and control--a
history of the technologies that made aviation possible’
-0.0869 1.2412 -19.8722 9.7798 0 Cambridge Aerospace Series, Cambridge University Press, UK,
-1.2293 -11.5214 0.9885 0 0 1997.
0.2730 -1.2495 -0.4208 0 0 [5] Ashraf, Z and Chaudhry M.A., “Dynamic Modeling of the airship
0 1.0000 0.0643 0 0 using analytical aerodynamic model”, internatiocal Conference on
0 0 1.0021 0 0 Emerging Technologies, pp. 188-193, 2009.
[6] Sahu, J., “Time-accurate aerodynamic modeling of synthetic jets for
projectile control”, Users Group Conference, pp. 144-150, April 2005.
Control matrix: B =
[7] Dong Sun, Huaiyu Wu, Rong Zhu and Ling Che Hung,
“Development of Micro Air Vehicle Based on Aerodynamic
0.0001 2.2759 Modeling Analysis in Tunnel Tests”, International Conference on
-104.4067 -3.9237 Robotics and Automation, pp. 2235-2240, 2005.
-9.3523 -10.9895 [8] Ahmad, S., “Aerodynamic Modeling and Real-time Control of a 1-
DOF TailPlane”, International Bhurban Conference on Applied
0 0 Sciences & Technology, pp.123-130, 2007.
0 0 [9] Zhong Lusheng, Fan Xiaoping and Yang Hui, “Identification
countercurrent rare earth extraction process based on state-space
model”, Control and Decision Conference, pp. 3266-3270, 2010.