JOURNAL OF GUIDANCE, CONTROL, AND DYNAMICS

Vol. 30, No. 3, May–June 2007

Flight Dynamics and Control Authority

of Flap-Controlled Open Boxes

Eric Beyer∗
Oregon State University, Corvallis, Oregon 97331
and
Mark Costello†
Georgia Institute of Technology, Atlanta, Georgia 30332
DOI: 10.2514/1.25301
The effectiveness of open-box micro air vehicles to deliver light, small payloads of high importance to specific
ground coordinates is investigated through dynamic simulation. The open box exhibits interesting and varied flight
dynamic behavior as key design parameters are changed. For example, the open box can achieve a coning behavior, a
corkscrewing behavior, or glide much like a conventional aircraft to the ground by merely shifting the mass center
location. The four rear flaps of the air vehicle can be used to control the box and affords the aircraft greater control
Downloaded by BEIHANG UNIVERSITY on June 2, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.25301

authority than dispersion caused by typical atmospheric winds. This control mechanism can also be used as a braking
system, which can greatly arrest the descent rate before ground impact. These dynamic qualities make the open box a
promising airdrop vehicle which can cut through atmospheric winds towards the target before decelerating and
gently landing.

Nomenclature 1 , 2 , 3 , 4 = flap deflection of flaps 1, 2, 3, and 4

CL , CM , CN = aerodynamic moment coefficients in body  = air density
reference frame , , = Euler roll, pitch, and yaw angles of box
CLP , CMQ , CNR = aerodynamic damping moment coefficients MW = azimuth angle of mean atmospheric wind
in body reference frame
CX , CY , CZ = aerodynamic force coefficients in body
reference frame
D = box reference diameter Introduction
Ixx , Iyy , Izz , Ixy , = aircraft inertia matrix terms
Ixz , Iyz
L, M, N = total applied moment components about
M ICRO air robots are small, autonomous, intelligent aircraft
designed to focus on a specific task. The range of applications
envisioned for future micro air robots in both the civilian and military
mass center in body reference frame
sectors is quite wide [1,2]. Micro air robots promise to provide
p, q, r = components of angular velocity vector in
unparalleled situation awareness and data-gathering opportunities in
body reference frame
many scenarios.
S = box reference area
Future micro air robots are being designed in different shapes and
u, v, w = components of velocity vector of mass
sizes, and are being highly optimized to meet narrow mission specific
center in body reference frame
requirements. Such unconventional configurations include the
u A , v A , wA = relative aerodynamic velocity components of
bioinspired flapping-wing, multirole morphing, and folding-wing,
mass center in body reference frame
tube-launched micro air vehicles described by Jones [3], Abdulrahim
VA = magnitude of relative aerodynamic velocity
[4], and Henry [5], respectively. The open-box micro air vehicle is
vector of mass center
another emerging configuration and is depicted in Fig. 1. These
VMW = magnitude of mean atmospheric wind
unpowered micro air vehicles are envisioned as cargo delivery
X, Y, Z = total applied force components in body
platforms which are released from a parent aircraft operating at
reference frame
relatively high altitudes. Once released from the parent aircraft, the
x, y, z = components of position vector of mass
box settles into a steady-state condition. Subsequent to achieving a
center in an inertial reference frame
steady-state condition, an autonomous flight control system is
x
p1!p2 , y
p1!p2 , = components of position vector from point p1
activated to steer the system to specific ground coordinates. Control
z
p1!p2 to p2 in body reference frame
of the aircraft is achieved by independent deflection of the box flaps

= aerodynamic angle of attack
in the rear. All payload, actuators, sensors, batteries, and associated
 = aerodynamic roll orientation
electronics are contained in the walls of the box. The air vehicle is
designed with the mass center forward of the aerodynamic center so
Presented as Paper 6326 at the AIAA Atmospheric Flight Mechanics that the box falls to the ground in a stable fashion.
Conference, Providence, RI, 16–19 August 2004; received 18 May 2006; The work reported here investigates the basic flight mechanics of
revision received 25 October 2006; accepted for publication 30 November unpowered, controllable open-box aircraft configurations using
2006. Copyright © 2006 by Mark Costello and Eric Beyer. Published by the simulations supported by wind-tunnel-obtained aerodynamic data.
American Institute of Aeronautics and Astronautics, Inc., with permission. The basic modes of motion are reported as a function of mass
Copies of this paper may be made for personal or internal use, on condition configurations and typical flight behavior is documented. Using
that the copier pay the $10.00 per-copy fee to the Copyright Clearance Center,
good mass management, the open box can achieve a stable gliding
Inc., 222 Rosewood Drive, Danvers, MA 01923; include the code 0731-5090/
07 $10.00 in correspondence with the CCC.
mode that can be autonomously controlled to deliver high-
∗
Graduate Research Assistant, Department of Mechanical Engineering. importance, small, light payloads. The rear flaps of the aircraft can
Member AIAA. also be used to dramatically arrest the rate of descent. Atmospheric
† wind dispersion is contrasted against control authority provided by
Sikorsky Associate Professor, School of Aerospace Engineering.
Associate Fellow AIAA. flap deflection.
827
 828 BEYER AND COSTELLO

where I is the mass moment of inertia matrix of the box evaluated at
the mass center with respect to body frame coordinates. The standard
shorthand is used for trigonometric functions: cos
 c
,
sin
 s
, and tan
 t
The applied loads contain
contributions from weight W and air loads A.
8 9 8 9 8 9 8 9 8 9
JB < X = < XW = < XA = < s = 1 < CX =
IB Y  Y  Y  mg s c  VA2 S CY
: ; : W; : A; : ; 2 : ;
Z ZW ZA c c CZ
KB −z (5)
II The aerodynamic forces are applied at the aerodynamic
y computation point, which is not necessarily the aerodynamic center.
The applied moments about the box mass center contain
x
contributions from two sources. Because the aerodynamic force
JI
KI given earlier is not located at the mass center, it produces a moment
about the mass center. Also, because the computation point is not the
Fig. 1 Schematic of the open box with associated reference frames.
aerodynamic center, an aerodynamic moment is also present. The
Box Dynamic Model applied moments about the box mass center is given by
The numerical simulation employed in this study consists of a 8 9 8 9
Downloaded by BEIHANG UNIVERSITY on June 2, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.25301

L > CL  pDC
2VA >
LP

rigid-body, 6-DOF model typically used in flight dynamic modeling < A>
> = 1 >
< >
=
qDC
of air vehicles. A schematic of an open-box configuration with major MA  VA2 SD CM  2VAMQ
>
: >
; 2 >
> >
>
elements of the system identified is given in Figs. 1 and 2. The : ;
degrees of freedom include three position components of the box NA CN  rDC
2VA
NR

mass center, as well as three Euler orientation angles of the body. The 2 38 9
0 z
cg!cp y
cg!cp > XA >
12 equations of motion describing the flight dynamics of an open box 6
7< =
are as follows: 64 zcg!cp 0 x
cg!cp 7
5> Y A > (6)
8 9 2 38 9


: ;
< x_ = c c s s c  c s c s c  s s < u = ycg!cp xcg!cp 0 ZA
y_  4 c s s s s  c c c s s  s c 5 v (1)
: ; : ; where x
cg!cp , y
cg!cp , and z
cg!cp are the components of the position
z_ s s c c c w
vector from the center of mass to the aerodynamic computation point
8 9 2 38 9 expressed in body frame coordinates.
< _ = 1 s t c t < p = The aerodynamic force and moment coefficients are dependent on
_  4 0 c s 5 q (2) aerodynamic angle of attack, aerodynamic roll orientation, and flap
: _; : ;
0 s =c c =c r deflection. The magnitude of the aerodynamic angle of attack is
given as


8 9 8 9 2 38 9 uA
< u_ = < X=m = 0 r q <u= j
j  cos1 p
 (7)
v_  Y=m  4 r 0 p 5 v (3) u2A  v2A  w2A
: ; : ; : ;
w_ Z=m q p 0 w
The angle of attack is defined as positive while wA is positive and
negative while wA is negative. The aerodynamic roll orientation of
8 9 28 9 2 3 8 93 the open box is defined as
< p_ = <L= 0 r q <p=

6 7
q_  I1 4 M  4 r 0 p 5I q 5 (4) v
: ; : ; : ;   tan1 A (8)
r_ N q p 0 r wA

δ1 δ3 The relative aerodynamic velocity components of the box mass

center used in the preceding equations are influenced by the
atmospheric winds. The mean atmospheric wind acts in the
Flap 4
Flap 3 horizontal ground plane and is directed at an angle MW from the II
axis. Thus, the relative aerodynamic velocity components of the box
Flap 1 mass center are
Flap 2 8 9 8 9
JB < uA = < u =
IB
v  v
KB : A; : ;
o wA w
Aerodynamic Computation Point 2 38 9
L4 c c c s s < VMW c =
δ4 MW
 4 s s c  c s s s s  c c s c 5 VMW s
: MW
;
L3
c s c  s s c s s  s c c c 0
o
(9)
L 3/2

δ2
L 1/2
L1 L4
t L 2/2 Aerodynamic Coefficient Determination
L2 Limited public domain information is available for aerodynamic
Fig. 2 Orthographic view of the open box with the dimensions and the force and moment coefficient data for box shapes. Hoerner
location of the aerodynamic computation point. documents drag coefficient data for various square and rectangular 2-
 BEYER AND COSTELLO 829

aerodynamic coefficients to the situation where several flaps are

Is simultaneously deflected, superposition is used. By subtracting the
KB Ks aerodynamic coefficients calculated for the box with one flap
IB
Js deflected by that associated with no flap deflection, the contribution
δ4 from one flap can be determined and generalized to each flap
JB α IW
JW
KW resulting in a model that predicts aerodynamic loads of a box with
γ each flap deflected by a different amount.
The aerodynamic damping moment coefficients were approxi-
mated using two different means. The roll damping coefficient CLP of
the open box was assumed to be similar to that of a projectile with a
Fig. 3 Schematic of the open box mounted in wind tunnel. bluff body. Such a projectile was found in PRODAS by Arrow Tech
Associates and its roll damping coefficient was used for the open box.
The pitch and yaw damping coefficients were approximated by
D shapes, and cube and rectangular prism 3-D shapes [6]. Drag data
considering a simple 2-D model for the box and calculating the
are not provided as a function of angle of attack and sideslip. Hoerner
moment produced when only two lifting surfaces are involved.
and Borst provide limited data on normal force characteristics of
Setting this moment equal to that of the standard aerodynamic
square and cube shapes for different Reynolds numbers [7].
damping moment produced the following equation:
However, like the drag data, detailed information on lift as a function
of angle of attack and sideslip are missing. Aerodynamic data were 2x
cg!cop 2
also found for missiles with square cross sections at various roll CMQ  CNR  CZ
(12)
D2
angles and angles of attack [8,9]. Unfortunately most of this work
Downloaded by BEIHANG UNIVERSITY on June 2, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.25301

involved missiles with noses, fins, and aspect ratios larger than that of where x
cg!cop is the distance along IB from the center of mass to the
the open box considered here. Most of these missiles were tested at center of pressure and CZ
is the slope of the normal force coefficient
supersonic speeds and none allowed air to flow through them, unlike at small angles of attack and zero roll orientation.
the open box. Other similar aerodynamics have been published on
square section cantilever beams in smooth flow [10,11].
Aerodynamic force and moment data for box shapes with deflected Results
flaps were not found in existing literature. To investigate the flight dynamics and control authority of an
Thus, the aerodynamic coefficient data were experimentally open-box micro air vehicle, the equations of motion described in the
obtained using a subsonic wind tunnel and a test model. The preceding sections were numerically integrated using a fourth-order
continuous flow, closed-circuit, low-speed wind tunnel located at Runge–Kutta algorithm. The box used in this study was a triple-
Oregon State University has a test section area of 1:334
1:524 m walled cardboard box with a mass of 345 grams (with no payload).
and is capable of air speeds up to 20 m=s. A six-component sting The dimensions of the open box are shown in Fig. 2, in which
balance is located in the middle of the test section and can be pitched L1  15:88 cm, L2  15:56 cm, L3  15:56 cm, L4  7:94 cm,
and yawed using a model positioning system (see Fig. 3). The open and t  1:43 cm. The mass center, roll inertia, and pitch inertia of the
box was mounted to the sting balance using a wire frame and hub open box are 15.88 cm from the front of the box, 0:002834 kg  m2 ,
assembly. The hub was designed to allow for rolling of the model. and 0:004312 kg  m2 , respectively. Because batteries, sensors, and
Once the box model was mounted to the sting, air was blown at the other electronics are an integral part of any controllable micro air
model while the forces and moments were tabulated for various vehicle, additional mass was added to the box model, enabling the
angles of attack, roll orientations, and flap deflections. The air loads mass center to shift while changing the mass moment of inertia terms.
were then transformed from the sting frame to the body frame and the The oversized wall thickness allows these electronic components to
aerodynamic coefficients were computed as be imbedded into the walls of the box without disrupting the airflow
8 9 8 9 through and around the aircraft. The mass properties (MP) and initial
< CX = 1 < XA = conditions (IC) used in this study are detailed in Tables 1 and 2,
CY  VA S YA 2
(10) respectively. Only the upper flaps of the outer layer of the box are
: ; 2 : ;
CZ ZA allowed to deflect, which creates a means of controlling the box while
changing the flight dynamics.
8 9 8 9
< CL = 1 < LA =
C  V 2 SD M (11) Aerodynamic Force and Moment Coefficient Results
: M; 2 A : A;
CN NA Aerodynamic coefficients of the open box were experimentally
obtained using a wind tunnel and a full-scale test model of the open
box. During testing, the air density varied between 1.16 and
Because of symmetry of the open-box configuration with no flap 1:22 kg=m3 while the air speed was maintained between 10 and
deflection, a moderate set of angle variations covers many 13 m=s. The roll orientation was varied from 90 to 90 deg in 15 deg
orientations. When a flap is deflected, aerodynamic symmetry of the increments and the angle of attack was varied between 60 and
box is largely lost and angle of attack and roll orientation must be 60 deg. Because of the symmetry of the box, a limited amount of roll
varied through a larger set of angles. To generalize these orientations and angles of attack can be used to determine other

Table 1 Mass properties of the featured open-box configurations

MP 1 2 3 4 5
Total mass, g 564.7 624.7 684.7 564.7 984.7
x
cg!o , cm 9.69 8.76 7.99 9.69 5.56
y
cg!o , cm 0.00 0.00 0.00 0.00 2:54
z
cg!o , cm 0.00 0.00 0.00 3:05 2:54
Ixx , kg  m2 0.0046046 0.0050876 0.0055706 0.0040806 0.00671425
Iyy , kg  m2 0.0085828 0.0093338 0.0099955 0.0085188 0.01190045
Izz , kg  m2 0.0085828 0.0093338 0.0099955 0.0081230 0.01190045
Ixy , kg  m2 0.0000000 0.0000000 0.0000000 0.0000000 0:00139034
Ixz , kg  m2 0.0000000 0.0000000 0.0000000 0:0016668 0:00139034
Iyz , kg  m2 0.0000000 0.0000000 0.0000000 0.0000000 0.00063568
 830 BEYER AND COSTELLO

4
Table 2 Initial conditions of the featured trajectories γ =-90 deg
γ =-75 deg
IC 1 2 3 3 γ =-60 deg
γ =-45 deg
x0 , m 0.0 0.0 0.0 γ =-30 deg
2
y0 , m 0.0 0.0 0.0 γ =-15deg
z0 , m Varied Varied Varied γ =0 deg
1 γ =15 deg
0 , deg 5:0 0.0 35.0
0 , deg 60:0 90:0 12:0 γ =30 deg
γ =45 deg

CZ
0
0 , deg 15.0 0.0 20.0
γ =60 deg
u0 , m=s 0.5 0.2 30.0 γ =75 deg
v0 , m=s 0:001 0.0001 2.0 -1
γ =90 deg
w0 , m=s 0.0002 0:0002 4:0
p0 , rad=s 0.01 0.00001 1.0 -2
q0 , rad=s 0.1 0.004 3.0
r0 , rad=s 0.002 0:003 1:5 -3

-4
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
orientations. Redundant data sets were taken and used to calculate an Angle of Attack: α (deg)
average load at those orientations. This helped rid any anomalies that Fig. 5 Aerodynamic normal force coefficient as a function of the
may have existed with the mounting of the model or asymmetry in aerodynamic angle of attack and roll orientation (
1

2

the model itself. The resulting aerodynamic coefficients expressed in
3

4
0 deg).
the body frame as a function of the aerodynamic angle of attack and
Downloaded by BEIHANG UNIVERSITY on June 2, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.25301

roll orientation are shown in Figs. 4–7 for the nominal open box with 0.25
no flaps deflected. Note that the side force coefficient CY and the γ =-90 deg
yawing moment coefficient CN as a function of angle of attack and 0.2 γ =-75 deg
γ =-60 deg
roll orientation are not reported because they can be determined from 0.15 γ =-45 deg
the plots of normal force and pitching moment coefficients, γ =-30 deg
respectively. The reference surface area and reference diameter were 0.1 γ =-15 deg
γ =0 deg
defined as L2
L3  242:04 cm2 and L1  15:88 cm, respectively. 0.05 γ =15 deg
The aerodynamic computation point was defined in the center of the γ =30 deg
γ =45 deg
CL

box (see Fig. 2). It should also be noted that during simulation, 0
γ =60 deg
aerodynamic coefficients associated with orientations not explicitly -0.05 γ =75 deg
measured were estimated using a bilinear interpolation. γ =90 deg
As is shown, the open box with no flap deflection has a maximum -0.1
axial force magnitude when at 15-deg angle of attack, whereas the
-0.15
maximum side and normal force occurs when the angle of attack is
between 40 and 50 deg for a given roll orientation. All three moment -0.2
coefficients also reach a maximum at 40-deg angle of attack. The
-0.25
coefficient of drag and lift (CDrag and CLift ) were calculated as a -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
function of angle of attack and roll orientation for the nominal no-flap Angle of Attack: α (deg)
deflection case. At 0-deg angle of attack the drag coefficient has a Fig. 6 Aerodynamic rolling moment coefficient as a function of the
minimum value of 0.43, and it increases with larger angles of attack aerodynamic angle of attack and roll orientation (
1

2

3

to a value between 3.1 (with   0 deg) and 4.0 (with   45 deg)
4
0 deg).
at 60-deg angle of attack. The lift coefficient vs angle of attack curve
is similar to that of symmetric airfoils with a lift slope of 0:1= deg The same procedure for determining the aerodynamic coefficients
while passing through CLift  0 at 0-deg angle of attack. The stall was also performed while flap 4 was deflected by 10, 20, 30, 50, 70,
angle of attack occurs at 40 deg with a CLiftmax value that varies and 90 deg. With the flap deflected, box symmetry was largely lost
between 2.5 (with   0 deg) and 3.5 (with   45 deg) depending and the aerodynamic coefficients were altered. The magnitude of the
on the roll orientation. This large stall angle of attack and CLiftmax is axial force coefficient increases when a flap is deflected, whereas the
attributed to the large thickness of the box panels as well as it being a normal force coefficient and the pitching moment coefficient become
finite wing with a low aspect ratio [12,13]. nonzero at 0-deg angle of attack. The coefficient of drag and lift vs

-0.05 1.5
γ =-90 deg γ =-90 deg
γ =-75 deg γ =-75 deg
-0.1 γ =-60 deg γ =-60 deg
1
γ =-45 deg γ =-45 deg
-0.15 γ =-30 deg γ =-30 deg
γ =-15 deg γ =-15 deg
γ =0 deg 0.5 γ =0 deg
-0.2 γ =15 deg γ =15 deg
γ =30 deg γ =30 deg
γ =45 deg
CM

γ =45 deg
CX

-0.25 0
γ =60 deg γ =60 deg
γ =75 deg γ =75 deg
-0.3
γ =90 deg -0.5 γ =90 deg

-0.35
-1
-0.4

-0.45 -1.5
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 -60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60
Angle of Attack: α (deg) Angle of Attack: α (deg)
Fig. 4 Aerodynamic axial force coefficient as a function of the Fig. 7 Aerodynamic pitching moment coefficient as a function of the
aerodynamic angle of attack and roll orientation (
1

2

3
aerodynamic angle of attack and roll orientation (
1

2

3


4
0 deg).
4
0 deg).
 BEYER AND COSTELLO 831

2 3
x* =0 cm
cg→o
x* =1.58 cm
1.5 cg→o
2 x* =3.17 cm
cg→o
x* =4.76 cm
1 cg→o
x* =6.35 cm
cg→o
Lift to Drag Ratio

1
0.5 x* =7.94 cm
cg→o
x* =9.52 cm
cg→o
x* =11.11 cm

CM
0 0 cg→o
x* =12.7 cm
δ 4=0 deg cg→o
-0.5 δ 4=10 deg x* =14.29 cm
cg→o
δ 4=20 deg -1 x* =15.88 cm
cg→o
-1 δ 4=30 deg
δ 4=50 deg
-2
-1.5 δ 4=70 deg
δ 4=90 deg
-2 -3
-60 -50 -40 -30 -20 -10 0 10 20 30 40 50 60 -60-50-40-30-20-10 0 10 20 30 40 50 60
Angle of Attack: α (deg) Angle of Attack: α (deg)
Fig. 8 Lift to drag ratio as a function of the aerodynamic angle of attack Fig. 9 Aerodynamic pitching moment coefficient as a function of the
for various flap 4 deflections (
0 deg). aerodynamic angle of attack for various xcg!o positions (ycg!o
zcg!o

0 cm, 
0 deg,
1

2

3

4
0 deg).
Downloaded by BEIHANG UNIVERSITY on June 2, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.25301

angle of attack were calculated when flap 4 was deflected by the

various amounts. Using this data, the lift to drag ratio was calculated and reaches a maximum at an angle of attack of 40 deg. Various
as a function of angle of attack while the roll orientation was set to angles of attack and roll orientations allow the open box to exhibit
0 deg and the results are plotted in Fig. 8. As shown, the best lift to unstable and stable static roll stability.
drag ratio occurs with no flap deflection at an angle of attack of 15 deg To better understand how the flight dynamics of the open box
with a value of 1.71. Increasing the flap deflection or the angle of change for various mass center locations, three different mass
attack past these values tends to decrease the lift to drag ratio. For a configuration sets were considered during a parametric study. The
flap 4 deflection of 10, 20, 30, 50, 70, and 90 deg, the maximum lift to first, case A, involved moving the center of mass along IB while
drag ratio occurs at 20-deg angle of attack with values of 1.59, 1.50, keeping it in the center of the box (JB –KB plane). The second mass
1.40, 1.23, 1.12, and 1.10, respectively. configuration set, case B, allowed mass to accumulate along one of
The roll damping coefficient was assumed to be 0:050 for all roll the front edges of the box so the mass center moved forward within
orientation and angles of attack. The pitch and yaw damping the IB –KB plane. The last mass configuration set, case C, involved
coefficients were calculated using Eq. (12) with CZ
 6:3. The increasing the mass along two front edges of the box so the mass
center of pressure location used in this equation was found by center shifted forward and toward one corner of the box. The flight
minimizing the pitching moment at low angles of attack and zero roll behaviors associated with the different mass configurations are
orientation. This center of pressure was found to exist at about 7.3 cm described in the following paragraphs as well as the dynamic stability
from the front of the box (x
cop!o  7:3 cm). Typical values for the of the system. To evaluate the dynamic stability of the open box, the
pitch and yaw damping coefficients were between 0:011 and equations of motion were numerically linearized about a steady-state
0:613 depending on the location of the mass center. condition [14,15]. The perturbations for angles were 0.0001 rad,
0:001 m=s for translational velocities, and 0:0001 rad=s for angular
velocities.
Stability and Flight Dynamics The flight dynamics of the open box are greatly altered while
The stability of the open box can, to a certain extent, be determined shifting the mass center forward by evenly distributing mass along
by considering the longitudinal and roll static stability [13,14]. To the front four edges of the box (case A). When the mass center is
understand longitudinal static stability, the aerodynamic pitching stationed further than approximately 11.0 cm from the front of the
moment coefficient CM vs angle of attack curve was examined about box (x
cg!o > 11 cm and y
cg!o  z
cg!o  0 cm), the box exhibits
the mass center of the box. The open box possesses longitudinal an unstable behavior and tumbles end over end until it impacts the
static stability as long as the slope of this curve (CM
) is negative. An ground, an expected result considering it is not longitudinally stable.
example of how the longitudinal static stability of the open box can If more weight is added to the front edges so the mass center lies
change is given in Fig. 9 for a box with no flap deflection and at zero between 9:2 cm < x
cg!o < 10:8 cm, the box rolls and cones. An
roll orientation. Here the center of mass was varied from the front of example of this behavior (MP 1, IC 1, and no atmospheric winds) is
the box (x
cg!o  0 cm) to the center of the box (x
cg!o  15:88 cm) shown in Fig. 10, in which the line represents the location of the mass
while keeping y
cg!o and z
cg!o equal to zero. As is shown, the open center and the box pictures depicts the orientation at different times
box possesses longitudinal static stability when the mass center is along the trajectory. Note that only the last portion of the trajectory is
near the front of the box, becomes neutrally stable when shown so the steady-state condition can be examined. The box
x
cg!o  7 cm, and becomes unstable as the mass center approaches reaches a steady-state condition after approximately 60 s, which is
the center of the box. Also notice that only one equilibrium point is characterized by the box rolling as it “cones” to the ground. The
achieved when the center of mass is near the front of the box, but coning behavior can be described by defining a coning angle that is
three equilibrium points exist when x
cg!o is between 7.3 and 12 cm. the angle between IB and KI . The coning angle oscillates at a
These curves can be shifted up or down by changing the amount of frequency of 3.5 Hz at a value around 24.5 deg. This condition is
flap deflection or by varying z
cg!o . Increasing the deflection of flap 4 associated with roll rate oscillations at 3.5 Hz between 11 and
will shift the curve up as will moving the center of mass in the KB 11:8 rad=s, making the box roll one full revolution every 1.14 s.
direction (increasing z
o!cg ). This allows the equilibrium point to The forward velocity of the mass center reaches a maximum value of
exist at a higher angle of attack and increases the y intercept (CMo ). 16:65 m=s before decreasing to around 15:3 m=s due to the
For static roll stability, the slope of the aerodynamic roll moment increased density of air at lower elevations. Although the forward
coefficient CL vs the roll orientation should be examined. Static roll velocity oscillates at a frequency of 3.5 Hz, the side and vertical
stability is achieved if the slope of this curve (CL ) is negative. velocity oscillate at only one quarter of this frequency at 0.875 Hz
Although not shown, the plot of CL vs roll orientation for various between 6:25 m=s. The pitch and yaw rate oscillate between
angles of attack is sinusoidal in nature with equilibrium points 2:65 rad=s with a frequency of 0.875 Hz. The aerodynamic angle
existing at 90, 45, 0, 45, and 90 deg. The rolling moment of attack maintains a value of around 22.4 deg with 3.5-Hz
coefficient is virtually zero for small angles of attack (
< 15 deg) oscillations.
 832 BEYER AND COSTELLO

20
50
15
u , v, w, q, r ψ , u , v, w, p
40
10 φ , θ ,ψ , v, r
ψ , u, p
Altitude (m)

Imaginary Axis
30 5

20 0

-5
10
-10
0
-25
-24 -23 15 16 -15
-22 13 14
Cross Range (m) -21 12
Range (m)
-20
Fig. 10 Last section of trajectory for open box with MP 1, IC 1, and no -2.5 -2 -1.5 -1 -0.5 0 0.5
atmospheric winds. Real Axis
Fig. 12 Root locus of corkscrewing behavior for mass configurations of
case A. (○: mass
668:7 g; ×: mass
708:7 g).
Once the mass center is brought between 8:7 cm<
x
cg!o < 9:1 cm, the open box tends to glide. An example case was
Downloaded by BEIHANG UNIVERSITY on June 2, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.25301

axis. The frequency of the third oscillatory mode associated with , ,

simulated with MP 2, IC 1, and no atmospheric winds. During steady , v, and r becomes greater as the center of mass is shifted forward.
state the rolling angle oscillates at a frequency of 0.6 Hz between 43.5 The mode associated with , u, v, w, and p becomes less damped
and 46.5 deg, the pitch angle oscillates around 12 deg, the roll rate with increased weight, whereas the other mode associated with , u,
oscillates between 0:1 and 0:1 rad=s, and the forward velocity and p becomes more damped.
decreases from 14:4 m=s to 13:25 m=s. The angle of attack oscillates When more mass is added in the front of the box so the mass center
around 18 deg while the roll orientation varies between 32 and lies between 6:94 cm < x
cg!o < 7:68 cm, the box becomes
58 deg. The associated glide ratio (distance/altitude) was calculated neutrally stable. This is explained in Fig. 9, in which the slope of
to be approximately 1.7, which matches up well with lift to drag ratio CM
is zero at the equilibrium point. If the mass is further increased so
at this particular angle of attack. This system possesses a neutrally the mass center is positioned closer than 6.9 cm from the front of the
stable pole associated with the body frame side and vertical velocity box (x
cg!o < 6:9 cm), the box will reach a steady-state condition in
as well as the body roll rate. The mass configuration is neutrally
which it falls straight down with 0-deg angle of attack. Once again,
stable when considering static roll stability, making it susceptible to
this behavior can be explained by noting that the slope of CM
is
wind gusts and other disturbances.
negative (see Fig. 9) with an equilibrium point at 0-deg angle of
The open box flies in a “corkscrew” pattern if the mass center is
attack.
brought between 7:7 cm < x
cg!o < 8:2 cm. The plot in Fig. 11
Shifting the mass center towards a front edge of the box (case B)
demonstrates this steady-state behavior for a box with MP 3, IC 1, gives similar results as case A when the mass center is near the center
and no atmospheric winds. Although this behavior looks much like of the box. The open box tends to display an unstable behavior by
that of the coning behavior of MP 1 and IC 1, the characteristics of the either tumbling or rolling and coning sideways. When more weight is
two trajectories are much different. The coning angle associated with added so the mass center lies between rcg!o  10:67IB 
the corkscrewing behavior does not oscillate like that of the previous.
2:56KB cm and rcg!o  8:04IB  3:86KB cm, the box tends to
At steady state the coning angle remains fairly constant at 17.2 deg
glide in a stable fashion with the weighted edge pointing towards the
while the roll angle increases slowly due to a 4:1 rad=s roll rate. The
ground. This behavior is shown in the plot of Fig. 13 for MP 4, IC 1,
forward velocity has increased to values of 25:7–23:5 m=s due to the
and no atmospheric winds. The associated pitch angle is 7:8 deg,
increased weight. The angle of attack maintains a value of 13.6 deg
the roll rate is zero, and the forward velocity decreases with lower
whereas the roll orientation is 22.6 deg. The eigenvalues associated
altitudes from 11.5 to 10:14 m=s. The angle of attack maintains a
with this corkscrewing behavior are shown in the root locus plot of
value of 27.3 deg during steady state and the associated glide ratio is
Fig. 12. In this case, the mass of the box was increased from 668.7 to
approximately 1.42 as predicted by the lift to drag ratio at this angle
708.7 g, which moved the mass center from x
cg!o  8:18 cm to
of attack. Heavier boxes in this category shift the center of mass
x
cg!o  7:72 cm. As is shown, these mass configurations are all
forward and downward, allowing for smaller steady-state angles of
stable. Two of the oscillatory modes are associated with the states u, attack and hence greater glide ratios. A root locus plot of these mass
v, w, q, and r. The higher oscillatory mode becomes more stable as configurations is given in Fig. 14, in which the mass has been
weight is increased, whereas the other approaches the imaginary increased from 512.7 to 683.7 g corresponding to the mass center
varying from rcg!o  10:67IB  2:56KB cm to rcg!o  8:00IB 
60
3:88KB cm. As shown, an open box that is weighted with enough
mass on one of the front edges is stable and has many of the same
50 modes as a conventional aircraft. A rolling mode and a spiral mode
associated with , , v, p, r are shown to exist along the real axis. As
40 more weight is increased, the rolling mode becomes more damped,
Altitude (m)

whereas the spiral mode becomes less damped as it approaches the

30 imaginary axis. Two longitudinal modes are also shown that involve
the states , u, w, q. The frequency of the short period mode as well as
20 the damping increases with increasing weight, whereas the Phugoid
mode experiences less damping but about the same damped natural
10 frequency. A Dutch roll mode is also shown associated with states ,
v, p, r. The eigenvalues of this mode become less stable as the center
0
-62 of mass is shifted forward and downward and the system eventually
-26 -25
-60 -27 becomes unstable. This instability occurs for boxes that are weighted
-58 -28
Cross Range (m) -29
Range (m) heavier than 682 g corresponding to mass center locations x
cg!o <
Fig. 11 Last section of trajectory for open box with MP 3, IC 1, and no 8:0 cm and z
cg!o < 3:86 cm. These heavier boxes will tend to
atmospheric winds. corkscrew towards the ground at high roll rates. It should also be
 BEYER AND COSTELLO 833

50
Table 3 Average dispersion radii of uncontrolled and controlled
trajectories at an altitude of 2000 m

40 MP IC Wind intensity, Average dispersion

m=s radius, m
Altitude (m)

30 Uncontrolled 1 2 2 266
Uncontrolled 1 2 4 414
Uncontrolled 1 2 8 941
20
Uncontrolled 4 3 2 273
Uncontrolled 4 3 4 582
10 Uncontrolled 4 3 8 1150
Controlled 4 3 0 1733
0
2740
2760 -260
2780
2800 -250 -255 0 and 350 deg in 10 deg increments. Because the lightly weighted
Range (m) 2820 -240 -245
Cross Range (m) box falls relatively slowly while coning, the applied wind is able to
push the box an average of 266, 414, and 941 m in a given direction
Fig. 13 Last section of trajectory for open box with MP 4, IC 1, and no
atmospheric winds. with a 2, 4, and 8 m=s wind speed, respectively.
The potential to control an open box with weight along one of the
front edges (case B) is great due to its conventional aircraft like flying
10 characteristics and high glide ratio, which allows the box to achieve
Downloaded by BEIHANG UNIVERSITY on June 2, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.25301

θ , u, w, q great control authority and cut through wind more effectively. Using
8
the rear, outer flaps as control mechanisms, the open box can be
6 guided in a given direction. A simple roll controller was implemented
φ ,ψ , v, p, r which deflected flap 1 or 3 by a small amount (<5 deg) depending on
4
φ ,ψ , v, p, r the direction of the roll rate so the maximum horizontal distance of
θ , u, w, q
Imaginary Axis

2 the box could be achieved. When this controller was applied to a box
with MP 4 and IC 3 with no mean atmospheric winds, the control
0
authority radius was found to be approximately 1733 m. Note that the
-2 initial conditions represent those of a parent aircraft drop and that
flap 4 was deflected by 30 deg to increase the stability of the system.
-4 φ , v, p , r Table 3 compares the average dispersion radii achieved by
-6 controlling the box with that of an uncontrolled box with the same
mass properties and initial conditions but with applied atmospheric
-8
winds. The uncontrolled case has flap 1 deflected by 10 deg, which
-10 allows the box to corkscrew down while mean atmospheric winds of
-3 -2.5 -2 -1.5 -1 -0.5 0 0.5
Real Axis 2, 4, and 8 m=s in magnitude are applied, resulting in wind
Fig. 14 Root locus of gliding behavior for mass configurations of dispersion radii of 273, 582, and 1150 m, respectively. As is shown
case B. (○: mass
512:7 g; ×: mass
683:7 g). the control authority radius dominates the effect of wind on an
uncontrolled open box.
Using the same simple roll controller, an open box weighted like
noted that the behavior of the mass configurations that glide well are
dependent on initial conditions. These boxes may exhibit a coning case B can achieve greater control authority by applying additional
mass to the front of the box or deflecting flap 4 by different amounts.
behavior if extreme initial conditions are used.
Figure 15 demonstrates the control authority variation when
The last mass configuration scheme, case C, involved increasing
controlling the open box while flap 4 is held at a constant angle
the mass along two adjacent edges of the box. With the mass center
(4  30 deg) but additional mass is added. Here the mass was
near the center of the box, the air vehicle will either tumble or roll and
cone sideways. When more mass is added, the box rolls and cones varied from 521.7 to 746.7 g, which moved the mass center from
rcg!o  10:49IB  2:65KB cm to rcg!o  7:33IB  4:21KB cm.
down much like in case A. If the weight is further increased so the
location of the mass center is between rcg!o  7:43IB  2:08JB  As is shown, greater control authority can be achieved when the mass
along the front edge of the box is increased to a certain point. This is
2:08KB cm and rcg!o  6:64IB  2:28JB  2:28KB cm, the open
because the steady-state angle of attack is decreased with heavier
box glides like that of case A while gently rolling back and forth.
boxes, which moves the lift to drag ratio towards the maximum (see
Once again a neutrally stable pole exists for this mass configuration
Fig. 8). The maximum lift to drag ratio for a flap deflection of 30 deg
associated with the body frame side and vertical velocity as well as
is 1.40, which occurs at 20-deg angle of attack. This is achieved when
the body roll rate. The rolling motion dampens as the weight is
the mass is 731.7 g. Further increases in the mass result in smaller
further increased, allowing the box to glide with the weighted edge
steady-state angles of attack and lowers the glide ratio as shown by
pointing towards the ground. Such a behavior occurs when MP 5,
the curve associated with a mass of 746.7 g in Fig. 15.
IC 1, and no atmospheric winds are simulated. The associated
To achieve the maximum glide ratio for a specific flap deflection,
steady-tate pitch angle is 26:6 deg, the forward velocity is about
the center of mass should be designed so the equilibrium point occurs
25:6 m=s, and the roll rate is zero. The steady-sate angle of attack is
at an angle of attack that corresponds to the maximum lift to drag
8.0 deg, making the glide ratio about 1.45. Further increases in
ratio. Figure 16 shows the location of x
cg!o and z
cg!o for various
mass along the front edge result in decreased angles of attack and
flap 4 deflections to achieve the maximum lift to drag or glide ratio.
glide ratios.
Deviating from these mass center locations will tend to decrease the
glide ratio. This is demonstrated by considering the points in Fig. 16
Control Authority and Other Interesting Properties that represent the mass center location of the box configurations used
Depending on the flight behavior of the open box during earlier (and in Fig. 15). As is shown, increasing the weight moved the
atmospheric flight, one might expect that the open box could cut location of the mass center towards the optimum, which occurred
through the air quite nicely while gliding, but be blown around by when the mass was 731.7 g and then passed it resulting in a lower
atmospheric wind while coning or corkscrewing. Average wind glide ratio.
dispersion radii are given in Table 3, in which MP 1 and IC 2 were Another interesting behavior of the open-box configuration is the
used. In this case, the open box was dropped at an altitude of 2000 m ability to drastically reduce the rate of descent. Figure 17 shows the
at various atmospheric wind speeds with an azimuth varying between forward velocity of the box mass center as it falls from an altitude of
 834 BEYER AND COSTELLO

6000 30
Mass = 521.7 g
Mass = 536.7 g
Mass = 551.7 g
5000 Mass = 566.7 g 25
Mass = 581.7 g
Control Authority Radius (m)

Forward Velocity, u (m/s)

Mass = 596.7 g
Mass = 611.7 g 20
4000 Mass = 626.7 g
Mass = 641.7 g
Mass = 656.7 g
15
Mass = 671.7 g
3000
Mass = 686.7 g
Mass = 701.7 g
Mass = 716.7 g 10 δ1 = δ2 = δ3 = δ4 = 0 deg
2000 Mass = 731.7 g δ1 = δ2 = δ3 = δ4 = 20 deg
Mass = 746.7 g δ1 = δ2 = δ3 = δ4 = 30 deg
5 δ1 = δ2 = δ3 = δ4 = 50 deg
1000 δ1 = δ2 = δ3 = δ4 = 70 deg
δ =δ =δ =δ =
1 2 3 4
90 deg
0
0 20 40 60 80 100 120 140
0 Time (sec)
0 1000 2000 3000 4000
Altitude (m) Fig. 17 Forward velocity of box mass center vs time for MP 3, IC 1, and
no atmospheric winds from an altitude of 2000 m (all four flaps deflected
Fig. 15 Average control authority radius vs altitude for various mass at 40 s).
configurations of case B with IC 2 and
4
30 deg.
Downloaded by BEIHANG UNIVERSITY on June 2, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.25301

make the open box a promising airdrop vehicle that can travel at high
11
δ4 = 10 deg speeds during most of the flight and avoid atmospheric wind effects
δ4 = 20 deg Mass Center that would normally create large dispersions such as for parafoils or
10 Locations
δ4 = 30 deg
of Boxes used parachutes. But at the same time, the open box has the ability to arrest
δ4 = 50 deg in Figure 15 impact velocity so fragile packages can be delivered.
9 δ4 = 70 deg
δ4 = 90 deg
x*cg→o (cm)

8
References
[1] Nordwall, B., “Micro Air Vehicles Hold Great Promise, Challenges,”
Aviation Week and Space Technology, Vol. 146, No. 15, April 1997,
7
p. 67.
[2] Mohan, S., and Sridharan, G., “Emerging Technologies for Micro-
6 Unmanned Air Vehicles,” Defense Science Journal, Vol. 51, No. 3,
July 2001, pp. 223–228.
5 [3] Jones, K. D., “Bio-Inspired Design of Flapping-Wing Micro Air
Vehicles,” The Aeronautical Journal, Vol. 109, No. 1098, Aug. 2005,
pp. 385–393.
4 [4] Abdulrahim, M., and Lind, R., “Control and Simulation of a Multi-Role
-8 -7 -6 -5 -4 -3 -2 -1 0
z*cg→o (cm) Morphing Micro Air Vehicle,” AIAA Paper 2006-6643, Aug. 2006.
[5] Henry, J., “Design, Construction, and Testing of a Folding-Wing, Tube
Fig. 16 xcg!o vs zcg!o for various
4 for maximum lift to drag ratio. Launched Micro Air Vehicle,” AIAA Paper 2005-1451, Jan. 2005.
[6] Hoerner, S. F., Fluid–Dynamic Drag, Hoerner Fluid Dynamics,
Bakersfield, CA, 1965.
2000 m (MP 3, IC 1, and no atmospheric winds). At 40 s, all four rear [7] Hoerner, S. F., and Borst, H. V., Fluid–Dynamic Lift, Hoerner Fluid
flaps are deflected to 0, 20, 30, 50, 70, or 90 deg to increase the Dynamics, Bakersfield, CA, 1975.
amount of drag. As is shown, with no flap deflection the open box [8] Daniel, D. C., Yechout, T. R., and Zollars, G. J., “Experimental
achieves speeds up to 24 m=s and with more flap deflection the Aerodynamic Characteristics of Missiles with Square Cross Sections,”
descent rate is arrested. When all the flaps are deflected by 90 deg, the Journal of Spacecraft and Rockets, Vol. 19, No. 2, March–April 1982,
pp. 167–172.
descent rate is reduced to approximately half of the descent rate
[9] Sigal, A., and Lapidot, E., “Aerodynamic Characteristics of
achieved with no flap deflection. This braking mechanism allows the Configurations Having Bodies with Square, Rectangular, and Circular
open box to travel at relatively high speeds to avoid unwanted wind Cross Sections,” Journal of Spacecraft and Rockets, Vol. 26, No. 2,
effects before slowing down before ground impact. March–April 1989, pp. 85–89.
[10] Mukhopadhyay, V., and Dugundji, J., “Wind Excited Vibration of a
Conclusions Square Section Cantilever Beam in Smooth Flow,” Journal of Sound
and Vibration, Vol. 45, No. 3, April 1976, pp. 329–339.
The open box is an emergent, unconventional micro air robot [11] Mukhopadhyay, V., “Galloping Oscillation of Square Section
envisioned to carry small flat cargo to specific ground coordinates. Suspended Cable in Smooth Flow,” Journal of Industrial
The results presented show that the flight mechanics of the open box Aerodynamics, Vol. 5, No. 2, Oct. 1979, pp. 35–51.
are largely dependent on the mass center location and are quite [12] Anderson, J. D., Fundamentals of Aerodynamics, McGraw–Hill, New
interesting, particularly the coning and corkscrewing behaviors. The York, 1991.
[13] Dole, C. E., and Lewis, J. E., Flight Theory and Aerodynamics: A
aircraft also has the capability to glide like that of a conventional
Practical Guide for Operational Safety, John Wiley & Sons, New York,
aircraft towards the ground in a stable manner. When the rear flaps 2000.
are used as control mechanisms, this open-box configuration can [14] Nelson, R. C., Flight Stability and Automatic Control, McGraw–Hill,
achieve high glide ratios that more than overcome typical wind New York, 1989.
disturbances. This unique aircraft also has the ability to drastically [15] Rowell, D., and Wormley, D. N., System Dynamics: An Introduction,
arrest descent rate at crucial moments. All these dynamic qualities Prentice Hall, Upper Saddle River, NJ, 1997.
 This article has been cited by:

1. Hao Sun, Qinglin Sun, Wannan Wu, Shuzhen Luo, Jin Tao. 2019. Flexible modelling and altitude control for powered
parafoil system based on active disturbance rejection control. International Journal of Systems Science 50:12, 2385-2408.
[Crossref]
2. Dongyang Li, Sijiang Chang, Zhongyuan Wang. 2017. Analytical Solutions and a Novel Application: Insights into Spin–
Yaw Lock-In. Journal of Guidance, Control, and Dynamics 40:6, 1472-1480. [Citation] [Full Text] [PDF] [PDF Plus]
3. Charles H. Murphy, William H. Mermagen. 2009. Spin-Yaw Lock-In of a Rotationally Symmetric Missile. Journal of
Guidance, Control, and Dynamics 32:2, 378-383. [Citation] [PDF] [PDF Plus]
4. Eric Beyer, Mark Costello. Autonomous Airdrop with a Flap Controlled Open Box . [Citation] [PDF] [PDF Plus]
Downloaded by BEIHANG UNIVERSITY on June 2, 2020 | http://arc.aiaa.org | DOI: 10.2514/1.25301