Proceedings of 2004 IEEElRSJ InternationalConference e n

Intelligent Robots and Systems

September 28 -October 2.2004, Sendai, Japan

Stabilization of a mini-rotorcraft having four

rotors
P. Castillo, R. Lozano A. Dzul
Heudiasyc - UTC U M R CNRS 6599 Instituto Tecnol6gico de la Laguna
Compikgne, France. Torrebn, Coahuila, Mkxico
Email: castillo@hds.utc.fr, rlozano@hds.utc.fr Email: dzul@itlalaguna.edu.mx

Abslmcl-In this paper we present a controller design and

implementation on a mini-rotorcraft having four rotan. A
Lagrangien model of the helicopter was used for the controller
synthesis. The propmed contmller is based on Lyapunov
ana1)sis. Experimental results show that the controller is able
to perform autonomously the tasks of taking-off,hovering and
landing.

I. INTRODUCTION
Flight control problems for small unmanned helicopters
and some special airplanes has attracted the attention of
control researches in the last decade. The classical control
strategies for helicopters basically assumed a linear model
obtained for a p d c u l a r operating point. The use of modem
nonlinear conuol theory should impmve the pelfomance
of the controller and allow the tracking of agressive trajec-
tories.
In this paper we obtain a nonlinear model for a four-rotor
mini-rotorcraft and propose a nonlinear control strategy
based on Lyapunov analysis. The problem is closely related
to the problem of controlling a PVTOL i.e. Planar Verti-
cal Take-off and Landing aircraft. Generally. the control
strategies are based on simplified models which have both
a minimum number of states and a minimum number
of inputs. These reduced models should retain the main
features that must he considered when designing control
laws for real aerial vehicles.
The classical helicopter is one of the most complex fly-
ing objects. The classical helicopter is basically equipoed
of a main rotor and a tail rotor.
Fig. 1. Riadcr romion direction in the four-m$orrotor"
Mi
" = X%%+f,
A=&.", i-2, ..., 4
However other type of helicopters exist including the
twin rotor or tandem helicopter and the co-axial rotor
helicopter. We are particularly interested in conuolling a
mini-rotorcraft having four rotors (see Fig. 1).
A. CHARACTERISTICS OF THE FOUR-ROTOR ROTOR-
CRAFT
In this type of helicopters the front and the rear mo-
tors rotate counter-clockwise while the other two rotate
clockwise (see Fig. I), gyroscopic effects and aerodynamic
torques tend to cancel in trimmed flight.
The four-rotor rotorcraft does not have a swashplate
as standard helicopters. In fact it does not need any
senromechanism. The main thrust is the sum of the thrusts
of each motor (see Fig. 2). Pitch movement is obtained
by increasing (reducing) the speed of the rear motor while
reducing (increasing) the speed of the front motor. The roll
movement is obtained similarly using the lateral motors
(see Fig. 3).
The yaw movement is obtained by increasing (decreas-
ing) the speed of the front and rear motors while decreasing
(increasing) the speed of the lateral motors. This should be
done while keeping the total thrust constant (see Fig. 3).
In view of its configuration, the four-rotor rotorcraft in
figure 1 has some similarities with the PVTOL (Planar
Vertical Take Offand Landing aircraft) problem. Indeed,
if the roll and yaw angles are set to zero, the four-rotor
rotorcraft reduces to a PVTOL (see Fig. 3a). In a way the
four-rotor rotorcraft can be seen as two PVTOL connected
such that their axes are orthogonal (see Fig. 3a and 3b).

0-7803-84636104/120.00 e2004 IEEE

i!693
 The model of the full rotorcraft dynamics is obtained
from the Euler-Lagange Equations with external general-

where Ft = RF^ is the translational force applied to

the rotorcraft due to the throttle control input, T E P3
represents the pitch, roll and yaw moments, and R denotes
the rotational matrix R($, Q,$) E SO(3) representing the
t' orientation of the rotorcraft relative to a fixed inertial frame
E) (see figure 5 ) . From figures (2) and (4) we have that

(C)
F is a vector given by

E= (e)
where U is the collective control input expressed as
4
Fig. 3. (a) Pitch. (b) roll and ( c ) yaw C O ~ V Oinputs
I
u=cri
i=1

This article begins with a introduction that describes and f i is the force produced by motor M i , i = 1,..,4
the main characteristics of the mini-rotorcraft having four as shown in (2). Typically f i = k,w: where ki is a
rotors. After, we present the dynamical model obtained via constant and w, is the angular velocity of the i-rh motor.
a Lagrange approach. A control strategy is proposed having The generalized torques are thus
in mind that the four rotor rotorcraft can be seen as the
interconnection of two PVTOL.
The control algorithm is based on saturation control
strategy. We then prove global stability of the proposed
controller. Furthermore the controller has been imple- where C is the distance between the motors i d the center
mented on a PC and real time experiments have shown of gravity and TA{, is the moment produced hy motor A{,,
that the proposed control strategy performs well in practice. i = 1,.., 4. around the center of gravity of the aircraft.
Finally, we discuss some conclusions.
11. DYNAMICAL 0
moue^
In this section we will describe the dynamical model we I,, ma
have used for the four-rotor mini-rotorcraft, This model
is basically obtained representing the mini-rotorcraft as a
solid body evolving in 3D and subject to one force and
3 momens [IO]. The four electric motors' dynamics is .. .
relatively fast and therefore it will he neglected as well as
the flexibility of the blades. The generalized coordinates of
the rotorcraft are
q=(z,y,=,$,8.$)ER6
where (x.y, 2 ) = E E @ denote the position of the center
of mass of the four-rotor rotorcraft relative to the frame Z
and ($,e,#) = q L3are the three Euler angles (yaw,
pitch and roll) [SI,[9]and represent the orientation of the
rotorcraft (see figure 4).
The Lagrangian of the rotorcraft is
U q .9') = Ttrma + Tvot - U
where T~,,,, = tiTi is the translational kinetic energy,
Trot= $jT.Jq is the rotational kinetic energy, U = mgz I EX
is the potential energy of the rotorcraft, z is the rotorcraft
altitude, m denotes the mass of the rotorcraft, the mamx
J
. represents the inertia matrix and g is the gravity. Fig. 4. Schema of the four-rotor rotorcraft.

2694
 Since the Lagrangian contains no cross terms in the Phore Name DC!SCrip,iM?

kinetic energy involving i and e,

the Euler Lagrange
1 Control Alcilude
21 is used to d c the
allirude w c h a desi-
equations can he pnnitioned into two parts. One part for red value
the dynamics of the E coordinates and the other part for i y . is used to set [he
the 1) cmrdinates. One obtains 2 Yaw C0"Ml yaw displacemen1 1"

mx = -usin0 (3)
my = ucosOsind (4)
mi = ucosBcosQ-mg (5)
4 = ?$ (6)
0 = ?e (7)
4.- TV (8) 'TABLE I
where x and y are the coordinates in the horizontal plane, CONTROL
STRATEGY SEQUENCE
and z is the vertical position (see figure 4). $ is the yaw
angle around the z-axis, 0 is the pitch angle around the
(new) y-axis, and $ is the roll angle around the (new) x-
axis. We use the notation cg for cos B and sg for sin 8. u is Finally the pitch angle 0 and the x-displacement are
the thrust directed out the hottom of the aircraft and4, ?e controlled using again a strategy designed for the PVTOL.
and ?+ are the moments (yawing moment, pitching moment The proposed control strategy is relatively simple. This
and rolling moment), which are related to the generalized property is important when performing real-time applica-
torques T*, Q,T+ as follows: tions. Furthermore, the four control inputs can indepen-
dently operate in either manual or automatic mode , For
flight safety reasons this feature is particularly important
when implementing the control strategy.
Roughly speaking each of the control inputs can be used
to conml one or two degrees of freedom as follows: The
where C(vl 6 ) is the Coriolis matrix control input U is essentially used to make the altitude
reach a desired value. The conuol input T+ is used to
111. CONTROL STRATEGY set the yaw displacement to zero. io is used to control
In this section w'e will develop a control strategy for the pitch angle and the horizontal movement in the x -
stabilizing the rotorcraft at hover. axis. Similarly ?+ is used to control the roll angle and the
The controller synthesis procedure regulates each of the horizontal displacement in the y-axis.
states variables in a sequence using a priority rule as
follows: A. Conrml of the altitude and yaw displacemenr
We first design a control to stabilize the yaw angular The control of the altitude is obtained by using the
displacement. We then control the roll angle 4 and the y- following control input.
displacement using a controller designed for a PVTOL.

where
TI = -uz,i - a;& - "d) (10)

where a x > ,oz2 are positive constants and z d is the desired

altitude. The yaw angular position can he controlled by
applying
?+ = -aq.,$ - a@2($- $ d ) (1 1)

Indeed, introducing (9)-(11) into (3)-(6) and provided

cec+ # 0, we obtain
tan 0
mx = -(r 1 + mg)-cos ,iJ

2695
 The control parameters a+>,a+*, a,,, a,, should be From the definition of u(s) we can see that 1 ui(s) /<
carefully chosen to ensure a stable well-damped response Mi. This implies that in a finite time, 3Tl such that 1 $ 15
of the rotorcraft.
From equation (141, (IO) and (15) it follows that lu $ d
and t i zd.
- Mm2 for t t 2'1. Therefore, f o r t 2 Tl, I $ + u ~ ~ ( <I<+ ~ )
ZA!,. It then follows that, V t 2 Tl

B. Control of the roll angle and the horizontal movement 0*,(4+~+~(Cm~))

= &+umz(Cc1) (26)
in the y-axis Using (24) and (26). we get
We will now find the input ?+ to control 4, and y. 4, i.1 = -om2 (6,) (27)
The control algorithm will be obtained step by step. The
final expression for ?* will be given at the end of this Then, from (7). (18) and (26) we obtain for t 2 Tl
section (see 46).Roughly speaking, for $ close to zero,
the (x,9) subsystem is represented by four integrators in
cascade.
ci F 4- u*,(<m, ) (28)
We will first consider the subsystem given by (8) and 2) Boundedness of $: To establish a hound for 8, define
(13). For simplicity we chose +bd 0. Therefore, from Q, as
-
(15) it follows that $ -t 0. Note also that from (IO) and
(14) that r1 0. For a time T large enough, r1 and $ are
arbitrarily small.
CBI = U1 +.m,(Q.)
Introducing the above in (27) it follows
(29)

To simply further the analysis we will chose the am- 4 = - 4 u 1 +oc(Q*)) (30)
plitude of the saturation function in the nested saturation
The upper bounds are assumed to satisfy
control law in such a way that after a some finite time,
141 < 1. Therefore, the difference tan($) - $ is also small. 22 4 , (31)
Thus, the subsystem (8), (13) reduces to
This implies that 3T2 such that I u1 15 A& f o r t > Tz.
U = g4 (16) From equation (22) we can see that 'dt >_ Tz. 1 $ 15 film3.
i =?Q (17) Mea should be chosen small enough such that tan($) 4.
which represents four integrators in cascade.
From (30) and (311, we have that for t 2 Tz,I u1 +
U , + + ~ ( < O1
~5
) ZAf,. It then follows that, V t 2 T2
I ) Boundedness of $: In order to establish a bound for
4let io be
+
~ B ~ ( U uo,(Cc.))
I = VI +~4.(<4~) (32)
?* = -0*,(4+.4,(6J) (18) Then, in view of the above, (30) reduces to
where ui(s)is a saturation function such that 1 oi(s) 15 Al.
f o r i = 0,1,... and <*,
will be defined later to ensure global v, = -" 1 -ag.(6,)
stability.
for 1 2 Tz.
We propose the following Lyapuno,. function
3) Bourwledness of y: To establish a bound for jl, let us
1. introduce the following function
V = -42 119)
2
Differentiating V, we obtain U2 = U1 + $ + -Y (33)
9
then
p = 43 (7.0)
U 2 = C1
y .
+4 + - (34)
and from the equation (17) and (18) we have 9
Using (161, (23, (30) and (32 ) into (34). we obtain
v = -duol(d+%(<41)) (21)
Note that if 14 /> "4. then V < 0, i.e. 3T1 such that
C2 = -.m,(Q.) (35)
I $' 15 At*, f o r t > 7-1. Now, define c+%as
We define
h = U2 + %(b.) (36)
U1 =$+$ (22) Let us rewrite (35) as
Differentiating (22) U* = -063(v2 +0O4(CC8)) * (37)
Cl = &+$ (23) We chose
= 4-%(d+4<+J) (24) hf43 t 2M+4 (38)
Let us choose We then have that in a finite time, 3T3 such that 1 uz 15
I&, 22 4 , (25) A?', for t 2 T3, this implies from (33) that y is bounded.

2696
 “h(Y+%(Q.)) = Y +%(Q,) (39)
Thus, after time T3,(37) reduces to
U2 = -U2 - U 6, (Cd (40)
4 ) Boundedness of y: To establish a bound for y, let us
define
U3 = U2
B +
2- + $ + -
Y
(41)
9 9
then
Y
+
1;s = 92 2- + $ + -
9
Ij
9
(42)
Finally using (16), (32), (37) and (39) into (42). we
obtain
U3 = -U $4 (Q,1 (43) Fig. 6. Pitch mnlrol scheme.
We propose cd, of the following form
sm, = v3 (44)
IV. EXPERIMENTAL
RESULTS
then
The control algorithm proposed above has been tested
U3 = -ua,(u3) (45)
in real-time experiments. The experimental platform is
4,
5 ) Cr,nverSence of $> y orid y to zem: From (45) it

-
composed of a Draganfly four-rotor rotorcraft, a Futaba
follows that for a time large enough u3 4 0. From ‘(37) 72 MHr radio, a PC pentium LI and a 3D tracker system
it follows that uz 0 and from equation (36) <+* + 0. (POLHEMUS) [3] for measuring the position (z, y, i)and
From (30) v, i0 then from (29) + 0: c+! orientation ($,@,$)of the rotorcraft. Tbe Polhemus is
We can see from equation (21) that 4 + 0. From

-
connected via RS232 to the PC (see Fig. 7).
equation (22) we get $ + 0. From (33) y + 0 and finally The radio and the PC (INTEL Pentium 11) are connected
from (41) y 0. using data acquisition cards (ADVANTECH PCL-818HG
The control input ?4 is given by and PCL-726).
TQ = + U$* (@+ i+
-O Q 1 ( i In order to simplify the experiments, the control inputs
+ -9e +
can be independently commuted between the automatic and
oma(2* + i the manual control modes (see Fig. 7).
The connection in the radio is directly made to the
joystick potentiometers for the gas, yaw, pitch and roll
. .
controls. The rotorcraft evolves freely in a 3D space
C. Control of rhe pitch angle und rhe horizontal displace-
without any flying stand. The control law also requires the
inelifin !he x - a i s
time derivatives of the position (x:y, z ) and the orientation
From equations (17) and (18) we obtain Q
from (12) gives
0, then
i
(ww.
i = -gtanB (47)
Finally we take the sub-system
x = -gtan8 (48)
0 = 70 (49)
As before, we assume that the control strategy will insure
a very small bound on I 0 I in such a way that tan(8) 0.
Therefore (48) reduces to
5 = -go (50)
Using a procedure similar to the one proposed for the
, B
roll control we obtain
i0 = -oas,(S+uas,(B+S+
x
ua,(28
9
+S -- +
x x
ua,(B+38-3- - -)I)) (51)
9 9 Fig. 7. RuJ-.lTme architecture or Ihc platform.

2697
 TABLE 11
G A I N VALUES USED I N THE CONTROL L A W

. .

l7m 1.4 n"*a;l n"-l

They are obtained in practice using the following a p
proximation qt % F, where q is a given variable and Fig. 9. Position and orientation of the rolarcdt. The dotted lines
T is the sampling period. The controller parameters that represent thc desired uajectory

were used in the experiment are given in Table II. These

parameters were tuned to obtain the best performance in
practice
The control objective in the practical experiment was to
make the mini-rotorcraft hover at an altitude of 30 [cm],
i.e. we wish to reach the position (z,y; z ) = (0,O: 30 [cm])
while ($: 8, $) = (0,O;0) (see Fig. 8).
Figures 8 and 9 show the performance of the controller
when applied to the rotorcraft.
V. CONCLUSION
We have proposed a stabilization control algorithm for
a minkatorcraft having four rotors. The dynamic model
of the rotorcraft was obtained via a Lagrange approach
and the proposed control algorithm is based on Lyaponov
analysis.
The proposed strategy has been successfully applied
to the min-rotorcraft, and the experimental results have
shown that the controller performs satisfactorily. We have
done some videos of the experiments.
REFERENCES
111 Barnes W. McCormick. Aemdymmicr Aemnouricr and Flighr Me-
chonics. John Wiles & Sons. INC., NEW York. 1995.
121 Hauser 1.. Sariry S . B Meyer ti.. Nonlincvr conuol dcsirn for
slightly nonminimum phase systcms: Application to V/STOL akmft.
Automorico, 28(4):665-679. 1992.
131 Psirack 3Space Palhcmur, User's Manual, Colchesier. Vcrmont,
USA.
141 Lorano K., Bmgliato 8..Egeland 0.. Maschkc E. Porrivio-based
corirmi qslem amiysir and design. Springer-Vcrlap. Communications
and Control EnginePring Series, 2030. ISBN 1-85233-285-9.
151 Fantoni I., Lomno K.. Contmi ofnonlineor medoriicol snde"uoied
SIPI IN. Springer-Verlag, Communications and Conlml Engineering
series, 2001.
161 Marconi L., lsidari A., S m n i A., Auionomour vertical landig on an
O s ~ i l l a l i nplatrom:
~ an intem&-modcl had approach, Aelomooca.
38: 21-32, 2w2.
171 Wayno P.. Gyroscopes that don't spin makc it cary to hover. New
York Times, 8th of August 2002.
[RI Alderetc T. S.. Simulator aem model implcmenlation, NASA Amcs
Research Center, MoNeu Field Wifomia.
191 Et!& B.. Djnamics oJFlighi, John Wiley and Sons, Inc.. New York
1959.
1101 Goldsrein H., Clolricol Mechanics, Second Edition, Addison Wes-
ley, 1980.

Fig. 8. Real-lime control of the rotaeraft

2698