Professional Documents
Culture Documents
Hybrid Backstepping Control For Rotorcraft Guidance: Antoine Drouin, Jules Ghislain Slama, Felix Mora-Camino
Hybrid Backstepping Control For Rotorcraft Guidance: Antoine Drouin, Jules Ghislain Slama, Felix Mora-Camino
Abstract— The purpose of this communication is to implementations appear of direct interest for the design of
display a non-linear control approach based on a new two-layer control structure based on backstepping
backstepping for the positioning and orientation for a control laws. The proposed control approach is compared
four-rotor aircraft. Realistic rotorcraft flight dynamics with a nonlinear inverse control approach introduced in
are introduced and the effectiveness of its control [10].
channels is analyzed. Then two complementary
implementations of the backstepping control approach II. ROTORCRAFT FLIGHT DYNAMICS
are considered. The compatibility of these control The considered system is shown in figure 1 where rotors
approaches with a two-layer control structure devoted one and three are clockwise while rotors two and four are
to the guidance of the rotorcraft is displayed. The counter clockwise. The main simplifying assumptions
resulting control laws are detailed and their expected adopted with respect to flight dynamics in this study are a
performances are discussed. A simulation study is rigid cross structure, constant wind, negligible aerodynamic
performed where the performances of the proposed contributions resulting from translational speed, no ground
control structure are compared with those of a classical effect as well as small air density effects and negligible
non linear inverse control solution. response times for the rotors. It is then possible to write the
rotorcraft flight equations as follows [7].
I. INTRODUCTION
I N the last years a large interest has risen for the four-
rotor concept since it appears to present simultaneously
hovering, orientation and trajectory tracking capabilities
of interest for many practical applications [1]. The flight
mechanics of this rotorcraft are highly non-linear and
different control approaches [2], [3], have been considered
with little success to achieve either only autonomous
hovering and orientation or also trajectory tracking.
In this paper we consider the flight dynamics of a four- Fig. 1. Four rotor aircraft
rotor aircraft with fixed pitch blades. The control problem
of interest is the design of flight control laws enabling A. Rotorcraft Flight Equations
autonomous positioning and orientation for this class of
i ∈ {1, 2, 3, 4}
The rotor forces and moments are given by:
Fi = f ωi
rotorcraft. This study investigates the solution of this
2
problem using a backstepping control approach. Here it is (1-1)
i ∈ {1, 2, 3, 4} (1-2)
required that a single continuous control law performs the
whole maneuver while to achieve it by manual control it M i = k Fi = k f ωi 2
i ∈ {1, 2, 3, 4}
applied to airships flight control law design [4], is constraints:
0 ≤ ωi i ≤ ωmax
introduced. Two different design techniques are developed
i ∈ {1, 2, 3, 4} (2-2)
(2-1)
0 ≤ Fi ≤ Fi max = f ωmax
following the main guidelines of this approach. These two
i
A. Drouin is with the Automation and Operations Research Laboratory-
LARA of ENAC, 7 Avenue Edouard Belin, 31055, Toulouse, France (e- Since the inertia matrix of the rotorcraft can be considered
mail:antoine.drouin@enac.fr). diagonal with Ixx = Iyy, the roll, pitch and yaw moment
p& = (l ( F4 − F2 ) + k2 q r ) / I xx
J. Slama is with LAVI-PEM/Programa de Engenharia Mecânica of equations may be written as:
COPPE/UFJ, Rio de Janeiro, Brazil (e-mail:julesslama@yahoo.com.br).
(3-1)
q& = (l ( F1 − F3 ) + k 4 p r ) / I yy
F. Mora-Camino is with the Automation and Operations Research
Laboratory-LARA of ENAC, 7 Avenue Edouard Belin, 31055, Toulouse, (3-2)
France (corresponding author, phone:0033562174358, fax:
0033562174403, e-mail:felix.mora@enac.fr).
r& = (k ( F2 − F1 + F4 − F3 )) / I zz (3-3) problem can be also considered as a first step towards the
design of more efficient trajectory tracking systems.
The manoeuvre under study is, when performed manually
rates. Here k2 = ( I zz − I yy ) and k 4 = ( I xx − I zz ) , where
Where p, q and r are the roll, pitch and yaw body angular through direct radio control of the four engine thrusts (see
picture 1), quite difficult to be achieved in one step.
Let φ, θ and ψ be respectively the bank, pitch and heading and θ specially when considering the control of its
the length of the four arms of the rotorcraft.
angles, then the Euler equations relating the derivatives of horizontal position error (x-xc, y-yc).
a y = (1 / m)((sin(ψ ) sin(θ ) cos(φ ) − cos(ψ ) sin(φ )) F − d y ) of the rotor forces appears as differences so, we define new
a z = g − (1 / m )(cos(θ ) cos(φ ) F + d z ) u p = F4 − F2
attitude inputs uq and up as:
uq = F1 − F3
(5-2)
(5-3) (10.1)
In the heading and position dynamics, the effects of rotor
where x, y and z are the centre of gravity coordinates, m is the forces and moments appear as sums, so we define new
guidance inputs uψ and uz as:
F = F1 + F2 + F3 + F4
total mass of the rotorcraft and:
( )
(6)
uψ = ( F2 + F4 ) − ( F1 + F3 ) uz = F = F1 + F2 + F3 + F4
Here the drag force d = d x
(10.2)
F = [ F1 F2 F3 F4 ] ' u = [u p u q uψ u z ] '
dy d z ' is given by:
with c = 1 / 2 ρ S C d where ρ is the volumetric mass of the Equations (3-1), (3-2) and (3-3) are rewritten:
p& = (l u p + k 2 q r ) / I xx
air, S is the reference surface for the aerodynamics of the
(12-1)
q& = (l uq + k4 p r ) / I yy
rotorcraft and Cd is its dimensionless drag factor. The
p = 0, q = 0, r = 0, x& = 0, y& = 0, z& = 0 (9-1) can be seen as virtual controls for the horizontal position of
θ e = arctg ((c w / mg )(cosψ w x + sin ψ w y ))
the rotorcraft. Here the attitude dynamics are considered to
(9-2)
φe = arctg ((c w / mg )(sinψ wx − cosψ wy ) cosθ e ) (9-3)
be the fast dynamics , they are at the heart of the control
system. The heading and height dynamics are intermediate
while the dynamics of the horizontal position coordinates
Y = h( X )
respect to u. The control objective here is to design a
cannot be written easily in the form (13), but it obeys to:
X& = g ( X , U ) with
control law such that the state x1 can be stabilized at x 1c .
Here also, x 2 can be regarded as a virtual control input for (24)
the dynamics of x1 while the dynamics of x 2 are where X∈Rn, U∈Rm, Y∈Rm, g is a smooth vector
r1 = r2 = L = rm = 1 , where the
possible choice is:
G ( x1 , x1c ) = − Λ ( x1 − x1c ) (15) Here we assume that
where Λ is a positive definite symmetric matrix. Then in jacobian of B with respect to the control inputs is invertible.
Z 2 = Y&
In that case, two auxiliary outputs can be defined:
Z 1 = L (Y − Y c ) + Y& and
this case:
V1 ( x1 − x1c ) = W ( x1 , x1c ) =
( x1 − x1c )' ( x1 − x1c ) (16)
1 (26)
2 where L is a positive definite symmetric matrix. A
The whole dynamics can be expressed as: candidate Lyapunov function is then given by:
x& 1 = G ( x 1 , x 1c ) + z z& = w V2 = (Z 1 Z 1 + Z 2 Z 2 )
1 ' '
and (17-1) (27)
2
z = x 2 − G ( x 1 , x 1c ) The time derivative of V2 is such as:
V&2 = Z 1 ' Z 1 + Z 2 ' Z 2
where (17-2)
(28-1)
and
V&2 = ( L (Y − Y c ) + 2 Y& ) ' ( L Y& + Y&&) − L Y& ' Y& ⎡ l l tgθ sinθ k tgθ cosφ ⎤
⎢ I ⎥
or (28-2)
⎢ ⎥
0
l cosφ − k sinφ
I xx I xx (33-2)
⎢ 0 ⎥
L Y& + Y&& = −Λ( L Y + 2 Y& )
Choosing a control such as : xx
⎢ ⎥
0
J (Y ) = ⎢ ⎥
(29)
where Λ is another symmetric positive definite matrix. We ⎢ l sinφ k cosφ
I yy I yy
⎥
⎢ cosθ I yy cosθ I zz ⎥
0 0
have: ⎢ − cosφ cosθ ⎥
V&2 = −( L (Y − Y c ) + 2 Y& ) ' Λ ( L (Y − Y c ) + Y& ) − Y& ' L Y& ⎢ 0 ⎥
⎣ ⎦
(30) 0 0
m
and it is straightforward to show that the system is globally
cos 2 φ tgθ sin φ
and the above matrix is invertible if:
− (sin θ sin φ + cos2 φ ) ≠ 0
asymptotically stable. The corresponding control law is (34)
U = − B −1 ( X )(Λ L (Y − Y c ) + ( L + 2Λ) Y& + A( X )) (31)
given by :
which is the case when φ and θ remains small with respect
I zz I yy
x& = v x
corresponding to the second case. The above approaches of by:
y& = v y
backstepping are now applied to each of the control layer (36-1)
necessary to perform attitude control and guidance of the
(36-2)
v& x = (1 / m)((cos(ψ ) sin(θ ) cos(φ ) +
rotorcraft.
A. Control of Rotorcraft Attitude and Level
sin(ψ ) sin(φ )) F − d x )
The attitude and altitude dynamics can be given by the state (36-3)
+ Pq ( p , q , r , φ , θ )
(32-4) corresponding backstepping approach, we get with V1
φc = arcsin((sin ψ (m ε x + d x )
(32-5)
ψ&& = k cosφ uψ /(cosθ I zz ) + l sin φ u p /(cosθ I yy ) (32-6)
− cosψ (m ε y + d y ) / uz )
(37-1)
+ gψ ( p, q, r , φ , θ ,ψ )
θ c = arcsin((cosψ (mε x + d x ) / u z
z& = v z
+ sinψ (mε y + d y ) / u z / cos φc )
(37-2)
&z& = g − (1 / m)(cos(θ ) cos(φ ) u z + d z )
(32-7)
⎡ε x ⎤ ⎡ x − xc ⎤ ⎡ x& ⎤
Pp ( p, q, r , φ , θ ) , ⎢ε ⎥ = −( I 2 + Ω Λ ) ⎢ ⎥ − Ω⎢ ⎥
⎣ y − yc ⎦ ⎣ y& ⎦
where the exact expressions of
⎣ y⎦
(37-3)
Pq ( p, q, r ,φ ,θ ) and g ψ ( p, q, r , φ , θ ,ψ ) can be derived from
where Λ and Ω are symmetric positive definite matrices.
relations (3-i) and (4-i) .
Then, the horizontal position of the rotorcraft follows the
the form (24) with Y=[φ θ ψ z]’, U = u with
The outputs dynamics (32-2), (32-4), (32-6) and (32-8) take
linear dynamics:
⎡ &x&⎤ ⎡ x& ⎤ ⎡ x − xc ⎤
u = (u p , u q , uψ , u z ) ' . Here X = ( p, q, r , φ , θ ,ψ , z&, z ) ' ⎢ &y&⎥ + Ω ⎢ y& ⎥ + ( I 2 + Ω Λ ) ⎢ y − y ⎥ = 0
⎣ ⎦ ⎣ ⎦ ⎣ c⎦
(38)
B( X , u ) = J (Y ) u
and B(X, u) is such as:
(33-1) Since the actuator settings are determined by the inner
L = diag ( μ1 , μ 2 , L , μ m )
and (40-2)
TABLE II
λi and the μi are positive real, we get always real negative ζθ = 0.8 ωθ = 10 rad/s
and where s is the Laplace variable. In this case, since the SELECTED DYNAMIC PARAMETERS
roots. In the general case, the dynamics modes of the ζφ = 0.8 ωφ = 10 rad/s
outputs will be characterized by the solutions of the global ζψ = 0.8 ωψ = 2 rad/s
− Im
ζz = 0.8 ωz = 1.5 rad/s
characteristic polynomial:
⎡ sI ⎤
det( ⎢ m ⎥) = 0 ζx = 0.8 ωx = 1.5 rad/s
⎣Λ L sI m + ( L + 2Λ )⎦
(41)
ζy = 0.8 ωy = 1.5 rad/s
Since this last relation is independent of the application it is
possible to study once for all the reachable pole sets within
the left half complex plane. The results (see figures 2 to 5) show that the two control
Remark: In the case of an horizontal wind, a necessary laws, in both levels, present equivalent performances.
θ cc θ
Decision and Control, Piscataway, NJ, 1978.
Backstepping
[5] R.Ghosh and Tomlin,C. J., “Nonlinear Inverse Dynamic Control for
Model-based Flight”, Proceeding of AIAA, 2000.
[6] R. Asep, R., Shen, T.J., Achaíbou, K. and Mora-Camino, F., An
application of the nonlinear inverse technique to flight-path
θcc θ
supervision and control, Proceedings of the 9th International Non Linear Inverse
Conference of Systems Engineering, Las Vegas,NV, 1993.
~
[7] B. Etkin, B. and Reid L. R., Dynamics of Flight-Stability and Control.
John Wiley & Sons. New York, NY, 1996.
[8] W.C.Lu, Mora-Camino, F. and Achaibou, K. , “Flight Mechanics and Fi
Differential Flatness”. Dincon 04, Proceedings of Dynamics and mg/4
Control Conference, Ilha Solteira, Brazil, pp. 830-839, 2004.
[9] T. Miquel, « Contribution à la synthèse de lois de guidage relatif, t
approche non linéaire ». PhD Thesis, Université Paul Sabatier,
Toulouse, 2004.
[10] A. Drouin, Miquel T. and Mora-Camino F. , Nonlinear Control
(with θ, θc, ωx=1.5, rad/s, ωθ=15 rad/s)
Fig.4. Step response of x
Structures for Rotorcraft Positionning, AIAA-GNC, Honolulu,
August 2009.
FIGURES
zc
φ
φc
BS
NLI
NLI
BS
Fi Fi
Fi NLI
mg/4
BS
t mg/4
t
( NLI : non linear inverse, BS : backstepping )
Fig.2 Step response of φ ( ωφ=10 rad/s) Fig. 5. Step response of z ( ω z = 1.5 rad / s )
φ
φc
NLI
BS
Fi
mg/4