Sunteți pe pagina 1din 5

Trans. Japan Soc. Aero. Space Sci.

Vol. 49, No. 165, pp. 187191, 2006


Technical Note

Robust Adaptive Variable Structure Control of a Flexible Spacecraft


Containing Input Nonlinearity/Dead-Zone

By Qinglei H U and Guangfu M A


Department of Control Science and Engineering, Harbin Institute of Technology, Harbin, 150001, China

(Received December 13th, 2005)

Key Words: Variable Structure Control, Vibration Suppression, Flexible Spacecraft, Attitude Maneuver, Input Nonlinearity

1. Introduction be accomplished using the derived controller for both


asymptotically and exponentially stable design cases.
Fine attitude control and stabilization are key technolo-
gies required for modern exible spacecraft, whose mis- 2. System Statement
sions, such as stereoscopic mapping, require high pointing
accuracy and stabilization. However, in the realistic envi- A general description of nonlinear systems containing
ronment, knowledge about system parameters such as iner- nonlinearity/dead-zone in the input function is given in
tia matrix and modal frequencies are usually not known, and the form of
various disturbances also occur. Therefore, disturbance re- x_t A Axt Bu f t
jection control strategies that are also robust to parametric 1
y Cx;
uncertainty are of great interest in spacecraft applications.
Variable structure control (VSC) is an ecient control where xt 2 Rn is a vector of state variables, u 2 Rm is a
technique that is applicable to a wide class of nonlinear sys- vector of control inputs of the system, A 2 Rnn is the state
tems subject to modeling uncertainty and external distur- matrix of the normal part of the uncertain dynamic system,
bances.1,2) In those works,1,2) VSC is limited to systems with B 2 Rnm is the input matrix of the uncertain dynamic
full-state feedback. In practical application, however, full- system, f t 2 Rm stands for external disturbances of the
state measurement might be neither possible nor feasible. system, A is the bounded uncertainty matrix of A, u 2
The problem of designing output feedback controllers via Rm is a vector of nonlinear input functions, and 0 0.
VSC to stabilize multi-variable plants has been investigated Throughout the remainder of this paper, the following as-
by Wang3) and Yallapragada.4) In addition, the inputs of a sumptions are made:
system are usually restricted by its physical structure and Assumption 1: For the nominal part of the uncertain dy-
energy consumption, which also gives the system inputs namic system with nonlinearities indicated in Eq. (1), the
such nonlinear characteristics as saturation, dead-zone, triplet A; B; C of the nominal system is controllable and
etc. The existence of input nonlinearity is a source of degra- observable.
dation or even worse, instability in system performance. Hsu Assumption 2: The uncertainty matrix and the disturb-
et al.5) have proposed VSC schemes for uncertain dynamics ance A; f meet the following matching condition
systems with dead-zone nonlinearity for solving regulation A BH; kHk  1 2a
problems. These control schemes can possess insensitivity
to matching uncertainties and disturbances. However, infor- f t Bd; kdk < 2 ; 2b
mation on the upper boundaries of uncertainty is required, where H and d are appropriate dimensions, and 1 and 2
which is generally not available or too expensive to assess. the unknown bounds of uncertainty and disturbance, respec-
Adaptive approaches may oer a simple and eective tool tively.
to solve this problem.1,6) Assumption 3: The nonlinear input function i ui ap-
The contribution of this paper is the design of a nonlinear plied to the system is mathematically dened as follows:
8
controller to achieve attitude maneuvering of a three-axis >  i ui  ui ui > ui
<
stabilized exible spacecraft under parametric uncertainty,
i ui 0 ui  ui  ui ; 3
external disturbances and control input nonlinearity/dead- >
:
zone. The proposed controller ensures the global reaching i ui ui ui < ui
condition of the sliding mode of an uncertain system in where   i > 0 and   i > 0 are nonlinear functions of ui ,
the presence of input nonlinearity/dead-zone without the and ui > 0 and ui > 0 are constants. In addition, for the
limitation of knowing the boundaries of uncertainty and ith input ui , the nonlinear input function i ui is assumed
the perturbations in advance. Numerical simulations show to satisfy the following inequalities:
that precise attitude control and vibration suppression can
ui  ui i ui  i ui  ui 2 for ui > ui 4a
2006 The Japan Society for Aeronautical and Space Sciences
188 Trans. Japan Soc. Aero. Space Sci. Vol. 49, No. 165

and with the adaptive gain pi > 0, then the reaching condition
ui ui i ui  i ui ui 2
for ui < ui ; 4b ST S_ < 0 is satised for fx : kxk k  g.
Before proving Theorem1, the following lemma has to be
 i and i  
where i    i for all ui , and all are con- introduced.
stants. Lemma: For all the input nonlinearities u that satisfy
Assumption 3, if the controller is selected as Eq. (6), then
3. Adaptive Variable Structure Output Feedback Con-
S~ T u  kS~ T ky; t: 9
trol Design
Proof of Theorem 1: From Eqs. (4) and (6), we can obtain
A linear switching surface is dened as
s~i 2
s~i i ui   y; t; 10
St Gy GCxt; 5 kS~k
where G g1 ; g2 ; . . . ; gm T 2 Rmn is a specied constant Therefore, the following inequality is held,
matrix such that detGCB 6 0. Xm
s~i 2
As a consequence, asymptotically and exponentially sta- S~ T u   y; t kS~ T ky; t 11
kS~k
i1
ble design methods are analyzed for the development of the
adaptive variable structure output feedback control law Proof of Theorem 1: Consider the following Lyapunov
without knowing the upper boundaries of uncertainty and function candidate
disturbance in advance.
1 2 
3.1. Asymptotically stable design V kSk p1 1 2 p2 2 2 pi > 0; 12
Without loss of generality, suppose that the dead-zone 2
function i ui is anti-symmetric with respect to origin where i ^i  i denotes the adaptation error.
(i.e., i i i , ui ui ui0 ). Taking the derivative of V with respect to time t,
Dening the singular value decomposition of GC as
V_ t ST S_ p  _ p  _
1 1 1 2 2 2
GC UV1T with U 2 Rmm ,  2 Rmm and V1 2 Rnm ,
T
and letting S~ ~s1 ; s~2 ; . . . ; s~m T BT CT GT S and  S GCAx Bu BHx d
minfi g, the adaptive variable structure output feedback
control can then be p1 1 _1 p2 2 _2 : 13
8 s~ By Eq. (11) in Lemma, we have
>
> 
i
y; t  ui0 if s~i > 0
>
> ~
< k S k  kS~ T kx; t  ST GCBNy   kSk: 14
ui 0 if s~i 0 ; 6
>
>
> s~i If x is decomposed as x xk xp , where xk 2 N GC and
>
: y; t ui0 if s~i < 0 xp 2 N ? GC, substituting Eq. (14) into Eq. (13), and using
kS~k
Eq. (7) with assumption GCB I, the upper boundary for
where the expression in Eq. (13) is
1 V_  ST GCAxk  ST S   kSk  kkSk ST d
y; t #y; t; >

ST HV1 1 U T S p1 1 _1 p2 2 _2
and
 kSkkGCAk  ST S 1 k1 kST S
T
S GCBNy  kSk
#y; t
kBT C T GT Sk
k:   kSk  kkSk 2 kSk p1 1 _1 p2 2 _2

Theorem 1: Consider the uncertain nonlinear system < kkSk 1  ^1 k1 kST S 2  ^2
Eq. (1) containing input nonlinearity and a dead-zone which
 kSk p1 1 _1 p2 2 _2
are subject to assumptions 13. For the control law in
Eq. (6), use kkSk  1 k1 kST S  2 kSk
N I GCAV1 1 U T G; 7 p1 1 _1 p2 2 _2 : 15
where  > ^1 k1 k is a scalar, and k is any positive con- Since i is assumed to be constant, we have _i 0 and the
stant. If  is chosen to satisfy  > kGCAk ^2 , expression _i ^_i holds. Inserting _i ^_i with Eq. (8) into
#y; t  0, and > 0, where ^i i 1; 2 is the estimate the right-hand side of the inequality Eq. (15) yields
of i i 1; 2, and can be obtained from the adaptive law
V_  kkSk: 16
1 T
^_1 kS GCBkk1 kkSk 8a Since k > 0 is chosen, V_ < 0. Using Barbalats Lemma,7)
p1
one concludes that St ! 0 as t tends to be innity.
1 T
^_2 kS GCBk 8b 3.2. Exponentially stable design
p2 If the control law Eq. (7) is modied as
Nov. 2006 Q. HU and G. M A: RAVSC of Flexible Spacecraft Containing Input Nonlinearity/Dead-Zone 189

8
> s~i
>
>   y; t  ui0 if s~i > 0
>
< k S~k
ui 0 if s~i 0 ; 17
>
>
>
> s~
:  i  y; t ui0 if s~i < 0
kS~k
where

q
 y; t y; t ST S=kS~k;
2
q > 0 with the adaptive law in Eq. (8).
Theorem 2: Consider an uncertain nonlinear dynamic
system Eq. (1), and suppose assumptions 13 are satised.
The variable structure output feedback control law is chosen
as Eq. (17) with the adaptive law in Eq. (8). Then the sys-
tem state globally and exponentially converges to the sliding Fig. 1. Response of disturbance torques.
mode.
Proof: Consider the same Lyapunov function candidate
Vt as given in Eq. (12). Using Eq. (17) and the same ma- ces and the natural frequencies are taken from Di Gennaro.8)
nipulations as Eqs. (1417), its derivative is given Suppose the initial values of the three attitude angles are
0 6 , 0 4 , and 0 4 . It is required to con-
1
V_ <  qkSk2 qV  0: 18 trol each of the attitude angles approaching zero from their
2 initial value within tf 150 s, and each of the vibration
Therefore, the system trajectories globally and exponential- mode coordinates satises
ly converge to the sliding mode.
i _ i i tf _ i tf 0;
4. Application to a Flexible Spacecraft i 1; 2; 3; 4. fs t is usually imprecise and satises
Assumption 2. Assume that d is a random disturbance
4.1. Mathematical model torque with mean 0 and standard deviation 1, whose
The dynamics of spacecraft with exible appendages maximum absolute dmax has been xed to 0.25 Nm, as
can be obtained from a Lagrangian approach.8) With the shown in Fig. 1.
assumption of small Euler angle rotations, the dynamic In order to reduce chattering in an actual implementation,
model of Di Gennaro8) can be approximated as the discontinuous control given as Eq. (6) is often replaced
" # " # " # with a smoothed control of the form:
JT  JR
T 0 0 0 0

Z _
Z Z 8 s~i

I 0 D 0 K >
>  y; t  ui0 if s~i > 0
>
> ~
S "
" # < k k
I ui 0 if s~i 0 21
; ; ; t  us 19 >
>
0 >
> s~i
: y; t ui0 if s~i < 0
 T kS~k "
where Z T , , and are roll, pitch and
yaw attitude angles, respectively, and ; ; ; t can be where " > 0 is small.
obtained by equating similar terms in Eq. (19) and the cor- For the controller in Eq. (21), four design parameters
responding spacecraft model given in Di Gennaro.8) ( , ,  and ") are involved. A considerable number of
Dening the state variable xs Z Z_ T 2 R2N3 , the simulations have been done for determining parameters
state-space spacecraft model describing the dynamics be- , ,  and ", while here only one case of those
havior of the exible spacecraft can be described by numerical studies is given for the space limitation:
x_s t As As xs t Bs s us fs t 10,  10,  5, k 0:1 and " 0:01. In addition,
20 the following initial values are arbitrarily chosen; that is
ys Cs xs :
^1 ^2 0.
Here, the nonlinearities in the input s us are given as With the designed controller in Eq. (21), the require-
saturation and dead-zone whose saturated and dead-band ments for the inner torque of each ywheel are illustrated
values are 10 Nm and 0.5 Nm, respectively. in Fig. 2(a). From this gure, we can see that the inner
4.2. Simulation results torque of each ywheel approaches 0 at the time of
In this section, a numerical application of the proposed 50 s, and the maximum magnitudes of the torque do not
control schemes to the attitude control of a exible space- exceed the saturated value, 10 Nm, but with chattering
craft is presented using MATLAB/SIMULINK software. slightly. The responses of roll, pitch and yaw angles
The nominal moments of inertia parameters, coupling matri- corresponding to the initial attitude values are shown in
190 Trans. Japan Soc. Aero. Space Sci. Vol. 49, No. 165

(a) Inner-torque of flywheels


(a) Inner-torque of flywheels

(b) Attitude angle and rate


(b) Attitude angle and rate

(c) Coordinate of 1st-4th vibration modes


(c) Coordinate of 1st-4th vibration modes
Fig. 2. The case for asymptotically stable design method.
Fig. 3. The case for exponentially stable design method.

Fig. 2(b). We can see that each of the three attitude angles
response almost approaches 0 at the time of 150 s, and less than 0.002 /s. However, with control input dead-
the steady error of each angle is less than 0.002 . The bot- zone/nonlinearities and external perturbations, fast and
tom plot of Fig. 2(b) shows the responses of angular veloc- precise attitude control was achieved for the current design
ities of the exible spacecraft. It is noted that an acceptable system. Meanwhile, responses of the rst four modes are il-
angle rate response was also achieved with a steady error lustrated in Fig. 2(c). The maximum induced vibration co-
Nov. 2006 Q. HU and G. M A: RAVSC of Flexible Spacecraft Containing Input Nonlinearity/Dead-Zone 191

ordinate of the 1st mode is eectively suppressed below variable structure output feedback controller can derive
0.1 kg1=2 m. It is noted that all of the vibration modal coor- the trajectories of an uncertain system with input
dinates and vibration rates approach 0 by the end of the nonlinearity/dead-zone onto the sliding mode, and (ii)
required control time. the variable structure output feedback control can also
For the second case, the controller Eq. (17) can also be be achieved through adaptive control without knowing
modied into a form similar to Eq. (21) by introducing the boundaries of the uncertainty and perturbation terms
the boundary layer. Here, parameters q1 q2 15 are in advance.
selected through a considerable number of numerical
simulations. The parameters N, ,  , ",  and  remain Acknowledgments
the same for a fair comparison. The same initial condition
and disturbances as above are also used. The simulations This work was supported by the Research Fund for the Doctoral
are shown in Fig. 3. The maximum torque of the control Program of Higher Education of China (Project Number:
torque is 8 Nm, which also does not exceed the saturated 20050213010). The authors wish to thank Prof. Gangbing Song
from the University of Houston, Prof. Brij N. Agrawal from the
value, 10 Nm. The three-axis attitude response almost
Naval Postgraduate School, and Seung-Bok Choi from the Inha
approaches 0 at the time of 100 s, and the steady error
University, for their useful designs and insightful comments. The
of each angle is also less than 0.001 as shown in the top plot authors would like to thank the reviewers and the Editorial Board
of Fig. 3(b). Figure 3(b) also shows the responses of angular for many suggestions that substantially improved the quality of the
velocities of the exible spacecraft with a steady error of paper.
less than 0.002 /s. It is clear that, from a comparison of
Figs. 2(b) and 3(b), the attitude and rate responses under References
the exponential design case are faster than the asymptotical
case, even if input dead-zone/nonlinearity is considered. In
1) Boskovic, J. D., Li, S. M. and Mehra, R. K.: Robust Adaptive Variable
addition, the maximum induced vibration coordinate of the Structure Control of Spacecraft under Control Input Saturation, J. Guid.
1st mode is eectively suppressed below 0.1 kg1=2 m as Control Dynam., 24 (2001), pp. 1422.
shown in Fig. 3(c). 2) Hu, Q. L. and Ma, G. F.: Variable Structure Control and Active Vibra-
tion Suppression of Flexible Spacecraft during Attitude Maneuver,
Aerospace Sci. Technol., 9 (2005), pp. 307317.
5. Conclusions 3) Wang, W. J. and Fan, Y. T.: New Output Feedback Design in Variable
Structure Systems, J. Guid. Control Dynam., 17 (1994), pp. 337340.
In this paper, an approach to the vibration suppression 4) Yallapragada, S. V., Heck, B. S. and Finney, J. D.: Reaching Conditions
for Variable Structure Control with Output Feedback, J. Guid. Control
of a three-axis stabilized exible spacecraft is investigated
Dynam., 19 (1996), pp. 848853.
during attitude maneuvering in the presence of bounded 5) Hsu, K. C., Wang, W. Y. and Lin, P. Z.: Sliding Mode Control for
model uncertainty, external disturbances and control input Uncertain Nonlinear Systems with Multiple Inputs Containing Sector
nonlinearity/dead-zone. The approach of the current study Nonlinearities and Dead-zone, IEEE Trans. Syst., Man, Cybernetics,
Part B: Cybernetics, 34 (2004), pp. 374380.
is the construction of an adaptive variable structure output
6) Haddad, W. M. and Kapila, V.: A Sliding Mode Controller with
feedback controller to stabilize uncertain dynamics sys- Improved Adaptation Laws for the Upper Bounds on the Norm of
tems with explicitly considering nonlinearity/dead-zone Uncertainties, Automatica, 34 (1998), pp. 16571661.
in the control input. It is shown that the controller guaran- 7) Popov, V. M.: Hyperstability of Control System, Spinger-Verlag,
Berlin, 1973.
tees the global reaching condition of the sliding mode
8) Di Gennaro, S.: Adaptive Robust Tracking for Flexible Spacecraft in
in uncertain systems. The signicant advantages of this Presence of Disturbance, J. Optim. Theory Appl., 98 (1998), pp. 545
control include the following: (i) the presented adaptive 568.

S-ar putea să vă placă și