Professional Documents
Culture Documents
http://dx.doi.org/10.1007/s12555-017-0294-9 http://www.springer.com/12555
Abstract: A new closed-loop control allocation method is proposed to distribute the force and moment increment
commands among multiple actuators at the same time. An innovative feedback configuration for the force and
moment increment produced by the actuators deflections is added into the trajectory and attitude control loop,
respectively. In addition, the trajectory and attitude controllers designed by nonlinear dynamic inversion are used
to derive the desired force and moment increment commands. Then, the stability of the closed-loop system is
guaranteed theoretically when the actuators are healthy and fail, where there is no a fault detection method to
identify the failure. Finally, the proposed method is used to achieve direct force control for decoupling the attitude
and flight trajectory in the longitudinal flight. The nonlinear simulation using Canard Rotor/Wing (CRW) aircraft
longitudinal model in fixed-wing mode demonstrates that this method can achieve two direct force control modes:
pitch pointing and direct lift, whether the actuators are healthy or fail.
Keywords: Closed-loop control allocation, decoupling control, direct force control, flight control, redundant actu-
ators.
Manuscript received May 22, 2017; revised September 12, 2017; accepted November 17, 2017. Recommended by Associate Editor Pinhas
Ben-Tzvi under the direction of Editor Fuchun Sun. This work is supported by the National Natural Science Foundation of China (No.
61603220, 61333005, 61473177), the Shandong Provincial Natural Science Foundation (ZR2014FQ008, ZR2017BEE058), and the Qingdao
Post-doctor Applied Research Foundation (2015183).
Wendong Gai, Jie Liu, Jing Zhang, and Yuxia Li are with the College of Electrical Engineering and Automation, Shandong University of Sci-
ence and Technology, Qingdao 266590, China (e-mails: gwd2011@sdust.edu.cn, {master_sdust_2014, zjing133, yuxia_sdust}@163.com).
* Corresponding author.
⃝ICROS,
c KIEE and Springer 2018
1356 Wendong Gai, Jie Liu, Jing Zhang, and Yuxia Li
V̇x = −qVz + Fx /m, Define Fz0 = −L0 cos α − D sin α + mg cos θ and ∆Fz =
V̇z = qVx + Fz /m, −∆L cos α , equation (6) can be simplified to
q̇ = My /Jy ,
(3) Fz = Fz0 + ∆Fz . (7)
θ̇ = q,
ẋg = Vx cos θ +Vz sin θ , Equation (7) is the same as the first equation in (2), there-
fore, the longitudinal force increment ∆Fz can be derived.
ḣ = Vx sin θ −Vz cos θ , According to the second equation of (3), (6) and (7), the
where q is the pitch angular rate. Vx and Vz are the compo- lift increment ∆L is described as
nents of flight velocity along the body longitudinal axes. ( cos α )−1 [ F 0 ( )
]
θ is the pitch attitude angle. m is the mass of the aircraft. ∆L = z
− V̇z − qVx . (8)
Jy is the moment of inertia. xg and h are the north and m m
vertical components of the aircraft position in the locally Therefore, the actual longitudinal force increment ∆Fz
level geographic frame on the surface of the Earth. produced by the actuators deflections is
In (3), Fx and Fz are the components of the resultant
[ ]
force along the body longitudinal axes, which is composed ( ) Fz0
of the engine thrust, aerodynamic forces and gravity. And ∆Fz = m V̇z − qVx − , (9)
m
they are defined by
{ where V̇z is measured by the longitudinal acceleration sen-
Fx = P + L sin α − D cos α − mg sin θ ,
(4) sor.
Fz = −L cos α − D sin α + mg cos θ , In addition, according to (3), the actual pitching mo-
where α is the angle of attack. The engine thrust P = ment increment produced by the actuators deflections can
Kp δ p , and Kp is the thrust coefficient, δ p is the throttle be derived by
setting. D and L are the drag and lift, respectively. q̇ − fq
According to (4), the longitudinal force Fz will be ∆My = , (10)
Jy
changed when the lift L or drag D change. The lift L is
defined by where q̇ is measured by the angular acceleration sensor,
L = L0 + ∆L, (5) and fq is given by
where ρ is the air density, S is the wing reference area, V 2.3. Attitude controller design
is the flight velocity, c̄ is the wing mean geometric chord, In Fig. 1, the purpose of attitude controller design is
C(∗) is the aerodynamic derivatives. q̄ and α̇¯ are defined to derive the desired moment increments ∆M M d from the
by attitude command r a . Having separated the longitudinal
{
q̄ = qc̄/(2V ), attitude states into fast and slow dynamics, the feedback is
(13) used to provide the system with desirable dynamics. Then,
α̇¯ = α̇ c̄/(2V ).
the desired pitch angular rate dynamic is specified by
2.2. Trajectory controller design q̇d = ωq (qc − q), (21)
In Fig. 1, the purpose of trajectory controller design is
to derive the desired force increments ∆F F d from the trajec- where ωq is the design parameter. qc is the commanded
tory command r t . According to (6), the changes of longi- pitch angular rate given by the slow dynamic loop.
tudinal force Fz is mainly determined by the lift L. There- Replacing the q̇ in the left of (10) by the q̇d from (21),
fore, the trajectory command is altitude command hc , and the desired pitch moment increment is
the desired longitudinal force increment ∆Fzd is desired lift
increment ∆Ld . q̇d − fq
∆Mdy = . (22)
To simplify the expressions of aerodynamic forces, the Jy
longitudinal force equations are given in the wind axes
coordinates. In addition the desired pitch attitude dynamic is similar to
(21)
dV
m = P cos α − D − mg sin γ ,
dt (14) θ̇d = ωθ (θc − θ ), (23)
mV d γ = P sin α + L − mg cos γ ,
dt where ωθ is the design parameter, and θc is the com-
where γ is the flight-path angle, and satisfies the following manded pitch attitude angle given by designer.
equation. Replacing the θ̇ in the fourth equation of (3) by the θ̇d
dh from (23), the commanded pitch angular rate is
= ḣ = V sin γ . (15)
dt
qc = θ̇d . (24)
The desired fight altitude dynamics is specified by
ḣd = kh (hc − h), (16) 2.4. Closed-loop force and moment dynamic control
allocation
where h is the actual altitude, and kh is the design param-
eter. The control allocation problem is to determine the actual
According to (15) and (16), the flight-path angle com- control vector u (t) ∈ R m according to the virtual control
mand γc is given by vector v (t) ∈ R l . Mathematically, given v (t), then the u (t)
is derived by
ḣd
γc = arcsin . (17)
V f (uu(t)) = v (t), (25)
Then the flight-path angle dynamics is defined as
where f : Rm 7→ Rl is the nonlinear mapping from u (t) to
γ̇d = kγ (γc − γ ), (18) v (t). The actuator dynamic can be defined as
where γ is the actual flight-path angle, which is computed
by γ = θ − α , and kγ is the design parameter. δ̇δ (t) = g (δ (t),uu(t)) , (26)
In (5), L0 determined by the aircraft configuration and
flight states is in the form of where δ (t) is the actuator deflection. In a digital flight
control system, u(t) subjects to the following constraints
1
L0 = ρ V 2 S(CL0 +CLα α +CLq̄ q̄), (19) [32]
2
where CL0 , CLα , and CLq̄ are the aerodynamic derivatives. u (t) ≤ u (t) ≤ ūu(t),
Submitting (5) and (19) into (14), the desired lift incre-
u (t) = max(δ min , (uu(t − T ) − δ̇δ max T )),
ment ∆Ld is
ūu(t) = min(δ max , (uu(t − T ) + δ̇δ max T )), (27)
∆Ld =mV γ̇d − P sin α + mg cos γ
1
− ρ V 2 S(CL0 +CLα α +CLq̄ q̄), (20) where δ min , δ max , and δ̇δ max are the lower and upper ac-
2 tuator position and rate constraints, respectively. T is the
where γ̇d is derived by (18). sample time.
A New Closed-loop Control Allocation Method with Application to Direct Force Control 1359
The dynamic control allocation can be expressed as the Because the change of B r (k) is small during the sample
following sequential quadratic-programming problem if time T , the following equation is derived.
no actuators are saturated.
{ B r (k) = B r (k − 1). (37)
W 1 [uuv (t) −uus (t)]∥2
u v (t) = arg min ∥W
Ω
u v (t)∈Ω Next, substituting (33), (36) and (37) into ( 35), equa-
} tion (35) is changed to
W 2 [uuv (t) −uuv (t − T )]∥2
+ ∥W
∆Y
Y (k) = ∆Y
Y (k − 1) +V R(k) − ∆Y
V (k) [∆R Y (k)] , (38)
Ω= arg min ∥W Bu v (t) −vv(t))∥ ,
W Ω (B (28)
u (t)≤uuv (t)≤ūu(t) where
[ ]
where us (t) ∈ Rm is the desired steady-state control in- F )−1 E (B
V (k) = B r (k)(II −F Br (k))† +G
G , V ∈ R l×l .
put, and u v (t) ∈ R m is the dynamic control allocation out-
(39)
put; W 1 , W 2 and W Ω are the square matrices of proper
dimensions.
√ ∥∗∥ denotes the Euclidean norm defined by The z transform of (38) is
∥x∥ = xT x.
∆Y
Y (z) V (z)
zV
Lemma 1 [9]: The explicit solution to (28) is = . (40)
∆R
R(z) z(II +VV (z)) −II
F u v (k − 1) +G
u v (k) = E u s (k) +F Gv (k), (29)
3. STABILITY ANALYSIS
where
The stability for the closed-loop dynamic system is sig-
E = (II −G W −2W 21 , F = (II −G
GB )W W −2W 22 ,
GB )W nificantly important. In this part, the closed-loop dynamic
( )† √
G = W −1 BW −1 , W = W 21 +W W 22 , (30) control allocation system stability is analyzed when the
actuators are healthy and fail.
where I is the identity matrix, and the symbol † denotes
the pseudoinverse operator A † = A T (A AA T )−1 . 3.1. Stability without actuator failures
In the steady-state, the dynamic control allocation out- Lemma 3 [33]: Given A ∈ C n×n , and A is normal ma-
put u v (k) = u v (k − 1) and the desired steady-state control trix, there is a unitary matrix U ∈ U n×n , such that
input u s (k) = B †v (k), therefore, equation (29) reduces to U H AU = diag(λ1 , λ2 , · · · , λn ), (41)
[ ]
u v (k) = (II −FF )−1 E B †v (k) +G
Gv (k) . (31) where U HU = U U H = I , and λ1 , λ2 , ..., λn are the eigen-
Lemma 2: Let explicit solution to (28) is defined in values of A .
Lemma 1, then the input-output relation of the dash-dotted Theorem 1: Let the input-output relation of closed-
frame in Fig. 1 is described as loop dynamic control allocation be defined in Lemma 2,
and suppose the matrix V (k) in (39) is a real normal ma-
∆Y
Y (z) zVV (z) trix with eigenvalues λ1 , λ2 , ..., λn . The closed-loop force
= , (32)
∆R
R(z) z(II +V V (z)) −II and moment dynamic control allocation system described
by (40) is stable, if |1 + λi | > 1 is satisfied.
where ∆YY (z) = [∆FF (z), ∆MM (z)]T , ∆R M d (z)]T .
F d (z), ∆M
R(z) = [∆F Proof: According to (40), the closed-loop characteris-
Proof: The actual force and moment increments ∆Y Y (k) tic equation of the closed-loop allocation system in the z
can be derived by plane is
∆Y
Y (k) = B r (k)uu(k), (33) V (z)) −II = 0 .
z(II +V (42)
where B r (k) is the actual control effectiveness matrix, and Considering Lemma 3, the following equation is de-
u (k) is the input of actuator in Fig. 1. rived.
Substituting (31) and (34) into (33), ∆YY (k) is where U is unitary matrix.
[ ] Substituting (43) into (42), we obtain
∆YY (k) =B F )−1 E (B
Br (k)(II −F Br (k))† +G G v (k)
λ1 0 · · · 0
Br (k)uu(k − 1),
+B (35) 0 λ2 · · · 0
H
UU H + zU
zU U . .. . . U −U
UU H = 0 .
.. . . 0
where the input of dynamic control allocation v (k) is
0 ··· 0 λn
v (k) = ∆R
R(k) − ∆Y
Y (k). (36) (44)
1360 Wendong Gai, Jie Liu, Jing Zhang, and Yuxia Li
Fig. 9. Deflections of actuators in direct lift mode (no fail- Fig. 10. Responses of attitude and trajectory in pitch
ures). pointing mode (failure).
and Systems, vol. 9, no. 6, pp. 1203-1208, December 2011. [18] X. Yu and J. Jiang, “Hybrid fault-tolerant flight control sys-
[click] tem design against partial actuator failures,” IEEE Trans-
actions on Control Systems Technology, vol. 20, No. 4, pp.
[4] W. Durham, “Constrained control allocation,” Journal of
871-886, July 2012. [click]
Guidance, Control, and Dynamics, vol. 16, no. 4, pp. 717-
725, July 1993. [click] [19] M. Qian, B. Jiang, and H. H. Liu, “Dynamic surface active
[5] S. Tang, S. Zhang, and Y. Zhang, “A modified direct allo- fault tolerant control design for the attitude control systems
cation algorithm with application to redundant actuators,” of UAV with actuator fault,” International Journal of Con-
Chinese Journal of Aeronautics, vol. 24, no. 3, pp. 299- trol, Automation, and Control, vol. 14, No. 3, pp. 723-732,
308, June 2011. [click] June 2016. [click]
[6] A. R. Teel and J. Buffingtion, “Anti-windup for an F-16’s [20] X. Yu, Y. Zhang, and Z. Liu, “Fault-tolerant flight control
daisy chain control allocation,” AIAA Guidance, Naviga- design with explicit consideration of reconfiguration tran-
tion and Control Conference, New Orleans, LA, 1997. sients,” Journal of Guidance, Control, and Dynamics, vol.
39, No. 3, pp. 556-563, December 2016.
[7] J. Petersen and M. Bodson, “Constrained quadratic pro-
gramming techniques for control allocation,” IEEE Trans- [21] B. Xiao and S. Yin, “Velocity-free fault tolerant and uncer-
actions on Control Systems Technology, vol. 14, no. 1, pp. tainty attenuation control for a class of nonlinear systems,”
91-98, January 2006. IEEE Transactions on Industrial Electronics, vol. 63, no.
7, pp. 4400-4411, July 2016. [click]
[8] T. A. Johansen, T. I. Fossen, and S. P. Berge, “Constrained
nonlinear control allocation with singularity avoidance us- [22] X. Yu, Y. Fu, and Y. Zhang, “Aircraft fault accommodation
ing sequential quadratic programming,” IEEE Transactions with consideration of actuator control authority and gyro
on Control Systems Technology, vol. 12, no. 1, pp. 211-216, availability,” IEEE Transactions on Control Systems Tech-
January 2004. [click] nology, DOI: 10.1109/TCST.2017.2707378, June 2017.
[9] O. Härkegrrard, “Dynamic control allocation using con- [23] X. Yu, and J. Jiang, “A survey of fault-tolerant controllers
strained quadratic programming,” Journal of Guidance, based on safety-related issues,” Annual Reviews in Control,
Control, and Dynamics, vol. 27, no. 6, pp. 1028-1034, vol. 39, No. 1, pp. 46-57, April 2015.
November 2004. [click]
[24] A. Cristofaro and T. A. Johansen, “Fault tolerant control al-
[10] A. Zhang, Q. Hu, and X. Huo, “Dynamic control alloca- location using unknown input observers,” Automatica, vol.
tion for spacecraft attitude stabilization with actuator un- 50, no. 7, pp. 1891-1897,June 2014. [click]
certainty,” Proceedings of the Institution of Mechanical En-
gineers, Part G: Journal of Aerospace Engineering, vol. [25] M. T. Hamayun, C. Edwards, H. Alwi, and A. Bajodah, “A
228, no. 8, pp. 1336-1347, June 2014. fault tolerant direct control allocation scheme with integral
sliding modes,” International Journal of Applied Mathe-
[11] Q. Hu, B. Li, and A. Zhang, “Robust finite-time control matics and Computer Science, vol. 25, no. 1, pp. 93-102,
allocation in spacecraft attitude stabilization under actuator March 2015.
misalignment,” Nonlinear Dynamics, vol. 73, no. 1, pp. 53-
71, January 2013. [26] Q. Hu, B. Li, and Y. Zhang, “Nonlinear proportional-
derivative control incorporating closed-loop control allo-
[12] F. A. Almeida, “Robust off-line control allocation,”
cation for spacecraft,” Journal of Guidance, Control, and
Aerospace Science and Technology, vol. 52, no. 5, pp. 1-
Dynamics, vol. 37, no. 3, pp. 799-812, March 2014. [click]
9, May 2016.
[27] A. Tomczyk, “Aircraft maneuverability improvement by
[13] J. Tjønnrras and T. A. Johansen, “Adaptive control alloca-
direct lift control system application,” Aerospace Science
tion,” Automatica, vol. 44, no. 1, pp. 2754-2765, November
and Technology, vol. 9, no. 8, pp. 692-700, November
2008. [click]
2005.
[14] F. Liao, K. Y. Lum, J. L. Wang, and M. Benosman, “Adap-
tive control allocation for non-linear systems with internal [28] G. Sachs and F. Holzapfel, “Predictor-tunnel display and
dynamics,” IET Control Theory Applications, vol. 4, no. 6, direct force control for improving flight path control,”
pp. 909-922, June 2010. [click] AIAA Guidance, Navigation and Control Conference and
Exhibit, Honolulu, Hawaii, August 2008.
[15] Y. Zhang and J. Jiang, “Bibliographical review on recon-
figurable fault-tolerant control systems,” Annual Reviews [29] G. M. Siouris, J. G. Lee, and J. W. Choi, “Design of a mod-
in Control, vol. 32, no. 2, pp. 229-252, December 2008. ern pitch pointing control system,” IEEE Transactions on
[click] Aerospace and Electronic System, vol. 31, no. 2, pp. 730-
736, April 1995.
[16] M. Bkanke, M. Kinnaert, J. Lunze, and M. Staroswiecki,
Diagnosis and Fault-tolerant Control, 2nd ed., Springer- [30] Z. Liu, Y. Wang, and X. Hao, “Coordinated landing control
Verlag, Berlin, Heidelberg, 2006. of unmanned aerial vehicle,” International Conference on
Electronics, Communications and Control, Ningbo, China,
[17] M. Blanke, R. Izadi-Zamanabadi, S. A. Bøgh, and C. P.
1965-1970, September 2011.
Lunau, “Fault-tolerant control systems - a holistic view,”
Control Engineering Practice, vol. 5, no. 5, pp. 693-702, [31] B. Stevens and F. Lewis, Aircraft Control and Simulation,
May 1997. [click] John Wiley & Sons, New York, 2003.
1366 Wendong Gai, Jie Liu, Jing Zhang, and Yuxia Li
[32] A. Kishore, S. Sen, G.Ray, and T. K. Ghoshal, “Dynamic Jie Liu received the B.Eng. degree in
control allocation for tracking time-varying control de- Automation from Shandong University of
mand,” Journal of Guidance, Control, and Dynamics, vol. Science and Technology, China, in 2015.
31, no. 4, pp. 1150-1157, July 2008. She is currently a M.Eng. candidate in
[33] R. Horn and C. Johnson, Matrix Analysis, Cambridge Uni- Control Engineer, Shandong University
versity Press, Cambridge, 1999. of Science and Technology. Her research
interests include nonlinear flight control,
[34] J. Chen, Q. Li, C. Liu, P. Li, and Z. Ren, “Integrated guid- system identification and adaptive con-
ance and control design based on a reference model,” In- trol.
ternational Journal of Control, Automation, and Systems,
vol. 14, no. 5, pp. 1299-1308, September 2016. [click]
Jing Zhang received the Ph.D. degree in
[35] L. J. Liu, C. H. Zhu, and Z. Yu, “Guidance and ignition
Pattern Recognition and Intelligent Sys-
control of lateral-jet-controlled interceptor missiles,” Jour-
tem from South China University of Tech-
nal of Guidance, Control, and Dynamics, vol. 38, no. 12,
nology, China, in 2011. She joined the
pp. 2455-2460, July 2015. College of Electrical Engineering and Au-
[36] W. Gai, H. Wang, and T. Guo, “Modeling and LPV flight tomation, Shandong University of Science
control of the canard rotor/wing unmanned aerial vehi- and Technology, Qingdao, China, in 2011.
cles,” Proceedings of 2nd Artificial Intelligence, Manage- Her research interests include intelligent
ment Science and Electronic Commerce, pp. 2187-2191, control theory, image process and pattern
September 2011. recognition, and the application of advanced control and opti-
[37] C. Mitchell and B. Vogel, “The canard rotor wing (CRW) mization technique to robots.
aircraft: a new way to fly,” AIAA International Air and
Space Symposium and Exposition: The Next 100 Years, Yuxia Li received the B.S. degree from
AIAA-2003-2517, July 2003. Shenyang Jianzhu University, China, in
1990, and the Ph.D. degree from Guang-
dong University of Technology, China, in
Wendong Gai received the Ph.D. degree 2005. She has been a Professor at the
in Navigation, Guidance and Control from College of Electrical Engineering and Au-
Beihang University, China, in 2013. He tomation, Shandong University of Science
joined the College of Electrical Engineer- and Technology, Qingdao, China, since
ing and Automation, Shandong Univer- 2008. Her current research interest cov-
sity of Science and Technology, Qingdao, ers memristor-based circuits and systems, nonlinear circuits and
China, in 2013. His research interests in- systems, intelligent robot.
clude modeling and nonlinear flight con-
trol of unmanned aerial vehicle, fault tol-
erant control, sense and avoid in multiple UAVs.