an Input-Redundant Aircraft
L. Cui, Ph.D. 1; Y. Yang 2; and Y. H. Liu 3
Downloaded from ascelibrary.org by University of California, San Diego on 03/07/16. Copyright ASCE. For personal use only; all rights reserved.
Abstract: A new scheme of robust fault-tolerant control allocation is designed for a discrete-time aerodynamic model in a research envi-
ronment (ADMIRE) aircraft model in this paper. The control strategy has two steps. In the first step, a discrete-time robust virtual controller is
designed to produce three axis moments. In the second step, the fault information will be used to reconfigure the robust fault-tolerant control
allocation once an actuator fault is detected. A fault-detection scheme is assumed to exist for estimating the actuator fault. The robust virtual
controller and the proposed robust fault-tolerant control allocation are applied to a discrete-time ADMIRE aircraft model. Simulation results
demonstrate that the reference signals can be well tracked when the control surface is damaged. Thus the proposed robust fault-tolerant
control allocation scheme is valid for input-redundant systems to recover from control surface failure, and simultaneously guarantee robust-
ness to the uncertain control effectiveness matrix and disturbance. DOI: 10.1061/(ASCE)AS.1943-5525.0000528. © 2015 American Society
of Civil Engineers.
Author keywords: Fault tolerant control; Control allocation; Flight control; Robust control.
Introduction actuator, the virtual controller remains unchanged and the virtual con-
trol signal is redistributed to the available redundant control surfaces.
Over the last three decades, the growing demand for safety and Previously, researchers have intensively investigated the AFTC,
reliability in control engineering system, especially in overactuated incorporating the controller design with CA. Marwaha and Valasek
flight control systems, has attracted many scholars to study fault- (2011) presented an approach to fault-tolerant control allocation for
tolerant control (FTC) systems (Patton 1997; Zhang et al. 2007; a Mars entry vehicle by using adaptive control. Alwi and Edwards
Zhang and Jiang 2008; Lombaerts et al. 2009). The main objective (2008) studied a sliding mode controller and an online control
of FTC is to maintain a certain performance of a closed-loop con- allocation module for AFTC. Fault-tolerant control allocation for
trol system when a fault or failure occurs. Two approaches can be a thrust controller floating platform using parametric programming
distinguished in this area: passive and active approaches. In the pas- has also been studied (Spjtvold and Johasen 2009). Casavola and
sive FTC (PFTC) approach, a fixed controller is designed so that it Garone (2010) proposed a fault-tolerant adaptive control allocation
will maintain stability when fault or failure occurs. In contrast, a scheme for overactuated systems by using an ad hoc online param-
fault detection and isolation scheme is introduced in active FTC eter estimator to perform online control reconfiguration. An and
(AFTC). Whenever a fault is detected and isolated, a supervisory Kwon (2010) designed robust control allocation, considering
system takes action and modifies the structure and/or the parame- the uncertainties in the system as an unknown input disturbance,
ters of the closed-loop control system. and a sliding mode controller is used to be robust to the uncertain-
Modern flight control systems are designed with redundant ac- ties with the existing method quadratic optimization employed to
tuators to achieve safety and reliability (Ducard 2009; Zhang et al. design control allocation module. Ye et al. (2009) showed a control
2008). This introduces more inputs to design AFTC systems for allocation scheme of robust fault-tolerant control, where the robust-
maintaining stability and certain performance in the presence of ness to uncertainty is also considered in the controller design, and
faults and failures. Control allocation (CA) is one of the significant the control reallocation is designed by the cascaded generalized
approaches to manage actuator redundancy for different control strat- inverse method (Bordignon 1996). The purpose of the cited liter-
egies of handling actuator faults (Alwi and Edwards 2008). The atures concerning AFTC using CA is to design a virtual controller
advantage of using CA for AFTC is that when fault occurs in the and use the estimated fault signal to tune the parameters in CA for
guaranteeing stability and required performance. The virtual con-
1
Engineer, Beijing Institute of Control and Electronic Technology, State troller remains unchanged even when a fault occurs. Previous lit-
Key Laboratory for Turbulence and Complex Systems, Dept. of Mechanics erature only considers robustness to uncertainty in the controller
and Aerospace Engineering, College of Engineering, Peking Univ., Beijing design; however, it is directly considered in the control allocation
100871, P.R. China. E-mail: cuilei2007@pku.edu.cn model of our new scheme.
2
Associate Professor, State Key Laboratory for Turbulence and Com- The aerodynamic parameter variation can be considered as a
plex Systems, Dept. of Mechanics and Aerospace Engineering, College parameter uncertainty for an aircraft. Thus, uncertainty exists in
of Engineering, Peking Univ., Beijing 100871, P.R. China (corresponding the control effectiveness matrix of control allocation. To deal
author). E-mail: yy@mech.pku.edu.cn with the problem of FTC for a faulty aircraft, a linear robust fault-
3
Engineer, Beijing Institute of Spacecraft System Engineering, Beijing
tolerant CA (RFTCA) is proposed in this paper. In a previous study
100094, P.R. China. E-mail: carilina122@163.com
Note. This manuscript was submitted on May 7, 2012; approved on May
(Cui and Yang 2011), robust least squares CA (RLSCA) was in-
29, 2015; published online on August 11, 2015. Discussion period open until vestigated. Further studies of RLSCA, a new AFTC architecture,
January 11, 2016; separate discussions must be submitted for individual pa- including robust virtual controller and RFTCA, is presented for
pers. This paper is part of the Journal of Aerospace Engineering, © ASCE, a discrete-time aerodynamic model in a research environment
ISSN 0893-1321/04015042(10)/$25.00. (ADMIRE) aircraft system. The novelty of this paper is to extend the
then d1 ðkÞ ¼ VX −1 xðkÞ is the H2 =H∞ optimal guaranteed perfor- where the uncertainty is put into MðuÞ, and the nominal control
mance virtual control law, and the poles of the closed-loop system effectiveness matrix is put into NðuÞ.
are assigned to a circle ð0; rÞ, 0 < r < 1. Theorem 2: The reconfigurable RFTCA problem has an
Proof: The proof is similar to the results in Yu (2002), thus it is optimal solution ðλ; τ ; urRFTCA Þ if the following matrix inequality
omitted here. is solved:
Remark 1: TraceðNÞ is related to the H2 performance index. " #
The Eq. (5) can be rewritten as TraceðBTd XBd Þ < TraceðNÞ, and τ I − ρ2 M T ðuÞMðuÞ −ρMT ðuÞNðuÞ
minimizing TraceðNÞ is to minimize the upper bound of H 2 min λ s:t: >0
performance index.
u≤u≤ū;λ;τ −ρN T ðuÞMðuÞ λ − τ − N T ðuÞNðuÞ
Downloaded from ascelibrary.org by University of California, San Diego on 03/07/16. Copyright ASCE. For personal use only; all rights reserved.
ð9Þ
Feedforward Controller Design
where the variable u ∈ Rm and the variables λ ∈ R; τ ∈ R are
The nominal system is represented as
positive.
Proof: In the presence of actuator fault, the squared
xðk þ 1Þ ¼ AxðkÞ þ Bddes ddes ðkÞ yðkÞ ¼ CxðkÞ worst-case residual is obtained by using the definitions NðuÞ
and MðuÞ
If the outputs have tracked the reference signals and the aircraft per-
forms at an equilibrium, the nominal system must satisfy the following:
r2f ðuÞ ¼ max kBðΘÞWuðkÞ − ddes ðkÞk2 ¼ max kθMðuÞ þ NðuÞk2
kθk⩽ρ kθk⩽ρ
xs ðk þ 1Þ ¼ Axs ðkÞ þ Bddes ddes ðkÞ yðkÞ ¼ Cxs ðkÞ ð6Þ ¼ max ðθMðuÞ þ NðuÞÞT ðθMðuÞ þ NðuÞÞ
θT θ≤ρ2
where the xs ðkÞ = stable trim point; xs ðk þ 1Þ ¼ xs ðkÞ, ¼ max ðθ2 M T ðuÞMðuÞ þ 2θM T ðuÞNðuÞ þ N T ðuÞNðuÞÞ
and yðkÞ ¼ rðkÞ. θT θ≤ρ2
T
Substituting the virtual control law ddes ðkÞ ¼ dref ðkÞ þ KxðkÞ M ðuÞMðuÞ M T ðuÞNðuÞ θ
into Eq. (6) yields ¼ max ½ θT 1
θT θ≤ρ2 N T ðuÞMðuÞ N T ðuÞNðuÞ 1
Cð−Bddes K − A þ IÞ−1 Bddes dref ðkÞ ¼ rðkÞ
With θ~ ¼ θ=ρ, it yields that
Moreover, it is obtained that 2 T
ρ M ðuÞMðuÞ ρM T ðuÞNðuÞ θ~
r2f ðuÞ ¼ max ½ θ~ T 1
dref ðkÞ ¼ ½Cð−Bddes K − A þ IÞ−1 Bddes −1 rðkÞ T
θ~ θ≤1
~ ρN T ðuÞMðuÞ N T ðuÞNðuÞ 1
ð10Þ
Denote that K ref ¼ ½Cð−Bddes K − A þ IÞ−1 Bddes −1 , then the
feedforward controller dref ðkÞ ¼ K ref rðkÞ is obtained. In order to minimize the upper bound of squared worst case
Finally, the stability and tracking property of the closed-loop residual (10), it must satisfy r2f < λ, i.e.
system are guaranteed by the robust virtual control law ddes ðkÞ ¼
K ref rðkÞ þ KxðkÞ.
½ θ~ T
ρ2 M T ðuÞMðuÞ ρM T ðuÞNðuÞ θ~ < λ
1
ρN T ðuÞMðuÞ N T ðuÞNðuÞ 1
Robust Fault-Tolerant Control Allocation
there exists a scalar τ ≥ 0 such that
As is well known, the uncertainty of the aerodynamic coefficient
in the control effectiveness matrix has a known structure in flight
control system (Frank and David 1999). Therefore, the following
½ θ~ T
τ I − ρ2 M T ðuÞMðuÞ −ρM T ðuÞNðuÞ θ~ > 0
1
reconfigurable RFTCA problem in the presence of actuator fault is −ρN ðuÞMðuÞ
T
λ − τ − N T ðuÞNðuÞ 1
considered.
Problem: In the presence of actuator fault, find the
optimal reconfigurable RFTCA solution urRFTCA in the interval This condition can be rewritten as
of ½u; ū as
τ I − ρ2 M T ðuÞMðuÞ −ρM T ðuÞNðuÞ
urRFTCA ¼ arg min max kBðΘÞWuðkÞ − ddes ðkÞk ð7Þ >0 ð11Þ
u⩽u⩽ū kθk⩽ρ
−ρN T ðuÞMðuÞ λ − τ − N T ðuÞNðuÞ
where ρ = the upper bound of the uncertainty Θ; and ū and u = the then the conclusion is obtained.
upper and lower bounds of the control vector u. The matrix inequality in Theorem 2 is nonlinear. Therefore, the
In order to obtain the optimal solution urRFTCA to the problem following Theorem 3 is obtained, and the optimal urRFTCA can be
of reconfigurable RFTCA, the following definitions are introduced: found by solving the linear matrix inequality (LMI).
Theorem 3: The reconfigurable RFTCA has an optimal solu-
MðuÞ ≜ ½B1 Wu; : : : ; Bq Wu ð8aÞ tion ðλ; τ ; urRFTCA Þ if the following is satisfied:
where the control vector is uT ¼ ½u1 ; : : : ; um ; the upper and Using the Schur complement lemma, it is easy to obtain
lower bounds of the control vectors are ūT ¼ ½ū1 ; : : : ; ūm ,
uT ¼ ½u1 ; : : : ; um ; and ξ i = an unit column vector and satis-
2 3
fies ½ξ 1 ; · · · ; ξ m ¼ I m×m . I ρMðuÞ NðuÞ
Proof: Eq. (11) can be rewritten as 6 7
4 ρM T ðuÞ τI 0 5>0
τ I − ρ2 M T ðuÞMðuÞ −ρM T ðuÞNðuÞ N T ðuÞ 0 λ−τ
−ρN T ðuÞMðuÞ λ−τ − N T ðuÞNðuÞ
2
τI 0 ρ MðuÞT MðuÞ ρMT ðuÞNðuÞ thus the non-LMI is transformed to an LMI.
¼ −
0 λ−τ ρM T ðuÞN T ðuÞ N T ðuÞNðuÞ In order to add the constraints to u by LMIs, the identity matrix
is given as I m×m ¼ ½ξ 1 ; : : : ; ξ m . Rewrite the constraints u < u < ū
τI 0 ρM T ðuÞ
¼ − ½ ρMðuÞ NðuÞ > 0 into two parts −ū < −u and u < u. Then it can be transformed into
0 λ−τ N T ðuÞ two LMIs:
2 3 2 3
ū1 − u1 0 ··· 0 ðū − uÞT ξ 1 0 ··· 0
6 0 ū2 − u2 ··· 0 7 6 0 ðū − uÞT ξ 2 ··· 0 7
6 7 6 7
−ū < −u ⇔ 6 .. .. .. .. 7>0⇔6 .. .. .. .. 7>0
4 . . . . 5 4 . . . . 5
0 0 ··· ūm − um 0 0 ··· ðū − uÞT ξ m
2 3 2 3
u1 − u1 0 ··· 0 ðu − uÞT ξ 1 0 ··· 0
6 0 u2 − u2 ··· 0 7 6 0 ðu − uÞT ξ 2 ··· 0 7
6 7 6 7
u<u ⇔6 .. .. .. .. 7>0⇔6 .. .. .. .. 7>0
4 . . . . 5 4 . . . . 5
0 0 ··· um − um 0 0 ··· ðu − uÞT ξ m
2 3
0.9947 1.364 × 10−4 1.51 × 10−7 0.009727 −6.591 × 10−7
Downloaded from ascelibrary.org by University of California, San Diego on 03/07/16. Copyright ASCE. For personal use only; all rights reserved.
6 7
6 0 0.9987 2.207 × 10−3 0 −9.638 × 10−3 7
6 7
6 7
A¼6 0 −0.1045 0.99 0 0.006644 7;
6 7
6 0.02609 −2.812 × 10 −5
−3.177 × 10 −8
0.9951 1.387 × 10−7 7
4 5
−3 −4
0 7.112 × 10 −9.255 × 10 0 0.9978
2 3 2 3
1.0 5.039 × 10−10 4.872 × 10−5 −2.199 × 10−9
6 7 6 7
6 4.967 7 6 1.105 × 10−5
0 −4.823 × 10−5 7
6 7 6 7
6 7 −3 6 7
Bd ¼ 6 7.724 7 × 10 ; Bddes ¼ 6 9.95 × 10−3
0 3.244 × 10−5 7;
6 7 6 7
6 6.116 7 6 −1.071 × 10−10
9.975 × 10−3 4.675 × 10−10 7
4 5 4 5
8.005 −4.65 × 10−6 0 9.989 × 10−3
2 3
0.6031 −0.6031 −2.2849 −1.9574 1.9574 2.2849 1.4871
6 7
B ¼ 4 0.8266 0.8266 −0.4628 −0.8107 −0.8107 −0.4628 0.0024 5; B̄ ¼ Bddes B;
−0.2615 0.2615 −0.0944 −0.1861 0.1861 0.0944 −0.8823
2 3
1 0 0 0 0 0
6 7
C0 ¼ C ¼ 6
40 1 0 0 0 05
7
0 0 1 0 0 0
The control input matrix B̄ is 5 × 7, the dimension of input ex- disturbance is set to be d ∼ Nð0,0.3Þ. The upper bound of uncer-
ceeds the degree of freedom, which means that the ADMIRE model tainty is chosen to be ρ ¼ 0.1.
is redundant in the control inputs and the reconfigurable RFTCA Fault Case 1: Assume that the control surface δ rie is completely
can be applied. The AFTC for the discrete-time linear ADMIRE damaged at the time t ¼ 6.5 s, the control surface δ r is 20% dam-
aircraft model is realized by redistributing the virtual control mo- aged at the time t ¼ 6.5 s and others stay healthy as
ments to the healthy control inputs.
The feedback and feedforward gain matrices are designed as
1; t ∈ ½0,6.5s; 1; t ∈ ½0,6.5s;
f4 ¼ f7 ¼
2 3 0; t ∈ ð6.5,15s: 0.8; t ∈ ð6.5,15s:
0.0059 10.5150 0.9924 −0.0107 −0.5987
6 7
K ¼ 4 −2.9612 0.0131 −0.0029 1.1171 0.0126 5
Fault Case 2: Assume that the control surface δ lie is completely
0.0064 −0.6908 0.0622 −0.0115 0.3510 damaged at the time t ¼ 8.5 s, the control surface δ r is 30% dam-
2 3
0.0110 0.0196 1.0000 aged at the time t ¼ 8.5 s and others stay healthy as
6 7
K ref ¼ 4 0.9520 −0.0009 0 5
0.0117 −1.0140 0.2293 1; t ∈ ½0,8.5s; 1; t ∈ ½0,8.5s;
f6 ¼ f7 ¼
0; t ∈ ð8.5,15s: 0.7; t ∈ ð8.5,15s:
and the robust virtual control law ddes ðkÞ ¼ KxðkÞ þ K ref rðkÞ is
obtained. Figs. 2 and 3 show the simulation results of tracking perfor-
mance to angle of attack α, sideslip angle β, and roll rate p; the
reference tracking performance in the presence of the actuator fault
Simulation Results
is similar to the one without fault.
The designed robust virtual controller and reconfigurable RFTCA Figs. 4 and 5 show that the reconfigurable RFTCA redistributes
are testified in a discrete-time linear ADMIRE aircraft model. The the virtual control moments to other healthy control surfaces δ rc ,
6 0.5
α (deg)
β (deg)
4
0
2
−0.5
0
Downloaded from ascelibrary.org by University of California, San Diego on 03/07/16. Copyright ASCE. For personal use only; all rights reserved.
−1
−2
−1.5
0 10 20 30 40 50 0 10 20 30 40 50
(a) Time(s) (b) Time(s)
4 3.5
Ref Fault free
3 Fault free 3 Fault
Fault
2.5
2
2
q (deg/s)
p (deg/s)
1
1.5
0
1
−1
0.5
−2 0
−3 −0.5
0 10 20 30 40 50 0 10 20 30 40 50
(c) Time(s) (d) Time(s)
0.6
Fault free
0.5 Fault
0.4
0.3
r(deg/s)
0.2
0.1
−0.1
−0.2
0 10 20 30 40 50
(e) Time(s)
Fig. 2. States of the ADMIRE aircraft using proposed robust fault-tolerant control allocation: Fault Case 1
δ lc , δroe , δ lie , and δ r when the faults occur in the control surfaces control effectiveness matrix, and the reference tracking perfor-
δ rie and δ loe . mance is guaranteed. Thus the robust controller and reconfigurable
According to the simulation results of two fault cases, the pro- RFTCA are valid to the discrete-time linear ADMIRE aircraft
posed reconfigurable RFTCA is robust to the uncertainty in the for AFTC.
0.5
−2
α (deg)
β (deg)
0
−4
−0.5
−6
Downloaded from ascelibrary.org by University of California, San Diego on 03/07/16. Copyright ASCE. For personal use only; all rights reserved.
−8 −1
−10 −1.5
0 10 20 30 40 50 0 10 20 30 40 50
(a) Time(s) (b) Time(s)
4 0.5
Ref Fault free
Fault free Fault
0
3 Fault
−0.5
2
−1
p (deg/s)
q (deg/s)
1
−1.5
0
−2
−1
−2.5
−2 −3
−3 −3.5
0 10 20 30 40 50 0 10 20 30 40 50
(c) Time(s) (d) Time(s)
0.2
Fault free
Fault
0.1
−0.1
r (deg/s)
−0.2
−0.3
−0.4
−0.5
−0.6
0 10 20 30 40 50
(e) Time(s)
Fig. 3. States of the ADMIRE aircraft using proposed robust fault tolerant control allocation: Fault Case 2
5 5
δrc(deg)
δlc(deg)
0 0
−5 −5
−10 −10
Downloaded from ascelibrary.org by University of California, San Diego on 03/07/16. Copyright ASCE. For personal use only; all rights reserved.
−15 −15
0 10 20 30 40 50 0 10 20 30 40 50
(a) Time(s) (b) Time(s)
20 20
Fault free Fault free
15 Fault 15 Fault
Bound Bound
10 10
5 5
δroe(deg)
δrie(deg)
0 0
−5 −5
−10 −10
−15 −15
−20 −20
0 10 20 30 40 50 0 10 20 30 40 50
(c) Time(s) (d) Time(s)
20 20
Fault free Fault free
15 Fault 15 Fault
Bound Bound
10 10
5 5
δloe(deg)
δlie(deg)
0 0
−5 −5
−10 −10
−15 −15
−20 −20
0 10 20 30 40 50 0 10 20 30 40 50
(e) Time(s) (f) Time(s)
20
Fault free
15 Fault
Bound
10
5
δr(deg)
−5
−10
−15
−20
0 10 20 30 40 50
(g) Time(s)
δrc(deg) 5 5
δlc(deg)
0 0
−5 −5
−10 −10
Downloaded from ascelibrary.org by University of California, San Diego on 03/07/16. Copyright ASCE. For personal use only; all rights reserved.
−15 −15
0 10 20 30 40 50 0 10 20 30 40 50
(a) Time(s) (b) Time(s)
20 20
Fault free Fault free
Fault Fault
15 Bound 15 Bound
10 10
5 5
δroe(deg)
δrie(deg)
0 0
−5 −5
−10 −10
−15 −15
−20 −20
0 10 20 30 40 50 0 10 20 30 40 50
(c) Time(s) (d) Time(s)
20 20
Fault free Fault free
Fault Fault
15 Bound 15 Bound
10 10
5 5
δloe(deg)
δlie(deg)
0 0
−5 −5
−10 −10
−15 −15
−20 −20
0 10 20 30 40 50 0 10 20 30 40 50
(e) Time(s) (f) Time(s)
20
Fault free
Fault
15 Bound
10
5
δr(deg)
−5
−10
−15
−20
0 10 20 30 40 50
(g) Time(s)
tuator fault, and (4) the closed-loop system is robust to uncertainty, flight control algorithm.” J. Guidance Control Dyn., 32(6), 1747–1765.
especially uncertainty in the control effectiveness matrix. Marwaha, M., and Valasek, J. (2011). “Fault tolerant control allocation for
mars entry vehicle using adaptive control.” Int. J. Adapt. Control Signal
Process., 25(2), 95–113.
Acknowledgments MATLAB v7.0 [Computer software]. Natick, MA, MathWorks.
Patton, R. J. (1997). “Fault tolerant control: The 1997 situation.” Proc.,
This work is supported by the National Basic Research Program of
3rd IFAC Symp. on Fault Detection, Supersion and Safety for Technical
China (973 Program) 2012CB821202, National Natural Science Processes, New York, 1033–1055.
Foundation of China under Grants 61174052, 61473004, and Spjtvold, J., and Johasen, T. A. (2009). “Fault tolerant control allocation for
61433001. a thruster-controlled floating platform using parametric programming.”
Proc., 48th IEEE Conf. on Decision and Control (CD-ROM), IEEE,
Piscataway, NJ, 3311–3317.
References Ye, S., Zhang, Y., Wang, X., and Rabbath, C. (2009). “Robust fault-tolerant
control using on-line control re-allocation with application to aircraft.”
Alwi, H., and Edwards, C. (2008). “Fault tolerant control using sliding
American Control Conf., 2009. ACC’09, IEEE, Piscataway, NJ,
modes with online control allocation.” Automatica, 44(7), 1859–1866.
An, S., and Kwon, D. (2010). “Robust control allocation of redundantly 5534–5539.
actuated variable structure systems.” 2010 Int. Conf. on Control Auto- Yu, L. (2002). Robust control-linear matrix inequality, Tsinghua University
mation and Systems (ICCAS), IEEE, Piscataway, NJ, 491–496. Press, Beijing.
Bordignon, K. A. (1996). “Constrained control allocation for systems with Zhang, Y., and Jiang, J. (2008). “Bibliographical review on reconfigurable
redundant control effectors.” Ph.D. thesis, Virginia Polytechnic Institute fault tolerant control systems.” Annu. Rev. Control, 32(2), 229–252.
and State Univ., VA. Zhang, Y., Rabbath, C., and Su, C. (2008). “Reconfigurable control allo-
Casavola, A., and Garone, E. (2010). “Fault tolerant adaptive control allo- cation applied to an aircraft benchmark model.” 2008 IEEE American
cation schemes for overactuated systems.” Int. J. Robust Nonlinear Control Conf., IEEE, Piscataway, NJ, 11–13.
Control, 20(17), 1958–1980. Zhang, Y., Suresh, V. S., Jiang, B., and Theilliol, D. (2007). “Reconfigur-
Cui, L., and Yang, Y. (2011). “Disturbance rejection and robust least able control allocation against aircraft control effector failures.”
squares control allocation in flight control system.” J. Guid. Control 16th IEEE Int. Conf. on Control Applications, IEEE, Piscataway,
Dyn., 34(6), 1632–1643. NJ, 1197–1202.