SS symmetry
Article
Design of Thrust Vectoring Vertical/Short Takeoff and Landing
Aircraft Stability Augmentation Controller Based on L1
Adaptive Control Law
Zan Zhou 1 , Zian Wang 2, * , Zheng Gong 1 , Xiong Zheng 2 , Yang Yang 2 and Pengcheng Cai 1
Citation: Zhou, Z.; Wang, Z.;
Gong, Z.; Zheng, X.; Yang, Y.; Cai, P.
Design of Thrust Vectoring
Vertical/Short Takeoff and Landing
Aircraft Stability Augmentation
Controller Based on L1 Adaptive
Control Law. Symmetry 2022, 14, 1837.
https://doi.org/10.3390/
sym14091837
Academic Editor: Jan Awrejcewicz
Received: 10 July 2022
Accepted: 26 August 2022
Published: 4 September 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Symmetry 2022, 14, 1837. https://doi.org/10.3390/sym14091837
1 Department of Aerospace Engineering, Nanjing University of Aeronautics and Astronautics,
Nanjing 210016, China
2 China Academy of Launch Vehicle Technology, Beijing 100076, China
* Correspondence: wangzian@nuaa.edu.cn
Abstract: Aiming at the conversion process of thrust vectoring vertical/short takeoff and landing
(V/STOL) aircraft with a symmetrical structure in the transition stage of takeoff and landing, there
is a problem with the coupling and redundancy of the control quantities. To solve this problem,
a corresponding inner loop stabilization controller and control distribution strategy are designed.
In this paper, a dynamic system model and a dynamic model are established. Based on the outer
loop adopting the conventional nonlinear dynamic inverse control, an L1 adaptive controller is
designed based on the model as the inner loop stabilization control to compensate the mismatch
and uncertainty in the system. The key feature of the L1 adaptive control architecture is ensuring
robustness in the presence of fast adaptation, so as to achieve a unified performance boundary in
transient and steady-state operations, thus eliminating the need for adaptive rate gain scheduling.
The control performance and robustness of the controller are verified by inner loop simulation and
the shooting Monte Carlo approach. The simulation results show that the controller can still track the
reference input well and has good robustness when there is a large parameter perturbation.
Keywords: L1 adaptive control; V/STOL aircraft; Monte Carlo simulations
1. Introduction
Thrust vectoring technology can directly change the thrust magnitude and thrust direc-
tion of aircraft, which is an important technical scheme to achieve the high maneuverability
of modern aircraft. Unlike conventional fighter jets, vertical/short takeoff and landing
(V/STOL) aircraft is a new type of aircraft [1–3]. It can not only realize the vertical takeoff
and landing, but also carry out a high-speed cruise in a conventional aircraft configuration
and will be widely used in the military field in the future. Therefore, the thrust vectoring
vehicle has the characteristics of a large flight space, complex flight action, and diverse
flight tasks, resulting in the strong nonlinearity of its control system and severe changes in
the external environment. How to design a control scheme that can deal with this large
uncertainty is a key problem in the control design of a thrust vectoring vehicle.
Regarding the aspect of mathematical model establishment and simulation, Songlin
Ma and Weijun Wang [4] established a longitudinal model in a transition flight phase for a
new concept of vertical/short takeoff and landing aircraft. Xiaomeng Zhang and Weijun
Wang [5] analyzed the dynamics of the vertical/short takeoff and landing of unmanned
aerial UAVs and obtained a nonlinear dynamics model. References [6,7] investigated the
thrust vector control of thrust vectoring V/STOL aircraft using a six-degree-of-freedom
flight dynamics simulation. For controller design, Yang, Xili et al. [8] proposed two non-
linear approaches for the autonomous transition control of two vertical/short takeoff and
landing aircraft. Walker, G. and Allen, D. [9] presented an overview of the X-35B control
https://www.mdpi.com/journal/symmetry
Symmetry 2022, 14, 1837 2 of 14

law requirements, design, analysis, and summary of the STOVL flight test results. In Refer-
ence [10], a control scheme comprising the dynamic characteristics of the thrust vectoring
system was developed for V/STOL aircraft. Simulation and experimental results were
presented. Zhiqiang Cheng et al. [11] proposed an optimal trajectory transition controller.
Reference [12] showed an application of L1 adaptive control theory for the attitude con-
trol of UAVs. Chiang R Y, et al. [13] presented an H-∞ flight control system design case
study for a supermaneuverable fighter flying quality of the Herbst maneuver, which may
provide some reference for hovering state flying quality. Zian, Wang et al. [14] used an
L1 adaptive inner loop controller and designed a roll-horizon landing deceleration and
landing strategy for hybrid-wing vehicles. Seshagiri S, et al. [15] considered the application
of a conditional integrator-based sliding mode control design for robust regulation of
minimum-phase nonlinear systems for the control of the longitudinal flight dynamics of
an F-16 aircraft. Liu, N, et al. [16] used the nonlinear active disturbance rejection controller
(ADRC) to control the tilt wings. In terms of control allocation, Min, B.M, et al. [17] focused
on applying various control allocation schemes to the SAT-II UAV system. Tan J, et al. [18]
studied attitude tracking UAVs with the terminal sliding mode based on the extended
state observer and with the multi-objective nonlinear control allocation. For the problem
of simulation verification, References [19,20] provided a practical Monte Carlo method to
verify the robustness of the controller.
In the transition stage of the thrust vectoring V/STOL aircraft, the tilt angle of the
vectoring nozzle and the opening of the lift fan will change greatly. In this power conversion
process, both the aerodynamic rudder surface and the power system of the aircraft can
control the six degrees of freedom of the aircraft. At this stage, the aircraft will face the
problems of strong nonlinearity, control quantity coupling, and redundancy. The design of
the control law is a great challenge. Under the physical constraints of the thrust vectoring
vehicle, the aerodynamic control inputs and thrust vectoring control inputs are allocated
according to the virtual control variables. To solve this problem, based on the F35B scale
model, the dynamic equation modeling is given in this paper. On this basis, the inner loop
controller and control distribution method are designed. Finally, the performance and
robustness of the controller are verified by inner loop simulation and the shooting Monte
Carlo approach.
The remainder of this paper is organized as follows. Section 2 introduces the com-
position of the thrust vector V/STOL aircraft power system and establishes the dynamic
equation modeling by combining the whole vehicle dynamics equation, kinematic equation,
and moment equation. Section 3 introduces the design of the inner loop and outer loop
controllers in detail, and the control allocation method is given. The robustness of the
inner loop controller is simulated and verified by Monte Carlo simulation in Section 4, and
conclusions and recommendations for future work are stated in Section 5

2. Dynamic Equation Modeling

The thrust vectoring vertical/short takeoff and landing aircraft used in this paper is a
self-made F35B scale model with a total weight of about 13 kg, a cruising speed of 30 m/s,
and a cruising altitude of less than 100 m.
The F35B consists of a fuselage, wings, tail, and power system as shown in Figure 1.
The F35B mainly has the fixed-wing mode and vertical takeoff and landing (VTOL) mode.
The platform’s aerodynamic control surfaces include a full-motion horizontal tail, a full-
motion vertical tail, and flaperons. The power system consists of two subsystems: the lift
fan and auxiliary motor subsystem installed in the front section of the fuselage; the main
ducted fan and the main motor subsystem installed in the rear section of the fuselage; a
three-bearing swivel duct (3BSD) nozzle is connected behind the main ducted fan.
Symmetry 2022, 14, x FOR PEER REVIEW 3 of 17
Symmetry 2022, 14, 1837 3 of 14

Figure 1. F35 model.

Figure 1. F35 model.
2.1. Whole Vehicle Dynamics Model
2.1. Whole Vehicle Dynamics Model
We use [ Tx Ty Tz ] to denote the thrust components of the engine under the three shafts.
We
Theuse [Tx Ty Tforce
aerodynamic z ] toAdenote the in
is defined thrust components
the airflow of thesystem,
coordinate engine andunder the
considering
the lift loss in the transition section, the total aerodynamic force
three shafts. The aerodynamic force A is defined in the airflow coordinate system, andcan be expressed as
Equation (1), where D is the drag force, C is the side force, and L is the lift
considering the lift loss in the transition section, the total aerodynamic force can be ex- force.
pressed as Equation (1), where
 D is the drag force, C is the side force, and L is the lift
Ax − D cos α cos β − C cos α sin β + L sin α
force.  Ay  =  − D sin β + C cos β  (1)
− sin cos −
 A  z − D cos α cos β − C cosα sin β + L sin α
A D C sin sin − L cos
x  α β α β
 α
   
 Aof
The direction y= the − D sinisβgiven
 aircraft’s gravity in β
+ C cos (1)
the z-axis of the ground coordinate
system, which isalso converted
 Az   − D sin α cos β − C sin α sin β − L cos α 
to the airframe coordinate system.
     
gx gravity is given
The direction of the aircraft’s − gofsin
0 in the z-axis θ ground coordinate
the
system, which is also converted
 = Lbg m 0  = m g sin φ cos θ 
m gtoy the airframe coordinate system. (2)
gz g g cos φ cos θ
xg  0   − g sin θ 
  (1) and
Combined with Equations
 (2), the  glinear motion equation of the aircraft in the

body coordinate system y =
m  gcan beLdetermined
bg m 0  =mfrom  sin φ cos θ (3).
Equation  (2)
g  du

 zm dt = Tx −gD cos α gcos
cos φ cos θ 
β − C cos α sin β
− − mwq

Combined with Equations
 + L sin α mg sin θ + mvr
 (1)dv and (2), the linear motion equation of the aircraft in the



m =
body coordinate system can be determined
dt T − D sin β + C
y from Equation (3). cos β
(3)

 +mg sin φ cos θ − mur + mwp
dw
 m dt = Tz − D sin α cos β − C sin α sin β




 − L cos α + mg cos φ cos θ + muq − mvp

where [u, v, w] are the velocity components of the three coordinate axes in the body coordi-
nate system and [ p, q, r ] are roll angle velocity, pitch angle velocity, and yaw angle velocity
in the body coordinate system, respectively.
Symmetry 2022, 14, 1837 4 of 14

2.2. Kinematic Equations and Moment Equation

The kinematic equations and moment equation of the aircraft are [9]:
 .
 φ. = p + (r cos φ + q sin φ) tan θ

θ = q cos φ − r sin φ (4)
 .
ψ = cos1 θ (r cos φ + q sin φ)

Symmetry 2022, 14, x FOR PEER REVIEW 5 of 17
 .
 p. = (c1 r + c2 p)q + c3 L + c4 N

q = c5 pr − c6 ( p2 − r2 ) + c7 M (5)
 mu r= Tx(−
c8 D c2 rα)qcos
cos + cβ − C cos α sin β
 .
= p− 4 L + c9 N

 α − mg
+ L sinpitch sin θand+ mvr
where [φ, θ, ψ] are the rollangle, angle, yaw−angle
mwqof the aircraft, respectively;
 mv = Tycoefficients;
c1 ∼ c9 are the moment equation − D sin β +[ C M, Nβ] are roll moment, pitch moment,
L, cos
and yaw moment in the body  coordinate system, respectively.
+ mg sinthe
Based on Equations (3)–(5), φ cos θ − murdynamics
nonlinear + mwp model of the aircraft can be
derived:   mw .  = T − D sin α cos β − C sin α sin β
mu = Tx z− D cos α cos β − C cos α sin β
+−LLsincosα α− +mgmg θ +φmvr θ+ muq − mvp

sincos cos−

mwq




 .
mpv ==(Tc1yr−+Dc2sinp )qβ + 3 + c4 N
+ cC L



 cos β (6)
−

+ mg sin φ cos θ mur + mwp


2 2
mqw==c5Tpr z −−D − r β)−+ Cc7sin
.
c6sin
( pα cos M α sin β





 −rL=cos

(c8 p − c2 r ) q + c4 L + c9 N − mvp
+ mg cos cos + muq

 α φ θ
.
p = ( c1 r + c2 p ) q + c3 L + c4 N (6)
.φ = p + ( r cos2φ + q φ θ
q = c5 pr − c6 ( p − r ) + c7 M sin ) tan


 2
. 


r.θ==(cq8cos
p − φc2− + cφ

r )rq sin

4 L + c9 N




φ. = p +1(r cos φ + q sin φ) tan θ



θψ==q cos φ −(rrcos φ + q sin φ )


sin φ

 . 1 θ
cos




ψ = cos θ (r cos φ + q sin φ)

3.3.L1
L1Stabilization
StabilizationController
ControllerDesign
Design
The
Thestability
stabilityaugmentation
augmentationcontroller
controllerininthis
thispaper
paperconstitutes
constitutesaacontrol
controland
andstability
stability
augmentation
augmentationsystemsystem(CSAS)
(CSAS)by bycontrolling
controllingthe theroll,
roll,pitch,
pitch,and
andyaw
yawofofthetheaircraft.
aircraft.The
The
CSAS
CSAScan cangenerate
generateangular
angularvelocity
velocityand
andacceleration
accelerationcontrol
controlcommands
commandsand andprovides
provides
torque
torquecommands
commandstotothe thecontrol
controldistributor.
distributor.Then,
Then,aasingle-input
single-inputsingle-output
single-output(SISO)
(SISO)L1L1
adaptive controller is designed for each of the three angular velocity channels
adaptive controller is designed for each of the three angular velocity channels to compen- to compen-
sate
satefor
formismatch
mismatchuncertainties
uncertaintiesininthe
thedynamics
dynamics[21–23].
[21–23].ForForouter-loop
outer-loopdesigns
designssuchsuchasas
pitch
pitchand
androll,
roll,nonlinear
nonlinear dynamic
dynamicinversion
inversion(NDI) provides
(NDI) providesthe the
tracking of the
tracking of required dy-
the required
namics.
dynamics.Figure 2 shows
Figure 2 showsthe the
adaptive control
adaptive flowflow
control diagram using
diagram NDINDI
using and and
the L1
theadaptive
L1 adap-
control law. law.
tive control

Figure2.2.Flow
Figure Flowchart
chartofofL1
L1adaptive
adaptivecontroller.
controller.

InInFigure
Figure2,2,the theouter
outerloop
loop state
state variable
variable x1 = φ [θφψθ] Tψand
x1 [= ]T and the inner
the inner loop loop
state state var-
variable
T
2 = [ p q rx] 2 ;=the
xiable T
[ pcontrol
q r ] inputs
.; the of thecontrol
V/STOLinputs
aircraft are ofU1 the= [δa δV/STOL
e δr δT δL δazaircraft ]T ;
δac δTl δTrare
U 1 = [δ a δ e δ r δ T δ L δ az δ ac δ Tl δ Tr ]T ; δ a , δ e , δ r are the aileron deflection, elevator deflec-
tion, and rudder deflection, respectively; δ L is the ratio of the lift fan thrust to the maxi-
mal lift fan thrust; δ T is the ratio between the maximum three-bearing swivel duct noz-
 Symmetry 2022, 14, 1837 5 of 14

δa , δe , δr are the aileron deflection, elevator deflection, and rudder deflection, respectively;
Symmetry 2022, 14, x FOR PEER REVIEWδL is the ratio of the lift fan thrust to the maximal lift fan thrust; δT is the ratio 6 of 17
between
the maximum three-bearing swivel duct nozzle thrust and the three-bearing swivel duct
nozzle thrust; δac is the three-bearing swivel duct nozzle’s pitch angle, when the aircraft
δac =3;0◦δ, and in vertical takeoff and landing mode, δac = 90◦ ; δaz
ductisnozzle, Tl , δ Tr are the thrust of the left and right roll nozzles,
in fixed-wing
as shown mode,
in Figure
is the yaw
respectively. angle
The of thecharacteristics
actuator three-bearing are swivel duct nozzle,
described as 1.
in Table shown in Figure 3; δTl , δTr are
the thrust of the left and right roll nozzles, respectively. The actuator characteristics are
described in Table 1.

Figure 3. Power system diagram.

Figure 3. Power system diagram.

TableTable 1. Effector
1. Effector characteristics
characteristics ofF35B.
of the the F35B.

Effector
Effector Position
Position LimitLimit Rate Rate
LimitLimit
δ a , δ eδ,aδ, rδe , δr [ −30，
30]30,deg
[− 30] deg ±30 deg/ s
±30 deg/s
δ ,δ [0, 1] ±0.4
δ T , δ LTδac L [0，1][0, 90] ±0.4
±40 deg/s
δ ac δaz [0，
90] deg12] deg
[−12, ±40 deg/ s
±40 deg/s
δ ,δ [0, 1000] N ±4000 N/s
δ az Tl Tr [−12，
12] deg ±40 deg/ s
δTl , δTr [0，
1000] N ±4000 N / s
The pitch and roll channels realize the decoupling of the angle and angular velocity
according to the timescale separation principle. The attitude angle is controlled by the NDI
The pitch
method, and the
while rollL1
channels
adaptiverealize the decoupling
controller controls theofangular
the angle and angular
velocity. velocity law
The L1 adaptive
according to the timescale separation principle. The attitude angle is controlled by the law.
can be divided into four parts: control object, state predictor, adaptive law, and control
NDI method, while the L1 adaptive controller controls the angular velocity. The L1 adap-
tive 3.1.
law Inner
can be divided
Loop into Design
Controller four parts: control object, state predictor, adaptive law, and
control law.
3.1.1. NDI Controller Design of Roll Loop
First, the NDI controller needs to be designed for the roll loop. The kinematic equation
3.1. Inner
of theLoop
outerController
loop is: Design
.
3.1.1. NDI Controller Design of Roll xLoop
1 = f 1 (x1 , t ) + g1 (x1 , t )x2 (7)
First, the
f 1 (NDI
x1 , t),controller
g1 (x1 , t) are needs to be designedsatisfying
for the rollf (loop. The kinematic equa-
where affine functions 1 x1 , t ) = 03×1 and g1 (x1 , t ) =
tion 
of1thesin
outer loop
φ tan θ is:
cos φ tan θ

0 cos φ sin φ x 1. = f1 ( x1 , t ) + g1 ( x1 , t ) x2 (7)
0 sin φ/ cos θ cos φ/ cos θ
where The f1 ( x1error g1 ( x1 , t )as ∆x
, t ) ,is defined x1c − x1functions
are1 =affine . Based on satisfying
the NDI method, , t ) =angular
f1 ( x1the 03×1 and
velocity
command pc can be obtained:
1 sin φ tan θ cos φ tan θ 

g1 ( x1 , t ) = 0 cos φ φ g1−1(x1 , t)(K1 ∆x1 − f 1 (x1 , t))
xsin
2c = (8)
0 sin φ / cosθ cos φ / cosθ 
where K1 ∈ R3×3 is the bandwidth coefficient . and is a diagonal matrix, which is determined
The
by error
the is defined
flight quality of Δx1aircraft
as the = x1c −and
x1 . Based on the
the outer NDI
loop method, the angular velocity
bandwidth.
command pc can be obtained:

x 2 c = g1−1 ( x1 , t )( K 1 Δ x1 − f1 ( x1 , t )) (8)

where K 1 ∈ R 3× 3 is the bandwidth coefficient and is a diagonal matrix, which is deter-

mined by the flight quality of the aircraft and the outer loop bandwidth.
Symmetry 2022, 14, 1837 6 of 14

3.1.2. L1 Adaptive Controller Design of Roll Angular Velocity Loop

According to the small perturbation theory, the kinematic equation of the outer loop
Equation (7) can be simplified as
 
L p. 0 0
 
Lβ L p  
. β α β  uc .
x2 =  Mα Mq  + 0 Mq. 0 x2uc (9)
p q r uc
Nβ Nr 0 0 Nruc
.

 
. Lβ L p
. . . T
where x2uc = puc quc r uc is the virtual angular acceleration,  Mα Mq  are the
Nβ Nr
aerodynamic coefficients, and L p. , Mq. , Nr. uc are the virtual control torque generated by
. . uc uc
the virtual output signal x2uc . x2uc can be solved by Equation (9) and expressed as:
. . .
x2uc = x2Tc + x2ac (10)
. .
where x2ac and x2Tc are the roll angular accelerations generated by the air surface and
the vector nozzle, respectively. At low airspeeds, the control torque provided by the
vectoring nozzle dominates. The vector nozzle provides the required pitch acceleration,
. . .
i.e., x2Tc , x2uc , x2ac = 0. At high airspeeds, the aerodynamic surfaces subject to the
. . .
required pitch angular acceleration, i.e., x2ac , x2uc , x2Tc = 0. During the transitional
flight, both the aerodynamic surfaces and the vectoring nozzle achieve the commanded
roll angular acceleration.
Considering the uncertainty of moment, Equation (9) can be expressed as:
 
L β + L̂ β L p + L̂ p  
. β α β
x2 =  Mα + M̂α Mq + M̂q 
p q r
Nβ + N̂β Nr + N̂r

L p. + L̂ p. 0 0
 .  (11)
p
uc uc
 . uc 
+ 0 Mq. + M̂q. 0  q. uc +σ

uc uc
0 0 Nr. uc + N̂r. uc r uc

where L̂ β , L̂ p , L̂ p. , M̂α , M̂q , M̂q. , N̂β , N̂r , N̂r. uc are the uncertain aerodynamic factors and
uc uc
 T
σ = σp σq σr is the disturbance factor. The structure of the first-order reference
model is as follows:
.
x2 = K2 ∆x2 = K2 (x2c − x2 ) (12)
Combining Equations (9)–(12), we can write the control model of the rolling loop as:
 .
 x2 = −K.2 x2 + K2 η
η = ωx2 x2 + f 2 (t, x2 ) (13)
f 2 (t, x2 ) = λx2 + σx2


 
L̂ p. M̂q. N̂r. uc
where K2 (k p , k q , kr ) ∈ R3×3 is the feedback gain matrix, ωx2 = 1 + L p.
uc
1+ Mq.
uc
1+ Nr. uc
uc uc
L̂ p. M̂q. N̂r. uc
is the virtual control factor, λ = diag( L̂ p − L p.
uc
L p , M̂q − Mq.
uc
Mq , N̂r − Nr. uc Nr ) is the aero-
uc uc
L̂ p. M̂q. N̂r. uc
dynamic factor, and σx2 = diag( L̂ β − L p.
uc
L β , M̂α − Mq.
uc
Mα , N̂β − Nr. uc Nβ )x2 + σ is the
uc uc
aerodynamic disturbance.
Three assumptions and a lemma are given.
Assumption 1. The unknown constant ωx2 is uniformly bounded, i.e., ωx2 ∈ Ω ⊂ R3×3 , Ω is a
compact convex set.
Symmetry 2022, 14, 1837 7 of 14

Assumption 2. f 2 (t, x2 ) is uniformly bounded in Equation (19), such as k f 2 (t, 0)k∞ ≤ b, b > 0,
where k•k∞ is the ∞-norm.
Assumption 3. The partial derivatives of f 2 are semi-globally uniformly bounded: for δ > 0, there
exist d f p (δ) > 0 and d f t (δ) > 0 independent of time t to ensure that the partial derivatives of
f 2 (t, x2 ) are piecewise continuous and bounded, as follows:

(t,x2 )
k ∂ f2∂x
(
k ≤ d f p (δ)
2 ∞ (14)
k ∂ f2 (∂tt,x2 ) k∞ ≤ d f t (δ)

The assumptions can be met since the uncertainty for the inner loop control system is
of a given magnitude.
.
Lemma 1. For τ > 0, if kxτ k L∞ ≤ ρ and kxτ k L∞ ≤ dx , where ρ and dx are positive constants, λ
and σx2 are continuous. Furthermore, their derivatives with respect to t ∈ [0, τ ] are:

 f 2 (t, x2 ) = λkxk.L∞ + σx2

k λ k < d f q ( ρ ), k λ k ≤ d λ (15)
 kσ k < b, kσ. k ≤ d

x2 x2 σ

where dλ and dσ are computable finite values; k•k L∞ is the L∞ -norm.

Based on the above assumptions and lemma, an adaptive law controller is designed,
which consists of a state predictor, an adaptive law, and a control law:
• State predictor:
According to Equation (13), the state predictor can be expressed as:
 .
 x̂2 = −K2 x̂2 + K2 η̂

.
η̂ = x2c ω̂x2 + λ̂x2 + σ̂x2 (16)

 ŷ = x̂
2

where ω̂x2 is the predicted level of control factor uncertainty, λ̂ is the expected degree of
the aerodynamic factor’s uncertainty, and σ̂x2 is the predicted degree of the aerodynamic
disturbance’s uncertainty; ω̂x2 denotes the estimated level of control factor uncertainty,
B the estimated level of aerodynamic factor uncertainty, and C the estimated level of
aerodynamic disturbance uncertainty.
• Law of adaptation:
 .
 λ̂. = ΓK proj (λ̂, −ex2 PK2 kx2 k∞ )


σ̂ x2 = ΓK proj (λ̂, −ex2 PK2 ) (17)
 ω̂. = ΓK (ω̂. , −e .


x2 proj x2 x2 PK2 x2 )
where Γ is the adaptive gain, e
x2 = x̂2 − x2 is the prediction error, and P is the solution of
the Lyapunov equation K2T P + PK2 = −Q (Q = QT > 0). K proj is the projection operator.
• Control law:
The control law is as follows:
.
x2c = Kd D(Kg x2 − η̂) (18)
Symmetry 2022, 14, 1837 8 of 14

where Kg is the adaptive feedback gain, D is the low-pass filter, and Kd is the adaptive
feedforward gain. The L1 adaptive control scheme is shown in Figure 3. Control laws
should be designed to ensure that the following transfer functions are strictly regular:

C(s) = ωx2 Kd D(s)(I + ωx2 Kd D(s))−1 (19)

where C(s) satisfies C(0) = I and I is a 3 × 3 identity matrix.

When obtaining Kd and D, the following conditions must be satisfied to ensure the
stability of adaptive control:

ρr − kH(s)C(s)Kg k L kx2c k L∞ − ρin

1
kG(s)k L1 < (20)
L pr ρ r + b

where ρin , H(s), G(s), and L pr are defined as:

ρin = ks(sI + K p )−1 k L ρ0



 1
H(s) = (sI + K p )−1 K p


(21)
 G(s) = H(s)[I − C(s)]
Symmetry 2022, 14, x FOR PEER REVIEW  10 of 17
 ρ +γ
L pr = r ρr 1 d f p [ ρr + γ1 ]


where γ1 is an arbitrary positive number.

InInthe
thecase
case
ofof satisfying
satisfying Equation
Equation (20),
(20), TheThe adaptive
adaptive controller
controller of inner
of the the inner
looploop con-
consists
sists of Equations (16)–(18). The block diagram of the adaptive controller system
of Equations (16)–(18). The block diagram of the adaptive controller system is shown in is shown
in Figure
Figure 4. 4.

Figure4.4.Adaptive
Figure Adaptivecontroller
controllersystem
systemblock
blockdiagram.
diagram.

3.2.
3.2.Control
ControlAllocation
Allocation
Equation
Equation(6)
(6)can
canbebeexpressed
expressedas:
as:
. 
X=X f=(X,
f (U
X2,,Ut)2 , t ) (22)
(22)
.  are the system state variables, U 2 are the system control .
where
where X areXthe system state variables, U2 are the system control variables, and X = [u vvariables,
w p q r φ θ and
ψ],
X = [u v w p q r φ θ ψ ] , UT = [δ δ δ δ δ δ δ δ ] . T
U2 = [δa δe δr δT δL δaz δTl δTr ] 2. a e r T L az Tl Tr
ExpandingEquation
Expanding Equation(22)
(22)bybyTaylor
Taylorseries
seriesatatthe
theequilibrium
equilibriumpoint,
point,keeping
keepingthe
thelinear
linear
part, and ignoring the high-order part, we
part, and ignoring the high-order part, we can obtain: can obtain:
 = AΔ X + B Δ U
. ΔX (23)
∆X = A∆X + B∆U2 2 (23)
∂ f∂f
A =∂X
whereA =
where isisthe
thesystem
systemstate
statematrix,
matrix,which
whichconsists
consistsofofthe
thepartial
partialderivatives
derivatives
∂XX =XX=trim
X trim
∂f
of forces and moments with respect to the state variables. B = ∂U∂ is the control
2 f U =U
of forces and
derivative moments
matrix, which with
is therespect to the state
aerodynamic changevariables.
generatedBby
= the unit control
is the
2 trim
control
amount,
∂U 2 U =U
which can reflect the control efficiency of the aircraft by the control input.
2 trimThe subscript

trim indicates the trim value. System control variables ∆U2 can be solved bycontrol
derivative matrix, which is the aerodynamic change generated by the unit amount,
Equation (23).
which can reflect the control efficiency of the aircraft by the control input. The subscript
trim indicates the trim value. System control variables ΔU 2 can be solved by Equation
(23).
 Symmetry 2022, 14, 1837 9 of 14

4. Inner Loop Controller Simulation Experiment

According to the inner loop controller and control distribution method designed in the
previous section, the simulation verification results are given in this section. The difficulty
in controlling the V/STOL aircraft is that there is a flight mode conversion process during
the takeoff and landing process. At this stage, the two sets of control mechanisms of the
aircraft are involved in the work of jointly controlling the position and attitude of the
aircraft. Therefore, in this state, the working point is selected for inner loop simulation and
the shooting Monte Carlo approach to verify the performance of the inner loop stabilization
controller and control distribution method designed in this paper.
Symmetry 2022, 14, x FOR PEER REVIEW 11 of 17
4.1. Monte Carlo Targeting at Nominal State
Symmetry 2022, 14, x FOR PEER REVIEW 11 of 17
An L1 controller with ω p = 8 and D = 1/s receives a 5◦ /s frequent square roll rate
command as the input. Figures 5 and 6 show the outcomes for various combinations of Γ
and k p .

Figure 5. Roll rate command tracking results at kp = 8 .

Figure5.5.Roll
Figure Rollrate command
rate tracking
command results
tracking results = 8.k p = 8 .
at Kpat

Figure 6. Roll
Figure 6. Rollrate
ratecommand
command tracking
tracking results
results at Γ at Γ = 2500 .
= 2500.
Figure 6. Roll rate command tracking results at Γ = 2500 .
We settled on k p = 8 and Γ = 2500 as the virtual control coefficient and virtual
state We
coefficient, on k p = 8 and
settled respectively, Γ = 2500
because as theavirtual
they strike balancecontrol
betweencoefficient
speed and and virtua
stability,
have a brief risetime, and exhibit zero overshoots.
state coefficient, respectively, because they strike a balance between speed and stability
haveAccording to theand
a brief risetime, inner loop zero
exhibit controller designed in the previous section, in fixed–
overshoots.
wing mode, set the height to 25 m and the airspeed to 24 m/s. In this state, given step
Symmetry 2022, 14, 1837 10 of 14

We settled on k p = 8 and Γ = 2500 as the virtual control coefficient and virtual state
coefficient, respectively, because they strike a balance between speed and stability, have a
brief risetime, and exhibit zero overshoots.
According
Symmetry 2022, 14, x FOR PEER
Symmetry 2022, 14, x FOR PEER REVIEW to the inner loop controller designed in the previous section, in fixed–wing
REVIEW 12
12 of
mode, set the height to 25 m and the airspeed to 24 m/s. In this state, given step signals for
the roll and pitch angles, the simulation results are shown in Figure 7.

Figure
Figure 7. Roll
7. Roll rate
rate andand pitch
pitch rate
rate command
command tracking
tracking results
results in fixed-wing
inmode.
fixed-wing mode.
mode.
Figure 7. Roll rate and pitch rate command tracking results in fixed-wing

It can be seenIt It
can can
from bebe
the seen
seenfigurefrom
from thethe
that thisfigure
figure that
that this
this
controller controller
controller
can track the can
can track
track
pitch thethe
angle pitch
pitch
command angle
angle comm
comma
well;
well; the adjustment the adjustment
well; the adjustment time
time of
time of the inner of
loop the
thepitchinner
innerangle loop
loopand pitch
pitch angle
angle
roll angle and
and roll angle
roll angle
command command
command
is about 1 s, is is ab
abou
s,
and there iss,no
andand
obviousthere is no
there isovershoot. obvious overshoot.
no obvious overshoot.
In VTOL mode, In
In VTOL VTOL
set the mode,
height
mode, setset
to the
25
the mheight
heightandto to
the 25
mm
25airspeed
andand tothe
the 16airspeed
m/s. In
airspeed to
1616
tothis m/s.
state,
m/s. In this
Ingiven
this state,
state, g
giv
step
step signalsstep
for the signals
roll and
signals for the
pitch
for the roll and
rollangles,
and pitchpitch angles,
the simulation the simulation
angles, the results results
are shown
simulation resultsin are shown
Figure
are shown 8: in Figure 8: 8
in Figure

Figure 8. Roll Figure

rate
Figure and 8.pitch
Roll
8. Roll rate
rate
rate and pitch
command
and pitch rate command
tracking
rate commandresultstracking
in VTOL
tracking results
in in
mode.
results VTOL
VTOL mode.
mode.

The
The findings
findings reveal
reveal that
that thethe curve’s
curve’s shape
shape is is
thethe same
same in in both
both modes,
modes, and
and theth
controller’s tracking performance is superb with a steady-state error of zero and
controller’s tracking performance is superb with a steady-state error of zero and a min am
temporal
temporal lag.lag.

4.2.
4.2. Target
Target Shooting
Shooting Monte
Monte Carlo
Carlo Approach
Approach in in Level
Level Flight
Flight
Symmetry 2022, 14, 1837 11 of 14

The findings reveal that the curve’s shape is the same in both modes, and the L1
controller’s tracking performance is superb with a steady-state error of zero and a minor
temporal lag.
Symmetry 2022, 14, x FOR PEER REVIEW 13 of 17
4.2. Target Shooting Monte Carlo Approach in Level Flight
The robustness of the inner loop controller was simulated and verified by Monte Carlo
simulation. Its parameter perturbation variables mainly include aerodynamic characteris-
characteristics,
tics, mass characteristics,
mass characteristics, and the
and the dynamic dynamic
system. The system. The
change of change of perturbation
perturbation parameters
parameters is shown
is shown in Table 2. in Table 2.

Table2.2.Main
Table Mainparameters
parametersofofthe
theV/STOL
V/STOLvehicle.
vehicle.

Parameters
Parameters Perturbations
Perturbations
CLLδ e,,CCmmδeδ e, ,CCS Sδ,r C, C
C ,C
nδrn,δ rClδa lδ a ±±10%
10%
δe δr
Cmα , CSβ , Clβ , Cn β
Cmα , CSβ , Clβ , Cnβ ±10%
Cl p , Cnr
±10%
±10%
CC ,C
lr l,pC n p nr ±±10%
10%
Jxx , Jyy , Jzz ±10%
Clr , Cnp ±10%
CLδe , Cmδe , CSδrJ,xxC,nJδryy, C
, Jl zz are the control derivatives; Cmα , CS ,±C10%
δa β l β , Cn β are the stability
derivatives; Cl p , Cnr are the damping derivatives; Clr , Cn p are the cross-damping derivatives;
Jxx , JyyC
, Jzz ,are
Lδ e Cmδ e , CSδ r , Cnδ r , Clδ a are the control derivatives; Cmα , CS β , Clβ , Cnβ are the stabil-
the inertia moments.
The above parameters are randomly changed, and the shooting Monte Carlo approach
isity derivatives;
performed on the Cl paircraft
, Cnr are in the damping
constant levelderivatives; Clr , Cnp are
flight and acceleration the cross-damping
conditions, de-
respectively.
At an altitude
rivatives; J xxof
, J100
yy , J zz are the inertia moments.
m, the aircraft is in constant level flight at a speed of 20 m/s. At 15 s,
a 10◦ pitch angle command is given to conduct a target shooting Monte Carlo approach
The above parameters are randomly changed, and the shooting Monte Carlo ap-
simulation test. Then, comparing it with the PID controller [24–26], the simulation results
proach is performed on the aircraft in constant level flight and acceleration conditions,
are shown in Figures 9–13.
respectively.
As can beAt an altitude
seen from theofabove 100 m,figure,
the aircraft is in constant
the control channels level flightdirection
of each at a speedof of 20
the
m/s. At 15 s, a 10° pitch angle command is given to conduct a target
aircraft are decoupled. The controller designed in this paper is obviously better than the shooting Monte Carlo
approach
PID simulation
controller test. Then,
for instruction comparing
tracking. Thisitshows
with the PID
that thecontroller
controller [24–26], the in
designed simula-
this
tion results are shown in Figures 9–13.
paper can still track the command well and has good robustness under the premise that the
parameters of the aircraft are taken.

(a) (b)
Figure9.9. The
Figure Theshooting
shootingMonte
MonteCarlo
Carloapproach at the
approach pitch
at the angle.
pitch (a) L1(a)
angle. adaptive controller;
L1 adaptive (b) PID
controller;
controller.
(b) PID controller.
 Symmetry 2022, 14, x FOR PEER REVIEW 14 of 17
Symmetry 2022, 14, x FOR PEER REVIEW 14 of 17
Symmetry 2022, 14, x FOR PEER REVIEW 14 of 17

Symmetry 2022, 14, 1837 12 of 14

(a) (b)
(a) (b)
Figure 10. The shooting Monte Carlo approach at the airspeed; (a) L1 adaptive controller; (b) PID
(a)
controller. (b)
Figure 10. The shooting Monte Carlo approach at the airspeed; (a) L1 adaptive controller; (b) PID
controller.
Figure 10. The shooting Monte Carlo approach at the airspeed; (a) L1 adaptive controller; (b) PID
Figure 10. The shooting Monte Carlo approach at the airspeed; (a) L1 adaptive controller;
controller.
(b) PID controller.

(a) (b)
(a) (b)
Figure 11. The target shooting Monte Carlo approach with the elevator; (a) L1 adaptive controller;
(b)(a)
PID11.
Figure controller. (b)
The target shooting Monte Carlo approach with the elevator; (a) L1 adaptive controller;
Figure 11. The target shooting Monte Carlo approach with the elevator; (a) L1 adaptive controller;
(b) PID11.
Figure controller.
The target shooting Monte Carlo approach with the elevator; (a) L1 adaptive controller;
(b) PID controller.
(b) PID controller.

(a) (b)
(a) (b)
(a) 12. The shooting Monte Carlo approach at the pitch angle
Figure (b)velocity; (a) L1 adaptive controller;
(b) PID controller.
 Symmetry 2022, 14, x FOR PEER REVIEW 15 of 17

Figure 12. The shooting Monte Carlo approach at the pitch angle velocity; (a) L1 adaptive controller;
Symmetry 2022, 14, 1837 (b) PID controller. 13 of 14

(a) (b)
Figure13.
Figure 13.The
The shooting
shooting Monte
Monte Carlo
Carlo approach
approachat
atthe
theangle
angleofofattack;
attack;(a)(a)
L1L1
adaptive controller;
adaptive (b)
controller;
PID controller.
(b) PID controller.

As can be seen from the above figure, the control channels of each direction of the
5. Conclusions
aircraft
Basedareondecoupled. The controller
the conventional designed
nonlinear in this
dynamic paper
inverse is obviously
control betterloop,
of the outer than an
the
PID controller for instruction tracking. This shows that the controller designed
L1 adaptive controller was designed as the inner loop stability augmentation control toin this pa-
per can stillthe
compensate track the command
uncertainty in thewell and The
system. has good robustness
expected under
verification the premise
results that the
were obtained
parameters of the aircraft are taken.
from the Monte Carlo simulations. The main contributions of this paper are as follows:
(1) This paper introduced the composition of the power system of the thrust vectoring
5. Conclusions
V/STOL aircraft and established the dynamic equation of the F35B scale model
Based on the conventional nonlinear dynamic inverse control of the outer loop, an L1
prototype.
adaptive
(2) controller
For the was designed
thrust vector V/STOLasaircraft
the inner
withloop stability
strong augmentation
coupling control to
and nonlinearity, com-
based
pensate the uncertainty in the system. The expected verification results were
on the conventional dynamic inverse control in the outer loop, an L1 adaptive stabi- obtained
fromlization
the Monte Carlo simulations.
controller was designedTheonmain contributions
the inner of this paper
loop to compensate are
for asmismatch
the follows:
(1) uncertainty. The designed
This paper introduced thecontrol structure
composition integrates
of the the fixed-wing
power system modevectoring
of the thrust and the
VTOL
V/STOL mode.
aircraft and established the dynamic equation of the F35B scale model pro-
(3) The uncertainty of modeling and possible input disturbances were fully considered
totype.
(2) For the thrustwith
and compared theV/STOL
vector PID controller.
aircraftItwith
was strong
verifiedcoupling
by simulation that the controller
and nonlinearity, based
quickly
on the conventional dynamic inverse control in the outer loop, an L1there
responds to the command and has good robustness when is a large
adaptive stabi-
parameter perturbation.
lization controller was designed on the inner loop to compensate for the mismatch
uncertainty. The designed control structure integrates the fixed-wing mode and the
AuthorVTOL mode. Conceptualization, Z.G. and Z.Z.; methodology, Z.G.; software, Z.G. and
Contributions:
Z.Z.; validation, Z.Z. andof
(3) The uncertainty Z.G.; formal analysis,
modeling Z.Z. and
and possible Z.W.;
input investigation,
disturbances Z.W.;
were resources,
fully X.Z.;
considered
data curation, Z.W.; writing—original draft preparation, Z.G. and Z.Z.; writing—review
and compared with the PID controller. It was verified by simulation that the control- and editing,
Z.W. and
ler Z.Z.; visualization,
quickly responds Y.Y.; supervision,
to the commandZ.W.;and project administration,
has good P.C. Allthere
robustness when authors
is ahave
large
read and agreed to the published version of the manuscript.
parameter perturbation.
Funding: This research received no external funding.
Author Contributions: Conceptualization, Z.G. and Z.Z.; methodology, Z.G.; software, Z.G. and
Data Availability Statement: Data available on request due to restrictions eg privacy or ethical. The
Z.Z.; validation, Z.Z. and Z.G..; formal analysis, Z.Z. and Z.W.; investigation, Z.W..; resources, X.Z..;
data presented in this study are available on request from the corresponding author. The data are not
data curation, Z.W.; writing—original draft preparation, Z.G. and Z.Z.; writing—review and edit-
publicly available due to commercial use.
ing, Z.W. and Z.Z.; visualization, Y.Y.; supervision, Z.W..; project administration, P.C. All authors
have read
Conflicts ofand agreed
Interest: to authors
The the published
declareversion of the
no conflict ofmanuscript.
interest.
Funding: This research received no external funding.
References
1. Su, W.; Gao, Z.H.; Xia, L. Multiobjective optimization design of aerodynamic configuration constrained by stealth performance.
Acta Aerodyn. Sin. 2006, 24, 137–140.
2. Jacobson, S.; Britt, R.; Freim, D.; Kelly, P. Residual pitch oscillation (rpo) flight test and analysis on the b-2 bomber. ICES J. Mar.
Sci. 2003, 67, 1260–1271.
3. Britt, R.T.; Arthurs, T.D.; Jacobson, S.B. Aeroservoelastic analysis of the B-2 bomber. J. Aircr. 2000, 37, 745–752. [CrossRef]
Symmetry 2022, 14, 1837 14 of 14

4. Ma, S.; Wang, W. The Longitudinal Modeling of a New Concept V/STOL UAV in Transition Flight by Using the Method of
System Identification. In Proceedings of the 2018 10th International Conference on Intelligent Human-Machine Systems and
Cybernetics (IHMSC), Hangzhou, China, 25–26 August 2018.
5. Zhang, X.; Wang, W. Dynamic modelling of the hovering phase of a new V/STOL UAV and verification of the PID control
strategy. IOP Conf. Ser. Mater. Sci. Eng. 2018, 408, 12–17. [CrossRef]
6. Wang, X.; Zhu, J.; Zhang, Y. Dynamics modeling and analysis of thrust-vectored V/STOL aircraft. In Proceedings of the 32nd
Chinese Control Conference, IEEE, Xi’an, China, 26–28 July 2013.
7. Nagabhushan, B.L.; Faiss, G.D. Thrust vector control of a v/stol airship. J. Aircr. 1984, 21, 408–413. [CrossRef]
8. Yang, X.; Fan, Y.; Zhu, J. Transition flight control of two vertical/short takeoff and landing aircraft. J. Guid. Control. Dyn. 2008,
31, 371–385.
9. Walker, G.; Allen, D. X-35B STOVL Flight Control Law Design and Flying Qualities. In Proceedings of the 2002 Biennial
International Powered Lift Conference and Exhibit, Williamsburg, VA, USA, 5–7 November 2002.
10. Wang, X.; Zhu, B.; Zhu, J.; Cheng, Z. Thrust vectoring control of vertical/short takeoff and landing aircraft. Sci. China 2020,
63, 218–220. [CrossRef]
11. Cheng, Z.; Zhu, J.; Yuan, X.; Wang, X. Design of optimal trajectory transition controller for thrust-vectored v/stol aircraft. Sci.
China Inf. Sci. 2021, 64, 139201. [CrossRef]
12. Mallikarjunan, S.; Nesbitt, B.; Kharisov, E. L1 adaptive controller for attitude control of multirotors. In Proceedings of the
Guidance, Navigation, and Control Conference, Minneapolis, MN, USA, 13–16 August 2012.
13. Chiang, R.Y.; Safonov, M.G.; Haiges, K.; Madden, K.; Tekawy, J. A fixed H∞ controller for a supermaneuverable fighter performing
the herbst maneuver. Automatic 1993, 29, 111. [CrossRef]
14. Wang, Z.; Mao, S.; Gong, Z.; Zhang, C.; He, J. Energy Efficiency Enhanced Landing Strategy for Manned eVTOLs Using L1
Adaptive Control. Symmetry 2021, 13, 21–25. [CrossRef]
15. Seshagiri, S.; Promtun, E. Sliding mode control of F-16 longitudinal dynamics. In Proceedings of the 2008 American Control
Conference, Seattle, WA, USA, 11–13 June 2008.
16. Liu, N.; Cai, Z.; Wang, Y. Fast level-flight to hover mode transition and altitude control in tiltrotor’s landing operation. Chin. J.
Aeronaut. 2020, 34, 181–193. [CrossRef]
17. Min, B.M.; Kim, E.T.; Tahk, M.J. Application of Control Allocation Methods to SAT-II UAV. In Proceedings of the Aiaa Guidance,
Navigation, & Control Conference & Exhibit, San Francisco, CA, USA, 15–18 August 2005.
18. Tan, J.; Zhou, Z.; Zhu, X.; Xu, M. Attitude control of flying wing uav based on terminal sliding mode and control allocation. J.
Northwestern Polytech. Univ. 2014, 32, 505–510.
19. Shakarian, A. Application of Monte-Carlo Techniques to the 757/767 Autoland Dispersion Analysis by Simulation. Guid. Control
Conf. 1983, 83, 181–194.
20. Chao, Z.; Chen, L.; Chen, Z. Monte Carlo Simulation for Vision-based Autonomous Landing of Unmanned Combat Aerial
Vehicles. J. Syst. Simul. 2010, 22, 2235–2240.
21. Xargay, E.; Hovakimyan, N.; Dobrokhodov, V.; Statnikov, R.; Kaminer, I.; Cao, C.; Gregory, I. L1 Adaptive Flight Control System:
Systematic Design and Verification and Validation of Control Metrics. In Proceedings of the AIAA Guidance, Navigation, and
Control Conference, Toronto, ON, Canada, 2–5 August 2010.
22. Leman, T.; Xargay, E.; Dullerud, G.; Hovakimyan, N.; Wendel, T. L1 Adaptive Control Augmentation System for the X-48b
AircrafAIA Guidance. In Proceedings of the Navigation, and Control Conference, Chicago, IL, USA, 10–13 August 2009; AIAA,
Urbana-Champaign: Champaign, IL, USA, 2009.
23. Jiang, G.; Liu, G.; Yu, H. A Model Free Adaptive Scheme for Integrated Control of Civil Aircraft Trajectory and Attitude. Symmetry
2021, 13, 347. [CrossRef]
24. Faruk, T.; Ali, A.; Naim, A.; Firas, A.; Basil Al, K.; Özyavaş, A. Robust Nonlinear Non-Referenced Inertial Frame Multi-Stage PID
Controller for Symmetrical Structured UAV. Symmetry 2022, 14, 689. [CrossRef]
25. Ma, J.; Xie, H.; Song, K.; Liu, H. Self-Optimizing Path Tracking Controller for Intelligent Vehicles Based on Reinforcement
Learning. Symmetry 2022, 14, 31. [CrossRef]
26. Zhong, C.Q.; Wang, L.; Xu, C.F. Path Tracking of Permanent Magnet Synchronous Motor Using Fractional Order Fuzzy PID
Controller. Symmetry 2021, 13, 1118. [CrossRef]