Professional Documents
Culture Documents
10 1109@tcyb 2020 2984791
10 1109@tcyb 2020 2984791
Abstract—This article focuses on finite-time adaptive fuzzy lots of valuable results have been acquired in the domain
output-feedback control for a class of nontriangular nonlin- of adaptive fuzzy control or adaptive NN control based on
ear systems with full-state constraints and unmeasurable states. the backstepping technique [5]–[12]. Tong et al. applied FLS
Fuzzy-logic systems and the fuzzy state observer are employed
to approximate uncertain nonlinear functions and estimate the to approximate the unknown nonlinear function in [9], and
unmeasured states, respectively. In order to solve the algebraic both adaptive fuzzy state-feedback and output-feedback con-
loop problem generated by the nontriangular structure, a vari- trol schemes were designed, which confirmed the stability of
able separation approach based on the property of the fuzzy the closed-loop system. It should be noted that there exists a
basis function is utilized. The barrier Lyapunov function is incor- drawback inherent in the backstepping method. The problem
porated into each step of backstepping, and the condition of
the state constraint is satisfied. The dynamic surface technique of explosion of complexity is inevitable as the order of the
with an auxiliary first-order linear filter is applied to avoid the system increases, which is generated by the repeated dif-
problem of an “explosion of complexity.” Based on the finite-time ferentiations of the virtual control signal. To overcome this
stability theory, an adaptive fuzzy controller is constructed to problem, the dynamic surface control (DSC) technology [13]
guarantee that all signals in the closed-loop system are bounded, was introduced. For the non-Lipschitz systems in [13], the
the tracking error converges to a small neighborhood of the
origin in a finite time, and all states are ensured to remain low-pass filter was applied to deal with the “explosion of
in the predefined sets. Finally, the simulation results reveal the terms,” and an adaptive tracking control scheme with the DSC
effectiveness of the proposed control design. technique was designed. This method was extended to strict-
Index Terms—Adaptive fuzzy control, finite-time control, full- feedback nonlinear systems in [14] and [15] with arbitrary
state constraints, nontriangular nonlinear systems. uncertainty.
In practice, the problem of constraint occurs in various
forms for engineering systems, such as input saturation, dead
zone, state constraints, and output constraint. For the output
I. I NTRODUCTION constraint, Tee et al. proposed the barrier Lyapunov func-
URING the past few years, the adaptive control method tion (BLF) in [16], and an important theorem was given to
D based on the backstepping technique serves as an effec-
tive tool to construct the controller for a class of nonlinear
ensure that the output constraint cannot be violated. Based
on [16], the application of the BLF was extended from the
systems [1]–[4]. For the nonlinear system with the func- output constraint to the state constraints. Considering the full-
tion being completely unknown, this method is not available. state constraints in [17] for pure-feedback systems, the BLF
Due to the excellent ability of approximating, the fuzzy- was incorporated into each step of the design procedure to
logic system (FLS) or neural networks (NNs) have been confirm the constrained states were satisfied. Partial state con-
widely applied to model uncertain nonlinearities. Moreover, straints were studied in [18], compared with the full-state
constraints, and the feasibility condition with BLF for the
Manuscript received December 4, 2019; accepted March 27, 2020. This partial constraints was relaxed. For the high-order nonlinear
work was supported in part by the National Key Research and Development systems in [19], the tan-type BLF was introduced with the sign
Program of China under Grant 2018YFA0702200, in part by the National function to prevent the violation of the output constraint. A
Natural Science Foundation of China under Grant 61627809 and Grant
61433004, and in part by the Liaoning Revitalization Talents Program under high-order tan-type BLF was first constructed in [20] to handle
Grant XLYC1801005. This article was recommended by Associate Editor the full-state constraints of the high-order systems. Moreover,
S.-F. Su. (Corresponding author: Huaguang Zhang.) with this kind of BLF and a reduced adaptive fuzzy control, a
Huaguang Zhang is with the State Key Laboratory of Synthetical
Automation for Process Industries, Northeastern University, Shenyang novel control scheme was designed in [21] for the high-order
110004, China, and also with the College of Information Science and stochastic nonstrict-feedback nonlinear systems with full-state
Engineering, Northeastern University, Shenyang 110004, China (e-mail: constraints.
hgzhang@ieee.org).
Yang Liu and Yingchun Wang are with the College of Information Science Compared with the asymptotic control, finite-time con-
and Engineering, Northeastern University, Shenyang 110004, China (e-mail: trol has received a lot of attention due to higher tran-
myliuyang@yeah.net; drwangyc@gmail.com). sient performance and better steady tracking property [22].
Color versions of one or more of the figures in this article are available
online at http://ieeexplore.ieee.org. Moreover, there are many applications in practical systems,
Digital Object Identifier 10.1109/TCYB.2020.2984791 such as teleoperation systems [23], [24]; aircraft flight systems
2168-2267
c 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
[25]; and so on. In [26], the finite-time Lyapunov stability the- II. P RELIMINARIES AND S YSTEM D ESCRIPTION
ory was proposed for non-Lipschitzian autonomous systems. A. Necessary Preparations
Taking the attitude control of rigid spacecraft into consider-
In order to obtain the desired result in this article, some
ation, a practical finite-time stable lemma was given in [27],
necessary definitions and lemmas are given in the following.
and the controller was designed based on this lemma to con-
Definition 1 [27]: Consider the nonlinear system ξ̇ =
firm the finite-time convergence of the system. In [28]–[30], a
f (ξ, u), where ξ is the system state vector, and u represents the
novel global fast finite-time stabilization was investigated for
system input. f : D → Rn is continuous on an open set D ⊂ R.
high-order nonlinear systems. Compared with the traditional
The nonlinear system ξ̇ = f (ξ, u) is SGPFS, if for every initial
finite-time stability theory, the fast finite-time theory provides
condition ξ(t0 ) = ξ0 , there exists ε > 0 and a settling time
a faster convergence rate. Because of the complete uncertainty
T(ε, ξ0 ) < ∞ to make ξ(t) < ε, for all t ≥ t0 + T.
of the systems, the criterion of semiglobal practical finite-time
Lemma 1 (Young’s Inequality) [40]: For a ∈ R and b ∈ R,
stability (SGPFS) was proposed in [31] for pure-feedback non-
the following inequality holds:
linear systems. Based on this criterion, an adaptive fuzzy con-
trol strategy was constructed, which made the tracking error lm m 1
converge to the neighborhood of the original in finite time. A ab ≤ |a| + n |b|n
m nl
practical finite-time stability corollary was developed in [32],
where l > 0, m > 1, n > 1, and (m − 1)(n − 1) = 1.
and an adaptive finite-time scheme was developed with the
Lemma 2 [41]: For any variables ϕ and ψ, and positive
help of a command filter. An SGPFS criterion was proposed
constants υ, ω, and ρ, the following inequality holds:
for the nontriangular system in [33], and the designed adaptive
υ ω υ
NN controller could acquire the faster tracking performance of |ϕ|υ |ψ|ω ≤ ρ|ϕ|υ+ω + ρ − ω |ψ|υ+ω .
the systems. By applying the mean value theorem of integrals, υ +ω υ +ω
a new SGPFS criterion in [34] was developed for the stochastic
Lemma 3 [31]: Consider the nonlinear system ξ̇ =
systems.
f (ξ ). Suppose that there is a C 1 positive-definite function
Inspired by the above results, this article emphasizes the
V(ξ ) : D → R with scalars c > 0, 0 < β < 1, and η > 0,
finite-time adaptive fuzzy tracking problem for nontriangu-
such that
lar systems subject to full-state constraints and unmeasurable
states. The main contributions in this article are summarized V̇ ≤ −cV β (ξ ) + η, t≥0
as follows.
1) Based on the finite-time stability criterion, an adaptive then the system ξ̇ = f (ξ ) is SGPFS.
fuzzy control scheme is designed for the nontrian- Lemma 4 [41]: For ai ∈ R, i = 1, 2, . . . , p, and a constant
gular nonlinear systems with constrained states. The 0 < σ ≤ 1, the following relation holds:
proposed control strategy guarantees that the practi- p σ p σ
p
cal stability of the systems can be realized in a finite σ
|ai | ≤ |ai | ≤ p1−σ
|ai | .
time. i=1 i=1 i=1
2) The property of the fuzzy basis function is applied
to solve the algebraic loop problem caused by the Lemma 5 [42]: For any positive constant kbi (i =
nontriangular structure, and the assumption condition 1, 2, . . . , n), let E := {e ∈ Rn | : |ei | < kbi } ⊂ Rn and
in [35]–[39] is removed, in which the system functions N := Rl × E ⊂ Rl+n be open sets. Consider the system
must be monotonically increasing. To circumvent the
ξ̇ = f (ξ, u)
obstacle caused by an unmeasurable state variables, a
fuzzy state observer is constructed, which is much more where ξ := (ω, e)T ∈ N and f : R+ × N → Rl+n are piecewise
precise than the general linear observer. continuous in t, and locally Lipschitz in ξ , uniformly in t, and
3) By utilizing the DSC technique based on the backstep- on R+ × N. Let Ei := {ei ∈ R : |ei | < kbi } ⊂ R. If there exist
ping framework, the difficulty of explosion of complex- positive functions U : Rl → R+ and Vi : Ei → R+ which are
ity in traditional backstepping is overcome. The log-type continuously differentiable and positive definite in Rl and Ei ,
BLF associated with error surface systems is incorpo- respectively, such that
rated into each step of the design for the controller and
all states are guaranteed not to transgress the predefined Vi (ei ) → ∞ as ei → kbi or ei → −kbi
sets. γ1 (ω) ≤ U(ω) ≤ γ2 (ω)
The remainder of this article is organized as follows. In n
Section II, some preliminaries and the description of the with γ1 , γ2 ∈ K∞ . Let V(ζ ) := i=1 Vi (ei ) + U(ξ ) and
system are presented. In Section III, a fuzzy observer is given. ei (0) ∈ Ei . If the following relationship holds:
Section IV provides the details of the design process for the ∂V
adaptive controller, and the stability analysis is presented. V̇ = f ≤ −cV β + η
∂ξ
Section V shows two simulation examples to illustrate the
effectiveness of theoretical results. The conclusion is given where c > 0, η > 0, and 0 < β < 1, then ei (t) ∈ Ei ∀t ∈
in Section VI. [0, ∞).
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
n
where i > 0 is a time constant, αi−1 is the virtual control
= Ax̃ + Bi θ̃iT ϕi x̂ + δ
signal serving as the input of the first-order filter (10), and ςi
i=1
is the error surface. Moreover, from the definition (9) and (10),
where δ = (δ1 , δ2 , . . . , δn )T , and θ̃i = θi∗ − θi is the adaptive we can obtain
parameter estimation error. ςi
Choose the Lyapunov function α̇i,L = − , i = 2, 3, . . . , n. (11)
i
1 T
V0 =
x̃ Px̃. (6) Remark 3: In the traditional backstepping method, the vir-
2
tual control signal αi−1 is repeatedly differentiated, which will
The derivative of V0 with respect to t is lead to the problem of explosion of complexity. In order to
n
avoid this problem, an auxiliary first-order filter (10) is intro-
V̇0 = −x̃ Qx̃ + x̃ P
T T
Bi θ̃i ϕi x̂ + δ .
T
(7) duced. The virtual control signal αi−1 passes the filter, a new
i=1 signal αi,L is generated without explosion of terms, and it will
According to Young’s inequality and the property of the be employed in the next step for the controller design. As the
fuzzy basis function 0 < ϕi (·)T ϕi (·) ≤ 1, one obtains results, that repeatedly differentiating is avoided.
n Step 1: According to the change of coordinates (9), the
derivative of e1 respect to t is
T
x̃ P Bi θ̃i ϕi x̂ + δ
T
where δ ∗ = (δ1∗ , δ2∗ , . . . , δn∗ )T . + θ1T ϕ1 (x̂1 ) + θ̃1T ϕ1 (x̂1 ) − ẏr + x̃2 + δ1 . (12)
Combining (8) with (7), yields
Since the states of system (1) are constrained in the set xi ,
2 P2
n
the symmetric log-type BLF is applied to guarantee that all the
V̇0 ≤ −λ0 x̃ + θ̃iT θ̃i + M0 states cannot violate those constraints. The Lyapunov function
2
i=1
is constructed as follows:
where λ0 = λmin (Q) − 1 and M0 = (P2 /2)δ ∗ 2 . 2
1 kb 1 T
Remark 2: For the unmeasured states, the linear state V1 = V0 + log 2 1 2 + θ̃ θ̃1 (13)
observer was introduced in [44]–[46] such as 2 kb − e1 2γ1 1
1
x̂i = x̂i+1 − li y − x̂1 , 1 ≤ i ≤ n − 1 where kb1 = kc1 − A0 , and γ1 > 0 is the design parameter.
In the set e1 = {e1 ∈ R : |e1 | < kb1 }, V1 is a continuous
x̂n = u − ln y − x̂1 .
function. Combine with (12), one can obtain
It is noteworthy that the linear observer does not depend on
the controlled plants and only relates to the output signal. For e1 ė1 1
V̇1 = V̇0 + − θ̃1T θ̇1
the fuzzy observer (3) in this article, the FLS is applied to kb21 − e21 γ1
approximate the uncertain function fi (x), and better estimation e1 α1 − ẏr + e2 + ς2 + x̃2 + δ1 + θ1T ϕ1 (x̂1 )
performance can be obtained. = V̇0 +
kb21 − e21
IV. A DAPTIVE F UZZY C ONTROL D ESIGN e1 θ1∗ T ϕ1 (x̂) − θ1∗ T ϕ1 (x̂1 )
AND S TABILITY A NALYSIS +
kb21 − e21
A. Adaptive Fuzzy Control Design
1 T γ1 ϕ1 (x̂1 )e1
In this section, by combining the backstepping method with − θ̃ θ̇1 − 2 . (14)
the DSC technology, a finite-time adaptive fuzzy control will γ1 1 kb1 − e21
be developed for the systems (1). The detailed design pro-
cess will be presented in the following. First, the change of The property of the fuzzy basis function (0 < ϕ1 (·)T ϕ1 (·) ≤ 1)
coordinates is given and Lemma 1 is applied and yields
e1 = x1 − yr e1 (x̃2 + δ1 ) e2 1 1
≤ 2 1 2 + x̃2 + δ1∗ 2
ei = x̂i − αi,L , i = 2, 3, . . . , n kb1 − e1
2 2 (kb1 − e1 )2 2 2
ςi = αi,L − αi−1 , i = 2, 3, . . . , n (9) e1 θ1∗ T ϕ1 (x̂) − θ1∗ T ϕ1 (x̂1 ) τ e21 2 ∗ 2
≤ 2 + θ1
where αi,L is the output of the first-order filter, which is kb1 − e1
2 2 τ
2 kb21 − e21
defined as
e1 ς2 e21 1 2
i α̇i,L + αi,L = αi−1 , αi,L (0) = αi−1 (0) ≤ 2 + ς2 (15)
kb21 − e12
2 kb21 − e21
2
i = 2, 3, . . . , n (10)
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
where τ > 0. Substituting (15) into (14), yields Similar to the analysis in step 1, we have
2 P2
n
V̇1 ≤ −λ0 x̃ + θ̃iT θ̃i + M0 e2 θ2∗ T ϕ2 (x̂) − θ2∗ T ϕ2 (x̂2 ) τ e22 2 ∗ 2
2 ≤ 2 + θ2
i=1
kb2 − e2
2 2 τ
2 kb22 − e22
e1 α1 − ẏr + θ1 ϕ1 (x̂1 ) + (τ + 3)e1 /2 kb21 − e21
T
+ e2 ς3 e22 1 2
kb21 − e21 ≤ 2 + ς3
kb2 − e2
2 2
2 kb22 − e22
2
e1 e2 1 T γ1 ϕ1 (x̂1 )e1
+ 2 − θ̃1 θ̇1 − 2
kb1 − e21 γ1 kb1 − e21 e2 θ̃2T ϕ2 x̂ e22 1 T
− 2 ≤ 2 + θ̃2 θ̃2 . (23)
1 1 1 2 2 kb2 − e2 2
2 kb22 − e22
2
+ ς22 + x̃2 + δ1∗ 2 + θ1∗ . (16)
2 2 2 τ
The virtual control signal α1 and adaptive law θ˙1 are With the help of the DSC technology and formu-
designed as follows: las (10) and (11), one can obtain ς̇2 = α̇2,L −α̇1 = −(ς2 /2 )+
2β−1
Y2 (·), where Y2 (·) = −[(∂α1 )/(∂ x̂1 )]x̂˙ 1 − [(∂α1 )/(∂θ1 )]θ̇1 −
c1 e1 (τ + 3)e1 [(∂α1 )/(∂yr )]ẏr − [(∂α1 )/(∂ ẏr )]ÿr .
α1 = − β−1 − θ1 ϕ1 (x̂1 ) − + ẏr
T
2 − e2 The virtual controller α2 and the adaptive law θ̇2 are
kb1 − e1
2 2 2 k b1 1
designed as
(17)
2β−1
γ1 ϕ1 (x̂1 )e1 c2 e2
θ̇1 = 2 − μ1 θ1 (18) α2 = − β−1 − l2 x̃1 − θ2 ϕ2 (x̂2 )
T
kb1 − e21 kb22 − e22
where c1 > 0 and μ1 > 0 are designed parameters.
By substituting (17) and (18) into (16), V̇1 is described as (τ + 2)e2 e1 kb22 − e22
− − + α̇2,L (24)
2 kb22 − e22 kb21 − e21
P2
n 2β
c1 e1
V̇1 ≤ −λ1 x̃2 + θ̃iT θ̃i − β
2
kb21 − e21 γ2 ϕ2 x̂2 e2
i=1 θ̇2 = 2 − μ2 θ2 (25)
kb2 − e22
μ1 T 1 e1 e2
+ θ̃ θ1 + ς22 + 2 + M1 (19)
γ1 1 2 kb1 − e21 where c2 > 0 and μ2 > 0 are positive design parameters.
According to (23)–(25), (22) can be rewritten as
where λ1 = λ0 −(1/2) and M1 = M0 +(1/2)δ1∗ 2 +(2/τ )θ1∗ 2 .
P2 T
n 2β
Step 2: The time derivative of e2 is c1 e1
V̇2 ≤ −λ1 x̃2 + θ̃i θ̃i − β
ė2 = x̂˙ 2 − α̇2,L 2
i=1 kb21 − e21
= x̂3 + l2 x̃1 + f̂2 x̂|θ2 − α̇2,L 2β
c2 e2 μ1 T μ2 T 1
= e3 + ς3 + α2 + l2 x̃1 + θ2∗ T ϕ2 x̂ − θ̃2T ϕ2 (x̂) − β + θ̃ θ1 + θ̃ θ2 + θ̃2T θ̃2
γ1 1 γ2 2 2
kb22 − e22
− θ2∗ T ϕ2 x̂2 + θ2T ϕ2 x̂2 + θ̃2T ϕ2 x̂2 − α̇2,L (20)
where x̂2 = (x̂1 , x̂2 )T . 1 2 1 2 ς22 e2 e3
+ ς 2 + ς3 − + |ς2 Y2 | + 2 + M2
The Lyapunov function candidate is selected as 2 2 2 kb2 − e22
kb 2 (26)
1 1 T 1
V2 = V1 + log 2 2 2 + θ̃ θ̃2 + ς22 (21)
2 kb − e2 2γ2 2 2 where M2 = M1 + [(2θ2∗ 2 )/τ ].
2
Step i (i = 3,4, . . . , n − 1): The derivative of ei with respect
where γ2 is the positive design parameter, kb2 > 0 and its to t is given as follows:
definition will be given later. It is obvious that V2 is continuous
in the set e2 = {e2 ∈ R : |e2 | < kb2 }. Then, one can obtain ėi = x̂˙ i − α̇i,L = x̂i+1 + li x̃1 + f̂i x̂|θi − α̇i,L
e2 ė2 1 = ei+1 + ςi+1 + αi + li x̃1 + θi∗ T ϕi x̂ − θ̃iT ϕi (x̂)
V̇2 = V̇1 + − θ̃2T θ̇2 + ς2 ς˙2
kb22 − e22 γ2 − θi∗ T ϕi (x̂i ) + θiT ϕi (x̂i ) + θ̃iT ϕi (x̂i ) − α̇i,L (27)
e2 α2 + l2 x̃1 + e3 + ς3 − α̇2,L + θ2T ϕ2 x̂2
= V̇1 + where x̂i = (x̂1 , x̂2 , . . . , x̂i )T .
kb22 − e22 Choose the Lyapunov function
e2 θ̃2T ϕ2 (x̂) e2 θ2∗ T ϕ2 (x̂) − θ2∗ T ϕ2 x̂2 kb 2
1 1 T 1
− + + ς2 ς˙2 Vi = Vi−1 + log 2 i 2 + θ̃i θ̃i + ςi2 (28)
kb22 − e22 kb2 − e22 2 kb − ei 2γi 2
2 i
1 T γ2 ϕ2 x̂2 e2 where γi > 0 is the design parameter, and kbi will be defined
− θ̃2 θ̇2 − 2 . (22)
γ2 kb2 − e22 later. In the set ei = {ei ∈ R : |ei | < kbi }, Vi is continuous.
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
As the same case of step 2, the following inequalities hold: Choose the positive-definite Lyapunov function
ei θi∗ T ϕi (x̂) − θi∗ T ϕi x̂i τ e2i 2 ∗ 2 1 kb 2
1 T 1
≤ 2 + θi Vn = Vn−1 + log 2 n + θ̃ θ̃n + ςn2 (34)
kbi − ei
2 2 τ 2 kb − e2n 2γn n 2
2 kb2i − e2i n
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Proof: According to the relationship θ̃i = θi∗ − θi , the Note that log [(kb2i )/(kb2i − e2i )] ≤ [(e2i )/(kb2i − e2i )] when
following inequality holds: |ei | < kbi . Using the previous results, (43) can be rewritten as
1 ∗T ∗ 1 T β
θ̃iT θi ≤ θ θ − θ̃i θ̃i 2λ1 1 T β n
1 e2i
2 i i 2 V̇n ≤ − x̃ Px̃ − σ
λmax (P) 2 2 (kbi
2 − e2 )
i
i=1
then n β n−1 β
1 1
n
μi
n
μi ∗ T ∗ μi T
n
−σ θ̃iT θ̃i − σ ς2 + η. (47)
θ̃iT θi ≤ θi θi − θ̃ θ̃i . (40) 2γi 2 i+1
γi 2γi 2γi i i=1 i=1
i=1 i=1 i=1
Define V = Vn , c = min{[(λ1 2β )/(λ
max (P))], σ }, and η =
For any constants δyr > 0 and ϒ > 0, the set yr =
Mn + 2σρ(1 − β) + ([2λ1 ]/[λmax (P)])(1 − β)ρ. We can obtain
{(yr , ẏr , ÿr )T : y2r + ẏ2r + ÿ2r ≤ δyr } and i := {x̃T Px̃ +
i i T i−1 2
j=1 log [(kbj )/(kbj − ej )] +
2 2 2
j=1 (1/γj )θ̃j θ̃j + j=1 ζj+1 ≤ V̇ ≤ −cV β + η. (48)
2ϒ} are compact in R and R , respectively. Thus, yr ×i is
3 3i
On the basis of Lemma 3, the inequality (48) indicates that
also compact in R3+3i . Therefore, Yi has a maximum i > 0,
the closed-loop system is SGPFS. ∀t ≥ T ∗ , we can obtain
such that |Yi | ≤ i , and the following inequality holds:
V β ≤ [η/(1 − κ)c], where 0 < κ ≤ 1, and the settling time
ςi2 2i τ ⎡ ⎤
1−β
|ςi Yi | ≤ + . (41) 1 η β
2τ 2 T∗ = ⎣V 1−β (0) − ⎦.
(1 − β)κc (1 − κ)c
Thus, (39) can be expressed as
β Furthermore, based on the definition of V, for ∀t ≥ T ∗ , we
λ1 n
e2i
β obtain
V̇n ≤ − x̃ Px̃ −
T
2 ci 2
λmax (P) 2 kbi − e2i
i=1 −β 1
1 −2 η 2
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
kb22 − e22
e2 e1 kb22 − e22
− − + α̇2,L
2 kb22 − e22 kb21 − e21
2β−1
c1 e1 (τ + 3)e1
α1 = − β−1 − θ1 ϕ1 x̂1 − + ẏr
T
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
Fig. 4. Trajectories of x1 , x̂1 , and yr and the tracking error e1 in Example Fig. 6. Trajectories of u in Example 2.
2.
Fig. 5. Trajectories of x2 , x̂2 , and x3 and x̂3 in Example 2. Fig. 7. Phase portraits of the observer error x̃1 , x̃2 , and x̃3 in Example 2.
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
control method. Further, the finite-time command filtered con- [21] W. Sun, S.-F. Su, Y. Wu, J. Xia, and V.-T. Nguyen, “Reduced
trol will be applied to the nontriangular nonlinear systems, and adaptive fuzzy tracking control for high-order stochastic non-
strict feedback nonlinear system with full-state constraints,” IEEE
the external disturbance can be considered simultaneously. Trans. Syst., Man, Cybern., Syst., early access, Mar. 5, 2019,
doi: 10.1109/TSMC.2019.2898204.
[22] H. Du, S. Li, and C. Qian, “Finite-time attitude tracking control of
R EFERENCES spacecraft with application to attitude synchronization,” IEEE Trans.
Autom. Control, vol. 56, no. 11, pp. 2711–2717, Nov. 2011.
[1] H. Ma, H. Liang, Q. Zhou, and C. K. Ahn, “Adaptive dynamic sur- [23] D. Zhai and Y. Xia, “Finite-time control of teleoperation systems with
face control design for uncertain nonlinear strict-feedback systems with input saturation and varying time delays,” IEEE Trans. Syst., Man,
unknown control direction and disturbances,” IEEE Trans. Syst., Man, Cybern., Syst., vol. 47, no. 7, pp. 1522–1534, Jul. 2017.
Cybern., Syst., vol. 49, no. 3, pp. 506–515, Mar. 2019. [24] H. Zhang, A. Song, H. Li, and S. Shen, “Novel adaptive finite-
[2] Z.-Y. Sun, C.-H. Zhang, and Z. Wang, “Adaptive disturbance attenuation time control of teleoperation system with time-varying delays and
for generalized high-order uncertain nonlinear systems,” Automatica, input saturation,” IEEE Trans. Cybern., early access, Jul. 18, 2019,
vol. 80, pp. 102–109, Jun. 2017. doi: 10.1109/TCYB.2019.2924446.
[3] L. Zhang, H. Liang, Y. Sun, and C. K. Ahn, “Adaptive event-triggered [25] H. Huang and Y. Jia, “Adaptive finite-time 6-DOF tracking control for
fault detection scheme for semi-Markovian jump systems with output spacecraft fly-around with input saturation and state constraints,” IEEE
quantization,” IEEE Trans. Syst., Man, Cybern., Syst., early access, Trans. Aerosp. Electron. Syst., vol. 55, no. 6, pp. 3259–3272, Dec. 2019,
May 7, 2019, doi: 10.1109/TSMC.2019.2912846. doi: 10.1109/TAES.2019.2906096.
[4] Z.-Y. Sun, C.-Y. Liu, S.-F. Su, and W. Sun, “Robust stabilization of high- [26] B. P. Sanjay and B. S. Dennis, “Finite-time stability of continu-
order nonlinear systems with unknown sensitivities and applications in ous autonomous systems,” SIAM J. Control Optim., vol. 38, no. 3,
humanoid robot manipulation,” IEEE Trans. Syst., Man, Cybern., Syst., pp. 751–766, 2000.
early access, Aug. 30, 2019, doi: 10.1109/TSMC.2019.2931768. [27] Z. Zhu, Y. Xia, and M. Fu, “Attitude stabilization of rigid spacecraft
[5] S. Yin, P. Shi, and H. Yang, “Adaptive fuzzy control of strict-feedback with finite-time convergence,” Int. J. Robust Nonlinear Control, vol. 21,
nonlinear time-delay systems with unmodeled dynamics,” IEEE Trans. no. 6, pp. 686–702, 2011.
Cybern., vol. 46, no. 8, pp. 1926–1938, Aug. 2016. [28] Z.-Y. Sun, Y. Shao, and C.-C. Chen, “Fast finite-time stability and
[6] Y. Wang, L. Zheng, H. Zhang, and W. X. Zheng, “Fuzzy observer-based its application in adaptive control of high-order nonlinear system,”
repetitive tracking control for nonlinear systems,” IEEE Trans. Fuzzy Automatica, vol. 106, pp. 339–348, Aug. 2019.
Syst., early access, Aug. 21, 2019, doi: 10.1109/TFUZZ.2019.2936808. [29] Z.-Y. Sun, M.-M. Yun, and T. Li, “A new approach to fast global finite-
[7] Y. Wang, H. Zhang, and Y.-Z. Wang, “Fuzzy adaptive control of stochas- time stabilization of high-order nonlinear system,” Automatica, vol. 81,
tic nonlinear systems with unknown virtual control gain function,” Acta pp. 455–463, 2017.
Automatica Sinica, vol. 235, pp. 170–178, Jan. 2014. [30] Z.-Y. Sun, Y.-Y. Dong, and C.-C. Chen, “Global fast finite-time par-
[8] R. Lu, J. Tao, P. Shi, H. Su, Z.-G. Wu, and Y. Xu, “Dissipativity-based tial state feedback stabilization of high-order nonlinear systems with
resilient filtering of periodic Markovian jump neural networks with quan- dynamic uncertainties,” Inf. Sci., vol. 484, pp. 219–236, May 2019.
tized measurements,” IEEE Trans. Neural Netw. Learn. Syst., vol. 29, [31] F. Wang, B. Chen, X. Liu, and C. Lin, “Finite-time adaptive fuzzy
no. 5, pp. 1888–1899, May 2018. tracking control design for nonlinear systems,” IEEE Trans. Fuzzy Syst.,
[9] S. Tong, Y. Li, and S. Sui, “Adaptive fuzzy tracking control design vol. 26, no. 3, pp. 1207–1216, Jun. 2018.
for SISO uncertain nonstrict feedback nonlinear systems,” IEEE Trans. [32] J. Yu, P. Shi, and L. Zhao, “Finite-time command filtered backstep-
Fuzzy Syst., vol. 24, no. 6, pp. 1441–1454, Dec. 2016. ping control for a class of nonlinear systems,” Automatica, vol. 92,
[10] P. Li, J. Lam, R. Lu, and K.-W. Kwok, “Stability and L2 synthesis of a pp. 173–180, Jun. 2018.
class of periodic piecewise time-varying systems,” IEEE Trans. Autom. [33] Y. Sun, B. Chen, C. Lin, and H. Wang, “Finite-time adaptive control for
Control, vol. 64, no. 8, pp. 3378–3384, Aug. 2019. a class of nonlinear systems with nonstrict feedback structure,” IEEE
[11] X.-M. Li, B. Zhang, P. Li, Q. Zhou, and R. Lu, “Finite-horizon H∞ Trans. Cybern., vol. 48, no. 10, pp. 2774–2782, Oct. 2018.
state estimation for periodic neural networks over fading channels,” [34] F. Wang, B. Chen, Y. Sun, Y. Gao, and C. Lin, “Finite-time fuzzy control
IEEE Trans. Neural Netw. Learn. Syst., early access, Jul. 1, 2019, of stochastic nonlinear systems,” IEEE Trans. Cybern., early access,
doi: 10.1109/TNNLS.2019.2920368. Jul. 18, 2019, doi: 10.1109/TCYB.2019.2925573.
[12] P. Du, H. Liang, S. Zhao, and K. C. Ahn, “Neural-based decentralized [35] H. Wang, X. P. Liu, X. Zhao, and X. Liu, “Adaptive fuzzy finite-time
adaptive finite-time control for nonlinear large-scale systems with time- control of nonlinear systems with actuator faults,” IEEE Trans. Cybern.,
varying output constraints,” IEEE Trans. Syst., Man, Cybern., Syst., early early access, May 8, 2019, doi: 10.1109/TCYB.2019.2902868.
access, Jun. 13, 2019, doi: 10.1109/TSMC.2019.2918351. [36] H. Li, L. Bai, L. Wang, Q. Zhou, and H. Wang, “Adaptive neural control
[13] D. Swaroop, J. K. Hedrick, P. P. Yip, and J. C. Gerdes, “Dynamic surface of uncertain nonstrict-feedback stochastic nonlinear systems with output
control for a class of nonlinear systems,” IEEE Trans. Autom. Control, constraint and unknown dead zone,” IEEE Trans. Syst., Man, Cybern.,
vol. 45, no. 10, pp. 1893–1899, Oct. 2000. Syst., vol. 47, no. 8, pp. 2048–2059, Aug. 2017.
[14] D. Wang and J. Huang, “Neural network-based adaptive dynamic sur- [37] Q. Zhou, L. Wang, C. Wu, and H. Li, “Adaptive fuzzy control for
face control for a class of uncertain nonlinear systems in strict-feedback nonstrict-feedback systems with input saturation and output constraint,”
form,” IEEE Trans. Neural Netw., vol. 16, no. 1, pp. 195–202, Jan. 2005. IEEE Trans. Syst., Man, Cybern., Syst., vol. 47, no. 1, pp. 1–12,
[15] T.-S. Li, D. Wang, G. Feng, and S.-C. Tong, “A DSC approach to robust Jan. 2017.
adaptive NN tracking control for strict-feedback nonlinear systems,” [38] H. Su and W. Zhang, “Adaptive fuzzy tracking control for a class of
IEEE Trans. Syst., Man, Cybern. B, Cybern., vol. 40, no. 3, pp. 915–927, nonstrict-feedback stochastic nonlinear systems with actuator faults,”
Jun. 2010. IEEE Trans. Syst., Man, Cybern., Syst., early access, Dec. 12, 2018,
[16] K. P. Tee, S. S. Ge, and E. H. Tay, “Barrier Lyapunov functions for the doi: 10.1109/TSMC.2018.2883414.
control of output-constrained nonlinear systems,” Automatica, vol. 45, [39] H. Wang, B. Chen, K. Liu, X. Liu, and C. Lin, “Adaptive neural
no. 4, pp. 918–927, 2009. tracking control for a class of nonstrict-feedback stochastic nonlinear
[17] Y.-J. Liu and S. Tong, “Barrier Lyapunov functions-based adaptive systems with unknown backlash-like hysteresis,” IEEE Trans. Neural
control for a class of nonlinear pure-feedback systems with full state Netw. Learn. Syst., vol. 25, no. 5, pp. 947–958, May 2014.
constraints,” Automatica, vol. 64, pp. 70–75, Feb. 2016. [40] H. Deng and M. Krstić, “Output-feedback stochastic nonlinear stabi-
[18] K. P. Tee and S. S. Ge, “Control of nonlinear systems with partial state lization,” IEEE Trans. Autom. Control, vol. 44, no. 2, pp. 328–333,
constraints using a barrier Lyapunov function,” Int. J. Control, vol. 84, Feb. 1999.
no. 12, pp. 2008–2023, 2011. [41] H. G. Hardy, E. Littlewood, and J. Polya, Inequalities. Cambridge, U.K.:
[19] C.-C. Chen and Z.-Y. Sun, “A new approach to stabilisation of a class Cambridge Univ. Press, 1995.
of nonlinear systems with an output constraint,” Int. J. Control, to be [42] H. Li, S. Zhao, W. He, and R. Lu, “Adaptive finite-time tracking control
published, doi: 10.1080/00207179.2018.1501162. of full state constrained nonlinear systems with dead-zone,” Automatica,
[20] W. Sun, S.-F. Su, Y. Wu, J. Xia, and V.-T. Nguyen, “Adaptive fuzzy con- vol. 100, pp. 99–107, Feb. 2019.
trol with high-order barrier Lyapunov functions for high-order uncertain [43] L.-X. Wang and J. M. Mendel, “Fuzzy basis functions, universal approx-
nonlinear systems with full-state constraints,” IEEE Trans. Cybern., early imation, and orthogonal least-squares learning,” IEEE Trans. Neural
access, Jan. 15, 2019, doi: 10.1109/TCYB.2018.2890256. Netw., vol. 3, no. 5, pp. 807–814, Sep. 1992.
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.
This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.
[44] B. Chen, H. Zhang, and C. Lin, “Observer-based adaptive neural network Yang Liu received the B.S. degree in mathematics
control for nonlinear systems in nonstrict-feedback form,” IEEE Trans. from Bohai University, Jinzhou, China, in 2009, and
Neural Netw. Learn. Syst., vol. 27, no. 1, pp. 89–98, Jan. 2016. the M.S. degree in fundamental mathematics from
[45] Q. Zhou, P. Shi, J. Lu, and S. Xu, “Adaptive output-feedback fuzzy Northeastern University, Shenyang, China, in 2011,
tracking control for a class of nonlinear systems,” IEEE Trans. Fuzzy where she is currently pursuing the Ph.D. degree in
Syst., vol. 19, no. 5, pp. 972–982, Oct. 2011. control theory and control engineering.
[46] J. Yu, L. Zhao, H. Yu, and C. Lin, “Barrier Lyapunov functions-based Her current research interests include adaptive
command filtered output feedback control for full-state constrained fuzzy control and neural-networks control.
nonlinear systems,” Automatica, vol. 105, pp. 71–79, Jul. 2019.
[47] M. Spong, S. Hutchinson, and M. Vidyasagar, Robot Dynamic and
Control. New York, NY, USA: Cambridge Univ. Press, 1989.
Authorized licensed use limited to: University of Canberra. Downloaded on May 01,2020 at 07:35:54 UTC from IEEE Xplore. Restrictions apply.