This item was submitted to Loughborough's Research Repository by the author.

Items in Figshare are protected by copyright, with all rights reserved, unless otherwise indicated.

Nonsmooth adaptive control for uncertain nonlinear systems: a non-recursive

design approach
PLEASE CITE THE PUBLISHED VERSION

https://doi.org/10.1109/lcsys.2021.3065187

PUBLISHER

Institute of Electrical and Electronics Engineers (IEEE)

VERSION

AM (Accepted Manuscript)

PUBLISHER STATEMENT

© 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other
uses, in any current or future media, including reprinting/republishing this material for advertising or
promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of
any copyrighted component of this work in other works.

LICENCE

All Rights Reserved

REPOSITORY RECORD

Zhang, Chuanlin, and Jun Yang. 2021. “Nonsmooth Adaptive Control for Uncertain Nonlinear Systems: A
Non-recursive Design Approach”. Loughborough University. https://hdl.handle.net/2134/16578584.v1.
 Nonsmooth adaptive control for uncertain nonlinear systems: a
non-recursive design approach
Chuanlin Zhang, Senior Member, IEEE and Jun Yang, Senior Member, IEEE

Abstract— In this paper, a one-step nonsmooth adaptive [11], it hence arouses a great deal of interests from the control
controller design framework is proposed the first time for community in recent decades. By incorporating an adding
a class of nonlinear systems with general non-parametric a power integrator approach, a systematic robust nonsmooth
uncertainties. By virtue of a novel non-recursive synthe-
sis philosophy, a nonsmooth adaptive stabilizer can be control design methodology has been well developed, see e.gs.,
constructed straightforwardly from the system in a possi- [12], [13] and references therein. Regarding the finite-time
ble simplest form, which facilitates practical implementa- adaptive stabilization design, under a polynomial hypothesis
tions. In reference to well-acknowledged recursive design of the system structure, the work in [3] proposes a nonsmooth
based approaches, direct improvements with this new non- adaptive design framework while a later result in [4] has
recursive methodology include largely reduced synthesis
complexity along with an essential detachment of control extended the finite-time adaptive control issue to a class of
law design and stability analysis. A numerical example is p-normal uncertain nonlinear systems. In a recent research,
provided to demonstrate both the simplicity and effective- [14] has proposed a finite-time stabilizing controller with time-
ness of the proposed algorithm. varying gains being tuned online by two dynamic equations.
Index Terms— adaptive stabilization, non-recursive de- Later on, [15] develops a finite-time stabilization result via
sign, homogeneous system theory, non-parametric uncer- the adding a power integrator technique. [16] shows the
tainty existence of a monotone dynamic scaling gain, whose explicit
construction procedure for practices, however remains open.
I. I NTRODUCTION If an uncertain high order system is considered, a recent
INCE the era of modern control theory, robust and study in [17] addresses the adaptive asymptotical stabilization
S adaptive control strategies have been well established for
systems in the presence of uncertain parameters, unknown
problem by incorporating a dual-layer update law with the
adding a power integrator technique. Using a similar dynamic
actuator nonlinearities, unmodelled dynamics and external handling strategy, the work [18] develops an adaptive finite-
disturbances, etc. Various control strategies can be found time stabilization result under a modified global nonlinearity
in the literature regarding the synthesis issue for uncertain restriction hypothesis.
nonlinear systems, such as high-gain control [1], adaptive It is worth pointing out here that most of the above
backstepping control [2], nonsmooth control [3], [4], [5], [6], mentioned adaptive design approaches involve a recursive
etc. As one classical design algorithm, adaptive backstepping determination of virtual controllers. Direct drawbacks may
design approach is well developed under the condition that the exhibit on the expanding complexity when the system order
system should subject to a parametric structure [2]. However, is growing (also be called as “explosion of complexity”).
for nonlinear systems with the presence of non-parametric Meanwhile, the gain selection guideline is made much conser-
uncertainties, an alternative synthesis tool is of significance, vative by going through a variety of necessary mathematical
while fewer results can be found. It is reported in [7], [8], estimations, see for instances, [14], [16], etc. These facts can
[9], [10], a time-varying scaling gain method can be found, be regarded as main obstacles for wide application of advanced
which could derive universal control laws by means of a dy- nonlinear control algorithms in real-life control engineering.
namic high-gain domination technique. However, a Lipschitz In the literature, there are some recent advances focusing on
continuous condition is required for the assurance of a global the relaxation of controller design complexity by proposing
asymptotical stabilization result. a one-step or non-recursive design strategy. For instance, by
By noting that in reference to popular asymptotical control integrating the dynamic surface control with neural networks,
strategies, nonsmooth control design could provide the closed- a one-step neuroadaptive control scheme is proposed in [19],
loop system a faster convergence rate and stronger robustness aiming to avoid the laborious repetitive design procedures. A
later result in [20] proposes an adaptive NN output feedback
This work is supported in part by National Natural Science Foundation
of China (No. 61503236 and 61973080), in part by the Program for Pro- control law for constrained pure-feedback systems in a non-
fessor of Special Appointment (Eastern Scholar) at Shanghai Institutions recursive manner. In [21], a non-recursive adaptive controller
of Higher Learning and in part by Shanghai Rising-Star program (NO. is constructed for an uncertain underactuated crane system. A
20QA1404000) (Corresponding author: J. Yang).
Chuanlin Zhang is with the College of Automation Engineering, novel C1 adaptive controller is also built in a non-recursive
Shanghai University of Electric Power, Shanghai, China, 200090 (e-mail: manner in [22], where the cumulative complexity in classical
[email protected]) backstepping design can be totally avoided.
Jun Yang is with the Department of Aeronautical and Automotive
Engineering, Loughborough University, Loughborough LE11 3TU, U.K. Regarding the nonsmooth control case, it is reported in
(e-mail: [email protected]) [23] that a global robust homogeneous stabilizer for a class
of inherent nonlinear systems can be built in a novel non- ii) [25] For a coordinate x ∈ ℜn , and positive real numbers
recursive synthesis manner. It is illustrated that the control (r1 , r2 , · · · , rn ) , r, a one-parameter family of dilation is a map
design procedure is largely simplified by neglecting recur- ∆r : ℜ+ × ℜn → ℜn , defined by ∆rε x = (ε r1 x1 , · · · , ε rn xn ) where
sive virtual controllers. Later on, based on this initiative, a ε > 0 is a constant parameter. For a given dilation ∆r and a
non-recursive finite-time exact tracking control algorithm for real number τ, a continuous function V : ℜn → ℜ is called
nonlinear systems with mismatched disturbances is proposed ∆r −homogeneous of degree τ, denoted by V ∈ Hτ∆r if V (∆rε ) =
in [24]. However, on one hand, these robust design meth- ε τ V . A continuous vector field f (x) is ∆r − homogeneous of
τ+r
ods with conservatively selected control gain may behave degree τ, if f j ∈ H∆r j , j ∈ N1:n . A homogeneous vector bxeτ∆r
τ τ
a deteriorate transient-time performance when the system is is denoted by bxeτ∆r , col(bx1 e r1 , · · · , bxn e ri ). A homogeneous
facing largely varying disturbances/uncertainties. On the other n
hand, the global non-recursive dynamic control design issue 2−norm is defined by kxk∆r = ( ∑ |xi |2/ri )1/2 .
i=1
for nonlinear systems with non-parametric uncertainties is Some useful lemmas are stated as follows for the conve-
not a trivial extension from its robust control counterpart. nience of readers.
Firstly, very few existing results can be referred to regarding Lemma 2.1: [25] Let Φ1 (x) ∈ Hτ∆1κ and Φ2 (x) ∈ Hτ∆2κ , re-
one-step nonsmooth synthesis approach for nonlinear systems spectively, then the following statements hold
with non-parametric uncertainties. Secondly, the online esti- i) Φ1 (x)Φ2 (x) ∈ Hτ∆1κ+τ2 .
mation/identification of the internal coupled uncertainties and ii) ∂ Φ∂1x(x)
i
∈ Hτ∆1κ−ri , i ∈ N1:n .
its implicit relation with control gain tuning mechanism are iii) If Φ1 (x), Φ2 (x) are positive definite, then there exists two
still very intricate. Thus, in order to present an explicit and τ2
τ
simple gain tuning mechanism, new design tools should be constants b1 ∈ ℜ+ and b2 ∈ ℜ+ , such that b1 Φ11 (x) ≤ Φ2 (x) ≤
τ2
investigated. τ
b2 Φ11 (x).
Inspired by the design facility of homogeneous system the-
Lemma 2.2: [11] Consider a dynamical system η̇ =
ory as illustrated in the existing literatures partially mentioned
f (η), f (0) = 0, where f : D → ℜn is non-Lipschitz continuous
above, a non-recursive nonsmooth adaptive controller design
on an open neighborhood D of the origin η = 0 in ℜn . If there
framework is proposed the first time to solve the dynamic
exist an open neighborhood U of the origin, a C1 positive-
stabilization problem for a class of nonlinear systems with the
definite and proper Lyapunov function V : U\{0} → ℜ+ and
presence of general non-parametric uncertainties. To assure
real numbers c ∈ ℜ+ and ι ∈ (0, 1), such that V̇ + cV ι ≤ 0 for
the theoretical justifications, rigourous stability analysis is
η ∈ U, then the origin η = 0 is a locally finite-time stable
presented by guaranteeing first the non-existence of a finite 1−ι (η ,t )

escape time and then a local finite time convergence rate. equilibrium with a settling time T ≤ V c(1−ι) 0 0
for any given
Later on, a numerical example and its simulation verification initial condition η0 = η(t0 ).
are offered to demonstrate the design simplicity and control Lemma 2.3: Let D = diag{d1 , d2 , · · · , dn } where di > 0, i ∈
performance of the proposed control scheme. In reference to N1:n is a constant and A ∈ ℜn×n , B ∈ ℜn be matrices in
existing related results, the main contributions of this paper the controllable canonical form. Then there exist a positive
can be summarized as the following two aspects: definite, symmetrical matrix P ∈ ℜn×n and a row vector K =
• As a direct benefit from the non-recursive design ap- [k1 , · · · , kn ], and two positive constants ν1 and ν2 , such that
proach, a simplest non-nested stabilizing controller can be (A − BK)> P + P(A − BK) ≤ −I, ν1 I ≤ DP + PD ≤ ν2 I.
constructed independent of the Lyapunov function based Proof: The proof follows straightforwardly from Theorem
stability analysis. A1 in [7], hence is omitted here.
• The control gain vector can be easily determined while
the dynamic scaling gain is online updated subjecting to
an explicit self-tuning adaption law. B. Problem Statement
The remainder of this paper is organized as follows. Some This paper revisits a class of lower-triangular nonlinear
notations, definitions, useful lemmas and problem statement system with non-parametric uncertainties, depicted by the
are listed in Section 2. The main result of this paper is following form:
presented in Section 3. In Section 4, a numerical simulation 
result is provided. Finally, a conclusion and a reference list ẋi = xi+1 + fi (θ , x̄i ), i ∈ N1:n−1 ,
(1)
end the paper. ẋn = u + fn (θ , x),
II. P RELIMINARIES AND P ROBLEM S TATEMENT where x̄i = col(x1 , x2 , · · · , xi ), x = x̄n and u are the system
A. Preliminaries partial state vector, full state vector and control input, respec-
The following definitions and notations are provided for tively, θ ∈ ℜr is an unknown bounded parameter vector which
briefness of expressions. could be either time-varying or constant, fi (·), i ∈ N1:n is a
i) For integers j and i satisfying 0 ≤ j ≤ i, denote N j:i := continuous nonlinear function satisfying a vanishing condition,
{ j, j + 1, · · · , i}. The symbol Ci denotes the set of all differen- i.e., fi (θ , 0) = 0 for i ∈ N1:n . Through out this paper, the initial
tiable functions whose first i-th time derivatives are continu- time is set as t0 while the initial state vector is denoted by
ous. ℜ+ denotes the set of positive real numbers. A continuous x0 , x(t0 ).
function bxer with r ∈ ℜ+ is defined by bxer , sign(x)|x|r . The control objective of this paper is to find a dynamic
stabilizing controller of the following non-nested form: systems, such as [3], [5], [16], etc., a distinguishable feature
 of the proposed controller is that both the proposed control
 u = u(`, x) ∈ C0 , u(`, 0) = 0,
law and design procedure are much simpler by totally ne-
`˙ = ψ(`, x) ∈ C1 [1, +∞) × ℜn → [0, +∞) ,


(2)
glecting the costly recursive calculations. More significantly,
`(t0 ) = 1,

the controller design procedure is essentially detached from
such that the following conclusions hold for the resulting a recursive determination process of virtual controllers so
closed-loop system (1)-(2). that the step-by-step calculations of the tuning functions are
i) All the signals are globally uniformly bounded. also avoided. This characteristic could endow the practitioners
ii) The origin x = 0 is a finite-time stable equilibrium. much more convenience to implement the controller in real-
life plants.
III. N ON -R ECURSIVE A DAPTIVE C ONTROLLER D ESIGN Remark 3.2: Inspired by the dynamic high gain technique
In this section, three handling procedures are presented to in [7], [8], [9] and the facility of homogeneous system
better illustrate the nonsmooth adaptive stabilization design. theory, we have found that if an indirect adaption law for
Step 1. Guideline for the Selection of Design Parameters the dynamic gain is settled, rather than directly estimating
With the denotation that A, B are matrices of controllable the uncertain parameters via classical tuning functions, a
canonical form, z(θ , x) = col( f1 (·), · · · , fn (·)), system (1) can non-recursive adaptive synthesis approach is promising. The
be rewritten as proposed method can be seen as an essential nonsmooth
extension of the result in [8], [9]. Indeed, in a special case
ẋ = Ax + Bu + z(θ , x). (3) by setting the homogeneous degree τ = 0, the control law (6)
Throughout this paper, the homogeneous dilation weight r is could reduce to a linear universal controller of the following
set as r1 = 1, ri = ri−1 + τ, i ∈ N2:n with a degree τ ∈ (− n1 , 0). form:

Let κ > 1nbe a pre-set design constant. Denote κ − τ/2 =oγ, u = `ρ+n v, v = −Kξ ,
o n (7)
Ξ1 , diag rγ1 , rγ2 , · · · , rγn and Ξ2 , diag 0, rγ2 , · · · , (n−1)γ . `˙ = µkξ k2 , `(t0 ) = 1.
rn
It is obvious that γ > 0 and ri > 0. Hence by following Remark 3.3: Owing to the proposed systematic non-
Lemma 2.3, there exist a positive definite, symmetrical matrix recursive synthesis approach, now the control gain parameter
P ∈ ℜn×n and a gain vector K, such that determination can be simplified as preliminary works. Similar
to the method in [14], one can find a constant gain vector K
(A − BK)> P + P(A − BK) ≤ −I, to render A − BK Hurwitz first, and then obtain P by solving
Ξ1 P + PΞ1 > 0. (4) LMI in (4). Therefore the value ρ can be selected. Noting that
the value of µ is critical to the increasing rate of `, hence, a
Based on the relation (4), we could define a suffi- proper choice of the parameter µ can be determined via “trial
n large constant odesign parameter ρ satisfying ρ >
ciently and error” method. Subsequently, the scaling gain ` will be
(Ξ2 P+PΞ2 )
max 0, − λλmin (Ξ . self-tuned online from its initial value `(0) = 1. Such a gain
min 1 P+PΞ1 )
Step 2. Change of Coordinates: Define a x−ξ coordination tuning mechanism can be regarded as a mediate role between
transform by the following form adaptive and robust methods. It provides the controlled system
some robustness, but the conservative selection guideline of `
ξi = xi /`ρ+i−1 , i ∈ N1:n , v = u/`ρ+n , in existing related robust methods, such as [13], [23], is no
where ` is a scaling gain function to be made precise by an longer required.
update law later on. By denoting An example as a demonstration of the simplicity and
effectiveness of the proposed control scheme is given below.
ξ = col(ξ1 , · · · , ξn ), D = diag{0, 1, · · · , n − 1}, Example 3.1: Consider the following uncertain nonlinear
z̃(`, θ , x) = col( f1 (·)/`ρ , · · · , fn (·)/`ρ+n−1 ), system:

and In as an identity matrix, it is not difficult to derive that  ẋ1 = x2 + ln(1 + (θ1 x1 )2 )
ẋ = x3 + sin(θ2 x2 ) (8)
the ξ -system dynamics should be expressed as the following  2
compact form: ẋ3 = u,

ξ˙ = `(Aξ + Bv) − (ρIn + D)ξ `/`

˙ + z̃(`, θ , x). (5) where θ1 and θ2 are unknown parameters.
As one of the most distinguishable feature in reference to
Step 3. Dynamic Control Law Construction: At this stage, existing recursive adaptive design strategies, such as back-
corresponding to (2), we are already able to design a dynamic stepping design, under the proposed framework, without go-
state feedback stabilizing controller of the following simplest ing through a series of recursive determination of virtual
form: controllers, one can design directly from system (8) of the
u = `ρ+n v, v = −Kbξ er∆nr+τ , following dynamic stabilizing controller:

(6)
`˙ = µkξ k2κ
∆r , `(t0 ) = 1,
(
>
u(t) = −`ρ+3 K bξ1 e4/7 , bξ2 e2/3 , bξ3 e4/5 , (9)
where µ ∈ ℜ+ is a design constant. `˙ = µkξ k2κ
∆r , `(0) = 1,
Remark 3.1: Compared to existing related results on uni- p
versal adaptive design problem for non-parametric nonlinear where kξ k∆r = |ξ1 |2 + |ξ2 |7/3 + |ξ3 |14/5 .
 IV. S TABILITY A NALYSIS Secondly, with Assumption 4.1 in mind, we have
fi (·) σ 1+iτ 1+iτ 1+iτ

In this section, we will show that global stabilization can ≤ |x1 | 1 + |x2 | 1+τ + · · · + |xi | 1+(i−1)τ
`ρ+i−1 `ρ+i−1
be realized by the proposed nonsmooth adaptive controller of σ 1+iτ 1+iτ
= ρ+i−1 |` ξ1 | 1 + |`ρ+1 ξ2 | 1+τ
ρ
the form (6) under the following condition. `
1+iτ

Assumption 4.1: There exists an unknown constant σ ∈ ℜ+ + · · · + |`ρ+i−1 ξi | 1+(i−1)τ
which is dependent on θ , such that the following relation ρ(1+iτ) 1+iτ 1+iτ 1+iτ
ri +τ ri +τ ri +τ =σ ` 1 −(ρ+i−1)
+ `(ρ+1) 1+τ −(ρ+i−1) |ξ2 | 1+τ
|ξ1 | 1
holds: | fi (θ , x̄i )| ≤ σ (|x1 |r1
+ |x2 |r2
+ · · · + |xi |
ri), i ∈ 1+iτ −(ρ+i−1) 1+iτ

(ρ+i−1) 1+(i−1)τ
N1:n . +···+` |ξi | 1+(i−1)τ . (12)
Remark 4.1: Assumption 4.1 can also be called as a ho- Note that ` ≥ 1, 0 < 1+iτ
< 1 and ρ > 0 which conclude
1+( j−1)τ
mogeneous growth condition, which is commonly utilized for (ρ+ j−1) 1+(1+iτ −(ρ+i−1)
a global stabilization objective in the sense of theoretical that ` j−1)τ < `0 = 1 for j ∈ N1:i . Hence
conscientiousness, see, e.gs., [13], [15], etc. However, in following (12), we have
real-life applications, Assumption 4.1 can also be removed fi (θ , x̄i ) 1+iτ 1+iτ 1+iτ

by exploiting a semi-global stabilization objective, see for ≤σ |ξ 1 | 1 + |ξ | 1+τ + · · · + |ξ | 1+(i−1)τ

2 i
`ρ+i−1
instance, [24], [26], etc. ≤c̄2,i σ kξ kr∆ir+τ , (13)
The main result of this paper can be stated as follows.
where c̄2,i ∈ ℜ+ is a constant independent of `.
Theorem 4.1: Consider the closed-loop system consisting 2γ
It is clear that V (ξ ) ∈ H∆r . By utilizing Lemma 2.1, there
of (1) satisfying Assumption 4.1 and the nonsmooth adaptive
exists a constant c2 ∈ ℜ+ which is independent of `, such that
control scheme (6). The following conclusions hold.
the following relations hold
i) All the signals in the closed-loop system are globally
n
uniformly bounded. ∂V (ξ ) ∂V (ξ )
z̃(`, θ , x) ≤ ∑ c̄2,i σ kξ kr∆ir+τ ≤ c2 σ kξ k2κ
∆r .
ii) The system state will converge to the origin within a finite ∂ξ> i=1 ∂ ξ i
time. (14)
Proof: Define a Lyapunov function candidate as V (ξ ) = ˙ ≥ 0, it is clear that
Thirdly, from the definition of ρ and `/`
γ
> γ
bξ e∆r Pbξ e∆r . From the fact that the homogeneous degree (Ξ1 P + PΞ1 )ρ + Ξ2 P + PΞ2 > 0. Then with Lemma 2.1, there
τ < 0 and κ > 1, we know that V (ξ ) ∈ C1 . With the dilation must exist a constant c3 ∈ ℜ+ such that the following relation
mapping ∆rε ξ = (ε r1 ξ1 , · · · , ε rn ξn ), one can easily verify that holds
V (∆rε ξ ) = ε 2γ V (ξ ). ∂V (ξ ) ˙
(ρIn + D)ξ `/`
Calculating the time derivative of V (ξ ) along system (5) ∂ξ>
gives γ
>
γ ˙
= bξ e r ∆ (Ξ1 P + PΞ1 )ρ + Ξ2 P + PΞ2 bξ e r `/` ∆
˙ 2γ
∂V (ξ ) ∂V (ξ ) ˙ ≥ c3 `/`kξ k∆r . (15)
V̇ (ξ ) = `(Aξ + Bv) − (ρIn + D)ξ `/`
∂ξ> ∂ξ> Substituting (11), (14) and (15) into (10), the time derivative
∂V (ξ ) of V (ξ ) along system (5)-(6) can be reduced to the following
+ z̃(`, θ , x). (10)
∂ξ> form
˙ 2γ
V̇ (ξ ) ≤ −c1 `kξ k2κ 2κ
∆r − c3 `/`kξ k∆r + c2 σ kξ k∆r
To proceed, a series of estimations on the terms in (10) will
be given. ≤ −(c1 ` − c2 σ )kξ k2κ
∆r . (16)
Firstly, it is not difficult to obtain the following relations i) In what follows, motivated by [9], we will use contra-
diction arguments to prove that there will be no finite escape

ξi+1 ◦ ∆rε = ε ri+1 ξi+1 = ε ri +τ ξi+1 , i ∈ N1:n−1 , time of the trajectory (ξ > , `)> starting from t0 . If the above
v ◦ ∆rε = −ε rn +τ Kbξ er∆nr+τ , statement is not true, there must exist a maximal time interval
[t0 ,t f ) where t f > t0 , such that (ξ > , `)> are well defined and
lim k(ξ > , `)> k = +∞.
which concludes that the vector field Aξ + Bv is ∆r - t→t f
homogeneous of degree τ. One can follow the corresponding Firstly, we show that ` cannot escape at t = t f . Otherwise,
proof in [25], [27] directly that the closed-loop system is the relation lim ` = +∞ holds. Noting that c1 , c2 , σ are all
t→t f
globally finite-time stable. bounded, hence there exists a time instant t1 satisfying t0 <
(ξ )
In addition, noting that ∂V∂ξ>
(Aξ + Bv) ∈ H2κ
∆r and V (ξ ) ∈ t1 < t f , such that c1 ` − c2 σ ≥ 1, t ∈ [t1 ,t f ]. This relation could
2γ
H∆r , by applying Lemma 2.1, there exists a constant c1 ∈ ℜ+ , lead to a fact that V̇ (ξ ) ≤ −kξ k2κ ∆r , t ∈ [t1 ,t f ].
such that the following relation holds for τ ∈ (−ς , 0) Therefore, with the update law (6), it is straightforward that
Z tf Z tf
+∞ = `(t f ) − `(t1 ) = ˙
`(s)ds = µkξ (s)k2κ
∆r ds
∂V (ξ ) t1 t1
(Aξ + Bv) ≤ −c1 kξ k2κ
∆r . (11)
∂ξ> ≤ µV (ξ (t1 )) < +∞, (17)
which is clearly a contradiction. Then one could claim that ` 4

is well defined in the time interval [t0 ,t f ]. 2

Secondly, we show that ξ cannot escape at t = t f . Integrat- 0

ing both sides of (16) over the time interval [t0 ,t] yields -2
Z t -4
V (ξ (t)) −V (ξ (t0 )) ≤ − c1 `(s)kξ (s)k2κ
∆r ds -6
t
Z 0t 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
time(sec) a)
+ c2 σ kξ (s)k2κ
∆r ds
t
Z 0t t c σ 3
c1
Z
˙ 2 ˙
≤− `(s)`(s)ds + `(s)ds
2.5
t0 µ t0 µ
c1 c2 σ
≤ − (`2 (t) − 1) + (`(t) − 1). 2
2µ µ
1.5
Hence one could deduce the relation that
c1 2 c2 σ 1
V (ξ (t f )) ≤ V (ξ (t0 )) − (` (t f ) − 1) + (`(t f ) − 1). 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
2µ µ time(sec) b)
From the fact that `(t f ) is bounded, it is obvious that
20
V (ξ (t f )) is also bounded. With Lemma 2.1, it is clear
10
that there exists a constant c4 ∈ ℜ+ such that kξ (t f )k∆r ≤
1 0
c4V (ξ (t f )), which implies that ξ (t f ) is bounded. This fact
2γ
-10

u
clearly contradicts the finite escape time claim. -20
In a conclusion, we have already proved that (ξ > , `)> is -30
well defined and globally uniformly bounded for t ∈ [t0 , ∞). -40
ii) Using the boundness of ξ˙ and ξ for t ∈ [t0 , ∞). it can be 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

straightforwardly deduced from Barbalat’s Lemma that ξ →

time(sec) c)
0, t → ∞. In addition, by recalling from the expression (6) that Fig. 1. Response curves under the proposed controller: a). system
` is monotone increasing and c1 , c2 , σ are bounded, we know states; b). dynamic gain `; c). time history of the input.
there must exist a time instant T1 ≥ t0 such that c1 ` − c2 σ >
2κ
0, ∀t ≥ T1 . Note that with Lemma 2.1, kξ k2κ ∆r ≥ c5V
2κ−τ (ξ )

with c5 being a positive constant. Then with (16) in mind, the quite easy, and by solving (A − BK)> P  + P(A − BK) ≤ −I, 
following relation holds 2.31 1.94 0.50
2κ one can determine a proper matrix P = 1.94 3.25 0.81.
V̇ (ξ ) + c5 (c1 ` − c2 σ )V (ξ ) 2κ−τ ≤ 0, t ≥ T1 . (18) 0.50 0.81 0.44
Based on (18), it can be deduced straightforwardly from Secondly, set κ = 1.2. Calculating the smallest eigenvalue of
Lemma 2.2 that the there exists a finite time T2 ≥ T1 , such Ξ1 P + PΞ1 gives 0.76, then the design parameter ρ can be
that ξ (t) = 0, ∀t ≥ T2 . This completes the proof of Theorem set as 2. Aiming to show impact on the control performance
4.1. from the design parameter µ, we select two different values
Remark 4.2: As implied by the proof, the closed-loop sys- in the simulation, i.e., µ = 2 and 20, respectively. To conduct
tem stability is directly dependent on a sufficiently large the simulation, the initial values are set as x(0) = [−2, 3, 4].
scaling gain ` which is subject to a simple self-tuning update It can be observed from Fig. 1. a) and b) that all the signals
law in (6). Indeed, it can be concluded from the initial in the closed-loop system (8)-(9) are uniformly bounded
value `(t0 ) = 1 that the closed-loop system could possibly be and the trajectories of the states converge to the origin in
unstable at the initial stage. However, ` is made monotone a satisfactory control performance. The time history of the
˙ ≥ 0, which is illustrated in the simulation
increasing as `(t) control signal is shown in Fig. 1. c). Note that a larger µ could
of Example 3.1 later on. After a short period, `(t) can be require a larger control effort, but lead to a faster response.
sufficiently large to render the closed-loop system stable.
VI. C ONCLUSIONS
V. N UMERICAL S IMULATION In this paper, a simple nonrecursive nonsmooth adaptive
Consider Example 3.1 in the case when θ1 = 0.2 sin(t) and stabilizing control design methodology for a class of nonlinear
θ2 = 0.1 are unknown parameters to the designer. systems with the presence of non-parametric uncertainties is
Denoting σ = max{2|θ1 |6/7 , |θ2 |5/6 }, one can verify the proposed. Distinguishable from popular methods including
following relations | f1 (θ , x1 )| = | ln(1 + (θ1 x1 )2 )| ≤ 2|θ1 |6/7 · adaptive backstepping and adding a power integrator, etc., a
|x1 |6/7 ≤ σ |x1 |6/7 , | f2 (θ , x2 )| = | sin(θ2 x2 )| ≤ σ |x2 |5/6 . Thus simplest non-nested controller form with an explicit one-step
system (8) satisfies Assumption 4.1 with τ = −1/7, r1 = gain tuning mechanism is obtained based on homogeneous
1, r2 = 6/7, r3 = 5/7. system theory. As a result, the control implementation could
Before doing simulation, K is firstly selected as in [1, 3, 3] be more promising with guaranteed control performance. The
to render its companion matrix A − BK Hurwitz which is stability analysis is also largely simplified from prior works.
Future works could focus on more general nonlinear systems [23] C. Zhang, J. Yang, and C. Wen, “Global stabilisation for a class of
subject to mismatched disturbances and more general nonlin- uncertain non-linear systems: a novel non-recursive design framework,”
Journal of Control and Decision, vol. 4, no. 2, pp. 57–69, 2017.
ear systems, etc. [24] C. Zhang, J. Yang, Y. Yan, L. Fridman, and S. Li, “Semi-global finite-
time trajectory tracking realization for disturbed nonlinear systems via
higher-order sliding modes,” IEEE Transactions on Automatic Control,
R EFERENCES vol. 60, no. 5, pp. 2185–2191, 2020.
[25] S. Bhat and D. Bernstein, “Geometric homogeneity with applications
[1] H. K. Khalil, Nonlinear Systems. New York: Macmillan Publishing to finite-time stability,” Mathematics of Control, Signals and Systems,
Company, 1992. vol. 17, no. 2, pp. 101–127, 2005.
[2] M. Krstic, I. Kanellakopoulos, and P. V. Kokotovic, Nonlinear and [26] Y. Yan, C. Zhang, A. Narayan, J. Yang, S. Li, and H. Yu, “Generalized
Adaptive Control Design. John Wiley, 1995. dynamic predictive control for nonparametric uncertain systems with
[3] W. Lin and C. Qian, “Adaptive control of nonlinearly parameterized application to series elastic actuators,” IEEE Transactions on Industrial
systems: a nonsmooth feedback framework,” IEEE Transactions on Informatics, vol. 14, no. 11, pp. 4829–4840, 2018.
Automatic Control, vol. 47, no. 5, pp. 757–774, 2002. [27] W. Perruquetti, T. Floquet, and E. Moulay, “Finite-time observers:
[4] Y. Hong, J. Wang, and D. Cheng, “Adaptive finite-time control of Application to secure communication,” IEEE Transactions on Automatic
nonlinear systems with parametric uncertainty,” IEEE Transactions on Control, vol. 53, no. 1, pp. 356 – 360, 2008.
Automatic Control, vol. 51, no. 5, pp. 858–862, 2006.
[5] J. Huang, C. Wen, W. Wang, and Y. Song, “Design of adaptive finite-
time controllers for nonlinear uncertain systems based on given transient
specifications,” Automatica, vol. 69, pp. 395–404, 2016.
[6] J. Yang, Z. Ding, S. Li, and C. Zhang, “Continuous finite-time output
regulation of nonlinear systems with unmatched time-varying distur-
bances,” IEEE Control Systems Letters, vol. 2, no. 1, pp. 97–102, 2018.
[7] P. Krishnamurthy and F. Khorrami, “Dynamic high-gain scaling: state
and output feedback with application to systems with iss appended
dynamics driven by all states,” IEEE Transactions on Automatic Control,
vol. 49, no. 12, pp. 2219–2239, 2004.
[8] L. Praly and Z. Jiang, “Linear output feedback with dynamic high gain
for nonlinear systems,” Systems and Control Letters, vol. 53, pp. 107–
116, Feb. 2004.
[9] H. Lei and W. Lin, “Universal adaptive control of nonlinear systems
with unknown growth rate by output feedback,” Automatica, vol. 42,
no. 10, pp. 1783–1789, 2006.
[10] M. Koo, H. Choi, and J. Lim, “Universal control of nonlinear systems
with unknown nonlinearity and growth rate by adaptive output feed-
back,” Automatica, vol. 47, no. 10, pp. 2211–2217, 2011.
[11] S. P. Bhat and D. S. Bernstein, “Continuous finite-time stabilization of
the translational and rotational double integrators,” IEEE Transactions
on Automatic Control, vol. 43, no. 5, pp. 678–682, 1998.
[12] X. Huang, W. Lin, and B. Yang, “Global finite-time stabilization of
a class of uncertain nonlinear systems,” Automatica, vol. 41, no. 5,
pp. 881–888, 2005.
[13] J. Polendo and C. Qian, “A generalized homogeneous approach for
global stabilization of inherently nonlinear systems via output feedback,”
International Journal of Robust and Nonlinear Control, vol. 17, no. 7,
pp. 605–629, 2007.
[14] X. Zhang, G. Feng, and Y. Sun, “Finite-time stabilization by state feed-
back control for a class of time-varying nonlinear systems,” Automatica,
vol. 48, no. 3, pp. 499–504, 2012.
[15] Z. Sun, L. Xue, and K. Zhang, “A new approach to finite-time adaptive
stabilization of high-order uncertain nonlinear system,” Automatica,
vol. 58, pp. 60–66, 2015.
[16] Z. Sun, T. Li, and S. Yang, “A unified time-varying feedback approach
and its applications in adaptive stabilization of high-order uncertain
nonlinear systems,” Automatica, vol. 70, pp. 249–257, 2016.
[17] Z. Liu and Y. Wu, “Universal strategies to explicit adaptive control
of nonlinear time-delay systems with different structures,” Automatica,
vol. 89, pp. 151–159, 2018.
[18] T. Li, J. Yang, S. Li, C. Wen, and C. Zhang, “Global adaptive finite-time
stabilization of uncertain time-varying p-normal nonlinear systems with-
out homogeneous growth nonlinearity restriction,” IEEE Transactions on
Automatic Control, vol. 64, no. 11, pp. 4637–4644, 2019.
[19] S. Zhou and Y. Song, “Neuroadaptive control design for pure-feedback
nonlinear systems: A one-step design approach,” IEEE Transactions on
Neural Networks and Learning Systems, vol. 31, no. 9, pp. 3389–3399,
2019.
[20] X. Huang, Y. Song, and C. Wen, “Output feedback control for con-
strained pure-feedback systems: A non-recursive and transformational
observer based approach,” Automatica, vol. 113, p. 108789, 2020.
[21] J. Gao, L. Wang, R. Gao, and J. Huang, “Adaptive control of uncertain
underactuated cranes with a non-recursive control scheme,” Journal of
the Franklin Institute, vol. 356, no. 18, pp. 11305–11317, 2019.
[22] C. Zhang and C. Wen, “A non-recursive c1 adaptive stabilization
methodology for nonlinearly parameterized uncertain nonlinear sys-
tems,” Journal of the Franklin Institute, vol. 355, no. 12, pp. 5099–5113,
2018.