A Novel Delay-Independent Robust Adaptive Controller Design For Uncertain Nonlinear Systems With Time-Varying State Delay

A Novel Delay-Independent
Robust Adaptive Controller

Design for Uncertain Nonlinear
Downloaded from https://asmedigitalcollection.asme.org/computationalnonlinear/article-pdf/15/1/011002/6442109/cnd_015_01_011002.pdf by Deenbandhu Chhotu Ram University of Science Tech user on 17 April 2020
Hadi Azmi
Electrical Engineering Department,
Shahid Beheshti University,
Systems With Time-Varying
Tehran 1719-16765, Iran
e-mail: h_azmi@sbu.ac.ir State Delay
Alireza Yazdizadeh1 In this paper, two novel adaptive control strategies are presented based on the linear
Electrical Engineering Department, matrix inequality for nonlinear Lipschitz systems. The proposed approaches are devel-
Shahid Beheshti University, oped by creatively using Krasovskii stability theory to compensate parametric uncer-
Tehran 1719-16765, Iran tainty, unknown time-varying internal delay, and bounded matched or mismatched
e-mail: a_yazdizadeh@sbu.ac.ir disturbance effects in closed-loop system of nonlinear systems. The online adaptive tun-
ing controllers are designed such that reference input tracking and asymptotic stability
of the closed-loop system are guaranteed. A novel structural algorithm is developed
based on linear matrix inequality (LMI) and boundaries of the system delay or uncer-
tainty. The capabilities of the proposed tracking and regulation methods are verified by
simulation of three physical uncertain nonlinear system with real practical parameters
subject to internal or state time delay and disturbance. [DOI: 10.1115/1.4045238]
1 Introduction in Ref. [15]. There is also a need for time-varying analysis in the
systems where the parameters of the nonlinear systems are
There are many delayed systems in various industries and
unknown and time-varying, though the structure of the system is
mathematical systems. One may refer to manufacturing, telecom-
known in real-time operating conditions. Namely a nonlinear
munication, transportation systems, and interactions between
model of an aircraft wing rock system and a machine model with
interconnected dynamics through the system energy and material,
friction [16,17]. The stability criteria for time delay systems have
and flow of its communication [1,2]. Analyzing of the stability of
been proposed to derive in Refs. [7,8], and [18].
delayed nonlinear systems, especially where the delay is time-
The time-delay systems’ stability criteria mostly are formulated
varying, to design a capable controller, have received significant
in the LMI form and be solved by MATLAB/YALMIP software
attention in recent years [3,4]. The signal or static image process-
[19–21]. The controller design and stability checking approaches
ing systems, pattern recognition approaches, and combined opti-
are mainly classified into two main categories for time delay sys-
mization plant are the most famous systems that include these
tem, which are related on the data of exist delays, (a) delay-
difficulties. Without considering the complexity of controller
independent approaches for stabilizing controller design criteria
design in the presence of the time delay in nonlinear systems
and stability checking as discussed in Refs. [21] and [22] and (b)
dynamic inherently and the unknown uncertain parameters, the
delay-dependent approaches for stabilizing controller design dis-
great challenge is to apply the recently published methods where
cussed in Refs. [23] and [24]. Most of the two last decades pub-
uncertainty and delay are associated with the states or inputs of
lished works do not consider the amount of the system delay;
the system [5,6]. Several of practical time delay systems are listed
however, recent papers use the boundaries of delays [25]. Gener-
below, and three of them are explored in this paper to evaluate the
ally, achieved delay-dependent criteria from the papers have less
merits and capabilities of the presented approach in practice.
conservative compared with independent methods, especially if
Wang and his colleagues proposed a controller for the linear
the size of delays is not noticeable. Therefore, the dependent sta-
model of real inverted nonlinear pendulum system with delay in
bility methods for the delay of nonlinear practical systems with
Ref. [7]. The novel work in Ref. [8] is developed about the fault
input or state (internal) delays are highly under focus [17–26].
tolerant control (FTC) method for nonlinear systems with delay
Most of the authors in the last decade developed stability crite-
and by simulating in software on pendulum system and nonlinear
ria for systems with unknown uncertainty and varying delay and
reactor recycle systems its capabilities is shown. Yuhong et al.
nonlinear perturbation based on Krasovskii stability by using
developed a neural network-based controller for the helicopter
LMIs [16,21,24,27–29]. The uncertainties origins may be time-
system [9]. Other Similar works were presented on power system
varying either directly or indirectly; therefore, it is possible to
in Ref. [10], the ratio control of air to fuel in delayed nonlinear
model uncertainties as the time-varying function [30,31]. A list of
system in Ref. [11], slow and delayed relaxation systems with
the main sources of uncertainty that have been discussed and
varying delay in system dynamic in Ref. [12], cutting machines
explained in Refs. [30] and [31] is given below and this paper
with variable spindle speed in Ref. [13], conversion systems of
presents what factors can lead to a time-varying uncertainty and
wind energy plants in Ref. [14], and delayed truck system trailer
how delayed practical system can be modeled with time-varying
uncertainties:
1
Corresponding author. (1) uncertainty due to the nonideal repeatability of measure-
Contributed by the Design Engineering Division of ASME for publication in the
JOURNAL OF COMPUTATIONAL AND NONLINEAR DYNAMICS. Manuscript received March 4,
ment (often called repeatability uncertainty),
2019; final manuscript received October 6, 2019; published online November 11, (2) uncertainty due to calibration of the system equipment
2019. Assoc. Editor: Zaihua Wang. (often called calibration uncertainty),
Journal of Computational and Nonlinear Dynamics JANUARY 2020, Vol. 15 / 011002-1

C 2020 by ASME
Copyright V
(3) uncertainty due to the temperature effect (often called as inputs and unmodeled dynamics. Comparing the proposed work,
temperature uncertainty), and they did not consider existence of time delay on system dynamic,
(4) application-specific uncertainty sources (often called as delay, and disturbance effects on closed-loop system performance
application uncertainty). and stability. Chang Huaa et al. developed a dynamic output feed-
back for nonlinear systems with time delays and uncertainties in
The works [16,21,24,27–29] have considered time-delay func- Ref. [39]. They also considered fuzzy adaptive output-dependent
tion with two main conditions such that the time-delay function feedback tracking control approach for uncertain nonlinear
should be bounded and differentiable and any of the mentioned switched systems with time-varying delay and unmodeled internal
works did not develop stabilizing controller method while this is dynamics in Ref. [40].
not similar to the our work. Adaptive stabilization method for sys- As one of the main references for robust controller design meth-
tems with nonlinear dynamic and fixed delays has been considered ods for nonlinear systems with time delay, we could refer to
in Ref. [18]. However, the stability for uncertain dynamic nonlin- Ref. [41]. In this book, the authors have considered most of the
ear systems with time-varying delays, which is more general than effective methods for controller design for nonlinear practical and
time-invariant delays and fixed dynamic nonlinear systems, has mostly mathematical systems with time delay. In contrast to the
not been considered in these works. proposed work, the published book did not develop any method to
A control approach, which is dependent on delay for the compensate fault effects on nonlinear systems. Chang Chun Hua
norm-bounded nonlinear systems with varying time delays, is con- et al. have investigated distributed output tracking consensus con-
sidered in Refs. [32] and [33]. Jiqing et al. developed a passive trol methods and adaptive fuzzy tracking control for high-order
controller for nonlinear systems with varying delay and they con- stochastic nonlinear multi-agent systems in Refs. [42] and [43],
sidered two numerical examples to show the merit of the method. respectively. Both of the papers considered output-based control
The performance of closed-loop system was not discussed in the methods to attenuate disturbance effects and track suitable input
presence of any type of square and sin disturbances to attenuate references. However, comparing with proposed work, the authors
disturbances effects besides providing a suitable control signal. did not consider time delay in system modeling and its effects on
Comparing Ref. [32], the proposed work considers the large system performance and closed-loop stability.
domain of state and input-dependent nonlinear systems and Chen et al. presented stabilization controller for linear time-
presents two adaptive controller design methods to guarantee delay systems with uncertainty and delay periods [44]. In
asymptotic stability, track the reference signal, and compensate Ref. [45], a fuzzy sliding-mode control method with adaptive tun-
any type of disturbances with an acceptable control signal and ing controller parameters for unknown nonlinear time-delay sys-
presents the merit of methods on three real practical systems. tems with input saturation has been proposed. The authors have
Another main difference between our work and the work of Jiqing considered fixed time delay for a special class of nominal nonlin-
et al. [32] is related to controller design structure. The proposed ear systems rather than time-varying delay and uncertain nonlin-
work presented a constructive algorithm for controller design ear systems, which is proposed in this paper. The papers [46,47]
based on LMI, and by doing according to algorithm steps, we have developed an output-dependent feedback controller to track
could design controllers for three real practical systems. arbitrary reference signals of nonlinear systems with varying input
Mobayen has considered uncertain linear systems rather than delay and bounded disturbances based on the back-stepping
nonlinear systems based on LMI robust tracker with multiple design approach, which still cannot be counted as a general
varying time delays in Ref. [20] and another fixed dynamic (pas- method. Our proposed control method is mainly different from
sive) nonlinear state-dependent feedback stabilizer for nonlinear recent papers. This paper’s innovations compared with recent
systems with Lipschitz nonlinearities and uncertainty has been works are as follows:
discussed in Ref. [19]. Exponentially stabilizing controller for
delayed uncertain nonlinear systems has been presented in (1) We develop a controller design method to compensate
Refs. [22,23], and [34]. However, these published papers assumed unknowns varying delay effects for nonlinear uncertain
controller with fixed dynamic, which could not result the stability systems with and construct the state-dependent controller
of systems with a wide varying or fixed delay range and to ensure global asymptotic stability of the system. Com-
disturbances. paring our paper with Refs. [7,10,13,14,19,27,44,48,49],
The delay-dependent control strategy for large-scale nonlinear the valid stability and result performance of an adaptive
systems with highly interconnection has been investigated in method are proved in Theorem 1.
Ref. [35]. Xiang et al. and Jia have developed the controller for (2) We proved adaption mechanism to tune controller parame-
stabilization of nonlinear systems with delay and uncertain param- ters for nonlinear Lipschitz systems in the presence of
eters but both Refs. [35,36] have considered only special form of bounded disturbance, which can be matched and
nonlinear systems with a more complex approach than the pro- unmatched too. These facts differ this work from other sim-
posed work. Song and Shuping have proposed the passive control- ilar works in the literature in the sense that they have only
ler for nonlinear uncertain systems with fixed time delays in proposed online tuning functions with more complex con-
Ref. [37] rather than varying time delays and uncertainties, which troller parameters than mine only for linear systems with
are considered in this paper. Comparing Ref. [37], the proposed disturbance [1,8,10,18,22,36].
work considers large domain of state and input-dependent nonlin- (3) A simple structural algorithm based on LMI by using
ear systems with varying time delay and develops two adaptive Krasovskii’s theory is presented to tune controller gain
controller methods to guarantee stability of closed-loop systems online, to stabilize asymptotically closed loop, and to
asymptotically, track the arbitrary reference signal, and compen- attenuate disturbance and uncertainty effects in this paper,
sate disturbances with an acceptable control signal. The proposed which is different from Refs. [1,5,7,12,14,33], and [50].
approach was applied on three real practical systems too. Another
main difference between our work and that of Song and He [37] is We organized the sections of this paper as follows: System
related to controller design structure. The proposed work pre- description and formulation of the problem for nonlinear system
sented a simple step forward-based constructive algorithm for are presented in Sec. 2. In Sec. 3, we developed adaptive regula-
controller design based on LMI. tion feedback controller. An algorithm is illustrated based on LMI
Chang Chun Hua’s papers solved most of the challenging prob- for an adaptive mechanic design in Sec. 4. In Sec. 5, the adaptive
lem related to time delay nonlinear practical and mathematical state feedback controller is extended to track arbitrary reference
systems. Based on the recursive method, the finite time decentral- states. The simulation on practical real systems with varying delay
ized control is developed for nonlinear time-varying intercon- and subject to the uncertainty and disturbances, the merits of the
nected systems in Ref. [38], where the subsystems have dead-zone presented approaches are demonstrated in Sec. 6.
011002-2 / Vol. 15, JANUARY 2020 Transactions of the ASME

2 System Description and Problem Formulation B2 ¼ B1 F (7)
In this section, we introduce a model for nonlinear system with
uncertainty and time-varying delay in system states. Consider a Equation (7) defines the matching condition for system disturban-
model of nonlinear Lipschitz system, which is modeled with the ces and physically means from practical point of view that the
following equation in state-space form: control and disturbance signals use identical channels. In this
paper, The following control problem is considered: Design a
_ ¼ ðA1 þ 䉭A1 ÞxðtÞ þ ðA2 þ 䉭A2 Þxðt dðtÞÞ
xðtÞ delay-independent adaptive controller u(t) for system (5) with
Assumptions 1–3 such that the solution of the closed-loop system
þ B1 uðtÞ þ B2 wðtÞ þ f ðxðtÞ; xðt dðtÞÞ; uðtÞ; tÞ (1)
states converges to zero in the presence of matched or mismatched
disturbances. 䊏
where xðtÞ 2 Rn and xðt dðtÞÞ 2 Rn are the state vector and the
delayed state vector, u 2 Rm is the control input vector and
wðtÞ 2 Rq is a the bounded disturbances, which are the continu- 3 Robust Adaptive State Feedback Controller
ous vector function. The function f ðxðtÞ; xðt dðtÞÞ; uðtÞ; tÞ of
the nonlinear dynamics of the model that is dependent on state, We develop an adaptive mechanism to tune the controller
delayed state, and control input vectors. The matrices A, B1, and parameters in this section while uncertainty and delay are
B2 are known real constants of system parameters with appropri- unknown. The direct adaptive controllers of nonlinear system
ate dimensions. Without loss of generality, it is assumed that the guarantee asymptomatic stability of closed-loop system in
matrix B1 is full column rank. d(t) indicates the varying and Theorem 1. Consider the following equation:
unknown time delay, which satisfies 0 dðtÞ d1 and
_ d2 , and it is assumed that d1 and d2 are known con- uðtÞ ¼ k1 xðtÞ þ k2 ðxðtÞ; tÞ (8)
0 dðtÞ
stant scalars. The uncertainties of system parameters 䉭A1 ðtÞ and
where k1 2 Rmn and k2 ðxðtÞ; tÞ 2 Rm are fixed and time-varying
䉭A2 ðtÞ are unknown real-valued functions, which are assumed
matrix gains, respectively, and c is any positive constant, which
to have the following form:
would be designed later.
½䉭A1 ðtÞ; 䉭A2 ðtÞ ¼ DHðtÞ½E1 ; E2 (2) T
ðxT PB1 Þ k^3 ðtÞ dk^3 ðtÞ
k2 ðtÞ ¼ ; 0 < a < 1; ¼ cjjxT PB1 jj
where D, E1, and E2 are assumed known constants and H(t) is an jjxT PB1 jja dt
assumed unknown function, which satisfies (9)
HðtÞT HðtÞ I (3) x_ ¼ ðA1 þ DHðtÞE1 ÞxðtÞ þ ðA2 þ DHðtÞE2 Þxd ðtÞ
þ B1 k1 xðtÞ þ B1 k2 ðxðtÞ; tÞ þ B2 wðtÞ þ f ðx; xd Þ (10)
The 䉭A1 ðtÞ and 䉭A2 ðtÞ are modeled based on the robust control
problems and define the matching condition on the uncertainties.
The general definition of uncertainties is such that jj䉭A1 ðtÞjj d1 THEOREM 1. The adaptive controller (8) stabilizes the closed-
and jj䉭A2 ðtÞjj d2 where d1 and d2 are two positive definite con- loop system asymptotically (10) if the positive symmetric matrices
stants. In another method, we have defined Eqs. (2) and (3). The R ¼ RT > 0; P ¼ PT > 0, constant matrix Z, and positive con-
nonlinear system dynamic (1) with Eqs. (2) and (3) is described stant scalars e > 0 and h > 0 exist such that
by 0 1
X1;1 PA2 PD PD P
B X2;2 0 0 0 C
_ ¼ ðA1 þ DHðtÞE1 ÞxðtÞ þ ðA2 þ DHðtÞE2 Þxd ðtÞ
xðtÞ B C
B C
þ B1 uðtÞ þ B2 wðtÞ þ f ðx; xd ; uÞ (4) B eI 0 0 C<0
B C
@ hI 0 A
1
We assumed the following assumptions to design controller and p I
to achieve stability and performance objectives. For easy notation
of the formulas xðt dðtÞÞ and f ðxðtÞ; xðt dðtÞÞ; uðtÞ; tÞ are where X1;1 ¼ Z þ Z T þ PA1 þ AT1 P þ I þ R þ eET1 E1 ; X2;2 ¼
shown by xd ðtÞ and f ðx; xd ; uÞ, respectively. hET2 E2 þ I ð1 d2 ÞR and p ¼ ðL2f þ L2g Þ; then controller
ASSUMPTION 1. The nonlinear part of system (1) should be parameters P and k1 are calculated by
locally Lipschitz about x xðtÞ and xd xðt dðtÞÞ uniformly, in
a region X containing the origin, Lf > 0 and Lg > 0 that is k1 ¼ B†1P1 Z
jjf ðx; xd ; uÞ f ð^
x ; x^d ; uÞjj Lf jjx x^jj þ Lg jjxd x^d jj (5) where B†1 shows Moore–Penrose Pseudo inverse of matrix B1.
Remark 2. The parameters a and c have no effect on LMI feasi-
where the function f ðx; xd ; uÞ depends on control signal u(t) and bility solution, but by changing these parameters as additional
delayed states xðtÞ; xðt dðtÞÞ. It is also considered as the nonlin- degrees-of-freedom, the designer may tune the controller to
ear perturbation affecting the system (see Refs. [51] and [52] for achieve a better performance for the closed-loop systems. In order
more details about Lipschitz systems). to select a proper value for the parameter, either in simulation or
ASSUMPTION 2. The disturbance vector w(t) is piecewise continu- in practical systems, we may define upper and lower boundaries
ous bounded function and it is assumed that there exists the known of control signals and then check for the best value that guarantees

positive scalar w the condition. In other words, we have 0 < a 1 boundary for a
as a scalar but according to the system’s condition and control sig-
jjwjj w
(6) nals’ boundaries, its proper and exact value is defined. 䊏
Remark 3. The proposed LMI can be solved using SCILAB and
ASSUMPTION 3. The couple of (A1, B1) is controllable. MATLAB. The computational complexity of the theorem is to find
Remark 1. In comparison with most of the recent works, we the feasible response for the LMI by MATLAB/YALMIP solver. It is
proposed adaptive controllers, which are appropriate for worth mentioning that the scalars e > 0; f > 0; h > 0; p > 0 and
mismatched disturbances but it can be extended to the matched matrix R ¼ RT > 0 are positive and calculated directly by the
disturbances as well. For system (4), the fixed matrix F with LMI too. The parameters a and c can be described as controller
appropriate dimensions should exist such that design degrees-of-freedom, which are selected by trial and error

according to the controller design objectives to provide a more 2xT Pf ðx; xd Þ L2f xT PPx þ xT x þ L2g xT PPx þ xTd xd
efficient and acceptable performance and time response. 䊏
Proof of Theorem 1: Closed-loop system, described by ðL2f þ L2g ÞxT PPx þ xT x þ xTd xd Þ
Eq. (10), and a Lyapunov–Krasovskii function candidate are
defined with We assume that p ¼ ðL2f þ L2g Þ; then using Eq. (16), it is shown
ðt
2
Vðx; k~3 Þ ¼ xT Px þ c1 k~3 þ xT ðsÞRxðsÞds (11) V_ xT ½ðA1 þ B1 k1 ÞT P þ PðA1 þ B1 k1 Þ þ e1 PDDT P
tdðtÞ
þ eET1 E1 þ h1 PDDT P þ pP2 þ I þ Rx þ xT PA2 xd
where k~3 in Eq. (11) is the error dynamics and the k3 is defined
known constant þ xTd AT2 Px þ xTd ½hET2 E2 þ I ð1 d2 ÞRxd þ 2c1 k~3 k~_ 3
k~3 ðtÞ ¼ k^3 ðtÞ k3 (12) 2jjxT PB1 jjk^3 þ 2jjxT PB2 jjw
(17)
According to Eqs. (9) and (10), derivation of the Lyapunov func- For matched form of disturbances, according to Eq. (7)
tion results
V_ xT ½ðA1 þ B1 k1 ÞT P þ PðA1 þ B1 k1 Þ þ e1 PDDT P
T
V_ ¼ x ½ðA1 þ B1 k1 Þ P þ PðA1 þ B1 k1 Þx þ 2x PDHE1 x
T T
þ eET1 E1 þ h1 PDDT P þ pP2 þ I þ Rx þ xT PA2 xd
T T
þ 2x PDHE2 xd þ x PA2 xd þ xTd AT2 Px T
þ 2x PB1 k2 ðtÞ
þ xTd AT2 Px þ xTd ½hET2 E2 þ I ð1 d2 ÞRxd þ 2c1 k~3 k~_ 3
T T
þ 2x PB2 wðtÞ þ 2x Pf ðx; xd Þ þ 2c1 k~3 k~_ 3
2jjxT PB1 jjk^3 þ 2jjxT PB1 Fjjw
(18)
_
þ xT Rx ð1 dðtÞÞx T
d Rxd (13)
then using norm properties
Multiplying jjxT PB1 jj by the numerator and denominator in
Eq. (9), it is obtained that V_ xT ½ðA1 þ B1 k1 ÞT P þ PðA1 þ B1 k1 Þ þ e1 PDDT P
ðxT PB1 ÞðxT PB1 ÞT jjxT PB1 jjk^3 þ eET1 E1 þ h1 PDDT P þ pP2 þ I þ Rx þ xT PA2 xd
T
2x PB1 k2 ðtÞ ¼ 2
jjxT PB1 jj2 a þ xTd AT2 Px þ xTd ½hET2 E2 þ I ð1 d2 ÞRxd þ 2c1 k~3 k~_ 3
jjxT PB1 jj2 jjxT PB1 jjk^3 2jjxT PB1 jjk^3 þ 2jjxT PB1 jjjFjjw
(19)
¼ 2
jjxT PB1 jj2 a
jjxT PB1 jjk^3 Assuming that
¼ 2 2jjxT PB1 jjk^3 (14)
a
2c1 k~3 k~_ 3 2jjxT PB1 jjk^3 ðtÞ þ 2jjxT PB1 jjjFjjw
0;
_ 1 then
For any scalar value for d2, we obtain 1 d2 1 dðtÞ
_
1 ð1 dðtÞÞ d2 1 and with considering Eq. (14), it is ) c1 ðk^3 k3 Þk~_ 3 ðjjxT PB1 jjÞðk^3 ðtÞ jFjjwÞ

easy to show that
and c1 k~_ 3 ¼ jjxT PB1 jj, then constant k3 satisfies Eq. (20) for
V_ xT ½ðA1 þ B1 k1 ÞT P þ PðA1 þ B1 k1 Þx þ 2xT PDHE1 x matched disturbances
þ 2xT PDHE2 xd þ xT PA2 xd þ xT AT Px 2jjxT PB1 jjk^3 ðtÞ
d 2 k3 jjFjjw
(20)
T T
þ 2x PB2 wðtÞ þ 2x Pf ðx; xd Þ þ 2c1 k~3 k~_ 3 For mismatched disturbances
þ xT Rx ð1 d2 ÞxTd Rxd (15)
2c1 k~3 k~_ 3 2jjx PB1 jjk^3 ðtÞ þ 2jjx PB2 jjw
T T
0;
where the parts xT PDHE1 x; xT PDHE2 xd ; xT PB2 wðtÞ; xT Pf ðx; xd Þ
) c1 k~3 k~_ 3 jjxT PB1 jjk^3 ðtÞ jjxT PB1 B1 B2 jjw
†
are scalar; then for any positive constants e and h, we have
2xT PDHE1 x xT ½e1 PDDT P þ eET1 E1 x;

†
where B1 shows Moore–Penrose Pseudo inverse of matrix B1
2xT PDHE2 xd h1 xT PDDT Px þ hxTd ET2 E2 xd
(16) c1 ðk^3 k3 Þk~_ 3 jjxT PB1 jjðk^3 ðtÞ jjB1 B2 jjwÞ;
†
2xT PB2 wðtÞ 2jjxT PB2 jjjjwðtÞjj 2jjxT PB2 jjw;

†
2xT Pf ðx; xd Þ ðL2f þ L2g ÞxT PPx þ xT x þ xTd xd ) k3 jjB1 B2 jjw

(21)
To prove 2xT Pf ðx; xd Þ ðL2f þ L2g ÞxT PPx þ xT x þ xTd xd , it is More detailed description is explained in Ref. [53]. Considering
shown Eq. (20), the adaptive law is verified
2xT Pf ðx; xd Þ 2jjxT Pf ðx; xd Þjj d k^3

¼ cjjxT PB1 jj (22)
2jjxT Pjjjjf ðx; xd Þjj dt
䊏
2jjxT PjjðLf jjxðtÞjj þ Lg jjxd ðtÞjjÞ Remark 4. The constant k3 should be assumed such that it satis-
B2, and B1 are
fies Eqs. (21) and (20) inequalities. It means that w,
Now, from the basic mathematics, it can be seen that for real num- known and B1 is full column rank, then there always exists a con-
bers a and b, the inequality holds 2ab a2 þ b2 . Using this constant k3 to satisfy Eqs. (20) and (21). Hence, the Lyapunov func-
cept, one gets the following relation: tion derivation should be

T ð t T
V_ xT ½ðA1 þ B1 k1 ÞT P þ PðA1 þ B1 k1 Þ þ e1 PDDT P M xT ðtÞPB1 jjx ðsÞPB1 jjds
0
þ eET1 E1 þ h1 PDDT P þ pP2 þ I þ Rx þ xT PA2 xd lim k2 ðtÞ ¼ lim
t!1 t!1 jjxT ðtÞPB1 jj
þ xTd AT2 Px þ xTd ½hET2 E2 þ I ð1 d2 ÞRxd (23)
where M ¼ ðc=aÞ. According to the norms, Lemma [51].
Hence, we have Proof.
ðt
T
dV T T R11 PA2 x xT ðtÞPB1 jjxT ðsÞPB1 jj jjxT ðsÞPB1 jjds
x xd (24)
dt AT2 P R22 xd
ð t 0

where

x T
ðsÞPB ds
1
0
ð
T t T
R11 ¼ ðA1 þ B1 k1 ÞT P þ PðA1 þ B1 k1 Þ þ e1 PDDT P M xT ðsÞPB1 jjx ðsÞPB1 jjds
0
cte lim
þ eET1 E1 þ h1 PDDT P þ pP2 þ I þ R t!1 T
jjx ðsÞPB1 jj

ð t
R22 ¼ hET2 E2 þ I ð1 d2 ÞR MjjxT ðsÞPB1 jj

x T
ðs ÞPB ds
0
lim
The Lyapunov derivation should be negative then t!1 jjxT ðsÞPB1 jj

ð t

R11 PA2 x lim M

x T
ðs ÞPB ds
1
cte
Q¼ T < 0; V_ ½xT xTd Q <0 (25) t!1 0
A2 P R22 xd
So, it is proved that limt!1 k2 ðtÞ ¼ cte. 䊏
We apply the Schur complement Lemma in Eq. (25) and the LMI
form of Eq. (25) is
0 1 4 Controller Design Algorithm
X1;1 PA2 PD PD P
B C The following algorithm summarizes robust adaptive controller
B X2;2 0 0 0 C
B C design:
B eI 0 C
B 0 C<0 (26) Step 1. Lipschitz constants Lf, Lg should be computed for the
B C
@ hI 0 A nonlinear part of model f ðxðtÞ; xd ðtÞ; uðtÞ; tÞ by using Eq. (5) and
p1 I Lipschitz Lemma.
Step 2. Assumptions (1)–(3) should be satisfied.
where X1;1 ¼ Z þ Z T þ PA1 þ AT1 P þ I þ R þ eET1 E1 ; X2;2 ¼ Step 3. After feasibility solution of LMI (26) by MATLAB-YALMIP
hET2 E2 þ I ð1 d2 ÞR and p ¼ ðL2f þ L2g Þ, then the controller or SCILAB software P and Z are computed and then the controller
parameters P and k1 are calculated by parameters P; k1 are calculated by using Eq. (26). For the feasible
solution of LMI (26), it is not necessary to assume e > 0; h > 0
† and any positive matrix R because these parameters are computed
k1 ¼ B1 P1 Z (27)
by LMI (26) directly.
† Step 4. By assuming c as a positive constant, the gain k^3 ðtÞ
where B1 shows Moore–Penrose Pseudo inverse of matrix B1. should be computed using Eq. (9).
Considering Eq. (25), we have Step 5. By assuming a as a positive constant, the gain k2 ðtÞ
should be computed using Eq. (9)
dV x; k~3 Step 6. Compute adaptive controller with Eq. (8).
0 Step 7. Simulate closed-loop system with Eq. (8), if the time
dt
responses are not acceptable, go back to steps 4 and 5.
Hence, the adaptive controller design problem is solvable and
closed-loop system is asymptotically stable and x(t) converges to
zero. 䊏 5 Tracking Robust Adaptive State Feedback
Notation. There are two main cases where values of all states Controller
can be zero, namely, for t ¼0 as the initial conditions and when In this section, we consider the tracking property of the
t ! 1 as the steady-state values. When t ¼0, it is necessary that proposed method. Consider a nonlinear Lipschitz system with the
the system has a nonzero initial condition, where the system starts following equation:
to work and the obscure mode for the controller would not be cre-
ated. In order to take into account the limitations in practice, and z_ ¼ ðA1 þ 䉭A1 Þðz þ xref Þ þ ðA2 þ 䉭A2 Þðzd þ xref Þ
for implementation of this method in real applications, a very
small positive constant like g is added to the denominator of k2 ðtÞ, þ B1 u þ B2 wðtÞ þ f ðz; zd ; uÞ (29)
which has no effect on the value of k2 as below:
where xref 2 Rn is the reference input, zðtÞ ¼ xðtÞ xref and
T ð t zd ðtÞ ¼ zðt dðtÞÞ ¼ xðt dðtÞÞ xref are the state and delayed
M xT ðtÞPB1 jjxT ðsÞPB1 jjds state vectors, respectively. Then the dynamics of system is
k2 ðtÞ ¼ 0
(28) described by
jjxT ðtÞPB1 jj þ g
_ ¼ ðA1 þ DHðtÞE1 ÞzðtÞ þ ðA2 þ DHðtÞE2 Þzðt dðtÞÞ
zðtÞ
When t ! 1, it is proved analytically in the Theorem 1 and guar- þ B1 uðtÞ þ B2 wðtÞ þ Gðz; zd ; uÞ (30)
anteed that the closed-loop system is asymptotically stable and all
states converge asymptotically to zero. It is also proved that where Gðz; zd ; uÞ ¼ f ðz; zd ; uÞ þ ðA1 þ A2 þ 䉭A1 ðtÞ þ 䉭A2 ðtÞÞxref .
limt!1 k2 ðtÞ ! cte analytically according to the Theorem 1 The nonlinear part of system (30) should be locally Lipschitz

about z(t) and zðt dðtÞÞ uniformly, in a region X containing the According to Eqs. (2) and (3), we have jjðA1 þ A2 þ 䉭A1 ðtÞ
origin, Lf > 0 and Lg > 0 that is þ䉭A2 ðtÞÞxref jj jjðA1 þ A2 þ 2EDÞxref jj. Then we assume that
R ¼ jjðA1 þ A2 þ 2EDÞxref jj and then
jjGðx; xd ; uÞ Gð^
x ; x^d ; uÞjj Lf jjz ^z jj þ Lg jjzd ^z d jj (31)
2zT PGðz; zd Þ 2jjzT PjjR þ 2jjzT Pjjjjf ðz; zd ; uÞjj
we developed an adaptive control law to update the parameters
zT P2 Rz þ 2jjzT PjjðLf jjzðtÞjj þ Lg jjzd ðtÞjjÞ
where uncertainty and varying time delay are unknown. The con-
troller design for the nonlinear system is such that guarantee zT P2 Rz þ ðL2f þ L2g ÞzT PPz þ zT z þ zTd zd
closed-loop system stability is presented in Theorem 2. Consider
Eq. (29) then the last LMI form will be
0 1
uðtÞ ¼ k1 zðtÞ þ k2 ðzðtÞ; tÞ (32) X1;1 PA2 PD PD P P
B C
where k1 2 Rmn and k2 ðzðtÞ; tÞ 2 Rm are fixed and the time- B X2;2 0 0 0 0 C
B C
varying matrix gains, respectively, and c is any positive constant B eI 0 0 0 C
B C
that would be designed later. Where B C<0
B hI 0 0 C
B C
B p1 I 0 C
T
ðzT PB1 Þ k^3 ðtÞ dk^3 ðtÞ @ A
k2 ðtÞ ¼ T
; 0 < a < 1; ¼ cjjzT PB1 jj R1 I
jjz PB1 jja dt
(33)
where X1;1 ¼ Z þ ZT þ PA1 þ AT1 P þ I þ R þ eET1 E1 ; X2;2 ¼ hET2 E2
The closed-loop system is þI ð1 d2 ÞR and p ¼ ðL2f þ L2g Þ. Thus, the adaptive controller
design problem is solved. Then closed-loop system is asymptoti-
_ ¼ ðA1 þ DHðtÞE1 ÞzðtÞ þ ðA2 þ DHðtÞE2 Þzd ðtÞ
zðtÞ cally stable and the state z(t) converges to input reference signals.
It means that x(t) track xref asymptotically. The merit of the pro-
þ B1 k1 xðtÞ þ B1 k2 ðzðtÞ; tÞ þ B2 wðtÞ þ f ðz; zd Þ (34) posed Theorem 2 in this paper is shown in Example 2. 䊏
THEOREM 2. The controller (32) stabilizes asymptotically the

closed-loop system (34) if the positive symmetric matrices R ¼ 6 Numerical Examples
RT > 0; P ¼ PT > 0 and constant matrix Z and positive constant In this section, we simulate three uncertain nonlinear Lipschitz
scalars e > 0; f > 0; h > 0 and p > 0 exist such that practical systems to show the capabilities of the proposed control-
0 1 lers. We follow the steps of algorithm, which is presented in
X1;1 PA2 PD PD P P Sec. 4 and P and k1 matrices are computed by MATLAB. Simulation
B C results are shown in Figs. 1–4, respectively.
B X2;2 0 0 0 0 C
B C
B eI 0 0 0 C
B C 6.1 Example 1: Inverted-Pendulum System. The inverted-
B C<0
B hI 0 0 C pendulum system [7] is considered. Denote that x1 ¼ x and
B C
B p I1
0 C x2 ¼ x_ 1 , then the motion equation is given by
@ A
R1 I
x_ 1 ðtÞ
where X1;1 ¼ Z þ Z T
þ PA1 þ AT1 P þ I þ R þ eET1 E1 ; X2;2 ¼ hET2 E2 ¼ x2 ðtÞ þ 0:08ðm þ MÞx2 ðt d ðtÞÞ
þI ð1 d2 ÞR and p ¼ ðL2f þ L2g Þ, then controller parameters P x_ 2 ðtÞ
and k1 is calculated by
g sinðx1 Þ 0:5amlx22 sinð2x1 Þ a cosðx1 Þu þ 1:2amlx2 ðt dðtÞÞ
¼
† 4
k1 ¼ B1 P1 Z (35) l aml cos2 ðx1 Þ
3
Proof of Theorem 2: The Lyapunov–Krasovskii function candi-
date is defined as where x and x_ are the angular displacement (in rad) and the angu-
lar velocity (in rad/s), respectively, g ¼ 9:8m=s2 is the gravity
ðt acceleration, a ¼ ð1=m þ MÞ, m ¼2 kg is the mass of the pendu-
2
Vðz; k~3 Þ ¼ zT Pz þ c1 k~3 þ zT ðsÞRzðsÞds (36) lum, M ¼8 kg is the mass of the cart, 2l ¼ 1 m is the length of the
tdðtÞ
pendulum, and u is the force applied to the cart. In order to com-
ply with the above nonlinear system (1) properties, we obtained
Considering Theorem 1, the proof is the same. There is a bit dif- nonlinear system model, B1 ¼ B2
ference as follows: Since zT PGðz; zd Þ are scalar and considering
Assumptions (1) and (2), it is easy to show that 0 1
0 1 8ðm þ MÞ
0 1 B 0
B C B 100 C C
2zT PGðz; zd Þ zT P2 Rz þ pzT PPz þ zT z þ zTd zd (37) A1 ¼ @ g A A2 ¼ B C
0 @ 1:2am A
4=3l aml 0
4=3 am
To prove 2zT PGðz; zd Þ zT P2 Rz þ ðL2f þ L2g ÞzT PPz þ zT z þ zTd zd , !
it is shown: 0
B1 ¼
0:1765
2zT PGðz; zd Þ
2jjzT PGðz; zd Þjj 2jjzT PjjjjGðz; zd Þjj The length of the pendulum l varies as the load is hoisted and
relatively mass of the pendulum will vary as well. In this case,
2jjz Pjjjjðf ðz; zd ; uÞ þ ðA1 þ A2 þ 䉭A1 þ 䉭A2 Þxref Þjj
T
it is assumed both values of the pendulum mass and center
2jjzT Pjjðjjf ðz; zd ; uÞjj þ jjðA1 þ A2 þ 䉭A1 þ 䉭A2 Þxref jjÞ of gravity are time-varying as m ¼ 2 þ 0:02 cosð50tÞkg and
l ¼ 0:5 þ 0:03 sinð10tÞ, then the matrices A1 and A2 are

Fig. 1 States, control, and disturbance signals of the closed-loop system
Fig. 2 State tracking error, control, and disturbance signals of closed-loop system
time-varying uncertainties. According to the mentioned points, we obtained HðtÞ ¼ 0:1 sinð10tÞ and assumed D ¼ I
! ! !
0 0 0 0:031 1
E1 ¼ E2 ¼ X0 ¼
8:52 0 0 0:015 1
0 1
0
B C
B g sinðx1 Þ 0:5amlx22 sinð2x1 Þ a cosðx1 Þu þ 1:2amlx2 ðt dðtÞÞ C
f ¼B C
@ 4 A
l aml cos2 ðx1 Þ
3

6.2 Example 2: Nonlinear Chemical Reactor Recycle
System. In this example, we consider a nonlinear chemical
delayed reactor recycle system with uncertainty as follows [8].
We apply Theorem 2 in this example to show that system tracks
reference vectors xref.
1 R1B 1
x_ 1 ¼ K1A x1 þ x2 ðxs3 þ xs1 Þ
V1A h1A
R1A 1 2 2C
x_ 2 ¼ K1B x2 þ xs1 x 1B xs2
V1B h1B d2 h1B
Fig. 3 Closed-loop state responses from Ref. [54] R1B X4

F1i
þ xs2 þ f1 ðxs Þ þ ui
V1B V
i¼1 1B
Multidisturbances are assumed and shown in Fig. 1. It is tried to 1 R2B 1

use all types of disturbance signals in this example to show the x_ 3 ¼ K2A x3 þ x4 ðxs1 Þ þ xs3
V2A h2A
merit of the proposed work compared with the recent works in
the literature for attenuating most type of disturbance effects. The R2A 1 2 2C R2B
matrices P; k1 using Eq. (26) and positive scalars e ¼ 3:01; x_ 4 ¼ K2B x4 þ xs3 xs4 2B xs4 þ xs4
f ¼ 2:94; h ¼ 2:99 and p ¼ 200 and positive matrix R are com- V2B h2B h2B V2B
puted R ¼ diagð331:2954½1; 1Þ
X
4
F2i
þ f2 ðxs Þ þ
0:1285 0:002 449:5944 V2B
P¼ k1T ¼ i¼1
0:002 0:1291 112:7205
The simulation results are shown in Fig. 1. These figures guaran- where the compositions C ¼ ½C1A ; C2A ; C1B ; C2B are the system
tee the closed-loop system stability with suitable performance and states; RiA ¼ RiB ¼ 0:5 are rates of flow recycling; hiA ¼ hiB ¼ 2
show that the proposed controller compensates all possible time- are the times for reactor residence; kiA ¼ kiB ¼ 0:5 are the con-
varying delay and uncertainty effects in the presence of bounded stants of reactor; F1i ; F2i are the rates of feed; ViA ¼ ViB ¼ 0:5 are
disturbances and all the states are convergent to zero. The simula- the reactors’ volumes; and HiA, HiB and fi ð:Þ for i ¼1, 2 are
tion results are robust against disturbances and uncertainties and nonlinear part of model, which represent the complexity of
show reasonable time-domain performance as well. The result of system behavior. With HiA ¼ C1A ðt sðtÞÞ þ C2A ðt sðtÞÞ and
the control signal is shown in the Fig. 1. It shows that the pro- HiB ¼ C2iB ðt sðtÞÞ, the reference input signals are shown by
posed method responds well to the time-varying delay and uncer- CiA ¼ 14=9 and CiB ¼ 7=3 and varying time delay is assumed
tainties and attenuates disturbances effect on closed-loop system sðtÞ ¼ 0:5 þ 0:5 sinðtÞ. We assume x1 ¼ C1A C1A ; x2 ¼
responses. The control signal achieves better performance without C1B C1B ; x3 ¼ C2A C2A , and x4 ¼ C2B C2B , the nonlinear
P
chattering, sudden falling, and rising. The proposed method guar- part of model is fi ðxðtÞÞ ¼ 4i¼1 x2i ðtÞ. Totally uncertain nonlinear
antees controller capabilities in the real application. system model parameters are as follows:
Fig. 4 States tracking error, control, and disturbance signals of closed-loop system

0 1 0 1 6.3 Example 3: The Two Inverted Pendulums Benchmark.
0:5 1 0 0 1 0 0 0
B C B C In this example, an uncertain nonlinear system, namely, the two
B 0 0:5 0 0 C B0 1 0 0C inverted pendulums with time-varying delay described in Ref. [8],
A1 ¼ B
B 0
C D¼B C
@ 0 0:5 1 C A
B0
@ 0 1 0CA is considered. We assumed that the pendulums’ positions will be
shown by a torque input ui and the result will be applied by a ser-
0 0 0 0:5 0 0 0 1
0 1 0 1 vomotor. We assumed that hi is measurable for i ¼1, 2th. Let
0:5 0 0:5 0 0:5 x1 ¼ h1 ; x2 ¼ h_ 1 ; x3 ¼ h2 , and x4 ¼ h_ 2 . Thus, the model can be
B C B C described as [8]
B 1 1:3 0 0 C B 1 C
A2 ¼ B
B 0:5
C X0 ¼ B B C
@ 0 0:5 0 C A @ 0:5 A
C
x_ 1 ¼ x2 ðtÞ
0 0 1 1:3 1 2
4m1 gr kr 2 kr xd1
0 1 x_ 2 ¼ G1;2 U1 þ sinðx1 Þ þ sinðx3 Þ þ
0:1069 0:0325 0:0102 0:0865 4J1 4J1 1 x2d1
B C
B 0:0809 0:0755 0:0241 0:0030 C x_ 3 ¼ x4 ðtÞ
E1 ¼ B
B 0:2944 0:1370
C 2
@ 0:0319 0:0165 C A 4m2 gr kr 2 kr xd3
x_ 4 ¼ G2;2 U2 þ sinðx3 Þ þ sinðx1 Þ þ
0:1438 0:1712 0:0313 0:0628 4J2 4J2 1 x2d3
0 1
0
0 1 BX C where G1;2 ¼ ½6; 6; 6; 6; G2;2 ¼ ½5; 5; 5; 5; sðtÞ ¼ 0:4ð1þ
0 0 0 0 B 4 2 C
B C B x ðx 2
=h 1B Þ C sin2 ðtÞÞ and U1 ¼ U2 ¼ ðu1 ; u2 ; u3 ; u4 ÞT . The model parameters
B1 1 1 1 C B i d2 C
B C assumed m1 ¼ 2 kg and m2 ¼ 2:5 kg; J1 ¼ 5 kg and J2 ¼
B1 ¼ B C f ðtÞ ¼ B i¼1
B0 C C
@ 0 0 0A B 0 C 6:25 kg; k ¼ 100 N=m; r ¼ 0:5 m; l ¼ 0:5 m and g ¼ 9:81 m=s2
B 4 C
1 1 1 1 BX C and b ¼ 0:5 m. The selected parameters in simulation, such that
@ x ðx =h1B Þ A
2 2
i d4 the model to be fit to the above nonlinear system with Eq. (1), are
i¼1
as follows: We assumed E1 ¼ E2, B1 ¼ B2, X0 ¼ ½0:1; 0:1;
i;j
0:1; 0:1T , D ¼ I and A2 ¼ fA2;1 4;3
2 ¼ A2 ¼ 1; else A2 ¼ 0; i;
As what we did in Example 1 to show the merits of the presented j ¼ 1; 2; 3; 4g
approach, the disturbance signal assumed a square signal and a 0 1 0 1
sinusoidal signal as shown in Fig. 2. By using MATLAB/YALMIP, the 0 1 0 0 0 0 0 0
matrices P; k1 are calculated and positive scalars e ¼ 1:01; B C B C
B 29:29 0 31:25 0 C B 6 6 6 6 C
f ¼ 2:04; h ¼ 1:1, and p ¼ 1:2 and positive matrix R are computed A1 ¼ B B C B2 ¼ B B C
@ 0 0 0 1CA @0 0 0 0 A
C
0 1 39:06 0 37:10 0 5 5 5 5
20:6401 0:2477 0:0151 0:1571 0 1
B C 1:5442 1:0616 0:1924 1:4224
B 0:2477 20:8316 0:2270 0:3016 C
R¼B B 0:0151
C B C
@ 0:2270 20:5235 0:1986 C A B 0:0859 2:3505 0:8886 0:4882 C
E2 ¼ B B 1:4916 0:6156 0:7648 0:1774 C
C
0:1571 0:3016 0:1986 20:5920 @ A
0 1 0:7423 0:7481 1:4023 0:1961
0:0860 0:0048 0:0064 0:0048 0 1
B C 0
B 0:0048 0:0619 0:0048 0:0006 C
P¼B B 0:0064 0:0048 0:0860 0:0048 C
C B C
B p1 sinðx1 Þ þ p2 sinðx3 Þ þ xd1 =ð1 þ x2d1 Þ C
@ A f ¼B C
B 0 C
0:0048 0:0006 0:0048 0:0619 @ A
0 1 2
p4 sinðx1 Þ þ p3 sinðx3 Þ þ xd3 =ð1 þ xd3 Þ
13:7610 28:5400 274:3345 4:9608
B C
B 13:7610 28:5400 274:3345 4:9608 C
k1 ¼ B
B 13:7610 28:5400 274:3345 4:9608 C
C where p1 ¼ 29:29; p2 ¼ 31:25; p3 ¼ 37:102; P4 ¼ 39:062.
@ A Multi disturbances are assumed and shown in Fig. 2. It is tried to
13:7610 28:5400 274:3345 4:9608 use all type of disturbance signals in this example to show the
merit of the proposed work compared to the recent works for
Figure 2 shows that, besides having varying delays, the tracking is attenuating all types of the disturbance effects. We calculate P; k1
ensured despite the presence of unknown uncertain nonlinearities and positive scalars e ¼ 0:119; f ¼ 2:04; h ¼ 0:12, and p ¼ 0:2
and all the measured outputs and system states converge to the and positive matrix R using MATLAB/YALMIP. P ¼ diag½0:0821;
origin and Fig. 2 shows the boundedness of the control signals. 0:0799; 0:0821; 0:0799
The shown simulation results in Fig. 2 compared with the 0 1
achieved results in Fig. 3 proved that the proposed method com- 1:1934 0:0001 0:0000 0:0002
pensates all varying delay and uncertainty effects and besides B C
B 0:0001 1:1931 0:0002 0:0001 C
acceptable performance in the presence of disturbances. As R¼B B 0:0000 0:0002 1:1932 0:0002 C
C
an important result, all of the tracking errors converged to zero. In @ A
Fig. 3 the system is simulated with a fixed time delay. It con- 0:0002 0:0001 0:0002 1:1938
cluded that our developed controller is robust against varying 0 1
uncertainty and disturbances and with an acceptable performance. 98:5871 387:6700 95:9813 232:7011
B C
Figure 2 shows a reasonable control signal compared with the B 28:7033 191:7777 14:3156 69:5338 C
B
k1 ¼ B C
results of Refs. [8] and [54] shown in Example 3. By streamlining C
the controller dynamic, the simulation has a better control signal @ 98:5871 387:6700 95:9813 232:7011 A
without chattering, sudden falling, and rising. More important 28:7033 191:7777 14:3156 69:5338
point is that the convergence to zero is achieved in a less time and
the application of our method in the real time would result in a Figure 4 shows that at the presence of varying delays, the state
better performance too. regulation is achieved and all the states of system converge to the

origin. Figure 4 shows that the control inputs of system will be [6] Ligang, W., Hak-Keung, L., Yuxin, Z., and Zhan, S., 2015, “Time-Delay Sys-
bounded too. tems and Their Applications in Engineering,” Math. Probl. Eng., 2015, p.
246351.
The proposed methods simulation results of nonlinear system [7] Wang, L., Shao, C., and Liu, X., 2011, “State Feedback Control for Uncertain
are demonstrated in Fig. 4. We compared our results with figures Switch Systems With Interval Time-Varying Delay,” Asian J. Control, 13(6),
in Ref. [8] and it is worth to mention that these figures show that pp. 1035–1042.
the proposed designed approach compensates varying time delay [8] Hashemi, M., 2017, “Adaptive Neural Dynamic Surface Control of MIMO
Nonlinear Time Delay Systems With Time-Varying Actuator Failures,” Int. J.
and unknown uncertainty effects and guarantees the closed-loop Adapt. Control Signal Process, 31(2), pp. 275–296.
system asymptotic stability with acceptable performance and the
[9] Yuhong, C., Xuebo, Y., and Xiaolong, Z., 2018, “Adaptive Neural Control of a
states are convergent to zero. The simulation in Ref. [8] has done 3 DOF Helicopter With Unknown Time Delay,” Neurocomputing, 307, pp.
with a fixed-time delay and the results proved that the proposed 98–105.
[10] Yazdizadeh, A., Mehrafrooz, A., Jouzdani, J., and Barzamini, R., 2009,
controller is robust in the presence of varying disturbances and “Adaptive Neuro-PID Controller Design With Application to Nonlinear Water
uncertainties and the results will be more reasonable in time Level in NEKA Power Plant,” J. Appl. Sci., 9(19), pp. 3513–3521.
domain as well. [11] Han, X., and Stobart, R., 2013, “Sliding Mode Fault Tolerant Control of Uncer-
Figure 4 demonstrates that the proposed method compared with tain Systems With Time Varying Delay: Application to AFR Control,” 52nd
IEEE Conference on Decision and Control, Florence, Italy, Dec. 10–13, pp.
Ref. [8] has a suitable performance and by streamlining the 2846–2851.
dynamic of the controller, the control signal will have more [12] Nadhem, E., 2018, “Observer Design and Practical Stability of Nonlinear Sys-
acceptable performance without chattering, sudden falling, and tems Under Unknown Time-Delay,” e-print arXiv: 1806.03166.
rising. The convergence of states to zero is got in a short time as [13] Wim, M., Vincent, V., and Silviu-Iulian, N., 2005, “Stabilization of Time-
Delay Systems With a Controlled Time-Varying Delay and Applications,”
well. The proposed method guarantees that in the implementation IEEE Trans. Autom. Control, 50(4), pp. 493–504.
of the proposed method in the real plants, we will have a suitable [14] Xin, W., and Alden, M. J., 2017, “Resilient and Robust Control of Time-Delay
performance too. Wind Energy Conversion Systems,” ASCE-ASME J. Risk Uncertainty Eng.
Syst. Part B: Mech. Eng., 3(1), p. 011005.
[15] Bing, C., and Xiaoping, L., 2004, “Reliable Control Design of Fuzzy Dynamic
Systems With Time-Varying Delay,” Fuzzy Sets Syst., 146(3), pp. 349–374.
7 Conclusion [16] Miao, X., and Li, L., 2015, “New Stability Criteria for Uncertain Nonlinear Sto-
chastic Time-Delay Systems Circuits,” Syst. Signal Process., 34(8), pp.
This paper developed two novel LMI-based robust adaptive 2441–2456.
state-dependent controllers, which track reference input signals [17] XIE, X. J., and ZHAO, C. R., 2014, “State Feedback Stabilization of Stochastic
using state feedback control strategies and applicable on the non- Feed Forward Nonlinear Systems With Input Time-Delay,” Acta Autom. Sin.,
40(12), pp. 2972–2976.
linear Lipschitz systems with uncertainty, unknown varying time [18] Tahoun, A. H., 2015, “Adaptive Stabilization of Neutral Systems With Nonlin-
delay, and bounded matched or mismatched disturbance. It is the- ear Perturbations and Mixed Time-Varying Delays,” Int. J. Adapt. Control Sig-
oretically proved that it is not needed to have knowledge about nal Process., 29(10), pp. 1328–1340.
the system’s time delay or uncertainty and a novel structural algo- [19] Mobayen, S., 2015, “An LMI-Based Robust Tracker for Uncertain Linear Sys-
tems With Multiple Time-Varying Delays Using Optimal Composite Nonlinear
rithm is presented based on LMI and Lyapunov–Krasovskii stabil- Feedback Technique,” Nonlinear Dyn., 80(1–2), pp. 917–927.
ity theory. The developed controllers by using an on-line adaptive [20] Mobayen, S., 2016, “Optimal LMI-Based State Feedback Stabilizer for Uncer-
tuning guarantee stability of the closed-loop system asymptoti- tain Nonlinear Systems With time-Varying Uncertainties and Disturbances,”
cally with attenuation of the disturbance and uncertainty effects. Complexity, 21(6), pp. 356–362.
[21] Mobayen, S., 2015, “Design of a Robust Tracker and Disturbance Attenuator
From practical point of views, our proposed controller has a sim- for Uncertain Systems With Time Delays,” Complexity, 21(1), pp. 340–348.
pler dynamic as compared with most of the existing methods. The [22] Sun, Z. Y., Li, T., and Yang, S. H., 2016, “A Unified Time-Varying Feedback
capabilities of the proposed schemes are proved by simulation on Approach and Its Applications in Adaptive Stabilization of High-Order Uncer-
real nonlinear systems. tain Nonlinear Systems,” Automatica, 70, pp. 249–257.
[23] Wang, D., and Wang, W., 2009, “Delay-Dependent Robust Exponential Stabili-
zation for Uncertain Systems With Interval Time-Varying Delays,” J. Control
Theory Appl., 7(3), pp. 257–263.
8 Future Recommendation [24] Liu, P. L., 2014, “Improved Delay-Range-Dependent Robust Stability for
Uncertain Systems With Interval Time-Varying Delay,” ISA Trans., 53(6), pp.
The following items may be considered as the future work of 1731–1738.
this paper: [25] Chen, Q. X., Yu, L., and Zhang, W. A., 2007, “Delay-Dependent Output Feed-
back Guaranteed Cost Control for Uncertain Discrete-Time Systems With Mul-
(1) To generalize the proposed procedure and decrease the con- tiple Time-Varying Delays,” IET Control Theory Appl., 1(1), pp. 97–103.
[26] Shi, X. C., and Zhang, T. P., 2013, “Adaptive Tracking Control of Uncertain
servativeness, the delay-dependent control design method MIMO Nonlinear Systems With Time- Varying Delays and Un-Modeled
may be introduced. As a result, the LMI condition is delay Dynamics,” Int. J. Autom. Comput., 10(3), pp. 194–201.
dependent and the results are less conservative. [27] Farnam, A., and Mahboobi, R., 2016, “Improved Linear Matrix Inequality
(2) It is better to modify the methodology such that the output Approach to Stability Analysis of Linear Systems With Interval Time-Varying
Delays,” J. Comput. Appl. Math., 294, pp. 49–56.
is used instead of states. [28] Zhai, J. V., 2014, “Global Output Feedback Stabilization for a Class of
(3) The authors are also trying to apply the procedure to a Nonlinear Time-Varying Delay Systems,” Appl. Math. Comput., 228, pp.
physical industrial system. 606–614.
[29] Wang, C., and Xu, Y., 2011, “Delay-Dependent Robust Stability of Uncertain
Nonlinear Stochastic Systems With Time-Varying Delay,” Advances in Com-
puter Science, Environment, Ecoinformatics and Education, Springer Berlin,
References pp. 424–429.
[1] Zhai, D., An, L., Li, J., and Zhang, Q., 2016, “Delay-Dependent Adaptive [30] Ellison, S. L., and Williams, A., 2012, Quantifying Uncertainty in Analytical
Dynamic Surface Control for Nonlinear Strict-Feedback Delayed Systems With Measurement, 2nd ed., Eurachem/CITAC.
Unknown Dead Zone,” J. Franklin Inst., 353(2), pp. 279–302. [31] Stewaet, I., 2012, “Sources of Uncertainty in Deterministic Dynamics an Infor-
[2] Zhou, X., and Zhong, R., 2013, “Delay-Probability-Distribution-Dependent Sta- mal Overview,” Philos. Trans. Then R. Soc., 369, pp. 4705–4729.
bility Criteria for Discrete-Time Stochastic Neural Networks With Random [32] Jiqing, Q., Zhifeng, G., Yuanqing, X., and Jinhui, Z., 2007, “Delay-Dependent
Delays,” Adv. Difference Equations, 314, pp. 2–18. Robust Passive Control for a Class of Nonlinear Systems With Time-Varying
[3] Rakkiyappan, R., Lakshmanan, S., and Balasubramaniam, P., 2013, “Delay- Delays,” Optimal Control Appl. Methods, 29(5), pp. 353–371.
Probability-Distribution-Dependent Stability of Uncertain Stochastic Genetic [33] Boren, L., and Guang, Y., 2016, “Robust Stabilization and H Inf Control of
Regulatory Networks With Time-Varying Delays,” Circuits, Syst. Signal Uncertain Stochastic Time-Delay Systems With Nonlinear Perturbation,” Int. J.
Process., 32(3), pp. 1147–1177. Robust Nonlinear Control, 26(15), pp. 3274–3291.
[4] Marquez-Lago, T., Leier, A., and Burrage, K., 2010, “Probability Distributed [34] Dong, Y., Wang, X., Mei, S., and Li, W., 2012, “Exponential Stabilization of
Time Delays Integrating Spatial Effects Into Temporal Models,” BMC Syst. Nonlinear Uncertain Systems With Time-Varying Delay,” J. Eng. Math., 77(1),
Biol., 4(1), pp. 1–9. pp. 225–237.
[5] Marquez, M., 2015, Analysis and Control of Oilwell Drilling Vibrations a [35] Ma, Y., Zhao, L., and Liu, D., 2016, “Delay-Dependent Decentralized Control
Time-Delay Systems Approach, 2nd ed., Springer International Publishing, for Nonlinear Similar Interconnected Large-Scale Systems,” Int. J. Mach.
Cham, Switzerland. Learn. Cybern., 7(4), pp. 665–678.

[36] Jia, X., Chen, X., Xu, S., Zhang, B., and Zhang, Z., 2017, “Adaptive Output [45] Markazi, A., Maadani, M., Zabihifar, S., and Doost-Mohammadi, N., 2018,
Feedback Control of Nonlinear Time-Delay Systems With Application to “Adaptive Fuzzy Sliding Mode Control of Under-Actuated Nonlinear Systems,”
Chemical Reactor Systems,” IEEE Trans. Ind. Electron., 64(6), pp. 4792–4799. Int. J. Autom. Comput., 15(3), pp. 364–376.
[37] Song, J., and He, S., 2015, “Finite-Time Robust Passive Control for a Class of [46] Khosrowjerdi, M. J., 2011, “Robust Sensor Fault Reconstruction for Lipschitz
Uncertain Lipschitz Nonlinear Systems With Time-Delays,” Neurocomputing, Nonlinear Systems,” Math. Probl. Eng., 2011, p. 146038.
159, pp. 275–281. [47] Wenxiang, D., Jianyong, Y., and Dawei, M., 2018, “Time-Varying Input
[38] Hua, C. C., Li, Y., and Guan, X., 2017, “Finite/Fixed Time Stabilization for Delay Compensation for Nonlinear Systems With Additive Disturbance:
Nonlinear Interconnected Systems With Dead-Zone Input,” IEEE Trans. An Output Feedback Approach,” Int. J. Robust Nonlinear Control, 28(1),
Autom. Control, 62(5), pp. 2554–2560. pp. 31–52.
[39] Hua, C.-C., and Guan, X.-P.,. 2016, “Smooth Dynamic Output Feedback Con- [48] Pagilla, P. R., and Zhu, Y., 2004, “Controller and Observer Design for Lipschitz
trol for Multiple Time-Delay Systems With Nonlinear Uncertainties,” Automa- Nonlinear Systems,” American Control Conference, Boston, MA, June 30-July
tica, 68(1), pp. 1–8. 2, pp. 2379–2384.
[40] Hua, C. C., Liu, G., Li, L., and Guan, X., 2018, “Adaptive Fuzzy [49] Azmi, H., and Khosrowjerdi, M. J., 2016, “Robust Adaptive Fault Tolerant
Prescribed Performance Control for Nonlinear Switched Time-Delay Control for a Class of Lipschitz Nonlinear Systems With Actuator
Systems With Un-Modeled Dynamics,” IEEE Trans. Fuzzy Syst., 26(4), pp. Failure and Disturbances,” Proc. Inst. Mech. Eng., Part I, 230(1), pp.
1934–1945. 13–22.
[41] Hua, C. C., Zhang, L., and Guan, X., 2018, Robust Control for Nonlinear Time- [50] Muhammad, R., Naeem, I., and Keum-Shik, H., 2016, “Delay-Range-Depend-
Delay Systems, Springer, Singapore. ent Control of Nonlinear Time-Delay Systems Under Input Saturation,” Int. J.
[42] Hua, C. C., Zhang, L., and Guan, X., 2017, “Distributed Adaptive Neural Robust Nonlinear Control, 26(8), pp. 1647–1666.
Network Output Tracking of Leader-Following High-Order Stochastic Nonlin- [51] Khalil, H., 2002, Nonlinear Systems, 3rd ed., Pearson, Michigan State Univer-
ear Multi-Agent Systems With Unknown Dead-Zone Input,” IEEE Trans. sity, East Lansing, MI.
Cybern., 47(1), pp. 177–185. [52] Herrmann, G., Turner, M. C., and Postlethwaite, I., 2007, Linear Matrix
[43] Zhang, L., Hua, C. C., Yu, H., and Guan, X., 2019, “Distributed Adaptive Inequalities in Control, Springer London.
Fuzzy Containment Control of Stochastic Pure-Feedback Nonlinear Multi- [53] Azmi, H., and Khosrowjerdi, M. J., 2017, “LMI-Based Adaptive Output Feed-
Agent Systems With Local Quantized Controller and Tracking Constraint,” back Fault-Tolerant Controller Design for Nonlinear Systems,” Int. J. Adapt.
IEEE Trans. Syst. Man Cybern.-Syst., 49(4), pp. 787–796. Control Signal Process., 31(12), pp. 1885–1902.
[44] Chen, X., Du, S. L., Wang, L. D., and Liu, L. J., 2016, “Stabilization for Linear [54] Mao, J., Xiang, Z., and Zhai, G., 2019, “Sampled-Data Output Feedback Stabi-
Uncertain Systems With Switched Time-Varying Delays,” Neurocomputing, lization for a Class of Switched Stochastic Nonlinear Systems,” Int. J. Robust
191, pp. 296–303. Nonlinear Control, 29(10), pp. 2844–2861.

A Novel Delay-Independent Robust Adaptive Controller Design For Uncertain Nonlinear Systems With Time-Varying State Delay

Uploaded by

Copyright:

Available Formats

A Novel Delay-Independent Robust Adaptive Controller Design For Uncertain Nonlinear Systems With Time-Varying State Delay

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Novel Delay-Independent Robust Adaptive Controller Design For Uncertain Nonlinear Systems With Time-Varying State Delay

Uploaded by

Copyright:

Available Formats

A Novel Delay-Independent

Robust Adaptive Controller

Journal of Computational and Nonlinear Dynamics JANUARY 2020, Vol. 15 / 011002-1

011002-2 / Vol. 15, JANUARY 2020 Transactions of the ASME

Journal of Computational and Nonlinear Dynamics JANUARY 2020, Vol. 15 / 011002-3

2xT PDHE1 x xT ½e1 PDDT P þ eET1 E1 x;

2xT PB2 wðtÞ 2jjxT PB2 jjjjwðtÞjj 2jjxT PB2 jjw;

2xT Pf ðx; xd Þ ðL2f þ L2g ÞxT PPx þ xT x þ xTd xd ) k3  jjB1 B2 jjw

2xT Pf ðx; xd Þ 2jjxT Pf ðx; xd Þjj d k^3

011002-4 / Vol. 15, JANUARY 2020 Transactions of the ASME

R22 ¼ hET2 E2 þ I ð1 d2 ÞR MjjxT ðsÞPB1 jj

R11 PA2 x lim M

Journal of Computational and Nonlinear Dynamics JANUARY 2020, Vol. 15 / 011002-5

THEOREM 2. The controller (32) stabilizes asymptotically the

011002-6 / Vol. 15, JANUARY 2020 Transactions of the ASME

Journal of Computational and Nonlinear Dynamics JANUARY 2020, Vol. 15 / 011002-7

Fig. 3 Closed-loop state responses from Ref. [54] R1B X4

Multidisturbances are assumed and shown in Fig. 1. It is tried to 1 R2B 1

011002-8 / Vol. 15, JANUARY 2020 Transactions of the ASME

Journal of Computational and Nonlinear Dynamics JANUARY 2020, Vol. 15 / 011002-9

011002-10 / Vol. 15, JANUARY 2020 Transactions of the ASME

Journal of Computational and Nonlinear Dynamics JANUARY 2020, Vol. 15 / 011002-11

You might also like

2xT Pf ðx; xd Þ ðL2f þ L2g ÞxT PPx þ xT x þ xTd xd ) k3 jjB1 B2 jjw