J Syst Sci Complex (2020) 33: 289–311

Robust Stabilization of Non-minimum Phase Switched

Nonlinear Systems with Uncertainty
KHALIL Jouili

DOI: 10.1007/s11424-020-8200-6
Received: 9 July 2018 / Revised: 30 November 2018
The
c Editorial Office of JSSC & Springer-Verlag GmbH Germany 2020

Abstract This paper investigates the problem of robust stabilization of a class of switched nonlinear
system with uncertain dynamics where each subsystem represents a non-minimum phase. The authors
first construct a stabilizing sliding mode controller for each subsystem to stabilize individually its own
unstable internal dynamics. Then, a switching strategy is introduced to select the most appropriate
diffeomorphism through an infinity of diffeomorphisms. Sufficient conditions are specifically given for
the exponential stability and the exponential upper bound of the trajectory of the switched subsystem,
which guarantees the global asymptotical stability of the resulting switched system. Obviously, the
proposed control approach can improvemore the transient state, compared to a feedback linearization
based on only one diffeomorphism. Simulation studies illustrate the effectiveness of the suggested
approach.
Keywords Exponential stability, sliding mode control, switching control, uncertain non-minimum
phase systems.

1 Introduction
In the past decades, switched systems have attracted since they can be used to describe a
large number of physical and engineering systemssuch as mechanical systems[1, 2] , networked
control systems[3, 4] circuit and power systems[5, 6] chemical processes[7, 8] , robot manipulators[9, 10]
and near space vehicle control systems[11, 12] . Stability is the most important issue in the study
of switched systems[13–16] . This problem is very difficult to solve due to the hybrid nature of
switched systems operation. Several approaches, such as common Lyapunov function, multi-
ple Lyapunov function, signal Lyapunov function, switched Lyapunov function, average dwell
timeand their variants, have been put forward to solve (robust) stability and stabilization
issues[17–21] . Nevertheless, the stabilization problem of the switched systems with unstable
subsystems remains a challenging issue which deserves further investigations. Concerning this
problem, two aspects should be considered:
KHALIL Jouili
Laboratory of Advanced Systems Polytechnic School of Tunisia (EPT), B.P. 743, 2078 Marsa, Tunisia.
Email: [email protected].
 This paper was recommended for publication by Editor JIA Yingmin.
290 KHALIL JOUILI

i) Stabilizing a switched system without control input by designing an appropriate switching

law[20]
ii) Stabilizing a switched control system via constructing both the controller and the switch-
ing law[22] .
Despite the importance of the results provided by previous studies, few attempts were
made to stabilize non-minimum phase nonlinear systems where each subsystem can be a non-
minimum phase. In fact, the stabilization of non-minimum phase nonlinear systems is a quite
challenging problem in the field of control[23–28] . Some contributions have also been devoted
to non-minimum phase switched nonlinear systems where each nonlinear mode may be non-
minimum phase. In [29], H∞ control goal was achieved for a class of non-minimum phase
cascade switched nonlinear systems where the internal dynamics of each mode is assumed
to be asymptotically stabilizable. A control approach for the stabilization of a class of non-
minimum phase switched nonlinear systems based on the concept of multi-diffeomorphism was
contemplated in [30]. Besides, the output tracking of non-minimum phase switched nonlinear
systems was considered in [31] where an approximated minimum phase model was utilized.
In [32], the stabilization of non-minimum phase switched nonlinear systems applied to multi-
agent systems was proposed. In these systems, the states of linearized dynamics of all modes,
which compose the whole state space and state-dependent stabilization switching laws, were
provided by considering both common and multiple Lyapunov functions. The same problem
was also investigated in [33] by means of an inversion-based control strategy. In addition, the
results provided by control synthesis of switched nonlinear system with uncertain dynamics,
where each subsystem represents a non-minimum phase, are rare[34, 35] .
In this direction, we have proposed a robust sliding mode control approach for a class
of uncertain switched nonlinear systems where each subsystem may be non-minimum phase.
This idea is based on designing an individual controller for each subsystem, to stabilize its
own unstable internal dynamics, and elaborating a switching strategy such that the resulting
uncertain switched nonlinear system is globally asymptotical stable.
In the control approach design, each subsystem of the switched nonlinear system with un-
certain dynamics is decomposed into an input output model and internal dynamics. A virtual
sliding variable is introduced to stabilize the unstable zero dynamics and realize the stabilization
of the associated subsystem. Then, a sliding mode controller, for each subsystem, is designed
to drive the virtual slip variable to converge to zero in finite time. The closed loop dynamics
of each subsystem are, therefore, guaranteed to converge asymptotically to zero on the sliding
mode surface. Thus, the controller developed for each subsystem has an excellent performance
that is robust to the uncertain dynamics.
Then, sufficient conditions are given for the exponential stability and the exponential upper
bound of the trajectory of the switched subsystem, which guarantees the globally asymptotical
stability of the resulting uncertain switched nonlinear system.
This paper is organized as follows. System description and problem formulation are intro-
duced in Section 2. The design of stabilizing sliding mode controllers is presented in Section 3.
Then, an analysis of stability is shown in Section 4. In Section 5, a numerical example is given
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 291

to illustrate the effectiveness of the theoretical results. Finally, conclusions and remarks are
summarized in Section 6.

2 System Description and Problem Formulation

We consider a class of uncertain nonlinear systems in the following form:
⎧
⎨ ẋ = f (x, θ) + G (x, θ) u,
(1)
⎩ y = h (x) ,

where the state variable x ∈ n and the control input u ∈ ; f (x, θ) and G (x, θ) are smooth
functions; h (x) is the output function and θ denotes the lumped perturbation.
For System (1), we make the following assumptions:
Assumption 1 The system (1) is locally input output linearizable and has a well-defined
relative degree r ≤ n at an equilibrium point x0 . More precisely, LG Ljf h (x) = 0 for all
(j−1)
0 < j < (r − 1), and LG Lf h (x) = 0 for all x in a neighborhood x0 , where the notation
Ljf h (x) = 0, j ≥ 0, represents the Lie derivative of the scalar function h (x) with respect to
the field f (x), and Ljf h (x) = ∂f (x)
∂x h (x).
Assumption 2 All the system states and the perturbation term in (1) are assumed to
be locally Lipschitz in the domain of interest.
The nonlinear system (1) can be transformed into the Byrnes Isidori normal form by a local
diffeomorphism Φ (x) as follows[23] :
⎡ ⎤ ⎡ ⎤
ξ1 h (x)
⎢ ⎥ ⎢ ⎥
⎢ ξ ⎥ ⎢ L h (x) ⎥
⎢ 2 ⎥ ⎢ f ⎥
⎢ ⎥ ⎢ ⎥
⎡ ⎤ ⎢ . ⎥ ⎢ . ⎥
⎢ .. ⎥ ⎢ .. ⎥
ξ ⎢ ⎥ ⎢ ⎥
⎢ ⎥ ⎢
⎣ ⎦ = ⎢ ξr ⎥ = Φ (x) = ⎢ Lr h (x) ⎥ ⎥ (2)
η ⎢ ⎥ ⎢ f ⎥
⎢ ⎥ ⎢ ⎥
⎢ η1 ⎥ ⎢ χ1 (x) ⎥
⎢ ⎥ ⎢ ⎥
⎢ .. ⎥ ⎢ .. ⎥
⎢ . ⎥ ⎢ . ⎥
⎣ ⎦ ⎣ ⎦
η(n−r) χ(n−r) (x)

with:
⎧
⎨ L χ (x) = 0,
G j
(3)
⎩ j = 1 , 2, · · · , n − r.

The dynamics of the system (1) can be decomposed into two parts in the new coordinate: Input
292 KHALIL JOUILI

output model and internal dynamics described by the following equations:

⎧
⎪
⎪ ξ˙ i = ξ i+1 , i = 1, 2, · · · , r − 1,
⎪
⎪
⎪
⎨ ξ˙ = a (ξ, η) + φ (ξ, η, θ) + b (ξ, η) u,
r
(4)
⎪ η̇ = Ψj (x) , j = 1, 2, · · · , n − r,
⎪
⎪
⎪
⎪
⎩
y = ξ 1,

where a (ξ, η, θ) = 0 , b (ξ, η) = 0, ξ ∈ r is the vector of input output dynamics, η ∈ n−r

represents the vector of internal states, y ∈  corresponds to a scalar output and φ (ξ, η, θ)
stands for the nonlinear perturbation of the system.
We can conclude that Equation (3) has not only one solution but infinite ones. However,
there exists an infinite number of diffeomorphisms.
If a function h (x) satisfies Assumption 1 and χ1, i (x), χ2, i (x), · · · , χ(n−r), i (x) are nonlinear
scalar functions of x which can be explicit by the following relation:
⎧
⎪
⎨ LG χj,i (x) = 0,
⎪
j = 1, 2, · · · , n − r, (5)
⎪
⎪
⎩
i = 1, 2, · · · , m.

Then, we will impose the following lemma on System (1).

Lemma 1 For each i ∈ M = {1, 2, · · · , m}, there exists a set of diffeomorphisms:
⎡ ⎤
ξ1
⎢ ⎥
⎢ ξ2 ⎥
⎢ ⎥
⎢ ⎥
⎢ .. ⎥
⎢ . ⎥
⎢ ⎥
⎢ ⎥
Φi (x) = ⎢ ξr ⎥ (6)
⎢ ⎥
⎢ ⎥
⎢ χ1, i (x) ⎥
⎢ ⎥
⎢ .. ⎥
⎢ . ⎥
⎣ ⎦
χ(n−r), i (x)

such that the system of Equations (4) takes the following form:
⎧
⎪
⎪ ξ̇1 = ξ2 ,
⎪
⎪
⎪
⎪ ..
⎪
⎪
⎪
⎪ .
⎪
⎪
⎨ ξ̇ = a (ξ, η) + φ (ξ, η, θ) + b (ξ, η) u ,
r i
(7)
⎪
⎪ η̇1 = Ψ 1, i (ξ, η) ,
⎪
⎪
⎪
⎪ ..
⎪
⎪
⎪
⎪.
⎪
⎪
⎩ η̇
(n−r) = Ψ (n−r), i (ξ, η) .
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 293

The system given by (1) and satisfying Lemma 1 is transformed into the following switched
nonlinear systems with uncertain dynamics:

ẋ = F (i) (x, θ) + G(i) (x, θ) ui (8)

with i is a set of indices specifying the active subsystem where x ∈ n are state variables. Let
M = {1, 2 , · · · , m} where m is the number of subsystems ∀ i ∈ M, ui ∈ , ∀ i ∈ M, ui ∈ 
is the input, F (i) (.) and G(i) (.) are smooth functions such that F (i) (0) = G(i) (0) = 0. For
each subsystem i ∈ M , we find a partition x = ξ η T where ξ ∈ r , η ∈ (n−r) .
Then, we rewrite the normal form (7) into System (8), which gives the following system
equation:
⎧
⎨ ξ˙ = a (ξ, η) + φ (ξ, η, θ) + b (ξ, η) u ,
i i i i
(9)
⎩ η̇ = Ψj, i (ξ , η) , i ∈ M = {1, 2 , · · · , m} ,

where m is the number of subsystems. ∀ i ∈ M, ui ∈  is the input, η̇ = Ψj, i (0 , η) is the

unstable zero dynamics, ai (ξ, η) ∈ (r×1) = 0, and φi (ξ, η, θ) = 0.
If the zero dynamics η̇ = Ψj, i (0 , η) are unstable, then the subsystem i will be a non-
minimum phase.
The major objective of this work is to design robust controllers to globally stabilize the
system (8).

3 Design of Stabilizing Sliding Mode Controllers

In this section, we design, for each active subsystem i of the uncertain nonlinear switched
system (8), a sliding mode controller to stabilize its unstable zero dynamics η̇ = Ψj, i (ξ , η).
First, the zero dynamics η̇ = Ψj, i (ξ , η) for each subsystem i of the switching system (9)
are linearized approximately around the equilibrium point (ξ, η) = 0 as:

η̇ = Ψj, i (ξ , η) = Aiξ ξ + Aiη η + δ i (ξ, η) , (10)

 
∂Ψj, i (ξ , η)  ∂Ψj, i (ξ , η) 
where δ i (ξ, η) is the higher order term and Aiξ = ∂ξ  (ξ, η) , Aiη = ∂η  (ξ, η) are
 
known matrices for which the pair Aiξ , Aiη is controllable.
The zero dynamics, for each subsystem i given by (10) of the switching system (9), are
completely unobservable. Thus, the behaviour as well as the stability of the zero dynamics, for
each subsystem i, have to be examined carefully.
In this paper, a class of uncertain nonlinear systems, where each subsystem represents a
non-minimum phase, is considered. More precisely, the matrices Aiη has eigenvalues on the
right half of the complex plane.
Remark 3.1 The zero dynamics, for each subsystem i given by (10) of the switching
system (9), are completely unobservable. Thus, the behaviour as well as the stability of the
zero dynamics, for each subsystem i, have to be examined carefully. In this paper, a class
of uncertain nonlinear systems, where each subsystem represents a non-minimum phase, is
294 KHALIL JOUILI

considered. More precisely, the matrices Aiη has eigenvalues on the right half of the complex
plane.
For each subsystem i, we define the sliding variable as:

S = ξr + (r−1), i ξr−1 + · · · + 1, i ξ1 , (11)

where the parameters 1, i , 2, i , · · · , (r−1), i are chosen such that: pr + (r−1), i p(r−1) + · · · +
T
1, i p is a Hurwitz polynomial. We define the new variable ζ = ξ1 ξ2 · · · ξr −1 η to obtain
the following relation:
⎡ ⎤
1 0 0 0 0
⎡ ⎤ ⎢ ⎥
⎢ .. .. .. ⎥
ζ ⎢ .⎥
⎣ ⎦ = ⎢ . . ⎥ ζ. (12)
⎢ ⎥
S ⎢ 0 0 ··· 1 0 ⎥
⎣ ⎦
1, i 2, i (r−1), i 1

For each subsystem i, we get diffeomorphism: Ti : x = (ξ, η) → (ζ, S),

⎡ ⎤
h (x)
⎢ ⎥
⎢ ⎥
⎢ Lf h (x) ⎥
⎢ ⎥
⎢ .. ⎥
⎢ . ⎥
⎡ ⎤ ⎢ ⎥
⎢ r−1 ⎥
ζ ⎢ Lf h (x) ⎥
Ti (x) = ⎣ ⎦ = ⎢⎢
⎥
⎥ (13)
S ⎢ χ1, i (x) ⎥
⎢ ⎥
⎢ .. ⎥
⎢ . ⎥
⎢ ⎥
⎢ ⎥
⎢ χ(n−r), i (x) ⎥
⎣ ⎦
S

such that the switched nonlinear system (8) takes the following form:
⎡ ⎤ ⎡ ⎤
i
ζ̇ Aiζ ζ + AiS S + δ (ζ, S)
⎣ ⎦ = ⎣ ⎦, (14)
Ṡ ai (ζ, S) + φi (ζ, S , θ) + bi (ζ, S) ui
r−1
where ai (S, ζ) is the expression of ai (ξ, η) = j=1 j, i ξj+1 + aj (ξ, η) under the coordi-
i
nate (ζ, S) and φi (ζ, S) , bi (ζ, S) , δ (ζ, S), is the expressions of φi (ξ, η) , bi (ξ, η) , δ i (ξ, η),
respectively, in the coordinate (ζ, S).
We rewrite the expression ai (S, ζ) of the sliding variable by separating its linear part from
its nonlinear one:
i i
ai (ζ, S) = Aζ ζ + AS S i + φi (ζ, S , θ) , (15)
 
i (S, ζ))  i ∂(ai (S, ζ)) 
where Aζ = ∂(ai∂ζ  (ζ, S) , As = ∂S  (ζ, S) and φi (ζ, S , θ) are nonlinear terms.
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 295

By combining (14) and (15), we obtain:

⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
i
ζ̇ Aiζ AiS ζ 0r−1 δ (ζ, S)
⎣ ⎦ =⎣ ⎦ ⎣ ⎦ +⎣ ⎦ ui + ⎣ ⎦, (16)
i i
Ṡ Aζ AS S bi (ζ, S) φi (ζ, S , θ) + φi (ζ, S , θ)

i (ζ, S , θ) + φ (ζ, S , θ), then System (14) can be written in the form
where φi (ζ, S , θ) = φ i
as:
⎡ ⎤ ⎡ ⎤
ζ̇ ζ
⎣ ⎦ = Ai ⎣ ⎦ + B i ui + Δi (ζ, S , θ) (17)
Ṡ S

with ⎡ ⎤ ⎡ ⎤ ⎡ ⎤
i
Aiζ AiS 0n−1 δ (ζ, S)
Ai = ⎣ i i
⎦, Bi = ⎣ ⎦, Δi (ζ, S, θ) = ⎣ ⎦.
Aζ AS bi (ζ, S ) φi (ζ, S, θ)

For each subsystem i, we define the sliding manifold σ i = S i − Li ζ in order to design

a virtual control to stabilize the uncontrollable zero dynamics η̇ = Ψj, i (ξ , η) where S i is
considered as a virtual control in the sliding modeand L i is a constant matrix.
When the internal mode of the system reaches the designed sliding surfaces, we have S i =
i
L ζ . Then, from the equation of the sliding mode of the system (17), we can write:

i ζ + Δi (ζ) ,
ζ̇ = A (18)
 
where the matrices A i = Ai − Ai Li and Δi (ζ) are the higher-order perturbation terms.
ζ S
 
Remark 3.2 For each subsystem i of the switching system (18), the pair Aiζ , AiS Li
 
is controllable. Thus, Li can be designed such that the eigenvalues of Aiζ − AiS Li are arbi-
trarily assigned in the left half-plane. Hence, each closed-loop active subsystem of the switched
system (18) will be asymptotically stable in the sliding mode.
Lemma 2 We consider the nonlinear switched system (18) where both the system dynam-
ics ζ and the perturbed dynamics Δi (ζ), for each active subsystem i, satisfy Assumption 2.
We suppose that the active subsystem ζ̇ = A i ζ is asymptotically stable at ζ = 0 and the
perturbation term Δi (ζ) which satisfies the following growth bound is rewritten as follows:

Δi ( ζ) ≤ ϕi (t) (ζ) , (19)

where ϕi (t) :  →  is non-negative and continuous for all t ≥ 0, and Δi (0) = 0.  

 T
Let Qi = Qi > 0, we solve the Lyapunov equation for each subsystem P i A i +
 T
i P i = −Qi for a unique positive definite solution P i .
A
λi (Qi )
For each subsystem i, if ϕi (t) < min
i i
, the active subsystem of the nonlinear switched
2λmax (P )
system (18) is exponentially stable.
296 KHALIL JOUILI

Proof Let V i (ζ) = ζ T P i ζ be a Lyapunov candidate function for each subsystem i of the
switched system (18). V i (ζ) is positive definite on n . In this case, the following inequalities
can be derived:
  2 i  
λimin P i ζ ≤ V i (ζ) ≤ 2λmax P i ζ , (20)
 i 
 V (ζ)   T i i  i
   
 ∂ ζ  = 2ζ P ≤ 2λmax P ζ . (21)

Its derivative is given by:

     
∂ V i (ζ) i i + A
T
i P i ζ
A ζ = ζT Pi A
∂ (ζ)
 
= −ζ T Qi ζ
  2
≤ −λimin Qi ζ . (22)

The time derivative of V i (ζ), with the help of (19) and (22), yields

∂ V i (ζ) i ∂ V i (ζ)
Aζ + Δi (ζ)
∂ (ζ) ∂ (ζ)
  i  
≤ − λimin Qi ζ 2 + 2ϕi (t) λmax P i ζ 2
. (23)

λimin (Qi )
If we choose ϕi (t) < i , then V̇ i (ζ) < 0.
2λmax (P i )
This means that the active subsystem with uncertain dynamics of the nonlinear switched
system (18) is exponentially stable at the origin in the sense of Lyapunov.
In the following, we design, then, a robust sliding mode controller in order converge the
sliding manifolds σ i to zero. Thus, we ensure, in the sliding mode, the stability for each
subsystem i of the switched nonlinear system (17).
For each subsystem i, we propose the sliding mode controller under the following form:

ui = ueq S
i + ui , (24)

where ueqi is the equivalent control which determines the dynamic property of the subsystem i
that reaches the sliding surfaces and uSi denotes the switching control that forces the subsystem
i trajectory to reach the sliding surfaces in finite time.
 
The application of derivative σ̇ i , with respect to time t along the trajectory of System (14),
leads to:
 i   i  i
σ̇ i = Aζ + L i Aiζ ζ + AS + L i AiS S i + δ (ζ, S) L i + φi (ζ, S , θ) + bi (ζ, S) ui . (25)

To ensure the sliding mode asymptotical stability of each closed-loop subsystem i, we deduce
the equivalent controller:
1  i   i  
ueq
i = − Aζ + L i Aiζ ζ + AS + L i AiS S i . (26)
bi (ζ, S)
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 297

By combining (17), (25) and (26), we obtain:

i
σ̇ i = Li δ (ζ, S) + φi (ζ, S , θ) + bi (ζ, S) usi . (27)

To compensate the effect of uncertainty and internal mode nonlinear high order item perturba-
tion, we design the controller usi based on the approximating technique to reduce the chattering
phenomenon resulting from discontinuous control and finite switching times:
1  i   
usi = − K sgn σ i + E i σ i , (28)
bi (ζ, S)

where E i = diag {ε i } , K i = diag {ki } , ε i ≥ 0 and ρi are the parameters that affect the
speed with which the sliding surfaces are reached, with ρi satisfies:
 
 i i 
L δ (ζ, S) + φi (ζ, S, θ) ≤ ρi . (29)

The parameters ki are responsible for the speed with which the sliding surfaces are reached. If
ki are constants that chosen rather large, the reaching time may be shorten, at the same time,
the high reaching speed would arise undesirable chattering. Dynamic parameters ki are chosen
instead, which decrease the reaching speed to weaken chattering influence, and proper large are
introduced to shorten the reaching time. Regulating these parameters properly could reduce
the chattering phenomenon and promote the dynamic performance of the approximating mode
of the system. To realize the sliding manifold could be reached in finite time, positive real
number ki satisfies:

ki ρi . (30)

By applying the following theorem, the developed concept is formulated:

Theorem 3.3 If we consider the switched nonlinear system with uncertain dynamics,
where each mode may be non-minimum
 phase (17), and we assume that matrix bi (ζ, S ) is
non-singular and Aiζ , AiS Li is completely controllable.
For each active subsystem i, we obtain the following sliding mode controllers:
1  i   i  
ui = − Aζ + L i Aiζ ζ + AS + L i AiS S i
bi (ζ, S)
1  i   
− K sgn σ i + E i σ i , (31)
bi (ζ, S)

which forces the trajectory of each subsystem to reach the sliding surfaces. The zero dynamics
η̇ = Ψj, i (ξ , η) could be asymptotically stabilized on the sliding manifolds σ i = 0.
Proof As we know, the principle of the sliding mode controller is to force the trajectory for
each subsystem i of the switched nonlinear system (17) to enter into the sliding mode in finite
time, i.e., to satisfy the reachable condition
 i T i  
σ σ̇ ≤ −ci σ i  , c i 0. (32)
298 KHALIL JOUILI

By substituting the sliding mode controller (31) into (25), we consider (29)
   i

σ̇ i = −εi σ i − ki sign σ i + Li δ (ζ, S) + φi (ζ, S , θ) (33)

and
 i T i  2    i
 
σ σ̇ = −εi σ i  − ki σ i  + Li δ (ζ, S) + φi (ζ, S , θ) σ i  . (34)

Then, the following inequality is satisfied

 i T i    2     
 i 
σ σ̇ = − λimin E i σ i  − ki −  Li δ (ζ, S) + φi (ζ, S , θ)  σ i  , (35)

where is the Euclidian norm and λimin (·) denotes the smallest eigenvalues.
 
As E = diag {ε i } and, ε i ≥ 0 the inequality λimin E i ≥ 0 is satisfied.
i

Moreover, with the aid of Inequality (30), the reachable condition (32) is satisfied.
Thus, under the action of the sliding mode controller (31), the sliding mode could be reached
in finite time and remains there. System (17) is written in the same form as system equation (8)
where
⎧
⎨ F (i) (x, θ) = A i x + Δ (x, θ) ,
i
(36)
⎩ G(i) (x) = B i .

The proof is finished.

4 Switching Strategy and Stability Analysis

In this section, we present a switching strategy and stability analysis of the uncertain non-
linear switched system (8). We also propose sufficient conditions for the exponential stability
and the exponential upper bound of the trajectory state of the resulting switched system. We
also introduce a strategy to select the most appropriate diffeomorphism, as there is an infi-
nite number of diffeomorphisms, and we develop switching rules for the global stabilization of
the uncertain nonlinear switched system (8). It is important to note that the stability of the
subsystem does not necessarily reflect the stable behavior of the overall closed loop system[36] .
However, the switch between the unstable subsystems can produce a stable system by applying
a stabilizing switching strategy. Thus, we may conclude that the global stability of the resulting
system depends on the type of the switching strategy.
In this framework, our objective is to develop switching rules for the global stabilization of
uncertain nonlinear switched system (8) through infinity of diffeomorphisms.
In this context, the suggested solution is based on using the following assumptions and
lemma:
Assumption 3 For tj ≤ t < tj+1 , with t is the time interval between two successive
switching, the following switched nonlinear system with uncertain dynamicsis active:
⎧
⎨ ẋ = F (kj ) (x, θ) + G(kj ) (x, θ) u ,
kj
(37)
⎩ kj ∈ [ 1 m] , j ∈ [0 ∞) .
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 299

Assumption 4 For all x ∈ Dx ⊂ n , the coordinate transformation

T  
Ti (x) = ξ1 · · · ξr−1 χ1,i · · · χ(n−r),i S . Let Υkj = Tkj (x) x , kj ∈ [ 1 m] and
j ∈ [0 ∞) checks the existence of a limit when x tends towards zero (x → 0).
Lemma 3 By taking Theorem 1 and Lemma 2 into account, the solution of each active
subsystem of the uncertain nonlinear switched system (8) is exponentially stable, which verifies
the following inequality:

 i    i 
λ Υi (x (t0 )) −
λmin (t−t0 )

x (t) ≤  max
x (t0 ) e 2λi
max , (38)
i Υi (x)
λ min

i i
where λimin is the minimal eigenvalues of Qi ; λmax and λmin represent respectively the maximum
and the minimum of P i .
Switching strategy
By considering Theorem 1 and Lemma 2, we propose switching rules to select the appropriate
diffeomorphism while respecting the following switching steps:
To preserve generality, we start with ẋ = F 1 (x, θ) + G1 (x, θ) u1 .
Step 1 We start by initializing the index i (i = 0) which specifies the active subsystem at
a given time t.
Step 2 By respecting Assumption 3 and Theorem 1, two cases are presented:
(i) If the following condition:
⎧  i −1
⎪
⎪ Υki+1 (x(ti )) λmax Υ1 (x(ti ))
⎪ min Υki+1 (x(t)) <
⎪
⎨ i
λmin Υ1 (x(t)) ,

ki+1 ∈ 2 m , (39)
⎪
⎪
⎪
⎪
⎩i ∈
1m

is verified, so for t ≥ ti+1 , the subsystem ẋ = F (ki+1 ) (x, θ) + G(ki+1 ) (x, θ) u(ki+1 ) will be
active.  i −1
Υki+1 (x(ti )) λmax Υ1 (x(ti ))
(ii) If the condition min Υk (x(t)) < i Υ1 (x(t)) is not verified, so for t ≥
i+1 λmin
1 1
ti+1 , the subsystem ẋ = F (x, θ) + G (x, θ) u1 will be active.
Step 3 The parameter i is incremented (i = i + 1), following the choice of the second active
subsystem, and the preceding step is resumed to change to another subsystem.
In the following part, we develop a systematic design procedure for the global stabilization
of uncertain nonlinear switched system (8).
For more generalization, we consider the development of decomposition in a specific time
interval.
Case 1 In the interval t ∈ t0 tn1 , we analyze the behavior of the switched nonlinear
system (8):
When t ∈ t0 t1 , the subsystem:

ẋ = F 1 (x, θ) + G1 (x, θ) u1 (40)

300 KHALIL JOUILI

is active.
For the subsystem (40), the inequality given by Equation (38) verifies the hypothesis of
Lemma 2 and is written in the following form:
⎛ i ⎞  i 
λ Υ (x (t )) −
λmin (t−t0 )

x (t) ≤ ⎝ i max 1 0 ⎠ x(t0 ) e i

2λmax
. (41)
λmin Υ1 (x)
! "# $
Γ0 (x)

When t ∈ [t1 t2 [, the subsystem:

ẋ = F k1 (x, θ) + Gk1 (x, θ) uk1 (42)

is active.
By taking Assumption 4 and the inequality (41) into account, we have:
%% i & &  i
λmin (t−t0 )

λmax Υk1 (x (t1 )) Υ1 (x (t0 )) − i
x (t) ≤ i x(t0 ) e 2λmax
. (43)
λmin Υk1 (x) Υ1 (x (t1 ))
! "# $
Γ1 (x)

Correspondingly, when t ∈ [tj , tj+1 [, the subsystem:

ẋ = F (kj ) (x, θ) + G(kj ) (x, θ) ukj (44)

is active.
Similarly, the solution x(t) of the subsystem (44) is expressed by
 
λi
min (t−t0 )
−
2λi
x (t) ≤ Γj (x) x(t0 ) e max (45)

with
% i
& j+1     %j−1 &
λmax
2
Υkj (x (tj )) Υ1 (x (t0 )) ' Υkl (x (tl ))
Γj (x) = i
. (46)
λmin Υkj (x) Υ1 (x (t1 )) Υkl (x (tl+1 ))
l=1

From the solution x(t) of the subsystem (44), we have the two following inequalities:
(i)
  % i
&−1  
Υkj (x (tj )) λmax Υ1 (x (tj ))
≤ i . (47)
Υkj (x (t)) λmin Υ1 (x (t))

(ii)
  % i
&−1  
Υki (x (ti )) λmax Υ1 (x (ti ))
≤ i . (48)
Υki (x (ti+1 )) λmin Υ1 (x (ti+1 ))
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 301

By combining the two inequalities (49) and (50), the expression Γj (x) is written as:
% i
&− ( j−1
2 )  
λmax Υ1 (x (t0 ))
Γj (x) < i . (49)
λmin Υ1 (x)

The solution x(t) of the switched system (8) given by the equation (38) of Lemma 3 is written
as:
% i
&−( j−1
2 )   
λi

min (t−t0 )
λmax Υ1 (x (t0 )) −
2λi
x(t) < i x (t0 ) e max

λmin Υ1 (x)
   
λi
min (t−t0 )
Υ1 (x (t0 )) −
2λi
≤ x(t0 ) e max . (50)
Υ1 (x)

Then, the subsystem described by the equation (44) exponentially converges on the time interval
t ∈ t0 tn1 .
Case 2 In the interval t ∈ tn1 tn2 , we analyze the behavior of the switched nonlinear
system (8):
When t ∈ t0 t1 , the subsystem ẋ = F 1 (x, θ) + G1 (x, θ) u1 is active.
For the active subsystem ẋ = F 1 (x, θ) + G1 (x, θ) u1 , the inequality given by Equation
(38) is rewritten in the following form:
⎛ i ⎞  i 
λ Υ (x (t )) −
λmin (t−t0 )

x (t) < ⎝  max 1 n1 ⎠ x (tn1 ) e 2λ i

max . (51)
i Υ1 (x)
λ min

Therefore, after some algebraic development, we can deduce that:

⎛ i ⎞  i 
λ Υ (x (t )) −
λmin (t−t0 )

x (t) < ⎝  max 1 0 ⎠

x (t0 ) e 2λi
max . (52)
i
λmin Υ1 (x)
! "# $
Γn1 (x)

Hence,
⎛ i    ⎞ ⎛ i ⎞
λ Υ x t  λ Υ (x (t ))
⎝ max 1 n1
⎠ = ⎝ max 1 0 ⎠
. (53)
i Υ1 (x) i
λmin λmin Υ1 (x)
! "# $ ! "# $
Γn1 (x) Γ0 (x)

We follow the same development as that presented in the interval t ∈ t0 tn1 :

When t ∈ tnk tnk +1 , the subsystem

ẋ = F (kn k ) (x, θ) + G(kn k ) (x, θ) ukn k (54)

is active.
302 KHALIL JOUILI

Based on Lemma 2 and the solution of Equation (38), we obtain:

⎛ i ⎞  i 
λ Υ (x (t )) −
λmin (t−t0 )

x (t) ≤ ⎝  max 1 n1 ⎠ x (t0 ) e 2λ i

max . (55)
i Υ1 (x)
λmin
! "# $
Γn k (x)

By following an algebraic development similar to that described for the two intervals t ∈ t0 tn1
and t ∈ tn1 tn2 , for all t ∈ tnk tnk+1 , we may conclude that:
⎧ ⎛ 
⎪     ⎞ ⎛ i ⎞
⎪  λi
⎪ Υ x t  λ Υ (x (t ))
⎪
⎪ ⎝ max 1 n1
⎠ = ⎝ max 1 0 ⎠
,
⎪
⎨ i Υ (x) i Υ (x)
λmin 1 λmin 1
! "# $! "# $ (56)
⎪
⎪
⎪
⎪ Γn1 (x) Γ0 (x)
⎪
⎪
⎩ k → ∞.

Therefore, for all t ∈ tnk tnk+1 , it is true that:

⎛
 i ⎞  
λ Υ (x (t )) −
λi
min (t−t0 )
x (t) ≤ ⎝ imax 1 0 ⎠ x (t0 ) 2λi
e max . (57)
λmin Υ1 (x)

This result is obtained after an algebraic manipulation which aims at presenting the expression
Γnk (x) in the following recurrent form:

 i
λ Υ1 (x (t0 ))
Γnk (x) = Γ0 (x) =  maxi . (58)
λ Υ1 (x)
min

By taking into account the already-presented Lemmas and assumptions, we may deduce that
whatever the final active system, its operation is exponentially stable.

5 Simulation Example
In this section, we present an illustrative example to demonstrate the applicability and
effectiveness of the proposed approach.
We consider the following uncertain nonlinear system described by [37]:
⎧
⎪
⎪ ẋ1 = −x1 + x2 ,
⎪
⎪  
⎪
⎨ ẋ2 = −3x2 + x3 + θ + (sin (x3 ))2 + 2 u,
1
(59)
⎪
⎪ ẋ3 = x1 − 2x3 ,
⎪
⎪
⎪
⎩
ẋ4 = − x4 + x23 .

This system (59) can be written under the same form of uncertain nonlinear system (1) where
θ (t) = 0.3 sin (3 t), and y = x1 − 3x3 represents the output. Hence, we have:
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 303

⎡ ⎤ ⎡ ⎤
−x1 + x2 0
⎢ ⎥ ⎢ ⎥
⎢ −3x + x3 + θ ⎥ ⎢ (sin (x ))2 + 2 ⎥
⎢ 2 1 ⎥ ⎢ 3 ⎥
F (x, θ) = ⎢ ⎥ and G (x, θ) = ⎢ ⎥.
⎢ x − 2x ⎥ ⎢ 0 ⎥
⎣ 1 3 ⎦ ⎣ ⎦
2
− x4 + x3 0
By simple calculations, the relative degree of the system (59) can be obtained as r = 2.
Thus, System (59) will be transformed into the normal form shown below:
⎡ ⎤
x1 − 3x3
⎢ ⎥
⎢ −4x + 6x + x ⎥
⎢ 1 3 2⎥
Ti (x) = ⎢ ⎥. (60)
⎢ χ1, i (x) ⎥
⎣ ⎦
χ2, i (x)

Using Lemma 1, the expressions of the synthesized dynamic compensator (χ1, i (x) and χ2, i (x))
are not unique because of their infinite solutions, which gives an infinite number of diffeomor-
phisms.
Using the method described to present the expression of the dynamic compensators, two
different diffeomorphisms are determined as follows:
⎡ ⎤
x1 − 3x3
⎢ ⎥
⎢ −4x + 6x + x ⎥
⎢ 1 3 2⎥
T1 (x) = ⎢ ⎥ (61)
⎢ x3 ⎥
⎣ ⎦
x4

and
⎡ ⎤
x1 − 3x3
⎢ ⎥
⎢ −4x + 6x + x ⎥
⎢ 1 3 2⎥
T2 (x) = ⎢ ⎥. (62)
⎢ x3 + x4 ⎥
⎣ ⎦
2x3

We apply the approach depicted in Section 3. The system given by (59) satisfying Lemma 1 is
transformed by the two following subsystems:
For the subsystem 1:
⎧
⎪
⎪ ξ̇1 = ξ2 ,
⎪
⎪
⎨ ξ̇ = 3ξ − 4ξ − 9η + (ξ + 3η )3 + θ + 2 + sin2 (η ) u ,
⎪
2 1 2 1 1 2 1 1
(63)
⎪
⎪ η̇1 = η1 + ξ1 ,
⎪
⎪
⎪
⎩
η̇2 = η12 − η2 .
304 KHALIL JOUILI

The approximately linearized model of the internal dynamic of the subsystem 1 can be obtained
as shown below:
⎡ ⎤ ⎡ ⎤
1 0 10
η̇ = ⎣ ⎦η +⎣ ⎦ ξ. (64)
2 −1 00

To introduce a virtual controller for the subsystem 1, let

S (1) = ξ1 + ξ2 . (65)

Thus, based on Assumption 3 and the development provided in Section 4, the subsystem (63)
can be rewritten as:
⎡ ⎤ ⎡ ⎤
ζ̇ ζ
⎣ ⎦ = A1 ⎣ ⎦ + B 1 u1 + Δ1 (ζ, S , θ) (66)
Ṡ S

with
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
1
A1ζ A1S T 03 δ (ζ, S)
A1 = ⎣ 1 1
⎦, ζ = ξ1 η1 η2 , B1 = ⎣ ⎦, Δ1 (ζ, S, θ) = ⎣ ⎦,
Aζ AS b1 (ζ, S ) φ1 (ζ, S, θ)

where
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
−1 0 0 1 0
⎢ ⎥ ⎢ ⎥ 1 1 1 ⎢ ⎥
A1ζ = ⎢ ⎥, A1S = ⎣0⎥
⎢ δ (ζ, S) = ⎣ 0 ⎥
⎢
⎣ 1 1 0 ⎦ ⎦, Aζ = 6 −9 0 , AS = [−3] , ⎦,
0 2 −1 0 0
φ1 (ζ, S, θ) = (ξ1 + 3η1 ) + θ.
3
b1 (ζ, S ) = 2 + sin2 (η1 ) ,

The virtual sliding surface is designed as:

⎡ ⎤
ξ1
⎢ ⎥
σ 1 = ξ1 + ξ2 + L1 ⎢ ⎥
⎣ η1 ⎦ . (67)
η2

We can model L1 = −3 −1 −6 to make the sliding mode have eigenvalues on the left half
of the complex plane.
Then, we ensure the approximate stability of the subsystem (63) in the neighborhood of the
equilibrium point. Subsystem (63) is written in the same form as that of equation system (14)
where:
⎧ ⎡ ⎤ ⎡ ⎤
⎪
⎪ −1 −1
⎪
⎪ ⎢
0 0
⎥
0
⎢ ⎥
⎪
⎪ ⎢ ⎥ ⎢ ⎥
⎪
⎪ 1 0 0 ⎥ ⎢ 0 ⎥
⎨ F 1 (x, θ) = ⎢
⎪ 1
⎢ ⎥ x + ⎢ ⎥,
⎢ 0
⎣ 0 −1 0 ⎥ ⎦
⎢ x2 ⎥
⎣ 3⎦ (68)
⎪
⎪
⎪
⎪ −12 −24 0 −6 3
x1
⎪
⎪
⎪
⎪ T
⎪
⎩ G1 (x) = 0 0 0 2 + sin2 (x4 ) .
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 305

Based on the design of stabilizing sliding mode control law (31) of the primary subsystem (68):
1
u1 =  
2 + sin2 (x3 )
   
· 8x1 − 6x2 − 36x3 + 6x4 + k 1 sgn + ε1 (−3x1 + x2 + 5x3 − 6x4 ) . (69)

For the subsystem 2:

⎧
⎪
⎪ ξ̇ = ξ2 ,
⎪ 1
⎪  3
⎪
⎪   η 
⎪
⎨ ξ̇2 = −3ξ1 − 4ξ2 +
3
η2 + ξ1 + θ + 2 + sin2
2
u2 ,
2 2 (70)
⎪
⎪ η 2
⎪
⎪ η̇1 = ξ1 − η1 + 2 ,
⎪
⎪ 4
⎪
⎩
η̇2 = 2ξ1 + η2 .

The approximately linearized model of the internal dynamic of the subsystem 2 can be obtained
as:
⎡ ⎤ ⎡ ⎤
−1 1 10
η̇ = ⎣ ⎦η +⎣ ⎦ ξ. (71)
0 1 20

To introduce a virtual controller for the subsystem 2, let

S (2) = ξ1 + 2ξ2 . (72)

Thus, based on Assumption 3 and the development provided in Section 4, the subsystem (70)
can be rewritten as:
⎡ ⎤ ⎡ ⎤
ζ̇ ζ
⎣ ⎦ = A2 ⎣ ⎦ + B 2 u2 + Δ2 (ζ, S , θ) (73)
Ṡ S

with
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
2
A2ζ A2S 03 T δ (ζ, S)
A2 = ⎣ 2 2
⎦, B1 = ⎣ ⎦, ζ = ξ1 η1 η2 , Δ2 (ζ, S, θ) = ⎣ ⎦,
Aζ AS b2 (ζ, S ) φ2 (ζ, S, θ)

where
⎡ ⎤ ⎡ ⎤
−1/2 0 0 1/2
⎢ ⎥ ⎢ ⎥ 2 2
A2ζ = ⎢
⎣ 1 −1 1 ⎥
⎦, A2S = ⎢ ⎥
⎣ 0 ⎦, AS = [−2] , Aζ = 0 0 −9/2 ,
0 12 0
⎡ ⎤
0
2 ⎢ ⎥
δ (ζ, S) = ⎣ 0 ⎥
⎢ φ2 (ζ, S, θ) = 2 + (ξ1 + η2 3/2) .
2 3
⎦ , b2 (ζ, S ) = 2 + sin ( η2 /2) ,
0
306 KHALIL JOUILI

The virtual sliding surface is designed as:

⎡ ⎤
ξ1
⎢ ⎥
σ 2 = ξ1 + 2ξ2 + L2 ⎢ ⎥
⎣ η1 ⎦ . (74)
η2

We can model L2 = −1 −2 −5 to make the sliding mode have eigenvalues on the left half of
the complex plane.
Then, we ensure the approximate stability of the subsystem (70) in the neighborhood of
the equilibrium point. The subsystem (70) is written in the same form as that of equation
system (13) where
⎧ ⎡ ⎤ ⎡ ⎤
⎪
⎪ −1/2 0 0 1/2 0
⎪
⎪ ⎢ ⎥ ⎢ ⎥
⎪
⎪ ⎢ 1 −1 1 0 ⎥ ⎢ x2 ⎥
⎪
⎪ ⎢ ⎥ ⎢
⎪ 3 ⎥
⎨ F (x, θ) = ⎢
2
⎥x +⎢ ⎥,
⎢ 0 −2 1 0 ⎥ ⎢ 4x2 ⎥ (75)
⎪ ⎣ ⎦ ⎣ 3 ⎦
⎪
⎪
⎪
⎪ 19/2 0 3 −13/2 2x13
⎪
⎪
⎪
⎪ T
⎩ G2 (x) = 0 0 0 2 + sin2 (x3 ) .

Based on the design of stabilizing sliding mode control law (31) of the secondary subsystem (72):
1
u2 = −
2 + sin2 (x3 )
   
· −8x1 − 5x2 + 18x3 + 2x4 + k 2 sgn + ε2 (−2x1 + 2x2 − 12x3 − 2x4 ) . (76)

In order to validate the proposed approach, we performed the simulations on Matlab.

The simulation results are presented below:
Using the first diffeomorphism (61), the sliding mode control u1 introduced in (69) was
applied to the subsystem (63). With the control parameters k 1 = 0.450, ε1 = 0.323 and since
limx→0 Tx1 (x)
= 1, Assumption 4 was satisfied. The results of the input output states, the
virtual sliding variable, the control input and the internal dynamics are shown in Figures 1(a),
1(b), 1(c), 1(d), respectively. It can be seen that, by using the stabilizing sliding mode control
law, the internal dynamics are completely stabilized to fulfill the stability of the subsystem 1.
Using the second diffeomorphism (62), the sliding mode control u2 , proposed in (76), was
applied to the subsystem (70). With the control parameters k 2 = 0.632, ε2 = 0.197 and since
limx→0 Tx2 (x)
= 1, Assumption 4 was satisfied. The results of the input output states, the
virtual sliding variable, the control input and the internal dynamics are shown in Figures 2(a),
2(b), 2(c), 2(d), respectively. It can be seen that, by using the stabilizing sliding mode control
law, the internal dynamics are completely stabilized to fulfill the stability of the subsystem 2.
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 307

5 0.4
x1
x2
4 0.2

0
3

−0.2
Sliding variable

x1 and x2
−0.4
1
−0.6

0
−0.8

−1 −1

−2 −1.2
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time [S] Time [S]

(a) Evolution of the virtual sliding variable σ1 that (b) Evolution of the input output states x1 and x2
corresponds to the diffeomorphism T1 (x) that corresponds to the diffeomorphism T1 (x)

0.6 0.5
x3
x4 0
0.5

−0.5
0.4

−1
0.3
Control input u1

−1.5
x3 and x4

0.2
−2
0.1
−2.5

0
−3

−0.1 −3.5

−0.2 −4
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time [S] Time [S]

(c) Evolution of the internal states x3 and x4 that (d) Evolution of the stabilizing sliding mode control
corresponds to the diffeomorphism T1 (x) signal u1 (x) that corresponds to the diffeomorphism
T1 (x)
Figure 1 (a) Virtual sliding variable σ 1 ; (b) input output states x1 and x2 ; (c) internal
states x3 and x4 ; and (d) stabilizing sliding mode control signal u1 (x)
0 0.9
x1
0.8 x2
−0.5
0.7
−1
0.6
−1.5
Sliding variable

0.5
x1 and x2

−2 0.4

0.3
−2.5
0.2
−3
0.1
−3.5
0

−4 −0.1
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time [S] Time [S]

(a) Evolution of the virtual sliding variable σ2 that (b) Evolution of the input output states x1 and x2
corresponds to the diffeomorphism T2 (x) that corresponds to the diffeomorphism T2 (x)
308 KHALIL JOUILI

0.5 9
x3
0.45 x4 8

0.4 7

0.35
6

Control input u2
0.3
5
x3 and x4

0.25
4
0.2
3
0.15
2
0.1

0.05 1

0 0
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time [S] Time [S]

(c) Evolution of the internal states x3 and x4 that (d) Evolution of the stabilizing sliding mode con-
corresponds to the diffeomorphism T2 (x) trol signal u2 (x) that corresponds to the diffeomor-
phism T2 (x)
Figure 2 (a) Virtual sliding variable σ 2 ; (b) input output states x1 and x2 ; (c) internal
states x3 and x4 and (d) stabilizing sliding mode control signal u2 (x)

Using multi-diffeomorphism, simulation results are shown in Figures 3(a), 3(b), 3(c), 3(d),
respectively. It is clearly observed that the proposed control scheme, using the concept of
multi-diffeomorphism, exhibits an excellent performance and strong robustness with respect
to uncertainty. Therefore, we generate the trajectory that has better transient by employing
multi-diffeomorphism, as shown in Figure 3(b). Indeed, an analysis of Figure 3(b), which
corresponds to the generated trajectory utilizing the concept of multi-diffeomorphism, shows a
clearer improvement of the stabilization speed around zero, compared with those resulting from
only one diffeomorphism represented in Figure 1(c) and Figure 2(c). Besides, we see that the
convergence rate of the internal state is faster and smoother, as shown in Figure 3, than that
obtained by applying only one diffeomorphism represented in Figure 1 and Figure 2.
0.25 0.6
x1 x3
x2 x4
0.2 0.5

0.15 0.4

0.1 0.3
x1 and x2

x3 and x4

0.05 0.2

0 0.1

−0.05 0

−0.1 −0.1
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time [S] Time [S]

(a) Evolution of the input output states x1 and x2 (b) Evolution of the internal states x3 and x4 by us-
by using the concept of multi-diffeomorphism (T1 (x) ing the concept of multi-diffeomorphism (T1 (x) and
and T2 (x)) T2 (x))
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 309

10

5 2

The i−th subsystem

1.5
−5
Control input

−10
1
−15

−20 0.5

−25

0
−30 0 1 2 3 4 5 6 7 8 9 10
0 1 2 3 4 5 6 7 8 9 10 Time [S]
Time [S]

(c) Evolution of the stabilizing sliding mode (d) Switching signal

control signal by using the concept of multi-
diffeomorphism (T1 (x) and T2 (x))
Figure 3 (a) input-output states x1 and x2 ; (b) internal states x3 and x4 ; (c) stabilizing
sliding mode control signal and (d) switching signal

6 Conclusion
In this paper, the stabilization problem for non-minimum phase switched nonlinear systems
with uncertain dynamics was addressed. Using the exact input output feedback linearization
method, sliding mode control technique and the Lyapunov stability theory, we designed a robust
sliding mode control approach that guaranteed the global robust asymptotic stability of the
resulting switched nonlinear system. We also elaborated a strategy of switching that recourse to
the concept of multi-diffeomorphism and we demonstrated that this strategy allowed obtaining
more improved the transient state, compared to a feedback linearization based on only one
diffeomorphism. Finally, we showed, by means of a numerical example and its simulation, that
the switching strategy can generate a trajectory with better transient behavior, compared to
that produced by using only one diffeomorphism, and a stronger robustness with respect to
uncertainty. In our future work, we will extend the proposed methodology to non-minimum
phase nonlinear multi-inputs and multi-outputs systems.

References

[1] Long L J, Wang Z, and Zhao J, Switched adaptive control of switched nonlinearly parameterized
systems with unstable subsystems, Automatica, 2015, 45: 217–228.
[2] He Z W and Xie W, Control of non-linear switched systems with average dwell time: Interval
observer-based framework, IET Control Theory Appl., 2016, 10: 10–16.
[3] Wu D, Sun X M, Zhao Y B, et al., Stability analysis of nonlinear switched networked control
systems with periodical packet dropouts, Circuits, Syst. Signal Process, 2012, 32: 1931–1947.
[4] Zhao J, Hill D J, and Liu T, Synchoronization of complex dynamical networks with switching
switching topology: A switched system point of view, Automatica, 2009, 45: 2502–2511.
310 KHALIL JOUILI

[5] Dayawansa W P and Martin C F, A converse Lyapunov function theorem for a class of dynamical
systems wich undergo switching, IEEE Trans. Autom. Control, 1999, 44(4): 751–760.
[6] Homaee O, Zakariazadeh A, and Jadid S, Real-time voltage control algorithm with switched
capacitors in smart distribution system in presence of renewable generations, Int. J. Electr.
Power Energy Syst., 2014, 54: 187–197.
[7] Ma R C and Zhao J, Backstepping design for global stabilization of switched nonlinear systems
in lower triangular form under arbitrary switchings, Automatica, 2010, 46: 1819–1823.
[8] Niu B, Zhao X D, Fan X D, et al., A new control method for state-constrained nonlinear switched
systems with application to chemical process, Int. J. Control, 2015, 88(9): 1693–1701.
[9] Wang X and Zhao J, Switched adaptive tracking control of robot manipulators with friction and
changing loads, Int. J. Syst. Sci., 2015, 46(6): 955–964.
[10] Zhang J F, Han Z Z, Zhu F B, et al., Absolute exponential stability and stabilization of switched
nonlinear systems, Syst. Control Lett., 2014, 66: 51–57.
[11] Zhang Q, Wu Q X, Jiang C S, et al., Robust adaptive control for a class of uncertain switched
nonlinear systems with application to near space vehicle, Proceedings of the 31st Chinese Control
Conf., 2012 (in Chinese).
[12] Zhai D, Lu A Y, Dong J X, et al., Dynamic output feedback H1 control for switched T-S fuzzy
systems via discretized Lyapunov function technique, Neurocomputing, 2016, 177: 651–669.
[13] Liu L, Zhou Q, Liang H, et al., Stability and stabilization of nonlinear switched systems under
average dwell time, Applied Mathematics and Computation, 2017, 298: 77–94.
[14] Li X, Lin X, Li S, et al., Globally smooth output feedback stabilization of a class of planar switched
systems with average dwell time, Nonlinear Analysis: Hybrid Systems, 2017, 24: 159–170.
[15] Liang X L, Hou M Z, and Duan G R, Output feedback stabilization of switched stochastic
nonlinear systems under arbitrary switchings, Int. J. Autom. Comput., 2013, 10: 571–577.
[16] Long L and Zhao J, Global stabilization for a class of switched nonlinear feedforward systems,
Syst. Control Lett., 2017, 60: 737–738.
[17] Hou Z G and Wu Y, Finite-time control for switched delay systems via dynamic output feedback,
International Journal of Innovative Computing Information and Control, 2012, 8: 4901–4913.
[18] Liang Y J, Ma R, Wang M, et al., Global finite time stabilisation of a class of switched nonlinear
systems, International Journal of Systems Science, 2015, 46: 2897–2904.
[19] Liberzon D, Switching in Systems and Control, Springer-Verlag, New York, 2003.
[20] Lin J, Fei S, and Gao Z, Stabilization of discrete-time switched singular time-delay systems under
asynchronous switching, Journal of the Franklin Institute, 2012, 349: 1808–1827.
[21] Valentino M C, Oliveira V A, Alberto L F C, et al., An extension of the invariance principle for
dwell time switched nonlinear systems, Systems and Control Letters, 2012, 61: 580–586.
[22] Yang H, Jiang B, Cocquempot V, et al., Stabilization of switched nonlinear systems with all
unstable modes: Application to multi-agent systems, IEEE Transactions on Automatic Control,
2011, 56: 2230–2235.
[23] Isidori A, Nonlinear Control Systems: An Introduction, Springer-Verlag, New York, 1995.
[24] Zhao X, Wang X, Zong G, et al., Fuzzy-approximation-based adaptive output-feedback control
for uncertain non-smooth nonlinear systems, IEEE Transactions on Fuzzy Systems, 2018, 56:
1–12.
[25] Zhao X, Wang X, Zong G, et al., Fuzzy adaptive control design and discretization for a class of
nonlinear uncertain systems, IEEE Transactions on Cybernetics, 2016, 46: 1476–1483.
 STABILIZATION OF NON-MINIMUM PHASE SWITCHED SYSTEMS 311

[26] Chang X H, Park J H, and Shi P, Fuzzy resilient energy-to-peak filtering for continuous time
nonlinear systems, IEEE Transactions on Fuzzy Systems, 2017, 25: 1576–1588.
[27] Chang X H and Wang Y M, Peak-to-peak filtering for networked nonlinear DC motor systems
with quantization, Transactions on Industrial Informatics, 2018, 25: 1–13.
[28] Chang X H, Li Z M, and Park J H, Fuzzy generalized H2 filtering for nonlinear discrete-time
systems with measurement quantization, IEEE Transactions on Systems, Man, and Cybernetics:
Systems, 2017, 1–10.
[29] Wang M, Dimirovski G M, and Zhao J, H∞ control for a class of non-minimum phase cascade
switched nonlinear systems, Proceedings of the American Control Conference (USA), Seattle,
2008.
[30] Jouili K and Benhadj B N, Stabilization of non-minimum phase switched nonlinear systems
with the concept of multi-diffeomorphism, Communications in Nonlinear Science and Numerical
Simulation, 2015, 23: 282–293.
[31] Oishi M and Tomlin, Switching in non-minimum phase systems: Applications to a VSTOL
aircraft, Proceedings of the American Control Conference (USA), Chicago, 2000.
[32] Yang H, Jiang B, and Zhang H, Stabilization of non-minimum phase switched nonlinear systems
with application to multi-agent systems, Syst. Control Lett., 2012, 61: 1023–1031.
[33] Benosman M and Lum K Y, Output trajectory tracking for a switched nonlinear non-minimum
phase system, Proceedings of the 16th IEEE International Conference on Control Applications
(Singapore), Singapore, 2007.
[34] Chiang M L and Fu L G, Adaptive stabilization of a class of uncertain switched nonlinear systems
with backstepping control, Automatica, 2014, 50: 2128–2135.
[35] Zhao X D, Zheng X L, Liu B, et al., Adaptive tracking control for a class of uncertain swicthed
nonlinear systems, Automatica, 2015, 52: 185–191.
[36] Khalil H K, Nonlinear Systems, Prentice, New Jersey, 2002.
[37] Do M T, Man Z, Jin J, et al., Sliding mode learning control of non-minimum phase nonlinear
systems, Int. J. Robust Nonlinear Control, 2016, 26: 2281–2298.