Proceedings of the 5th International Conference of Control, Dynamic Systems, and Robotics (CDSR’18)

Niagara Falls, Canada - June 7 - 9, 2018

Paper No. XXX (The number assigned by the OpenConf System)
ISSN: TBA
DOI: TBA
Static anti-windup compensator design for nonlinear parameter varying
systems

Najam us Saqib1 , Muntazir Hussain1 , Muhammad Siddique1 , Muhammad Rehan1 , Naeem Iqbal1
1 Department
of Electrical Engineering,
Pakistan Institute of Engineering and Applied Sciences, Islamabad.
nsaqib83@gmail.com; muntazir hussain14@yahoo.com; engr.siddique01@gmail.com; rehan@pieas.edu.pk;
naeem@pieas.edu.pk

Abstract - This article presents a method for designing the global anti-windup compensator (AWC) gain for nonlinear systems by using
the nomenclature of nonlinear parameter varying (NPV) theory. The sufficient conditions for the existence of static AWC for NPV system
under actuator saturation in the presence of exogenous input and actuator constraints guaranteeing the global asymptotic stability of the
overall closed-loop system are formulated. Linear matrix inequality (LMI)-based conditions are derived by using Lyapunov theory, NPV
theory, global sector condition, minimum and maximum bound on the saturation non-linearity, and parametric variations limits in order to
synthesis the AWC which ensure global asymptotic stability of the overall closed-loop system. The simulation example of the nonlinear
induction motor is provided to show the effectiveness of the suggested AWC methodology.

Keywords: Anti-windup compensator (AWC), Nonlinear parameter varying (NPV), Globally asymptotically stable, Nonlinear system

1. Introduction
In most situations, controller is designed without considering actuator saturation. Thereby reckoning the actuators
huge enough so that, the input energy demanded by the control law does not cross the saturation limits. However large
actuator assumption not only increase the cost of the overall control system, but it will also cause to increase the size of the
control system. Which may cause problem in an application such as in aerospace, where large actuators causes to increase
the mass of the aircraft and consequently fuel costs. Likewise, over sized actuators in power control systems lead to power
losses and efficiency degradation. Further, in a physical control system, it is not commonly possible to guarantee that all the
control signals are small and especially for extraordinary performance applications, substantial controller design is frequently
necessary. Unfortunately, the actuators cannot deliver unlimited control signal in practical applications and are invariably
subjected to magnitude limitations. Voltage restrictions in electrical actuators, deflection limitation in aircraft actuators, and
limits on a rate in hydraulic actuators are the common examples of actuator saturation.
Obviously, these limitations restrict the performance of the overall control system. If these constrained are not handled
carefully in the controller design, peculiar behavior may be observed [1, 2, 3]. Further, it may cause financial losses and
disaster (see [2] for detail). The controller design without taking actuator saturation into account causes to produce an
annoying effect commonly known as integral windup when the manipulated control signal crosses the saturation level. The
integral windup not only causes to affect the performance of the closed-loop system, but it also leads to destabilizing the
system. Historically, the saturation effect is initially faced in the 1940s and it was pointed out that saturation effect is one of
the core phenomena affecting the absolute stability of the closed-loop system.
So far, various techniques have been proposed to mitigate the windup effect in a linear system [4, 5], cascaded control
system [6, 7], a nonlinear system [8, 9], and nonlinear time-delay system [2, 10, 11]. In general, the saturation compensation
techniques can be classified into two categories. Firstly, one-step approach, as indicated by its name, a methodology to
controller design where a multi-objective controller ensures all nominal performance specifications and absolute stability
requirement along with handling the saturation restriction enforced by the actuators. However, it is usually picked apart
because of its lack of tuning rules, conservatism, and deficiency of applicability to some real-world control system. The
second approach usually known as anti-windup compensation (AWC) that provides a systematic way to compensate the
saturation effect in control system. In this approach, the nominal controller is designed to achieve the performance and
stability requirement, without taking into account the saturation effects. Then AWC is augmented in the closed-loop system
to mitigate the integral windup effect without affecting the overall stability of the closed-loop system. This approach is
attractive to the practicing engineers in practice because it does not restrict the performance of the nominal controller and,
it does not affect the closed-loops behavior in absence of saturation. The AWC become active only when saturation is
encountered and acts to adjust the controller behavior such that the effect saturation can be mitigated. The saturating inputs
cause the error signal grow large and the integral action of the controller accumulates the error and is called windup effect

XXX-1
[4, 12, 13]. The AWC is employed to remove the windup effects caused by the saturating effect. The AWC operates when
difference between the unsaturated control signal and the saturated control signal is nonzero. The difference is then amplified
through the proposed anti-windup compensator gain and is fed to the controller designed without considering the saturation
effects. Gigantic literature is available for AWC design for linear system. However, the AWC design for nonlinear system
is still a challenging task. Static AWC design for NPV is addressed by [3], however it uses a simple L2 reduction using
Lyapunov function. Also the matrix inequalities contain a nonlinear terms.
In this brief, the sufficient condition for the existence of static AWC are proposed for nonlinear parameter varying
(NPV) systems ensuring the global stability of the overall closed-loop system. AWC is employed in order to mitigate the
windup effects caused by the saturating inputs. Nominal output feedback controller is designed, which meets the desired
performance objective in the absence of input saturation. Sufficient conditions are derived in the form of LMI in order to
calculate the optimal values of the static AWC which ensures the closed-loop system is globally asymptotically stable. An
LMI-based conditions are derived for ciphering anti-windup gain for NPV system by using the Lipschitz condition, global
sector condition, Lyapunov function and L2 gain reduction for exogenous signals. To validate the results of the derived
LMI-based condition, a simulation example of nonlinear induction motor is also provided. The technique proposed here
is computationally simpler than the one provided in [3] because the nonlinear term have been reduced to linear term by
application of Schur complement. Furthermore, a simple algorithm is also provided to compute static AWC gain. A more
general term for L2 gain has been used which reduces the conservatism in the design.
The paper is organized as follow. Section 2 gives the system description along with assumption. Section 3 give the
proposed state feedback controller along with static AWC. Section 4 provides the necessary condition for the existence of
AWC and Section 4 provides a simulation results to validate results of the proposed methodology. Lastly Section 5 concludes
the article.
2. System description
The time-varying nonlinear systems is considered depicted by following state space representation
ẋ = A p (δ (t))x + B p (δ (t))usat + Bw (δ (t))w + f (t, x),
z = Cz (δ (t))x + Dz (δ (t))w, (1)
y = Cy (δ (t))x + Dy (δ (t))w,
where x ∈ Rn corresponds to the state vector, usat ∈ Rm represent the saturated input, w ∈ R p represent the exogenous in-
put (compromising of single or multiple reference signal, noise and disturbances), z ∈ Rk represents the exogenous output,
f (t, x) ∈ Rn is nonlinearity and y ∈ Rq symbolizes the measured output. A p (δ (t)), B p (δ (t)), Bw (δ (t)), Cz (δ (t)), Dz (δ (t)),
Cy (δ (t)), Dy (δ (t)) represent the parameter varying matrices of appropriated dimensions bearing parameter ϑ ∈ Rs fulfilling
Z = {ϕ ∈ Rs ; ϕh ∈ [ϑmin(h) , ϑmax(h) ]}, ∀h = 1, 2, · · · , s, (2)
where ϑmin(h) and ϑmax(h) are the time-varying parameter ϑ(h) upper and lower bounds, respectively. The δ (t) represents
a time-varying parameter of the nonlinear system (1). The vector ϕ corresponds to all potential values of parameter δ (t)
represented by the convex set Z.
Assumption 1. The function f (t, x) ∈ Rn ∀ x1 , x2 ∈ χ ⊆ Rn fulfills the conditions f (t, 0) = 0, ∀ t ≥ 0 and k f (t, x1 ) −
f (t, x2 )k ≤ kL(x1 − x2 )k, where L represents a constant matrix of suitable dimension.
Accordingly, the system (1) can be further re-written in generalized form as
ẋ = A p (ϕ)x + B p (ϕ)usat + Bw (ϕ)w + f (t, x),
z = Cz (ϕ)x + Dz (ϕ)w, (3)
y = Cy (ϕ)x + Dy (ϕ)w, ∀ϕ ∈ Z.
3. State feedback controller with AWC
For the nonlinear system (1), there exists an output feedback controller which guarantees craved closed-loop perfor-
mance and stability requirements without taking into consideration the input saturation nonlinearity is given as
ẋc = Ac xc + Bc y,
(4)
u = Cc xc + Dc y,
where xc ∈ Rc denotes the state of controller and Ac , Bc , Cc , and Dc are known and constant matrices of appropriate di-
mensions. However, when the control input u is subjected to constraints, to mitigate the windup consequences, an AWC is
augmented with the controller. As a result, the overall controller containing static AWC becomes
ẋc = Ac xc + Bc y + Ec (usat − u),
(5)
u = Cc xc + Dc y,

XXX-2
where AWC gain is represented by Ec . The closed-loop system compromising of (3) and (5) can be written as

ξ˙ =A(ϕ)ξ + T(ϕ)w − (B(ϕ) + REc )Ψ + F(ϕ) f (x),

z =J(ϕ)ξ + Dz (ϕ)w, (6)
u =K(ϕ)ξ + Dc Dy (ϕ)w, ∀ϕ ∈ Z
   
x A p (ϕ) + B p (ϕ)DcCy (ϕ) B p (ϕ)Cc B p (ϕ) I
h i h i
ξ = xc , A(ϕ) = , B(ϕ) = 0 , F(ϕ) = 0 , T(ϕ) =
BcCy (ϕ) Ac
 
Bw (ϕ) + B p (ϕ)Dc Dy (ϕ) 0
h i
,R= I , K(ϕ) = [ DcCy (ϕ) Cc ], J(ϕ) = [ Cz (ϕ) 0 ], Ψ = u − usat .
Bc Dy (ϕ) c×c
The classical global sector condition [12] for the saturation nonlinearity can be expressed as

ΨT W (u − Ψ) ≥ 0, (7)

where W ∈ Rm×m is a diagonal positive definite matrix. The condition (7) can be used for achieving the global stability of a
linear or a nonlinear closed-loop system be selecting a suitable value of W .

4. Anti-windup Design
In this section, various sufficient conditions for synthesis of the AWC gains for global stability of the overall closed-loop
system with exogenous input are derived.
Theorem 1: Consider the NPV system (3) satisfying the Assumption 1. Suppose there exist a symmetric matrix Q > 0,
S > 0, R > 0, diagonal matrix U > 0, and a matrix V of appropriate dimensions such that the set of LMIs given by

Γ1 (ϕ) T(ϕ) Γ2 (ϕ) F QJT QLT

 
 ? T T
−S Dy (ϕ)Dc 0 Dz (ϕ) T 0 
 ?
 ? −2U 0 0 0 
 < 0, ∀ ϕ ∈ Z, (8)
 ? ? ? −I 0 0 
? ? ? ? −R
 
0
? ? ? ? ? −I
where,
Γ1 (ϕ) = QAT (ϕ) + A(ϕ)Q,
Γ2 (ϕ) = QKT − B(ϕ)U − RV ,
is satisfied. Then there exists a controller of the form (5) such that
1. If w = 0, the closed-loop system is globally asymptotically stable,
2. If w ∈ L2 , the L2 gain from w to z is bounded.
The anti-windup gain matrix can be ciphered as Ec = VU −1 .
Proof :Consider a quadratic Lyapunov function as

V (t, ξ ) = ξ T Pξ , P > 0. (9)

For stability of (6), V̇ (t, ξ ) < 0 is required. Let us define Jz1 by using Assumption 1 and (7) to attain stability of closed-loop
system (6).
Jz1 = V̇ (t, ξ ) + zT R−1 z − wT Sw < 0. (10)
By application of Assumption 1, (7), and (10) and, further, defining

Jz2 = V̇ (t, ξ ) + ΨT W (u − Ψ) + (u − Ψ)T W Ψ + ξ T LT Lξ − f T (t, x) f (t, x) + zT R−1 z − wT Sw < 0, (11)

it can be validated that Jz2 < 0 ensures Jz1 < 0. Further simplification reveals

Jz2 = ξ T (AT (ϕ)P + PA(ϕ))ξ + ξ T PT(ϕ)w + wT TT (ϕ)Pξ − ξ T P(B(ϕ) + REc )Ψ − ΨT (B(ϕ) + REc )T Pξ
+ ξ T PF f (t, x) + f T (t, x)FT Pξ + zT R−1 z − wT Sw + ΨT W (Kξ + Dc Dy (ϕ)w − Ψ) + ξ T LT Lξ (12)
T T
+ (Kξ + Dc Dy (ϕ)w − Ψ) W Ψ − f (t, x) f (t, x) < 0,

XXX-3
 T
which further implies Jz2 = φ2T Ωφ2 < 0, where φ2 =

ξT wT ΨT f T (t, x) and
 
Γ3 (ϕ) Γ4 (ϕ) Γ5 (ϕ) PF
 ? Γ D T (ϕ)DT W 0  < 0, ∀ϕ ∈ Z
Ω = 6 y c (13)
? ? −2W 0

? ? ? −I

Γ3 (ϕ) = PA(θ ) + AT (θ )P + LT L + JT (ϕ)R−1 J(ϕ),

Γ4 (ϕ) = PT(ϕ) + JT (ϕ)R−1 Dz (ϕ),
Γ5 (θ ) = KT W − P(B(θ ) − REc ),
Γ6 (θ ) = −S + DTz (ϕ)R−1 Dz (ϕ).
By rearranging the terms in order to avoid the nonlinear terms after the application of the Schur complement and congruence
transform by employing pre and post multiplication of diag(P−1 , I,W −1 , I) to the set of inequalities in (13), and substituting
Q = P−1 and U = W −1 the LMIs in (8) are obtained. Integrating (10) from 0 to T → ∞, we have
Z T Z T Z T
T −1
Jz1 dt = (V (T, ξ ) −V (0, ξ )) + z R zdt − wT Swdt < 0, (14)
0 0 0

the following entailments can be drawn from the above inequality. Firstly, when w = 0, (10) implicates that V̇ (t, ξ ) < 0,
which in turns ensures asymptotic stability of the closed-loop system (6) at the origin. Secondly, when ξ (0) = 0 which
ensures V (0, ξ ) = 0, then V (T, ξ ) > 0 and (14) ensures, the L2 gain between w and z is always bounded, which completes
the proof. 
Remark 1: The recent studies like [8, 14, 15, 16] consider a different form of nonlinear systems and provide suffi-
cient design conditions for the existence of dynamic AWC. From performance and stability perspective, these proposed AWC
design schemes are attractive because their existence can be aboveboard for the stable systems and dynamic AWC can be
enforced to accomplish several performance and stability objectives. However, from implementation point of view these
schemes may not be practical due to extra software and hardware resources requirement and need of exhaustive implementa-
tion efforts from the researchers. In contrast, the present approach in Theorem 1 is straightforward in implementation owing to
the inherent simplicity of the static AWC. In addition, the conservatism’s like observability of the nonlinear plant as observed
in [14] are addressed in the present study.
Remark 2: The proposed AWC design methodology in Theorem 1 studies global stability of the closed-loop system
(6) against the exogenous input signal w. The L2 gain reduction from w to z is guaranteed. The exogenous input w can
contain the reference signal, noise and disturbance, therefore, the approach in Theorem 1 can also be enforced to design an
anti-windup gain for reference tracking, disturbance rejection and noise handling applications.
As seen in [3], an algorithm is provided for figuring out the LMIs of Theorem 1 to obtain the AWC gain matrix Ec .
By considering the variation in parameters be specified as (ϕ1 ∈ [ϕ11 , ϕ1q ], · · · , ϕr ∈ [ϕs1 , ϕsq ]). By reckoning the discrete
interval for all parameter as (ε1 , ε2 , ..., εs ) where εi = (ϕiq − ϕi1 )/q and qs refers to the total points counted.
Algorithm 1: The following steps are employed in figuring out the the matrix inequalities proposed in Theorem 1: (see
also [3])
1. Set i = 1, j = 1.
2. Build the matrix inequalities of Theorem 1 by portioning ϕ = ϕi j .
3. Increase j. If j ≤ q, cipher ϕi j = ϕi( j−1) + εi and proceed to Step 2), or else define j = 1.
4. Increase i. If i ≤ s, proceed to Step 2).
5. Figure out the matrix inequalities incurred in Step 2) for obtaining the variables R, Q, S, U, κ, V , µ and H(ϕ).

6. Calculate Ec by figuring out Ec = VU −1 .

Remark 3: The approach considered here is more general than the one originally presented in [3] by employing more
general L2 gain reduction inequality. Secondly, a computationally simple design approach is presented due to application of
Schur complement, as it eliminates the quadratic matrices containing the nonlinear parameter ϕ, which results in simplifica-
tion of Algorithm 1 compared to the one proposed in [3].
The global stability criterion is presented in the present study and the local stability result are considered as the extension of
the presented work.

XXX-4
 Remark 4: In the literature, various simultaneous dynamic controller and static AWC design approaches for attaining
the multi-objective synthesis are available (see [14, 17, 18]). These simultaneous design techniques are interesting as the
design approach can deal with the input saturation more efficiently owing to the consideration of both controller and AWC
for handling the windup consequences. However, the AWC-based controller design is challenging due to simultaneous com-
putation of the controller and compensator parameters and the technique cannot be applied to the complex controller and
AWC design scenarios. Additionally, the controller design becomes difficult for attaining various objectives like tracking
and robustness due to requirement of rigorous design efforts. In contrast, the present work provides a simple AWC design
approach, which can be applied to attain multi-objective synthesis and requires a simple unconstrained controller based on
the conventional control schemes. Due to availability of large number of unconstrained controller design techniques, the
proposed static AWC can be easily employed to the tracking and robustness applications. The approach in Theorem 1 can
also be employed to modify an existing controller of a practical control system against the windup phenomenon and it can be
employed along with the conventional control structures.

5. Simulation results
In this section, the proposed methodology is verified through the simulation results using the nonlinear induction motor.
5.1. Example 1
The induction motor nonlinear model can be represented by the following set of differential equations (see [19] and the
references therein)
h R
s Rr (1 − σ ) i e h Rr Lm e p i Lm Rr e 1 e
i̇eds = − + ids + ∗ iqs + ω rm ieqs + ψdr + v ,
σ Ls σ Lr Lr ψdr e 2 σ L L
s r
2 σ Ls ds
hR L
r m e p i Rs e h Rr Lm e p i L
m 1 e
i̇eqs = − ∗ i qs + ω i
rm ds
e
− i qs − ∗ i qs + ω rm
e
ψdr + v ,
Lr ψdre 2 σ Ls e
Lr ψdr 2 σ Lr Ls σ Ls qs
(15)
e Rr Lm e Rr e
ψ̇dr = i − ψ ,
Lr ds Lr dr
3pLm ψdr e?
Bm 1
ω̇rm = ieqs − ωrm − TL ,
2Lr Jm Jm Jm
where ieds represents stator current of d s -axis in the synchronous reference frame, ieqs denotes stator current of qs -axis in
the synchronous reference frame, ψdr e is the rotor flux of d e -axis in the synchronous reference frame, ω
rm stands for the
mechanical angular speed, Rs denotes the stator resistance, Ls represents the stator inductance, Rr refers to the rotor resistance,
Lr corresponds the rotor resistance, Lm is the mutual inductance of rotor and stator, Jm is the motors inertia, Bm symbolizes
the friction coefficient, TL represents load torque, p represents the number of poles pairs, veds is the stator voltage of d s -axis
in the synchronous frame, veqs denotes the stator voltage of qs -axis in the synchronous reference frame, ψdr e? is the rotor flux
2
command of d e -axis in the synchronous reference frame, and σ = 1 − LLs mLr is the total flux leakage coefficient, respectively.
The mechanical angular speed of motor is measured in rad/sec. The errors and the exogenous output of the system is defined
as follow i  e? 
h
e1 ψ − ψ e
z= e = dr
? −ω dr , (16)
2 ωrm rm

where ωrm? is the mechanical angular speed command. The various parameters of the nonlinear induction motor are repre-
sented in the Table. 1. Motor inertia is considered to be varied in the limits defined as Jm ∈ [ 0.00378, 0.00783 ]. The time
Table 1: Nonlinear induction motor parameters
Parameter Value Parameter Value
Rr (Ω) 0.775 Lm (H) 0.06547
Rs (Ω) 0.629 Bm (m/rad.sec) 0.00534
Ls (H) 0.06892 p(Pole pair) 2
Lr (H) 0.06892 usat (V) 130

varying parameter of the (15) is considered as δ (t) = Jm−1 . Hence the parameter ϕ which includes all potential appraises of
the parameter δ (t) for which the convex set (2) is determined as follow.
Zv = {ϕ ∈ R; ϕ ∈ [127.71, 264.55]}, (17)

XXX-5
by employing the time varying parameter ϕ, the nonlinear induction motor model (15) can be converted in to (3), and the
system matrices defined as follow
 −197.456 0 1587.9 0   148.648 0   0 0 0 
0 −93.499 0 0 0 148.648  0 0 0
A(ϕ) =  0.736 0 −11.2449 0
, B = 
0 0 , Bw =  0 0 0 , Dy =
0 0.284ϕ 0 −0.0053ϕ 0 0 0 0 −ϕ 
7.362x 2 +x x
2 2 4
0 0 0 1 0 0 0 0 1 0  −7.36x1 x2 − 1039x2 x3 − x1 x4 − 141.2x3 x4 
h i h i h i
0 0 0 , Dz = 0 1 0 , Cy = −Cz = 0 0 0 1 , f (t, x) =  0 .
0
The state feedback controller is designed by utilizing Matlab’s Simulink Response Optimization Toolbox, using the following
specifications (rise time 2 sec, settling time 4 sec, %overshoot 10%) for the mechanical angular speed output and (rise time
2 sec, settling time 4 sec, %overshoot 20%) for rotor flux output, respectively. The controller thus obtained of the form (4)
havingh the controller gain matrices hare as follow
−0.955 0.191 −1.714 0.053 0.199 0.522 −4.157 0.0112
i i h i h i
Ac = 0.148 −0.017 , Bc = −3.158 −0.342 , Cc = 0.0281 −0.621 , Dc = −0.499 −0.012 .
In order to ensure the tracking performance of the closed-loop system compromising of the nonlinear induction motor (15)

140
800
0

105
600
-35
(rad/sec)

(rad/sec)
x4 (rad/sec)

70
400
-70
vds

vds
e

e
35
200 -105

Reference Input
0
Mechanical angular speed (rad/sec)
0 -140
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Time (sec) Time (sec) Time (sec)

(a) ωrm (b) veds (c) veqs

Figure 1: Simulation response of nonlinear induction motor under square reference signal

and the controller obtained above of the form (4), a stair case signal is applied. The mechanical angular speed tracks the
reference input as represented in Figure 1a. The corresponding input control signals are represented in Figure 1b and 1c. The
LMI of the Theorem 1 is solved and the anti-windup gain is ciphered as

−0.75853 −4.8469
h i
Ec = −2.20766 91.7212 (18)

In order to observe the effects of windup the square wave reference input shown in Figure 2a is applied to the nonlinear
induction motor. The output without the proposed AWC shows the lag in following the reference input. The windup effect
can be clearly seen from Figure 2b and Figure 2c since the control signal are not following the reference signal. By using the
proposed AWC, clearly the performance of the nonlinear induction motor regarding the following of the reference input has
been improved, which is shown in Figure 2a. The windup effect has been taken out by when the proposed AWC is employed
which is clearly shown in the saturated control signal Figure 2b and Figure 2c. Furthermore, variation in the time varying
parameter is also considered, by changing the inertial load. The maximum inertial load is added to the motor shaft at t = 0sec
and then the inertial load is removed at t = 20sec. The constant reference signal of 700(rad/sec) is used. The mechanical
angular speed does not follow the reference input even when the load is removed. The input control signals remains saturated
even after the load is removed due to the windup effect, which is distinctly seen in the Figure 3b and Figure 3c. By using
the proposed AWC the mechanical angular speed follows the reference signal when the load is removed as shown in Figure
3a. However, when one of the control input gets saturated the proposed scheme extenuates the effects of saturation by not
allowing the other control inputs to get saturated as well. Similarly the windup effects is removed which can be clearly seen
from the Figure 3b and Figure 3c. Hence simulation study reveals that the suggested AWC removes the windup effects caused
from the saturating effects and the parametric variations.

XXX-6
 1200
140
Saturated control input without AWC
Saturated control input with AWC
0
1000

105

800 -35

(rad/sec)

(rad/sec)
x4 (rad/sec)

70
600
-70

vds

vds
e

e
400 35

-105
200
Reference Input
Mechanical angular speed without AWC
0 Saturated control input without AWC
Mechanical angular speed with AWC Saturated control input with AWC
0 -140
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Time (sec) Time (sec) Time (sec)

(a) ωrm (b) veds (c) veqs

Figure 2: Simulation response of nonlinear induction motor for square reference signal

140
Saturated control input without AWC
800
Saturated control input with AWC
0

105
600
-35
(rad/sec)

(rad/sec)
x4 (rad/sec)

70
400
-70
vds

vds
e

e
35
200 -105

Reference Input
Mechanical angular speed without AWC
0 Saturated control input without AWC
Mechanical angular speed with AWC Saturated control input with AWC
0 -140
0 10 20 30 40 0 10 20 30 40 0 10 20 30 40
Time (sec) Time (sec) Time (sec)

(a) ωrm (b) veds (c) veqs

Figure 3: Simulation response of nonlinear induction motor under load variation

6. Conclusion
This note presented a method for computing static anti-windup gain for parameter varying non-linear systems by
using the nomenclature of NPV system. The sufficient conditions for the existence of static AWC for NPV system under
actuator saturation in the presence of exogenous input were formulated by the application Lyapunov theory, global sector
condition, Lipschitz nonlinearity and NPV theory which ensures the global asymptotic stability of the closed-loop system.
An algorithm was also proposed for solving the obtained matrix inequalities. Furthermore, the simulation example of the
nonlinear induction motor was furnished to demonstrate the effectiveness of the suggested AWC scheme.

References
[1] S. Tarbouriech and M. Turner, “Anti-windup design: an overview of some recent advances and open problems,” IET
control theory & applications, vol. 3, no. 1, pp. 1–19, 2009.
[2] M. Hussain and M. Rehan, “Nonlinear time-delay anti-windup compensator synthesis for nonlinear time-delay systems:
A delay-range-dependent approach,” Neurocomputing, vol. 186, pp. 54–65, 2016.
[3] N. us Saqib, M. Rehan, N. Iqbal, and K.-S. Hong, “Static antiwindup design for nonlinear parameter varying systems
with application to dc motor speed control under nonlinearities and load variations,” IEEE Transactions on Control
Systems Technology, 2017.
[4] A. Saberi, Z. L. Lin, and A. R. Teel, “Control of linear systems with saturating actuators,” IEEE Transactions on
Automatic Control, vol. 41, no. 3, pp. 368–378, 1996.
[5] G. Grimm, J. Hatfield, I. Postlethwaite, A. R. Teel, M. C. Turner, and L. Zaccarian, “Antiwindup for stable linear systems
with input saturation: an LMI-based synthesis,” IEEE Trans. on Autom. Control, vol. 48, no. 9, pp. 1509–1525, 2003.
[6] N. Mehdi, M. Rehan, F. M. Malik, A. I. Bhatti, and M. Tufail, “A novel anti-windup framework for cascade control
systems: An application to underactuated mechanical systems,” ISA transactions, vol. 53, no. 3, pp. 802–815, 2014.
[7] M. Hussain, N. us Saqib, M. Rehan, and M. Siddique, “Anti-windup compensator synthesis for cascaded linear control
system,” ICCAS-2017, 2017.

XXX-7
 [8] M. Rehan and K.-S. Hong, “Decoupled-architecture-based nonlinear anti-windup design for a class of nonlinear sys-
tems,” Nonlinear Dynamics, vol. 73, no. 3, pp. 1955–1967, 2013.
[9] M. Hussain, N. us Saqib, and M. Rehan, “Nonlinear dynamic regional anti-windup compensator (RAWC) schema for
constrained nonlinear systems,” in Emerging Technologies (ICET), 2016 International Conference on, pp. 1–6, IEEE,
2016.
[10] N. us Saqib, M. Rehan, M. Hussain, N. Iqbal, and H. ur Rashid, “Delay-range-dependent static anti-windup compensator
design for nonlinear systems subjected to input-delay and saturation,” Journal of the Franklin Institute, vol. 354, no. 14,
pp. 5919–5948, 2017.
[11] A. Akram, M. Hussain, N. us Saqib, and M. Rehan, “Dynamic anti-windup compensation of nonlinear time-delay
systems using LPV approach,” Nonlinear Dynamics, vol. 90, no. 1, pp. 513–533, 2017.
[12] M. C. Turner, G. Herrmann, and I. Postlethwaite, “Incorporating robustness requirements into antiwindup design,” IEEE
Transactions on Automatic Control, vol. 52, no. 10, pp. 1842–1855, 2007.
[13] Y. Li and Z. Lin, “Design of saturation-based switching anti-windup gains for the enlargement of the domain of attrac-
tion,” IEEE Trans. on Autom. Control, vol. 58, no. 7, pp. 1810–1816, 2013.
[14] M. Rehan, A. Q. Khan, M. Abid, N. Iqbal, and B. Hussain, “Anti-windup-based dynamic controller synthesis for non-
linear systems under input saturation,” Applied Mathematics and Computation, vol. 220, pp. 382–393, 2013.
[15] G. Herrmann, P. Menon, M. Turner, D. Bates, and I. Postlethwaite, “Anti-windup synthesis for nonlinear dynamic
inversion control schemes,” Int. J. of Robust and Nonlinear Control, vol. 20, no. 13, pp. 1465–1482, 2010.
[16] G. Herrmann, M. C. Turner, P. Menon, D. G. Bates, and I. Postlethwaite, “Anti-windup synthesis for nonlinear dynamic
inversion controllers,” in Rob. Control Desig., vol. 5, pp. 471–476, 2006.
[17] E. F. Mulder, P. Y. Tiwari, and M. V. Kothare, “Simultaneous linear and anti-windup controller synthesis using multiob-
jective convex optimization,” Automatica, vol. 45, no. 3, pp. 805–811, 2009.
[18] N. Wang, H. Pei, and Y. Tang, “Anti-windup-based dynamic controller synthesis for lipschitz systems under actuator
saturation,” IEEE/CAA J. of Auto. Sinica, vol. 2, no. 4, pp. 358–365, 2015.
[19] Y.-H. Chang, Y.-Y. Wang, M.-H. Hung, and P.-C. Chen, “Regional stability and H∞ performance control of an input-
saturated induction motor via LMI approach,” Asian Journal of Control, vol. 7, no. 4, pp. 368–379, 2005.

XXX-8