The Journal of Engineering

Jiangsu Annual Conference on Automation (JACA 2019)

Continuous finite-time TSM control for eISSN 2051-3305

Received on 9th September 2019
Accepted on 30th September 2019
electronic throttle system doi: 10.1049/joe.2019.1086
www.ietdl.org

Oluwatoyin Timothy Agbaje1, Shihua Li2 , Hao Sun1,2, Lu Zhang1,2

1Key Laboratory of Measurement and Control of Complex Systems of Engineering, School of Automation, Southeast University, Nanjing
210096, People's Republic of China
2Key Laboratory of Measurement and Control of Complex Systems of Engineering, Ministry of Education, People's Republic of China

E-mail: [email protected]

Abstract: This article examines the position tracking difficulties of the electronic throttle (ET) system and presents a continuous
finite-time terminal sliding mode (TSM) control method. The development of this control method involves two procedures: (i)
designing a global finite-time observer to estimate the derivative and higher order derivatives of the ET system output, and the
total disturbances; and (ii) construction of a continuous finite-time TSM controller based on observer estimations, to ensure the
realisation of fast finite-time convergence of system output with a comparatively smooth control action. Comprehensive stability
proof, simulation study, and experimental implementation are provided to affirm the applicability of the proposed method.

1 Introduction Since chattering effect is caused by unmodelled cascade

dynamics, which increases the system's relative degree and disturbs
The electronic throttle (ET) system is an indispensable part of the ideal sliding mode, then it is needful that the sliding variable, as
recent automobiles. It employs advanced electronic technology to well as its derivative tends to zero [17]. This informs the
control the movement of the throttle plate, for efficient air-fuel development of the higher order sliding-mode control. The
mixture [1, 2]. The property of the mechanical and electrical simplest form of higher order sliding-mode is the second-order
components of the ET system introduces non-linearities, whose SMC, i.e. twisting algorithm and super-twisting algorithm (STA)
values are usually unknown [3, 4]. Consequently, the closed-loop [18, 19]. The STA is designed to achieve finite-time convergence
system performance of the ET system is affected. Designing a with continuous control action. Whereas it does guarantee finite-
controller for the ET system requires that the non-linearites and time convergence of system output for uncertain systems with
uncertainties are considered. It is imperative to know that relative degree one. A novel full-order terminal sliding mode
controller design difficulties for the ET system are not only tied to (TSM) control that guarantees finite-time convergence for higher
its uncertainties but also the industrial requirements for the order systems with continuous control action is proposed in [20].
controllers. It must be simple enough for implementation on a low- In this paper, a continuous finite-time TSM control (CFTSMC)
cost automotive controller, yet robust for a range of process method for the ET system is discussed. The contributions made by
parameter variations [5]. this study are highlighted as follows: First, the controller for the ET
Significant attention has been drawn by researchers towards system is developed using the measured throttle position and
solving the control problems of the ET system. Numerous control observer estimations; therefore, the sliding surface estimation
algorithms have been proposed and implemented over the last algorithm proposed in [20] is not necessary. Second, the designed
decades. The most widely known and applied linear control control law is continuous; therefore, chattering effect is curtailed to
method is the proportional–integral–derivative (PID) control. It is a larger extent. Third, the total disturbances are efficiently rejected
easy to apply and requires less computation. However, linear by introducing the estimate of the disturbances into the control law,
control methods cannot guarantee superior performance for the thereby forming a composite controller. Fourth, fast finite-time
entire range of operation of the ET system. Therefore, non-linear convergence of output error is realised. The methodical
control methods and intelligent control methods were proposed for arrangement of the remaining part of this paper is as follows: in
the ET system, such as adaptive control [6], fuzzy logic control [7], Section 2, the system's problem is described; the control strategy is
backstepping control [8], and sliding mode control (SMC) [9]. described in Section 3; simulation study and experimental
These control methods have contributed their quota from different verification are presented in Section 4; while the conclusion is
facets, to enhancing the ET system dynamic performance. given in Section 5.
Among the mentioned control methods, the SMC is recognised
as one of the efficient methods to create robust controllers for non-
linear dynamic systems, operating under uncertainty conditions. 2 Problem formulation
This is not far-fetched from its unique advantages, which are This section describes the physical model of the ET system as
system decoupling, order reduction, insensitivity to matched shown in Fig. 1. The ET system is a DC motor-controlled system.
uncertainties, fast dynamic response, guaranteed stability, and It is made up of an electronic-driven DC motor that drives the
simple implementation. Unfortunately, high frequency oscillation throttle valve plate, gear sets that transmits the rotational torque
usually referred to as chattering limits its application. The generated by the motor to the throttle valve plate, and a return
chattering phenomenon is caused by high-frequency switching and spring. The throttle plate angular position is controlled by the DC
parasitic dynamics, which represent fast actuators and sensor motor and the gearbox, while the position sensor is used to
dynamics that are not included in the control design [10]. Control measure the plate's angular position [3]. The dynamic equation
solutions have been proposed to eliminate chattering, such as describing the model of the ET system is derived as follows: the
boundary layer approach [11], replacing the discontinuous control current equation of the DC motor is given as:
with an approximate continuous control [12], observer-based
approach [13], time-delay sliding-mode control [14], and higher dia
order sliding-mode control approach [10, 15]. A continuous control La + Raia + keθ̇m = u (1)
dt
approach has also been applied in [16] for the ET system.

J. Eng. 1
This is an open access article published by the IET under the Creative Commons Attribution-NonCommercial-NoDerivs License
(http://creativecommons.org/licenses/by-nc-nd/3.0/)
where Ra is motor resistance, La is motor inductance, ia is armature
current, ke is back-emf coefficient, θm is motor angular
displacement, and θ̇m is motor speed. A low-power DC motor with
small inductance La is often used to drive the throttle plate, and the
current response time constant is much less than the sampling time,
so the inductance dynamics can be ignored [21]. By setting La = 0,
motor armature current can be approximated, resulting into a
simplified form of (1) given as:
Fig. 1  Schematic structure of the electronic throttle system
u − keθ̇m
ia = (2)
Ra Table 1 ET system parameter values
Parameters Definition Values Units
The motor torque equation is expressed as: ks0 spring coefficient 0.1 N m rad−1
N gear transmission ratio 40 —
Jmθ̈m + Bmθ̇m + T m = ktia (3)
Ra0 armature resistance 2.01 Ω
where Bm is motor damping coefficient, Jm is motor inertia, T m is km0 back emf coefficient 0.0217 N m A−1
gear input torque, and kt is torque constant. By combining (2) and Jm0 motor inertia 3 × 10 −6
kg m2
(3), the standard DC motor dynamic equation can be obtained as: Jt 0 throttle torque inertia 2 × 106 kg m2

u − keθ̇m
Jmθ̈m + Bmθ̇m + T m = kt (4)
Ra
3 Control design
The throttle dynamic equation is expressed as follows: A global finite-time observer is implemented to estimate the
derivatives of the ET system output position and the total
Jtθ̈t + Btθ̇t + T sp + T f + T l = T out (5) disturbances, which are then used in the design of the CFTSMC
method. Let θtr be the desired output, then the tracking error is
where T sp, T f , and T out are the non-linear spring torque, coulomb defined as:
friction torque, and gearbox output torque, respectively. T l is load
x1 = θt − θtr
torque, Jt is throttle torque inertia, Bt is throttle damping (10)
coefficient, θt is throttle angular displacement, and θ̇t is throttle x2 = θ̇t − θ̇tr
speed. The non-linear spring torque T sp is expressed as:
Then, system (8) can be rewritten as:
T sp = ks(θt − θt0) (6)
ẋ1 = x2
(11)
and gearbox output torque ẋ2 = Gu − ψ θ(x1 + θtr) − ψ ω(x2 + θ̇tr) + D(t) − θ̈tr

T out = NT m + P(T m) (7) Equation (11) can be described by a general class of second-order
system given below:
where P(T m) is a bounded non-linear function of T m, and θt0 is the
default opening angle of the throttle plate called the limp-home ẋ1 = x2
(12)
position. By combining (4), (5), and (7), and taking into ẋ2 = Gu + ψ(x1, x2) + D(t)
consideration nominal parameter variations, the ET system-state
model can finally be obtained as: where ψ(x1, x2) = − ψ θ(x1 + θtr) − ψ ω(x2 + θ̇tr) − θ̈tr.

θ̇t 0 1 θt 0 0 3.1 Finite-time observer design

= + u+ D (8)
θ̈t −ψθ −ψω θ̇t G 1
A finite-time higher order observer based on the proposition in [22]
will be employed. The ET system problem is that of output
and D(t) is the lumped disturbances expressed as feedback stabilisation problem, and finite-time convergent
observers are efficient tools in dealing with output feedback
Nkm Nkm0 N 2km2 0 N 2km2 stabilisation problems, since they can provide exact values of the
D(t) = − u+ − θ̇
JRa J0Ra0 J0Ra0 JRa t unmeasurable states after a finite time, rather than only the
(9) estimates produced by the conventional observers.
ks ks P(T m) − Bθ̇t + ksθt0 − Fssgn(θ̇t) − T l  
+ 0 − θt +
J0 J J Assumption 1: The total disturbance D(t) is assumed to satisfy
the following conditions:
where J = Jt + N 2Jm, B = Bt + N 2Bm, kt = ke = km, N = (θm /θt),
• The total disturbance D(t) is (n − 1)th differentiable;
ψ θ = (ks0 /J0), G = (Nkm0 /J0Ra0), and ψ ω = (N 2km2 0 /J0Ra0). The
• The condition D(n − 1)(t) ≤ ϕ is valid for a non-negative
parameters km0, Ra0, J0, and ks0 are the ET system nominal
constant ϕ, i.e. ϕ > 0
parameters given in Table 1. Detailed description of the ET system
model can be found in [20].
  Then, the finite-time observer can be constructed as follows:
Remark 1: (9) describes the total disturbances, which are a
combination of nominal parameter variation, non-linearities, and
external disturbances. All these practically affect the general
performance of the ET system.

2 J. Eng.
This is an open access article published by the IET under the Creative Commons Attribution-NonCommercial-NoDerivs License
(http://creativecommons.org/licenses/by-nc-nd/3.0/)
 ż1 = z2 + E1⌈x1 − z1⌋ 1
γ σ = − un (19)
γ2
ż2 = Gu + ψ(x1, z2) + z3 + E2⌈x1 − z1⌋ Taking the derivative of the sliding surface
ż3 = z4 + E3⌈x1 − z1⌋γ3
(13) λ3sgn(σ)
… σ̇ = − (20)
λ1 + αλ2 σ α − 1
γ
żn − 1 = zn + En − 1⌈x1 − z1⌋ n − 1
γ Considering a bounded candidate Lyapunov function V 0(σ, x1, x2)
żn = En⌈x1 − z1⌋ n defined as:
γ
where ⌈x1 − z1⌋γi denotes ⌈x1 − z1⌋γi = x1 − z1 i sign (x1 − z1), 1 1 1
V 0 = σ 2 + x12 + x22 (21)
γi = 1 + iτ, τ ∈ ( − (1/n), 0), (i = 1, 2, …, n) are the observer 2 2 2
coefficients; z1, z2, z3, z4,......, zn − 1 and zn are estimation of x1, x2,
Taking the time derivative of V 0, one can obtain:
D(t), D(i)(t),…, D(n − 3)(t) and D(n − 2)(t); and E1, E2, E3, …, En − 1, En
are the observer gains. Let the estimation error be defined as V̇ 0 = σσ̇ + x1 ẋ1 + x2 ẋ2 (22)
e1 = x1 − z1, e2 = x2 − z2, and e3 = D(t) − z3. Then, the error
dynamics can be written as: ẋ2 can be obtained by combining (15) and (12) as follows:
γ1
ė1 = e2 − E1⌈e1⌋ ẋ2 = Gu + ψ(x1, x^ 2) − [ψ(x1, x^ 2) − ψ(x1, x2)]
γ2
ė2 = e3 − ψ ωe2 − E2⌈e1⌋ ^ ^
+D − [D − D(t)]
(23)
ė3 = e4 − E3⌈e1⌋γ3 = σ + e3 − ψ ωe2 − k2 x^ 2
β2
sgn(x^ 2) − k1 x1
β1
sgn(x1)
(14)
… β2 β1
≤ σ + e3 + ψ ω e2 + k2 x^ 2 + k1 x1
γn − 1
ėn − 1 = en − En − 1⌈e1⌋
γ Since 0 < β < 1, then
ėn = D(n − 2) − En⌈e1⌋ n
β2
According to [23], there exists a finite time T e, such that for x^ 2 < 1 + β2 x^ 2 < 1 + x^ 2
(24)
∀te > T e, ei (i = 1, 2, …, n) will converge to zero in finite time. β1
x1 < 1 + β1 x1 < 1 + x1
Thus, z2 = x2 and z3 = D(t) will be realised.
Based on the definition of the observer estimation error
3.2 Controller design
x^ 2 = x2 + e2 ≤ x2 + e2 (25)
3.2.1 Theorem 3.1: From system (12), based on the observer
estimations, the sliding surface is designed as: α
Also, note that σ < 1 + σ . Substituting (23), (24) and (25) into
^ β2 β1 (22), one obtains
σ = Gu + ψ(x1, x2) + D + k2⌈x2⌋ + k1⌈x1⌋
^ ^
(15)

β
V̇ 0 ≤ x1 x2 + x2σ + x2e3 + ψ ω x2e2
where ⌈x⌋ β = x sgn (x), ki is selected such that the polynomial
+k2(1 + e2 ) x2 + k1(1 + x1 ) x2
n x12 + x22 x22 + e32 x22 + σ 2 x2 + e22
≤ + + + ψω 2
P(λ) = λ + n
∑ kiλ i−1
(16) 2 2 2 2 (26)
i=1 2 2 2 2 2 2
1+e x +e
2 1+x
2 x +x
2 2 1 2
+k2 + k2 + k1 + k1
is Hurwitz. The value of βi is derived based on the following 2 2 2 2
conditions: ≤ Lv1V 0 + Lv2

β1 = β, n=1 where Lv1 = max {1, 3 + 2k1 + k2 + ψ ω, 1 + k1} and

βi βi + 1 (17) Lv2 = max {(k1 /2) + (e32 /2) + (ψ ω + k2)(e22 /2) + k2((1 + e22)/2)} are
βi − 1 = , i = 2, …, n ∀n > 1 bounded constants, due to the boundedness of e2 and e3. Based on
2βi + 1 − βi
the proof in (26), it can be concluded that V 0(σ, x1, x2), σ, x1,, and x2
with βn + 1 = 1, βn = β, β ∈ (1 − ε, 1), and ε ∈ (0, 1). will not escape to infinity before the convergence to zero of the
The control law u is designed as: observer errors in finite time [25].
Next is to show that the sliding surface will converge to zero in
β β ^ finite time. Considering a candidate Lyapunov function
u = − G−1(ψ(x1, x^ 2) + k2⌈x^ 2⌋ 2 + k1⌈x1⌋ 1 + un + D)
V 1 = (1/2)σ 2 and taking the derivative, one obtains
λ3sgn(σ) (18)
u̇n =
λ1 + αλ2 σ α − 1 λ3sgn(σ)
V̇ 1 = − σ
^
λ1 + αλ2 σ α − 1
where k1, k2 ∈ ℝ, k1, k2 > 0, z3 = D is the total disturbance estimated
by the observer, z2 = x^ 2, λ1, λ2, λ3 are positive constants, and λ3 σ 1 − α sgn(σ)
= −σ
0 < α < 1. If the parameters are carefully chosen, the states x1, and λ1 σ 1 − α + αλ2
x2 will converge to zero in finite time [24]. 2−α
(27)
λ3
≤ − 1−α
σ
λ1 σ + αλ2
3.2.2 Proof: To show that the system-state error will converge to
2−α
zero in finite time; first, the finite-time boundedness of system = −Q σ
states is shown. Substituting the control law in (18) into (15), one
obtains = − 2QV ((2 − α)/2)

J. Eng. 3
This is an open access article published by the IET under the Creative Commons Attribution-NonCommercial-NoDerivs License
(http://creativecommons.org/licenses/by-nc-nd/3.0/)
where Q = λ3 /ℓ, ℓ = (λ1 σ 1 − α + αλ2). For σ ≠ 0, α ∈ (0, 1), and
λ1, λ2, λ3 > 0, then ℓ ≠ 0. According to the following inequality

V̇ 1 ≤ − qV 1r (28)

where q and r are positive constants, with r ∈ (0, 1). It can be

deduced from [26] that finite time stability can be ensured.
The overall closed-loop system can be written as:

ẋ1 = x2 − e2
β β
ẋ2 = σ − k2⌈x2 − e2⌋ 2 − k1⌈x1⌋ 1 + e3 − ψ ωe2
Γ: σ = − un (29)
λ3sgn(σ)
u̇n =
λ1 + αλ2 σ α − 1

It has been discussed that at time t > T e, the observer errors will
converge to zero in finite time. Also, it has been proven in (26)
based on the proposition in [25], that σ, x1, and x2 will not escape to Fig. 2  Block diagram of the CFTSMC closed-loop control system
infinity in finite time. After t > T e, the closed-loop system
becomes:

ẋ1 = x2
β β1
ẋ2 = σ − k2⌈x2⌋ 2 − k1⌈x1⌋
Γ: σ = − un (30)
λ3sgn(σ)
u̇n =
λ1 + αλ2 σ α − 1

For x1 ≠ 0 and x2 ≠ 0, k1 and k2 > 0. Therefore, by selecting

appropriate gains k1 > 0, k2 > 0, λ1, λ2, λ3 > 0, α ∈ {0, 1}, and
β1, β2 ∈ {0, 1}, σ = σ̇ = 0 in finite time tσ ≥ T σ = σ̇ = 0, where T σ = σ̇ = 0
Fig. 3  Transient performance comparisons (simulation study)
is the reaching time defined by
(a) Step input command, (b) Sinusoidal input command, (c) Series of step input
command
1 2
T σ = σ̇ = 0 = V (0)1 − r = V (0)α /2 (31)
q(1 − r) 1 αQ 1
k1 = 104, k2 = 300, α = 0.2, β1 = 9/13, β2 = 11/13, λ1 = 1.5, λ2, and
Therefore, at t ≥ max {T e, T σ = σ̇ = 0}, the closed-loop system finally = 0.01
becomes: λ3 = 2.
The control law for the nominal SMC controller is designed as
ẋ1 = x2 follows:
β β1 (32)
ẋ2 = − k2⌈x2⌋ 2 − k1⌈x1⌋ σ = ẋ1 + c1 x1
(33)
Usmc = − G−1(ψ(x1, x2) + c1 ẋ1 + ksgn(σ))
It can, therefore, be concluded according to [24] that x1 and x2 will
converge to zero in finite time.
  The SMC parameters are chosen as c1 = 90 and k = 900.
Remark 2: The proposed control method not only guarantees The control law for the FOTSMC controller is designed as:
fast and finite time convergence to zero of the ET system state β β1
error but also its void of singularity problem peculiar to σ = ẋ2 + k2⌈x2⌋ 2 + k1⌈x1⌋
conventional TSM control (Fig. 2). β β
U = − G−1 ψ(x1, x2) + k2⌈x2⌋ 2 + k1⌈x1⌋ 1 + un (34)

4 Simulation study and experimental verification u̇n − Υun = (δ + Eτ + η)sgn(σ)

4.1 Simulation study where sgn(σ ) is estimated as:
The performance of the proposed control method is analysed based
∫ ∫
t t
on the transient response and disturbance rejection ability. The β β
h(t) = σdt = x2 + k2⌈x2⌋ 2 + k1⌈x1⌋ 1 dt (35)
proposed control method is investigated alongside with the 0 0
nominal SMC and full-order TSM control (FOTSMC) law in [20]
for comparison. In order to ensure fair comparison, the control and
parameters are carefully chosen, when satisfactory transient
performance and steady-state performance are realised, under the sgn(σ) = sgn(h(t) − h(t − τ)) (36)
same maximum control input magnitude. The parameter
specifications of the ET system are given in Table 1, while the The FOTSMC parameters are chosen as follows:
proposed CFTSMC control parameters are chosen as follows: the δ + Eτ + η = 500,
observer gains E1, E2, E3 and E4 are chosen as (s + E)4, where β1 = 9/13, β2 = 11/13, k1 = 10000, k2 = 300, Υ = 0.1. The
E = 50, γ1, γ2, γ3, γ4, are chosen as 10/11, 9/11, 8/11, and 7/11, simulation results are shown in Figs. 3–5. It is observed that the
respectively, proposed controller exhibits an incredible performance with faster
transient response and higher accuracy, compared with the SMC

4 J. Eng.
This is an open access article published by the IET under the Creative Commons Attribution-NonCommercial-NoDerivs License
(http://creativecommons.org/licenses/by-nc-nd/3.0/)
Fig. 4  Time history of control input (simulation study)
(a) SMC control input, (b) FOTSMC control input, (c) Proposed control input

Fig. 5  Transient performance comparisons under constant and sinusoidal input voltage disturbance (simulation study)
(a) Constant disturbance, (b) 4 Hz frequency sinusoidal disturbance, (c) 8 Hz frequency sinusoidal disturbance

Fig. 6  Experimental setup for an electronic throttle system

Fig. 7  Transient performance comparisons for step reference command (experimental case 1)
(a) Transient response, (b) Tracking error, (c) Control input

control. It is as well obvious in Fig. 4 that chattering, which exist in saturated within the input limit of the DC motor. The sampling
the SMC controller, is reduced to a large extent in both the period is selected as 1 ms.
FOTSMC controller and the proposed controller. The degree of
reduction is higher in the proposed controller than in the FOTSMC 4.2.2 Experimental results: Experimental case 1: Nominal
controller as seen from the zoom-in plot. This is due to the effect of performance evaluation
the designed continuous control law and low control gain. Fig. 5 The outcome of the first experimental case is plotted in Figs. 7
shows the performance of the three controllers, subjected to and 8. The proposed controller and FOTSMC controller have faster
constant and sinusoidal input voltage disturbances. The SMC is transient response. For the 4 Hz sinusoidal input command, the
invariant to matched uncertainties, provided the control gain is steady-state error of each controller is obtained as SMC = 1.57 deg,
higher than the upper bound of the disturbance. However, the high FOTSMC = 1.03 deg, and proposed = 0.34 deg. This result reveals
control gain needed to remove the offset increases chattering. The the superiority of the proposed control method with respect to
FOTSMC control law also requires high control gain to remove the tracking accuracy. It is also obvious from the plots that control
offset caused by the disturbances, which could result into steady- chattering is highly reduced in the FOTSMC and proposed control,
state fluctuation. On the contrary, the proposed CFTSMC control with the proposed control having the least chattering effect (Fig. 9).
law has the ability to remove the offset caused by the disturbances,
with small control gain. Experimental case 2: Disturbance rejection ability
In the second experimental test case, the disturbance rejection
4.2 Experimental verification ability of each controllers is verified in the presence of constant
input voltage and sinusoidal input voltage disturbances:
4.2.1 Experimental setup: The experimental substantiation of the
proposed control method is demonstrated in the setup configuration 12V(Nominal), for t ∈ [0, 2]s,
shown in Fig. 6. The set-up consists of a DS1104 dSPACE data
acquisition system and digital signal processor, power supply 7V(Decrease), for t ∈ [2, 8]s,
E= (37)
system, a H-bridge PWM driver, a throttle valve, and a set of high 12V(Increase), for t ∈ [8, 15]s,
processor speed computer on which the dSPACE ControlDesk 4sin(4t), for t ∈ [15, 20]s,
software and Matlab softwares are installed. The input range of the
throttle valve DC motor is ( ± 12V), so the control action is

J. Eng. 5
This is an open access article published by the IET under the Creative Commons Attribution-NonCommercial-NoDerivs License
(http://creativecommons.org/licenses/by-nc-nd/3.0/)
Fig. 8  Transient performance comparisons for 4 Hz frequency sinusoidal reference command (experimental case 1)
(a) Transient response, (b) Tracking error, (c) Control input

Fig. 9  Transient performance comparisons under constant and sinusoidal input voltage disturbance (experimental case 2)
(a) Transient response, (b) Tracking error, (c) Control input

Table 2 Transient performance comparisons under constant and sinusoidal input voltage disturbances (experimental case 2)
Disturbances Algorithms Max(e), deg Recovery time, s RMS(e), deg
constant SMC 2.93 — 1.33
FOTSMC 1.64 >3.0 0.59
proposed 1.59 0.18 0.16
sinusoidal SMC 2.30 — 0.87
FOTSMC 1.43 — 0.79
proposed 0.49 0.18 0.11

This experimental test case is aimed at testing the performance of 7 References

the controllers, when confronted with constant and recurring input
[1] Loh, R.N.K., Pornthanomwong, T., Pyko, J.S., et al.: ‘Modeling, parameters
voltage fluctuation. Fig. 5 shows clearly the response of each identification, and control of an electronic throttle control (ETC) system’.
controller, with the proposed controller having the ability to Proc. Int. Conf. Intell. Adv. Syst., Kuala Lumpur, Malaysia, 2007, pp. 1029–
eliminate completely the offset caused by the input voltage change. 1035
The summary of the performances is given in Table 2. [2] Robert, N.K.L., Witt, T., Jan, S.P., et al.: ‘Electronic throttle control system:
modeling, identification and model-based control designs’, Engineering,
  2013, 5, (7), pp. 587–600
Remark 3: Due to hardware limitations, it is practically [3] Ashok, B., Ashok, S.D., Kumar, C.R.: ‘Trends and future perspectives of
impossible to add the same torque disturbance to the ET system for electronic throttle control system in a spark ignition engine’, Ann Rev Control,
different controllers [16]. Therefore, input voltage disturbance is 2017, 44, pp. 97–115
[4] Aono, T., Kowatari, T.: ‘Throttle-control algorithm for improving engine
applied in this experiment, which alters the normal dynamics of the response based on air-intake model and throttle-response model’, IEEE Trans.
ET system. Ind. Electron., 2006, 53, (3), pp. 915–921
[5] Yuan, X., Wang, Y., Wu, L.: ‘SVM-based approximate model control for
electronic throttle valve’, IEEE Trans. Veh. Technol., 2008, 57, (5), pp. 2747–
5 Conclusion 2756
[6] Bernardo, M.D., Gaeta, A.D., Montanaro, U., et al.: ‘Synthesis and
This research has proposed a continuous finite-time TSM control experimental validation of the novel LQ-NEMCSI adaptive strategy on an
method for the ET system, with the sole aim of achieving a better electronic throttle valve’, IEEE Trans Control Syst Tech, 2010, 18, (6), pp.
performance in the presence of several sources of disturbances, 1325–1337
either internal or external sources. A global finite time observer is [7] Yuan, X., Wang, Y., Wu, L., et al.: ‘Neural network based selflearning control
strategy for electronic throttle valve’, IEEE Trans. Veh. Technol., 2010, 59,
designed to estimate precisely the speed information and total (8), pp. 3757–3765
disturbances. The provided estimation serves as a basis for the [8] Wan, J., Zeng, Q., Cheng, S., et al.: ‘Nonlinear control of electronic throttle
design of the continuous finite-time TSM controller. Experimental based on backstepping approach’, Amer J Electromag App, 2016, 4, (1), pp.
results reveal that the proposed control method has fast transient 1–7
[9] Pan, Y., Ozguner, U., Dagci, O.H.: ‘Variable-structure control of electronic
response and superb disturbance rejection ability. It is also essential throttle valve’, IEEE Trans. Ind. Electron., 2008, 55, (11), pp. 3899–3907
to recognise that the control chattering is greatly reduced, which [10] Shtessel, Y.B., Fridman, L., Zinober, A.: ‘High order sliding modes as the
makes the implementation of the proposed controller easier. It is, natural phenomenon in control theory’, in Garofalo, F., Glielmo, L. (eds):
therefore, needful to recommend this control method for other ‘Robust control via Variable structure and lyapunov techniques lecture notes
in control and information Sciences', (Springer, Berlin Heidelberg, Germany,
automotive application due to its demonstrated uniqueness in 1996, vol. 217), pp. 107–133
performance. [11] Cheng, N.B., Guan, L.W., Wang, L.P., et al.: ‘Chattering reduction of sliding
mode control by adopting nonlinear saturation function’, Adv. Mater. Res.,
2010, 143–144, pp. 53–61
6 Acknowledgments [12] Burton, J.A., Zinober, A.S.I.: ‘Continuous approximation of variable structure
control’, Int. J. Syst. Sci., 1986, 17, (6), pp. 875–885
This work was supported by the National Nature Science [13] Chen, M.S., Chen, C.H., Yang, F.Y.: ‘An ltr-observer-based dynamic sliding
Foundation of China (grant no. 61973081). mode control for chattering reduction’, Automatica, 2007, 43, (6), pp. 1111–
1116
[14] Efimov, D., Polyakov, A., Fridman, L., et al.: ‘Delayed sliding mode control’,
Automatica, 2016, 64, pp. 37–43
[15] Bartolini, G., Ferrara, A., Usai, E.: ‘Chattering avoidance by second-order
sliding mode control’, IEEE Trans. Autom. Control, 1998, 43, (2), pp. 241–
246

6 J. Eng.
This is an open access article published by the IET under the Creative Commons Attribution-NonCommercial-NoDerivs License
(http://creativecommons.org/licenses/by-nc-nd/3.0/)
[16] Wang, H., Shi, L., Man, Z., et al.: ‘Continuous fast nonsingular terminal [22] Du, H., Qian, C., Yang, S., et al.: ‘Recursive design of finite-time convergent
sliding mode control of automotive electronic throttle systems using finite- observers for a class of time-varying nonlinear systems’, Autom., 2013, 49,
time exact observer’, IEEE Trans. Ind. Electron., 2018, 65, (9), pp. 7160– (2), pp. 601–609
7172 [23] Wang, Z., Li, S., Yang, J., et al.: ‘Current sensorless finite-time control for
[17] Fridman, L., Moreno, J.A., Bandyopadhyay, B., et al.: ‘Continuous nested buck converters with time-varying disturbances’, Control Eng. Pract., 2018,
algorithms: the fifth generation of sliding mode controllers’ (Springer 77, pp. 127–137
International Publishing, Berlin Heidelberg, Germany, 2015) [24] Bhat, S.P., Bernstein, D.S.: ‘Geometric homogeneity with applications to
[18] Emel'Yanov, S.V., Korovin, S.K., Levant, A.: ‘High-order sliding modes in finite-time stability’, Math. Control Signals Syst., 2005, 17, (2), pp. 101–127
control systems’, Comput. Math. Model., 1996, 7, (3), pp. 294–318 [25] Li, S., Tian, Y.P.: ‘Finite-time stability of cascaded time-varying systems’, Int.
[19] Levant, A.: ‘Sliding order and sliding accuracy in sliding mode control’, Int. J. Control, 2007, 80, (4), pp. 646–657
J. Control, 1993, 58, (6), pp. 1247–1263 [26] Bhat, S.P., Bernstein, D.S.: ‘Finite-time stability of continuous autonomous
[20] Feng, Y., Han, F., Yu, X.: ‘Chattering free full-order sliding-mode control’, systems’, SIAM J. Control Optim., 2000, 38, (3), pp. 751–766
Autom., 2014, 50, (4), pp. 1310–1314
[21] Jiao, X., Zhang, J., Shen, T.: ‘An adaptive servo control strategy for
automotive electronic throttle and experimental validation’, IEEE Trans. Ind.
Electron., 2014, 61, (11), pp. 6275–6284

J. Eng. 7
This is an open access article published by the IET under the Creative Commons Attribution-NonCommercial-NoDerivs License
(http://creativecommons.org/licenses/by-nc-nd/3.0/)