This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS 1

Adaptive Sliding Mode Control for Uncertain Active

Suspension Systems With Prescribed Performance
Yan-Jun Liu , Senior Member, IEEE, and Hao Chen

Abstract—In this article, the adaptive sliding mode (ASM) con- the same time, the above unknown uncertainties may directly
trol scheme of half-car active suspension systems with prescribed affect the control performance of the system and even lead
performance is studied. Because of the affected by model to system instability [7], [8]. Hence, the active suspension
uncertainty, time-varying parameter, pavement roughness exci-
tation, etc., the study of suspension systems can be regarded systems can be regarded as the tracking control problem of
as the multivariable nonlinear control problem. First of all, the the nonlinear system with uncertainty and multivariable [9].
prescribed performance function (PPF) is applied to constrain the For the tracking control problem of nonlinear systems, an
displacement and pitch angle of the suspension systems to ensure indirect data-driven tracking control for a class of nonlin-
the transient and steady-state suspension responses. Second, an ear discrete systems [10], [11] with unknown dynamics was
integral terminal sliding mode control method with strong robust-
ness is put forward, which can make the system converge rapidly given, thus realizing the prediction and optimization of non-
in a finite-time when it is far from the equilibrium point, solve linear models. In the suspension system, a tracking control
the singularity problem in the control process, and reduce the method for active suspension systems [12] of nonideal actua-
chattering phenomenon in the traditional sliding mode control. tor based on nonlinear adaptive control algorithm is proposed.
Then, the neural networks (NNs) approximation characteristics In addition, many scholars have studied other nonlinear
are used to deal with unknown items in the design of the con-
troller, and the Lyapunov stability theory is employed to analyze systems [13]–[15] and some actual dynamic systems [16]–[18]
the stability of the closed-loop system. In the end, the comparative in many aspects, and proposed a series of adaptive control
simulation results demonstrate the feasibility and effectiveness of methods [19]–[21].
the proposed control scheme. In recent years, neural network [22], fuzzy logic
Index Terms—Active suspension systems, neural networks algorithm [23], and sliding mode control [24] have become
(NNs), prescribed performance, terminal sliding mode control. the preferred methods for dealing with the problems of
uncertainties and robust control in nonlinear systems. In
particular, the combination of fuzzy control and sliding mode
control was introduced into the active suspension systems
I. I NTRODUCTION in [25]–[27] to solve the nonlinear control problem of vehicle
N RECENT decades, with the continuous development
I of industrial technology, automobile manufacturing has
become one of the primary development goals of our coun-
suspension systems. In [28], the proportional integral sliding
mode control scheme is studied for the active suspension
system, and the experimental results show that the control
try. In addition, because the control performance of vehicle strategy has strong robustness. However, in the sliding mode
suspension systems is an important factor which affects vehi- control of [25]–[28], the problem of infinite time convergence
cle driving safety, riding comfort and handling stability, of tracking error after the system reaches the sliding mode
it has been highly valued by the extensive industrial and surface and chattering of the sliding mode controller in
academic circles, (see [1], [2]). Therefore, the design of sus- the control system are not considered. The terminal sliding
pension systems which is an important development goal of mode control method was first proposed in [29]. The basic
the automotive industry. As we all know, compared with idea was to introduce nonlinear terms into the sliding mode
passive suspension systems [3] and semi-active suspension design, so that the sliding mode could reach the equilibrium
systems [4], active suspension systems [5], [6] have the char- point in a finite-time after the system entered the sliding
acteristics of nonlinear spring and damping, model uncertainty, mode. In [30], the control scheme with terminal sliding mode
time-varying system parameters, and strong parameter uncer- was used to the manipulator systems, which enables the
tainty caused by pavement interference and other factors. At systems to converge in a finite-time, improves the tracking
performance. However, terminal sliding mode control leads
Manuscript received November 30, 2019; accepted December 12, 2019.
This work was supported in part by the National Natural Science Foundation to its disadvantages, that is, the singularity may occur in the
of China under Grant 61973147, Grant 61751202, and Grant U1813203. This control law. In [31], the existence of the terminal sliding mode
article was recommended by Associate Editor R. Cui. (Corresponding author: singularity problem was first found, and a two-stage control
Yan-Jun Liu.)
The authors are with the College of Science, Liaoning University strategy was adopted to avoid the problem. In addition, the
of Technology, Jinzhou 121000, China (e-mail: [email protected]; author proposes a high-order terminal sliding mode control
[email protected]). method on an active suspension systems [32] with output
Color versions of one or more of the figures in this article are available
online at http://ieeexplore.ieee.org. feedback manner, which avoids the singular problem in the
Digital Object Identifier 10.1109/TSMC.2019.2961927 nonlinear controller. Although the above control methods
2168-2216 c 2020 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Newcastle University. Downloaded on May 16,2020 at 11:43:27 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS

can guarantee the stability of the closed-loop system, the

constraint problem in the system is ignored. It is worth
mentioning that the vertical displacement, velocity, pitch
angle, and other parameters of the active suspension system
should be limited within the safety range. Once the limit of
safety constraints is exceeded, the performance of the system
may be reduced [33], and major accidents that endanger
the safety of passengers may occur. Therefore, in order to
improve the control performance of the active suspension
system and realize the safety and stability of the vehicle, it
is necessary to consider the constraint characteristics in the
design of the control scheme [34]. Fig. 1. Model of half-car of active suspension system.
For the constraint problem in nonlinear control
systems [35]–[37], the Barrier Lyapunov functions were so as to guarantee the smoothness, stability, and safety
introduced as a kind of effective tools for this problem. of the vehicle.
For example, an adaptive control method based on Barrier
Lyapunov function is proposed for the nonlinear system [38]
II. S YSTEM D ESCRIPTIONS
with time-varying full state constraints to ensure that the
whole states do not violate the time-varying constraint Consider a class of nonlinear half-car model (Fig. 1) which
boundary under the stable state of the system. In addition, the ideal dynamic equations of the sprung and unsprung
aiming at the nonlinear active suspension systems, an adaptive masses are expressed as follows:
⎧
neural network (NN) control method was given in [39] to ⎪
⎪ mh D̈c + Fdf + Fdr + Fsf + Fsr = uD
⎨ ¨
constrain the suspension displacement and speed, which I  + p Fdf + Fsf − q(Fdr + Fsr ) = u
(1)
not only ensured the stability of the suspension systems ⎪ mf D̈f 2 − Fsf − Fdf + Ftf + Fbf = −uf
⎪
⎩
but also improved the ride comfort and safety. At the same mr D̈r2 − Fsr − Fdr + Ftr + Fbr = −ur
time, for the active suspension systems with hard constraints where uD = uf + ur and u = puf − qur . Meanwhile, the
in [40], an adaptive backstepping control method was given parameter definitions as follows: mh and Dc are the mass and
to ensure the stability and safety of the vehicle in the case of vertical displacement of the half-car;  and I are the pitch
uncertain parameters. The so-called prescribed performance angle and mass moment of inertia for the pitch motions; p and
can be regarded as a special constraint problem. For the q are the distances of suspensions to the center of mass; mr
uncertain nonlinear system [41], an adaptive control method and mf are the unsprung masses of rear and front wheels; Dr1
with prescribed performance is developed to ensure the and Df 1 are the displacements of sprung mass; Dr2 and Df 2 are
stability of the system and achieve their respective control the displacements of unsprung mass; Dr3 and Df 3 are the road
objectives. In recent years, many scholars have applied the disturbance; ur and uf are the control inputs of the system;
prescribed performance control method to the actual control Fsr and Fsf are the spring forces produced by the springs of
systems [42]–[45], which can not only check the transient suspension system; Fdr and Fdf are the forces produced by
and steady-state performance but also carry out constraint the dampers of active suspension system; Ftr and Ftf are the
control to a large extent. elasticity forces of tires; Fbr and Fbf are the damping forces
This article introduces an adaptive sliding mode (ASM) con- of tires.
trol for the uncertain active suspension systems with prescribed The nonlinear stiffening springs, piecewise dampers and the
performance. The main contributions of this article are given tire obey are expressed as follows:
as follows.
Fsf = kf 1 yf + knf 1 y3f , Fsr = kr1 yr + knr1 y3r
1) For the nonlinear active suspension systems with uncer-  
be1 ẏf b ẏ
tain parameters, the prescribed performance function Fdf = b ẏ , Fdr = be2 ẏr
c1 f c2 r
(PPF) is used to constrain the vertical displacement and  
pitch angle of this systems in combination with the Ftf = kf 2 Df 2 − Df 3 , Ftr = kr2 (D
 r2 − Dr3 )
sliding mode controller, so as to achieve the control Fbf = bf 2 Ḋf 2 − Ḋf 3 , Fbr = br2 Ḋr2 − Ḋr3
purpose. where kf 1 and kr1 represent the spring stiffness coefficients
2) An integral terminal sliding mode control method front and rear suspension; knf 1 and knr1 represent the spring
is applied in the half-car active suspension systems. stiffness coefficients of cubic terms; bei and bci , i = 1, 2
Compared with the existing adaptive control method represent the damping coefficients front and rear suspension;
with prescribed performance, this method can make the kf 2 , kr2 , bf 2 , and br2 represent the stiffness and damping
systems converge to the equilibrium position quickly coefficients of the tires; yf and yr represent the displace-
when encountering uncertain pavement disturbance. ments of front and rear suspension, and they are expressed as
3) The proposed control scheme can make the suspension follows:
systems produce a smooth signal that can accept buffet-
ing, weaken the buffeting phenomenon of sliding mode yf = Dc + p sin  − Df 2 = x1 + p sin x3 − x5
control, and reduce the singular problem in the system, yr = Dc − q sin  − Dr2 = x1 − q sin x3 − x7 .

Authorized licensed use limited to: Newcastle University. Downloaded on May 16,2020 at 11:43:27 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LIU AND CHEN: ASM CONTROL FOR UNCERTAIN ACTIVE SUSPENSION SYSTEMS WITH PRESCRIBED PERFORMANCE 3

The definitions of the state variables are given as From (2), the governing equations of the heave motion are

x1 = Dc , x2 = Ḋc , x3 = , x4 =  ˙ ẋ1 = x2
(4)
ẋ2 = m1h (−FD + uD )
x5 = Df 2 , x6 = Ḋf 2 , x7 = Dr2 , x8 = Ḋr2 .
where FD = Fdf + Fdr + Fsf + Fsr .
Furthermore, (1) can be expressed as In order to study the transient performance of vertical dis-
⎧ placement Dc quantitatively, the PPF is introduced into the
⎪ ẋ1 = x2
⎨ ẋ = 1 −F − F − F − F + u 
⎪ control design as follows:
2 mh df dr sf sr D
(2)
⎪ ẋ3 = x4 
⎪ ζi (t) = (ζi0 − ζi∞ )e−εi t + ζi∞ , i = 1, 3 (5)
⎩  
ẋ4 = 1I −p Fdf + Fsf + q(Fdr + Fsr ) + u where εi , i = 1, 3 represent the positive design parameters,
⎧
⎪
⎪ ẋ5 = x6 while ζi (t) : R+ → R+ , i = 1, 3 are the decreasing smooth
⎨ ẋ = 1 F + F − F − F − u 
⎪
functions with limt→0 ζi (t) = ζi0 >0, limt→∞ ζi (t) = ζi∞ > 0,
6 mf sf df tf bf f
(3) and ζi0 > ζi∞ . Then, the state variable x1 with the PPF is
⎪
⎪ ẋ7 = x8
⎪
⎩ ẋ = 1 (F + F − F − F − u ).
8 mr sr dr tr br r − ρ1 min ζ1 (t) < x1 < ρ1 max ζ1 (t) ∀t > 0 (6)
Assumption 1 [40]: It is assume that the mass of the half- where ρ1 min , ρ1 max > 0 are the design parameters.
car mh and the moment of inertia I are bounded, i.e., there are To overcome the condition in (6), the constraint control
positive constants mh min , mh max , Imin , and Imax such that will be transformed into the unconstrained one. And for that,
a smooth and strictly increasing function H(τ1 ) of the trans-
mh ∈ mh = {mh : 0 < mh min ≤ mh ≤ mh max } formed signal τ1 will be defined to realize that for ∀τ1 ∈ L∞ ,
I ∈ I = {I : 0 < Imin ≤ I ≤ Imax }. there is −ρ1 min < H(τ1 ) < ρ1 max , limτ1 →+∞ H(τ1 ) = ρ1 max ,
and limτ1 →−∞ H(τ1 ) = −ρ1 min .
The control objective of this article is to propose controllers Further, (6) can be rewritten as
that meet the following performance requirements.
1) Ride Comfort: The controller is designed to stabilize x1 = ζ1 (t)H(τ1 ). (7)
the vertical displacement and pitch motion of the car Lemma 1 [41]: Since system (2) is invariant under trans-
body and to isolate the forces transmitted to the pas- formation τ1 = H −1 (x1 /ζ1 (t)) with H(τ1 ), the stability of
senger as much as possible to make the passenger feel signal z1 can ensure the regulation of x1 under the prescribed
comfortable. performance (6).
2) Road Holding: The dynamic tire load of the wheels Define the function H(τ1 ) as
should not exceed the static ones, i.e., |Ftf + Fbf | < Ff ,
|Ftr + Fbr | < Fr , while Ff and Fr are the static tire ρ1 max eτ1 − ρ1 min e−τ1
H(τ1 ) = (8)
loads which satisfy Ff + Fr = (mh + mr + mf )g, eτ1 + e−τ1
Ff (p + q) = mh gq+ mf g(p + q), where g is the x1 1 ν1 (t) + ρ1 min
τ1 = H −1 = ln (9)
acceleration of gravity. ζ1 (t) 2 ρ1 max − ν1 (t)
3) Suspension Movement Limitation: The suspension space where ν1 (t) = x1 /ζ1 (t).
shall be within the allowable range by mechanical struc- Then, the derivative forms of τ1 can be expressed as
ture constraints, i.e., yf < Df max and yr ≤ Dr max .
Remark 1: Compared with the passive suspension system x1 ζ̇1
τ̇1 = μ1 x2 − (10)
and semi-active suspension system, active suspension system ζ1
can design different control scheme to adapt different road x1 ζ̇1 μ1
surface and unknown environment, which further makes the τ̈1 = μ̇1 x2 − + (−FD + uD )
ζ1 mh
riding comfort and handling stability of vehicle always in the
x2 ζ̇1 x1 ζ̈1 x1 ζ̇ 2
optimal state, and the passive suspension and semi-active sus- − μ1 + − 21 (11)
pension can only adapt few road and car conditions. Although ζ1 ζ1 ζ1
active suspension has the disadvantage of high cost, it has where μ1 = (1/2ζ1 )(1/(ν1 + ρ1 min ) − 1/(ν1 − ρ1 max )).
more practical research value because of its great advantage. Define the error variable as follows:
e1 = 1 τ1 + τ̇1 (12)
III. C ONTROLLER D ESIGN AND S TABILITY A NALYSIS
A. Adaptive Sliding Mode Control With PPF where 1 > 0 denotes a design parameter which is used to
ensure the error τ1 is bounded when e1 is bounded.
In this section, the problem of nonmatching interference Further, the integral terminal sliding mode surface is
in the controller design process is solved by combining the selected as
backstepping method with the sliding control for the non- t
e11 / 1 dι
a b
matching uncertain nonlinear active suspension systems (1), s1 = e1 + γ1 (13)
while the Lyapunov stability theory is adopted to analyze the 0
boundedness of the parameters in the control. where a1 , b1 > 0 are odds with 1 < a1 /b1 < 2, and γ1 > 0.

Authorized licensed use limited to: Newcastle University. Downloaded on May 16,2020 at 11:43:27 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS

σ1  
 2 σ1
Remark 2: In order to ensure the suspension systems error θ̃1  +
σ1 θ̃1T θ̂1 ≤ − θ1 2 . (24)
converges to zero in the finite-time, improve the convergence 2 2
accuracy, and remove the singular problem in the control pro- Substituting (23) and (24) into (22), it is easy to obtain that
cess, this article adopts the synovial surface of the integral σ1  
 2 σ1 1
terminal used in [46] for the control design. V̇1 ≤ −k1 s21 − θ̃1  + θ1 2 + η1∗2 . (25)
2 2 2
The derivative of s1 can be easily derived as
Theorem 1: Taking the controller in (18) and the adaptive
+ τ̈1 + γ1 e11 /
a b1
ṡ1 = 1 τ̇1 law (21) into account, if the initial condition −ρmin ζ1 (0) <
1 x1 ζ̇1 x1 (0) < ρmax ζ1 (0) is satisfied, and the following properties can
= mh ( 1 μ1 + μ̇1 ) x2 − + μ1 (−FD + uD ) be ensured: the prescribed performance in (6) is true when the
mh ζ1
estimate error θ̃1 converge to a small set around zero.
x2 ζ̇1 x1 ζ̈1 x1 ζ̇ 2
+ mh γ1 e11 /
a b1
− μ1 mh + − 21 . (14) Proof: From (19) and (25), it can be derived that
ζ1 ζ1 ζ1
V̇1 ≤ −A1 V1 + B1 (26)
Define the unknown function as
x1 ζ̇1 where A1 = min{2k1 /mh min , σ1 λmin (1 )} and B1 =
1 (Z1 ) = mh ( 1 μ1 + μ̇1 ) x2 − σ1 θ1 2 /2+ η1∗2 /2 are positive constants.
ζ1
Multiplying both sides of (26) by eA1 t , it has d(eA1 t V1 )/dt ≤
x2 ζ̇1 x1 ζ̈1 x1 ζ̇ 2 e t B1 , and integrating it over [0, t] has
A1
+ mh γ1 e11 / 1 .
a b
− μ1 mh + − 21
ζ1 ζ1 ζ1     
V1 (t) ≤ V1 (0) − B1 A1 e−A1 t + B1 A1 ≤ V1 (0) + B1 A1 .
(15)
(27)
Using the NN approximation property, it has
According to (27), V1 is exponential convergence, i.e., s1 is
1 (Z1 ) = θ1T S1 (Z1 ) + η1 (Z1 ) (16) exponential convergence. Then the state variables will reach
the sliding surface s1 = 0 for a finite-time.
where Z1 = [x1 , x2 , ζ1 , ζ̇1 , ζ̈1 ]T , θ1 is the ideal weight vector
If the time taken to go from e1 (0) = 0 to e1 (ts ) = 0 is ts ,
with θ̃1 = θ1 − θ̂1 , S1 (Z1 ) denotes the radial basis vector with
there is s1 = ṡ1 = 0 at this stage. Taking (13) into account,
Gaussian function, η1 (Z1 ) is the NN approximation error with a /b
|η1 (Z1 )| ≤ η1∗ , and η1∗ is a positive constant. there is de1 /e11 1 = −γ1 , and then integrate both sides of
0 a /b t
Accordingly, we have it is e1 (0) e11 1 de1 = 0s −γ dt. Hence, the time from any
initial state to zero is ts = b1 |e1 (ts )|1−(a1 /b1 ) /(γ1 (b1 − a1 )) on
1  T 
ṡ1 = θ S1 (Z1 ) + η1 (Z1 ) + μ1 (−FD + uD ) . (17) the terminal sliding mode. Under the above analysis, the state
mh 1 variables can reach the sliding mode surface s1 and the error
Then, the controller is designed as follows: signal e1 converges to zero in finite-time by setting γ1 , a1 ,
1 1 and b1 .
uD = −k1 s1 − θ̂1T S1 (Z1 ) − s1 + FD (18) In addition, the sliding mode surface s1 and weight vector
μ1 2
error θ̃1 will converge to the following compact set:
where k1 > 0 is the design parameter.   
 
Consider the Lyapunov function candidate as  = s1 , θ̃1 |s1 | ≤ 2V1 (0)e−A1 t + 2B1 A1
mh 2 1 T −1    
V1 = s + θ̃  θ̃1 (19)     
2 1 2 1 1 θ̃1  ≤ 2λmax (1 ) V1 (0)e −A 1 t + B1 A1 .
where 1T = 1 > 0 is the constant gain matrix.
Then, the time derivative of V1 can be written as Then the boundedness of τ1 , τ̇1 , and uD can be ensure, which
implies −ρmin < H(τ1 ) < ρmax can be guaranteed, while there
V̇1 = s1 −k1 s1 + θ̃1T S1 (Z1 ) − s1 + η1 (Z1 ) − θ̃1T 1−1 θ̂˙1 .
1
is −ρmin ζ1 (t) < x1 (t) < ρmax ζ1 (t) ∀t > 0. Therefore, the ver-
2 tical vehicle displacement x1 is retained within the prescribed
(20) bound. The proof is completed.
The adaptive control law is defined as Based on (2), the dynamic equations for the resultant motion
  are given as
θ̂˙1 = 1 S1 (Z1 )s1 − σ1 θ̂1 (21) 
ẋ3 = x4
(28)
where σ1 is the positive design parameter. ẋ4 = 1I (−F + u )
Substituting (21) into (20) leads to
where F = p(Fdf + Fsf ) − q(Fdr + Fsr ).
1 In the same way, the state variable x3 with the PPF is
V̇1 = −k1 s21 + σ1 θ̃1T θ̂1
+ s1 η1 (Z1 ) − s21 . (22)
2
− ρ2 min ζ2 (t) < x3 < ρ2 max ζ2 (t) ∀t > 0. (29)
Using the Young’s inequality, the following inequalities hold:
1 2 1 ∗2 In addition, a smooth and strictly increasing function H(τ2 )
s1 η1 (Z1 ) ≤ s + η (23) for the transformed variable τ2 will be defined to achieve that
2 1 2 1

Authorized licensed use limited to: Newcastle University. Downloaded on May 16,2020 at 11:43:27 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LIU AND CHEN: ASM CONTROL FOR UNCERTAIN ACTIVE SUSPENSION SYSTEMS WITH PRESCRIBED PERFORMANCE 5

 
−ρ2 min < H(τ2 ) < ρ2 max , limτ2 →+∞ H(τ2 ) = ρ2 max , and θ̂˙2 = 2 S2 (Z2 )s2 − σ2 θ̂2 (42)
limτ2 →−∞ H(τ2 ) = −ρ2 min ∀τ2 ∈ L∞ .
Therefore, (29) can be rewritten as where k2 and σ2 are the positive design parameters.
Based on (41) and (42), (40) can be expressed as
x3 = ζ2 (t)H(τ2 ). (30)
1
Define V̇2 = −k2 s22 + σ2 θ̃2T θ̂2 + s2 η2 (Z2 ) − s22 . (43)
2
ρ2 max eτ2 − ρ1 min e−τ2 Using the Young’s inequality leads to
H(τ2 ) = (31)
eτ2 + e−τ2
1 2 1 ∗2
x3 1 ν2 (t) + ρ2 min s2 η2 (Z2 ) ≤ s + η (44)
τ2 = H −1 = ln (32) 2 2 2 2
ζ2 (t) 2 ρ2 max − ν2 (t) σ2  
 2 σ2
σ2 θ̃2T θ̂2 ≤ − θ̃2  + θ2 2 . (45)
where ν2 (t) = x3 /ζ2 (t). In addition, the definitions of τ̇2 , τ̈2 , 2 2
and μ2 are similar to (10) and (11). Substituting (44) and (45) into (43) yields
Define the error variable as follows:
σ2  
 2 σ2 1
e2 = 2 τ2 + τ̇2 (33) V̇2 = −k2 s22 − θ̃2  + θ2 2 + η2∗2 . (46)
2 2 2
where 2 > 0 denotes a design parameter. Further, the following inequality hold:
The integral terminal sliding mode surface is designed as
follows: V̇2 ≤ −A2 V2 + B2 (47)
t
s2 = e2 + γ2 e22 / 2 dι
a b
(34) where A2 = min{2k2 /I, σ2 λmin (2 )} and B2 = σ2 θ2 2 /2+
0 η2∗2 /2 are positive constants.
where a2 , b2 > 0 are odds with 1 < a2 /b2 < 2, and γ2 > 0. Remark 3: Similar to Theorem 1, the stability analysis
Further, the time derivative of s2 is of subsystem (28) can also be obtained. Based on (47),
V2 is exponential convergence. Furthermore, the bounds of
x3 ζ̇2 μ2 the sliding mode surface s2 and the weight vector error θ̃2
ṡ2 = I( 2 μ2 + μ̇2 ) x4 − + (−F + u )
ζ2 I can be known. Then according to (33), (34), and (41), the
x4 ζ̇2 x2 ζ̈2 x2 ζ̇ 2 boundness of τ2 , τ̇2 , and u can be confirmed. At the same
+ Iγ2 e22 / 2 .
a b
− μ2 I + − 22 (35) time, −ρ2 min < H(τ2 ) < ρ2 max and −ρ2 min ζ2 (t) < x3 <
ζ2 ζ2 ζ2
ρ2 max ζ2 (t) can be guaranteed, that is to say, the pitch angle
Define the unknown function as displacement x3 is within the prescribed bound. In addition,
x3 ζ̇2 similar to the analysis of Theorem 1, the sliding mode surface
2 (Z2 ) = I( 2 μ2 + μ̇2 ) x4 − s2 and the error signal e2 can converge to zero in finite-time.
ζ2
Based on the results of uD and u , the actual control inputs
x4 ζ̇2 x3 ζ̈2 x3 ζ̇ 2 of uf and ur holds
+ Iγ2 e22 / 2 . (36)
a b
− μ2 I + − 22
ζ2 ζ2 ζ2 quD + u puD − u
uf = , ur = . (48)
The NN is used to approximate 2 (Z2 ), it has p+q p+q

2 (Z2 ) = θ2T S2 (Z2 ) + η2 (Z2 ) (37)

B. Zero Dynamic Analysis
where Z2 = [x3 , x4 , ζ2 , ζ̇2 , ζ̈2 ]T and |η2 (Z2 )| ≤ η2∗ . It is worth noting that the half-car active suspension systems
Then, (35) can be rewritten as include the upper spring subsystem (2) of the front and rear
1  car body and the lower spring subsystem (3). The above slid-
ṡ2 = θ2T S2 (Z2 ) + η2 (Z2 ) + μ2 (−F + u ) . (38) ing mode control schemes with prescribed performance are
I
mainly aimed at subsystem (2). Thus, the remaining fourth-
The Lyapunov function candidate is selected as
order subsystems (3) will constitute the zero dynamic system.
I 2 1 T −1 From Theorem 1, the boundedness of subsystem (2) can be
V2 = s + θ̃  θ̃2 (39)
2 2 2 2 2 known. Next, the boundedness of sprung subsystems (3) will
where 2T = 2 > 0 is the constant gain matrix. be analyzed.
The time derivative of V2 is given by Substitute (48) into (3) results in
 
V̇2 = s2 θ2T S2 (Z2 ) + η2 (Z2 ) + μ2 (−F + u ) − θ̃2T 2−1 θ̂˙2 . Ẋ = AX + BD0 + Br ω0 (49)
(40) where
⎤⎡ ⎡ ⎤
Similarly, the controller u for the pitch motion and adaptive x5 0 1 0 0
control law can be designed as ⎢ x6 ⎥ ⎢ k
− mf 2f
b
− mf f2 ⎥
X=⎢ ⎥, A = ⎢
⎢
0 0 ⎥
⎥
1 1 ⎣ x7 ⎦ ⎣ 0 0 0 1 ⎦
u = −k2 s2 − θ̂2T S2 (Z2 ) − s2 + F (41) x8 − kmr2r − bmr2r
μ2 2 0 0

Authorized licensed use limited to: Newcastle University. Downloaded on May 16,2020 at 11:43:27 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS

⎡ ⎤ ⎡ ⎤ TABLE I
Df 3 0 1 0 0
H ALF -C AR M ODEL PARAMETERS
⎢ Ḋf 3 ⎥ ⎢ kf 2 bf 2 ⎥
⎢ mf 0 0 ⎥
D0 = ⎢ ⎥
⎣ Dr3 ⎦, B = ⎢
mf
⎥
⎣ 0 0 0 1 ⎦
Ḋr3 kr2 br2
0 0 mr mr
⎡ ⎤ ⎡ ⎤
0 0 0
⎢ 1 ω ⎥ ⎢ − qmh − mf (p+q)
I ⎥
⎢ mf (a+b) 1 ⎥ ⎢ mf (p+q) ⎥
ω0 = ⎢ ⎥, Br = ⎢ ⎥
⎣ 0 ⎦ ⎣ 0 0 ⎦
qmh
1
ω
mr (a+b) 2
− mr (p+q)
I
mr (p+q)

with
p x3 ζ̇2
ω1 = − − θ2T S2 (Z2 ) − η2 (Z2 ) + I( 2 μ2 + μ̇2 ) x4 −
q ζ2
x3 ζ̇ 2
υ1 > 2λmax (P−1/2 BBT P−1/2 )/λmin (Q), υ2 >
x4 ζ̇2 x3 ζ̈2 −1/2
+ Iγ2 e22 / 2 Br BTr P−1/2 )/λmin (Q), which can guarantee:
a b
− μ2 I + − 22 2λmax (P
ζ2 ζ2 ζ2
 1  1  1 
x1 ζ̇1 1
α ≤ λmin P− 2 QP− 2 − λmax P− 2 BBT P− 2
1
+ p −θ1T S1 (Z1 ) − η1 (Z1 ) + mh ( 1 μ1 + μ̇1 ) x2 −
ζ1 υ1
1  1 
x2 ζ̇1 x1 ζ̈1 x1 ζ̇ 2 1
− μ1 mh + − 21 + mh γ1 e11 /
a b1 − λmax P− 2 Br BTr P− 2 (54)
ζ1 ζ1 ζ1 υ2

+
q 1
k1 s1 + θ̂1T S1 (Z1 ) + s1 +
1 1
k2 s2 + θ̂2T S2 (Z2 ) + s2 where α > 0. In addition, DT0 D0 ≤ D0 max and ω0T ω0 ≤
μ1 2 μ2 2 ω0 max are defined, and there is β = υ1 D0 max + υ2 ω0 max .
p x3 ζ̇2 Furthermore, V̇ ≤ −αV + β, which means that Lyapunov
ω2 = − −θ2T S2 (Z2 ) − η2 (Z2 ) + I( 2 μ2 + μ̇2 ) x4 −
q ζ2 function has a known bound. Similar to (26), the following
x4 ζ̇2 x3 ζ̈2 x3 ζ̇ 2 inequality is true:
+ Iγ2 e22 /
a b2
− μ2 I + − 22
ζ2 ζ2 ζ2  
V(t) ≤ V(0)e−αt + β 1 − e−αt . (55)
x1 ζ̇1
+ q −θ1T S1 (Z1 ) − η1 (Z1 ) + mh ( 1 μ1 + μ̇1 ) x2 −
ζ1 Further, the above derivation shows that the signals xi , i =
x2 ζ̇1 x1 ζ̈1 x1 ζ̇12 5, 6, 7, 8 are bounded as
− μ1 mh + − + mh γ1 e11 /
a b1
 
ζ1 ζ1 ζ12
|xi (t)| ≤ V(0) + β λmin (P), i = 5, 6, 7, 8. (56)
p 1 1 1
+ k1 s1 + θ̂1T S1 (Z1 ) + s1 + k2 s2 + θ̂2T S2 (Z2 ) + s2 .
μ1 2 μ2 2 Since the road disturbance Df 3 , Ḋf 3 , Dr3 , and Ḋr3 are bounded,
It is understood that the matrix A is stable, that is, there are and boundedness of Ftf , Fbf , Ftr , and Fbr can be obtained by
positive definite matrices P and Q, so that AT P + AP = −Q (56), i.e., |Ftf + Fbf | < Ff and |Ftr + Fbr | < Fr . Therefore,
is hold. Choose the Lyapunov function as V = X T PX, and its the tire static load Ff + Fr and Ff (p + q) are bounded,
derivative with respect to time is which means that the driving safety of vehicles is guaran-
teed. Accordingly, the suspension spaces are restricted by
V̇ = ẋT Px + xT Pẋ yf ≤ Df max and yr ≤ Dr max , further the stability of the
 
= xT AT P + AP x + 2xT PBD0 + 2xT PBr ω0 . (50) suspension is guaranteed.

By using the Young’s inequality, 2xT PBD0 and 2xT PBr ω0 can IV. S IMULATION E XAMPLE
be rewritten into the following form as:
In order to verify the feasibility of the proposed ASM con-
1 T trol approach, a simulation example is given for the uncertain
2xT PBD0 ≤ x PBBT Px + υ1 DT0 D0 (51)
υ1 active suspension systems with prescribed performance. In
1 T addition, the adaptive control approach in [41] with prescribed
2xT PBr ω0 ≤ x PBr BTr Px + υ2 ω0T ω0 (52)
υ2 performance is used in the simulation to show the advantage
of the proposed ASM control approach.
where υ1 and υ2 are the adjustable parameters.
The initial values of the controlled system are set as:
Substituting (51) and (52) into (50) has
x1 (0) = 0.01, x3 (0) = 0.04, θ̂1 (0) = θ̂2 (0) = 0, and
 1 1
 1  1 1
 other initial state values are selected as zero. At the same
V̇ ≤ − λmin P− 2 QP− 2 + λmax P− 2 BBT P− 2
υ1 time, the parameters of PPF in (5) are set as: ζ10 = 0.013,
1  1  ζ1∞ = 0.0018, ζ20 = 0.045, ζ2∞ = 0.0043, and ε1 = ε2 = 1;
−2 T − 12
+ λmax P Br Br P V + υ1 DT0 D0 + υ2 ω0T ω0 . the parameters the integral terminal sliding surface in (13)
υ2
(53) and (34) are given as: 1 = 2 =5, γ1 = 16, γ2 = 18, a1 =
a2 = 5, and b1 = b2 = 3; and other parameters are chosen as:
By choosing the proper matrices P and k1 = 150, k2 = 130, and 1 = 2 = 0.2 diag(1, . . . , 1)10×10 .
Q, the following inequality can be get as Some specific symbols of the half-car are shown in Table I. In

Authorized licensed use limited to: Newcastle University. Downloaded on May 16,2020 at 11:43:27 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LIU AND CHEN: ASM CONTROL FOR UNCERTAIN ACTIVE SUSPENSION SYSTEMS WITH PRESCRIBED PERFORMANCE 7

Fig. 5. Acceleration responses of vertical ẍ1 and pitch ẍ3 motions.

Fig. 2. Displacement response of vertical x1 motion.

Fig. 6. Suspension spaces of the front and rear.

Fig. 3. Displacement response of pitch x3 motion.

Fig. 7. Control inputs and adaptive laws.

Fig. 4. Velocity responses of vertical ẋ1 and pitch ẋ3 motions.

addition, the boundary values of half-car mass and the moment under the proposed control and adaptive PPF control, while
of inertia are given as: mh max = 1400 kg, mh min = 800 kg, proposed control has better convergence effect.
Imax = 800 kgm2 , and Imin = 400 kgm2 . At the same time, it can be seen from Figs. 4 and 5 that the
The comparison results of the two control methods are response of vertical and pitch for the velocity and acceleration
shown in Figs. 2–7. The time histories of vehicle vertical dis- under proposed control are smaller and has better convergence
placement and the pitch angle under collision input are shown than that under adaptive PPF control, which means that under
in Figs. 2 and 3, from which we can see that neither of these input disturbance, the proposed control can make the car body
two state variables violates the prescribed performance bounds vibrate slowly, that is, it has better stability.

Authorized licensed use limited to: Newcastle University. Downloaded on May 16,2020 at 11:43:27 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS: SYSTEMS

In addition, the suspension space of the front and rear are [10] Z.-L. Tang, S. S. Ge, K. P. Tee, and W. He, “Robust adaptive neural
plotted in Fig. 6, and it can be seen that the amplitude of tracking control for a class of perturbed uncertain nonlinear systems
with state constraints,” IEEE Trans. Syst., Man, Cybern., Syst., vol. 46,
car body vibration under proposed control is smaller than no. 12, pp. 1618–1629, Dec. 2016.
that under adaptive PPF control. The control inputs and adap- [11] Z. Wang and D. Liu, “Data-based controllability and observability anal-
tive laws are shown in Fig. 7. In conclusion, it can be seen ysis of linear discrete-time systems,” IEEE Trans. Neural Netw., vol. 22,
no. 12, pp. 2388–2392, Dec. 2011.
from Figs. 2–7 that compared with adaptive PPF control, [12] H. Pan, W. Sun, X. Jing, H. Gao, and J. Yao, “Adaptive tracking control
the proposed control the vehicle can recover its stability in for active suspension systems with non-ideal actuators,” J. Sound Vib.,
a relatively fast time after the occurrence of turbulence, and vol. 399, pp. 2–20, Jul. 2017.
it is obvious that proposed control eliminates the buffeting [13] Z. Wang, Y. Xu, R. Lu, and H. Peng, “Finite-time state estimation for
coupled Markovian neural networks with sensor nonlinearities,” IEEE
phenomenon in displacement, velocity and acceleration. Trans. Neural Netw. Learn. Syst., vol. 28, no. 3, pp. 630–638, Mar. 2017.
[14] X.-J. Xie and M. Jiang, “Dynamic state feedback stabilization of stochas-
tic cascade nonlinear time-delay systems with SISS inverse dynamics,”
V. C ONCLUSION IEEE Trans. Autom. Control, vol. 64, no. 12, pp. 5132–5139, Dec. 2019.
[15] M.-M. Jiang, X.-J. Xie, and K. Zhang, “Finite-time stabilization of
In this article, an ASM control method is introduced and stochastic high-order nonlinear systems with FT-SISS inverse dynamics,”
applied to a nonlinear active suspension system with uncertain- IEEE Trans. Autom. Control, vol. 64, no. 1, pp. 313–320, Jan. 2019.
ties and external disturbances. First, Newton’s second law is [16] C.-Y. Su and Y. Stepanenko, “Adaptive sliding mode coordinated control
of multiple robot arms attached to a constrained object,” IEEE Trans.
used to establish the dynamics equation of the half-car model, Syst., Man, Cybern., Syst., vol. 25, no. 5, pp. 871–878, May 1995.
and then the appropriate state variables are selected to establish [17] T. S. Li, R. Zhao, C. L. P. Chen, L. Y. Fang, and C. Liu, “Finite-
the state space expression. Second, PPF is used to constrain time formation control of under-actuated ships using nonlinear sliding
mode control,” IEEE Trans. Cybern., vol. 48, no. 11, pp. 3243–3253,
the vertical displacement and pitch angle to ensure the cor- Nov. 2018.
responding transient and steady control performance. At the [18] X. Y. Zhang, Y. Wang, C. L. Wang, C.-Y. Su, Z. Li, and X. K. Chen,
same time, the introduction of integral terminal sliding mode “Adaptive estimated inverse output-feedback quantized control for piezo-
improves the robustness of the system, and guarantees the con- electric positioning stage,” IEEE Trans. Cybern., vol. 49, no. 6,
pp. 2106–2118, Jun. 2019.
vergence accuracy and speed of the system. Third, through the [19] R. Lu, W. Yu, J. Lü, and A Xue, “Synchronization on complex networks
stability analysis of the Lyapunov function, the boundedness of networks,” IEEE Trans. Neural Netw. Learn. Syst., vol. 25, no. 11,
of all signals in the closed-loop system are assured, and the pp. 2110–2118, Nov. 2014.
[20] R. Lu, Y. Xu, A. Xue, and J. Zheng, “Networked control with state
vertical displacement and pitch angle do not exceed the spec- reset and quantized measurements: Observer-based case,” IEEE Trans.
ified performance limit. Finally, the simulation results reveal Ind. Electron., vol. 60, no. 11, pp. 5206–5213, Nov. 2013.
that compared with the nonsliding mode control scheme, this [21] Y. Wang, R. Wang, S. Wang, M. Tan, and J. Yu, “Underwater bio-
inspired propulsion: From inspection to manipulation,” IEEE Trans. Ind.
control method can stabilize displacement and speed of the Electron., to be published, doi: 10.1109/TIE.2019.2944082.
suspension systems in finite-time, and ensure the systems have [22] L. Liu, Y.-J. Liu, and S. C. Tong, “Neural networks-based adaptive
the characteristics of high precision, strong robustness, and fast finite-time fault-tolerant control for a class of strict-feedback switched
convergence speed. nonlinear systems,” IEEE Trans. Cybern., vol. 49, no. 7, pp. 2536–2545,
Jul. 2019.
[23] S. C. Tong and Y. M. Li, “Adaptive fuzzy output feedback tracking back-
stepping control of strict-feedback nonlinear systems with unknown dead
R EFERENCES zones,” IEEE Trans. Fuzzy Syst., vol. 20, no. 1, pp. 168–180, Feb. 2012.
[1] D. Cao, X. Song, and M. Ahmadian, “Editors’ perspectives: Road vehi- [24] D. Liu and G.-H. Yang, “Data-driven adaptive sliding mode control
cle suspension design, dynamics, and control,” Veh. Syst. Dyn., vol. 49, of nonlinear discrete-time systems with prescribed performance,” IEEE
nos. 1–2, pp. 3–28, 2011. Trans. Syst., Man, Cybern., Syst., vol. 49, no. 12, pp. 2598–2604,
[2] D. Hrovat, “Survey of advanced suspension developments and Dec. 2019.
related optimal control applications,” Automatica, vol. 33, no. 10, [25] H. Li, H. Liu, H. Gao, and P. Shi, “Reliable fuzzy control for active
pp. 1781–1817, 1997. suspension systems with actuator delay and fault,” IEEE Trans. Fuzzy
[3] I. Martins, M. Esteves, F. P. da Silva, and P. Verdelho, “Electromagnetic Syst., vol. 20, no. 2, pp. 342–357, Apr. 2012.
hybrid active-passive vehicle suspension system,” in Proc. IEEE 49th [26] H. Li, J. Yu, C. Hilton, and H. Liu, “Adaptive sliding-mode control for
Veh. Technol. Conf. (Cat. No. 99CH36363), vol. 3, 1999, pp. 2273–2277. nonlinear active suspension vehicle systems using T-S fuzzy approach,”
[4] M. Canale, M. Milanese, and C. Novara, “Semi-active suspension control IEEE Trans. Ind. Electron., vol. 60, no. 8, pp. 3328–3338, Aug. 2013.
using “fast” model-predictive techniques,” IEEE Trans. Control Syst. [27] N. Yagiz, Y. Hacioglu, and Y. Taskin, “Fuzzy sliding-mode control
Technol., vol. 14, no. 6, pp. 1034–1036, Nov. 2006. of active suspensions,” IEEE Trans. Ind. Electron., vol. 55, no. 11,
[5] S. Formentin and A. Karimi, “A data-driven approach to mixed- pp. 3883–3890, Nov. 2008.
sensitivity control with application to an active suspension system,” [28] Y. M. Sam, J. H. S. Osman, and M. R. A. Ghani, “A class of
IEEE Trans. Ind. Informat., vol. 9, no. 4, pp. 2293–2300, Nov. 2013. proportional-integral sliding mode control with application to active sus-
[6] Y.-J. Liu, Q. Zeng, S. C. Tong, C. L. P. Chen, and L. Liu, “Adaptive pension system,” Syst. Control Lett., vol. 51, nos. 3–4, pp. 217–223,
neural network control for active suspension systems with time-varying 2004.
vertical displacement and speed constraints,” IEEE Trans. Ind. Electron., [29] M. Zhihong, A. P. Paplinski, and H. R. Wu, “A robust MIMO terminal
vol. 66, no. 12, pp. 9458–9466, Dec. 2019. sliding mode control scheme for rigid robotic manipulators,” IEEE
[7] Q. Zhou, L. Wang, C. Wu, H. Li, and H. Du, “Adaptive fuzzy control Trans. Autom. Control, vol. 39, no. 12, pp. 2464–2469, Dec. 1994.
for nonstrict-feedback systems with input saturation and output con- [30] S. Yu, X. Yu, B. Shirinzadeh, and Z. Man, “Continuous finite-time con-
straint,” IEEE Trans. Syst., Man, Cybern., Syst., vol. 47, no. 1, pp. 1–12, trol for robotic manipulators with terminal sliding mode,” Automatica,
Jan. 2017. vol. 41, no. 11, pp. 1957–1964, Nov. 2005
[8] L. Liu, Y. J. Liu, S. C. Tong, and C. L. P. Chen, “Integral bar- [31] Y. Wu, X. Yu, and Z. Man, “Terminal sliding mode control design
rier Lyapunov function based adaptive control for switched nonlinear for uncertain dynamic systems,” Syst. Control Lett., vol. 34, no. 5,
systems,” Sci. China Inf. Sci., 2019, doi: 10.1007/s11432-019-2714-7. pp. 281–287, 1998.
[9] W. C. Sun, Z. L. Zhao, and H. J. Gao, “Saturated adaptive robust control [32] J. J. Rath, M. Defoort, H. R. Karimi, and K. C. Veluvolu, “Output feed-
for active suspension systems,” IEEE Trans. Ind. Electron., vol. 60, no. 9, back active suspension control with higher order terminal sliding mode,”
pp. 3889–3896, Sep. 2013. IEEE Trans. Ind. Electron., vol. 64, no. 2, pp. 1392–1403, Feb. 2017.

Authorized licensed use limited to: Newcastle University. Downloaded on May 16,2020 at 11:43:27 UTC from IEEE Xplore. Restrictions apply.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

LIU AND CHEN: ASM CONTROL FOR UNCERTAIN ACTIVE SUSPENSION SYSTEMS WITH PRESCRIBED PERFORMANCE 9

[33] Y. Q. Zhang, Y. J. Liu, and L. Liu, “Adaptive finite-time NN control for [45] G. Q. Zhu, L. L. Nie, Z. Lv, L. F. Sun, X. Y. Zhang, and C. L. Wang,
3-DOF active suspension systems with displacement constraints,” IEEE “Adaptive fuzzy dynamic surface sliding mode control of large-scale
Access, vol. 7, pp. 13577–13588, 2019. power systems with prescribe output tracking performance,” ISA Trans.,
[34] D.-P. Li, Y.-J. Liu, S. C. Tong, C. L. P. Chen, and D. J. Li, “Neural Sep. 2019, doi: 10.1016/j.isatra.2019.08.063.
networks-based adaptive control for nonlinear state constrained systems [46] C.-S. Chiu, “Derivative and integral terminal sliding mode control
with input delay,” IEEE Trans. Cybern., vol. 49, no. 4, pp. 1249–1258, for a class of MIMO nonlinear systems,” Automatica, vol. 48, no. 2,
Apr. 2019. pp. 316–326, 2012.
[35] D. P. Li, C. L. P. Chen, Y.-J. Liu, and S. C. Tong, “Neural network
controller design for a class of nonlinear delayed systems with time-
varying full-state constraints,” IEEE Trans. Neural Netw. Learn. Syst.,
vol. 30, no. 9, pp. 2625–2636, Sep. 2019.
[36] L. Liu and L. Tang, “Partial state constraints-based control for nonlinear Yan-Jun Liu (Senior Member, IEEE) received
systems with backlash-like hysteresis,” IEEE Trans. Syst., Man, Cybern., the B.S. degree in applied mathematics and the
Syst., to be published, doi: 10.1109/TSMC.2018.2841063. M.S. degree in control theory and control engineer-
[37] L. Ma and L. Liu, “Adaptive neural network control design ing from the Shenyang University of Technology,
for uncertain nonstrict feedback nonlinear system with state con- Shenyang, China, in 2001 and 2004, respectively,
straints,” IEEE Trans. Syst., Man, Cybern., Syst., to be published, and the Ph.D. degree in control theory and con-
doi: 10.1109/TSMC.2019.2922393. trol engineering from the Dalian University of
[38] Y.-J. Liu, S. M. Lu, D. J. Li, and S. C. Tong, “Adaptive controller Technology, Dalian, China, in 2007.
design-based ABLF for a class of nonlinear time-varying state con- He is currently a Professor with the College
straint systems,” IEEE Trans. Syst., Man, Cybern., Syst., vol. 47, no. 7, of Science, Liaoning University of Technology,
pp. 1546–1553, Jul. 2017. Jinzhou, China. His research interests include adap-
[39] Q. Zeng, Y.-J. Liu, and L. Liu, “Adaptive vehicle stability control tive fuzzy control, nonlinear control, neural network control, reinforcement
of half-car active suspension systems with partial performance con- learning, and optimal control.
straints,” IEEE Trans. Syst., Man, Cybern., Syst., to be published, Prof. Liu is currently an Associate Editor of the IEEE T RANSACTIONS ON
doi: 10.1109/TSMC.2019.2903203. S YSTEMS , M AN , AND C YBERNETICS : S YSTEMS.
[40] W. Sun, H. Gao, and O. Kaynak, “Adaptive backstepping control for
active suspension systems with hard constraints,” IEEE/ASME Trans.
Mechatronics, vol. 18, no. 3, pp. 1072–1079, Jun. 2013.
[41] Y. M. Li, S. C. Tong, L. Liu, and G. Feng, “Adaptive output-feedback
control design with prescribed performance for switched nonlinear
systems,” Automatica, vol. 80, pp. 225–231, Jun. 2017.
Hao Chen received the B.S. degree in information
[42] J. Na, Y. Huang, X. Wu, G. Gao, G. Herrmann, and J. Z. Jiang,
and computing science from the Liaoning University
“Active adaptive estimation and control for vehicle suspensions with
of Technology, Jinzhou, China, in 2017, where he
prescribed performance,” IEEE Trans. Control Syst. Technol., vol. 26,
is currently pursuing the M.S. degree in applied
no. 6, pp. 2063–2077, Nov. 2018.
mathematics.
[43] Y. Huang, J. Na, X. Wu, X. Liu, and Y. Guo, “Adaptive control of non-
His research interests include adaptive control,
linear uncertain active suspension systems with prescribed performance,”
active suspension control, neural network control,
ISA Trans., vol. 54, pp. 145–155, Jan. 2015.
and constraint control.
[44] W. Wang and C. Y. Wen, “Adaptive actuator failure compensation control
of uncertain nonlinear systems with guaranteed transient performance,”
Automatica, vol. 46, no. 12, pp. 2082–2091, 2010.

Authorized licensed use limited to: Newcastle University. Downloaded on May 16,2020 at 11:43:27 UTC from IEEE Xplore. Restrictions apply.