You are on page 1of 13

ISA Transactions ()

Contents lists available at ScienceDirect

ISA Transactions
journal homepage: www.elsevier.com/locate/isatrans

Parametric adaptive estimation and backstepping control


of electro-hydraulic actuator with decayed memory lter
Qing Guo a,c,n, Ping Sun a, Jing-min Yin a, Tian Yu b, Dan Jiang d
a

School of Aeronautics and Astronautics, University of Electronic Science and Technology of China, Chengdu, 611731, China
Center for Power Transmission and Motion Control, Department of Mechanical Engineering, University of Bath, Bath BA2 7AY, UK
c
The State Key Laboratory of Fluid Power Transmission and Control, Zhejiang University, Hangzhou, China
d
School of Mechatronics Engineering, University of Electronic Science and Technology of China, 611731 Chengdu, China
b

art ic l e i nf o

a b s t r a c t

Article history:
Received 26 October 2015
Received in revised form
4 January 2016
Accepted 8 February 2016
This paper was recommended for publication by Dr. A.B. Rad

Some unknown parameter estimation of electro-hydraulic system (EHS) should be considered in


hydraulic controller design due to many parameter uncertainties in practice. In this study, a parametric
adaptive backstepping control method is proposed to improve the dynamic behavior of EHS under
parametric uncertainties and unknown disturbance (i.e., hydraulic parameters and external load). The
unknown parameters of EHS model are estimated by the parametric adaptive estimation law. Then the
recursive backstepping controller is designed by Lyapunov technique to realize the displacement control
of EHS. To avoid explosion of virtual control in traditional backstepping, a decayed memory lter is
presented to re-estimate the virtual control and the dynamic external load. The effectiveness of the
proposed controller has been demonstrated by comparison with the controller without adaptive and
lter estimation. The comparative experimental results in critical working conditions indicate the proposed approach can achieve better dynamic performance on the motion control of Two-DOF robotic arm.
& 2016 ISA. Published by Elsevier Ltd. All rights reserved.

Keywords:
Parametric adaptive estimation
Electro-hydraulic system
Backstepping control
Decayed memory lter

1. Introduction
Electro-hydraulic servo systems (EHSs) are nowadays widely
used in mechatronic control engineering, due to their higher loadbearing and small size-to-power ratio [1]. One of the fundamental
difculties in hydraulic control and EHS operations is undesirable
dynamic behavior of the established controller due to the ignorance
of parametric uncertainties and unknown disturbance. Some parametric uncertainties are almost caused by unknown viscous
damping, load stiffness, physical characteristics of valve, bulk
modulus [2]. Since some unknown parametric variation may be
signicant in different working conditions (i.e., the phenomenon
such as oil temperature variations, pressure-ow characteristics,
hysteresis in ow gain characteristics, oil leakage, characteristics of
valves near null) [3], many state/output feedback controllers cannot
be well established to guarantee the dynamic performance of EHS.
Thus, the parametric estimation is one available method to obtain
unknown parameters by state observer construction integrated with
n
Corresponding author at: School of Aeronautics and Astronautics, University of
Electronic Science and Technology of China, No. 2006, Xiyuan Ave, West Hi-Tech
Zone, Chengdu 611731, China.
E-mail addresses: guoqinguestc@uestc.edu.cn (Q. Guo),
sunping0944@163.com (P. Sun), yinjm1028@163.com (J.-m. Yin),
ty267@bath.ac.uk (T. Yu), jdan2002@uestc.edu.cn (D. Jiang).

other nonlinear controller. However, different from parametric


uncertainties, the mainly disturbance is largely unknown torque/
force disturbances caused by external loadings on hydraulic actuator. So further references have been more focused on disturbance
rejection of EHS. For instance, Ref. [4] proposed a nonlinear controller in which the external load is treated as uncertain but
bounded disturbance. It has been shown that the closed loop stability can directly be analyzed by the Lyapunov technique. Yao and
Bu [5,6] assumed that the maximum relative uncertainty of the
external load disturbance is bounded by a known value and proposed a discontinuous projection-based adaptive backstepping
controller. Kim [7] presented a disturbance observer (DOB) with PI
form to estimate a biased sinusoidal external load. Then, Won [8]
developed a high-gain disturbance observer (HDOB) with backstepping to compensate for the unknown external load and guaranteed tolerance of the position tracking error. These references
denote that the external load with wide variations is an important
factor to decline dynamic response performance of a hydraulic
controller, especially in some critical condition where the external
load of the hydraulic actuator is close to the limitation. Therefore, to
improve the dynamic behavior of EHS, various advanced control
approaches were developed to estimate unknown parametric
uncertainties and unmeasured disturbance.
Recently, some robust H 1 control methods [912] and quantitative feedback theories [13] have been presented to overcome

http://dx.doi.org/10.1016/j.isatra.2016.02.009
0019-0578/& 2016 ISA. Published by Elsevier Ltd. All rights reserved.

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

parametric uncertainties and to guarantee the robustness of a


controller. A geometric control approach [14] was veried by
numerical simulation to realize the dynamic tracking position of a
single-rod cylinder. In the past two decades, backstepping control
was widely used in EHS [1517]. If the model of EHS is a strict
feedback system [18,19], the backstepping controller based on
state feedback can be well implemented. Then Yao and Bu [20]
presented a discontinuous projection-based adaptive backstepping controller to estimate some unknown parameters of
asymmetric hydraulic actuator. Subsequently, Guan [21,22] also
constructed a parametric adaptive estimation law to guarantee the
asymptotic convergence of backstepping controller. Kaddissi [23]
proposed an equivalent parameter identication method by the
least squares and obtained better performance than the linear
controller. Ahn [24] presented an adaptive position control for a
pump-controlled EHA based on an adaptive backstepping control
scheme. To address some unmeasured physical states, Sun [25]
proposed a perturbation observer to estimate the load pressure of
single-rod hydraulic actuator. Pi [27] designed an observer-based
cascade controller to estimate disturbance force in hydraulic
manipulator. In addition, some output observers [2630] were
used to estimate hydraulic states with less measured information
than state feedback observers. These various observers were veried in backstepping or other nonlinear control method.
However, to the best of authors knowledge, aforementioned
backstepping and adaptive controllers need to deal with some
derivatives of virtual control variables which exist in backstepping
iteration. These derivatives can be computed by the system state
error model and parametric estimated law established in recursive
controller design [18]. Since this computed process easily lead to
derivative explosion, few researches are focused on the quality of
these virtual control variables (i.e., data validation). If these virtual
control variables are not available, especially violent, both the stability and dynamic performance of EHS will be signicantly

decreased. Thus, this paper presents a parametric adaptive backstepping control method based on state feedback to estimate some
unknown parameters in hydraulic model. Then a decayed memory
lter is proposed to compute the derivatives value of virtual control
variables in the backstepping control design. The effectiveness of the
proposed control is veried by a comparative experimental study.
The remainder of this paper is organized as follows. The
dynamic models of EHS and external load are constructed in
Section 2. The parametric adaptive backstepping controller and
the decayed memory lter are designed in Section 3. Then the
related experimental results are demonstrated in Section 4. Finally,
the conclusion is drawn in Section 5.

2. Mechatronics plant description


2.1. Two-DOF robotic arm
The mechatronics plant is two-DOF Robotic Arm which is
similar to Robotic Bigdog. The motion control of Robotic Arm is
driven by an electro-hydraulic system (EHS) shown in Fig. 1. The
upper arm and forearm rotate around the shoulder and elbow
joint in sagittal plane respectively. The torso is xed on the
ground. The main external load is a disc on the top of forearm. The
electro-hydraulic system is comprised by two servo valves, two
hydraulic cylinders, two pistons, a xed-displacement pump and a
relief valve.
2.2. Dynamics of EHS
The load ow QL of a servo valve can be described as follows
s

1
ps  sgnxv pL ;
1
Q L C d wxv

13
Elbow

10

6
Shoulder

11

12

2-DOF Robotic Arm

8
9
Fig. 1. The electro-hydraulic servo control mechanism of Two-DOF robotic arm (1, 4servo valves, 2, 5hydraulic cylinders, 3, 6pistons, 7relief valve, 8xed displacement
pump, 9tank, 10forearm, 11upper arm, 12torso, 13external load).

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

where pL is the load pressure of the cylinder, xv is the spool


position of the servo valve, ps is the supply pressure of the pump,
Cd is the discharge coefcient, w is the area gradient of the valve
spool, and is the density of hydraulic oil.
The ow-pressure continuous equation of hydraulic cylinder is
Q L Ap

dy
V t dpL
C tl pL
;
dt
4e dt

where y is the displacement of cylinder, Ctl is the coefcient of the


total leakage of the cylinder, e is the effective bulk modulus, Ap is
the annulus area of cylinder chamber, and Vt is the half-volume of
cylinder.
If the major viscous friction of hydraulic oil is simplied as
Coulomb friction, the mechanical dynamic equation driven by an
hydraulic actuator can be constructed as follows
Ap pL my by_ Ky F L ;

where m is the load mass, K is the load spring constant, b is the


viscous damping coefcient, and FL is the external load on the
hydraulic actuator.
The dynamics of the servo valve describes the relationship
between the spool position and the input control voltage. It is one
order linear model [5] as follows
T sv x_ v xv K sv u;

where Ksv is the gain of the control voltage u, Tsv is the response
time constant of servo valve.
In Eq. (1), the function sgn(  ) should be smoothed in the
derivation of backstepping control, which is replaced by the
hyperbolic tangent function tanh(  ) as follows
sgnxv  tanhkxv

ekxv  e  kxv
;
ekxv e  kxv

k0;

where k is a positive constant.


_ pL ; xv T , the output
If the state vector X x1 ; x2 ; x3 ; x4  y; y;
variable y x1 and the control input u(t) are dened, the dynamics
of EHS is constructed as the 4th state space model as follows
8
x_ 1 x2
>
>
>
>
>
1
>
>
x_ 2  Kx1  bx2 Ap x3  F L
>
>
m
>
<
4 Ap
4 C
4 C w p
6
ps  tanhkx4 x3 x4
x_ 3  e x2  e tl x3 epd
>
>
Vt
V
V
>
t
t
>
>
>
>
1
K sv
>
>
x4
u:
> x_ 4 
:
T sv
T sv

vector respectively, 1 and 2 are 2 joint angles, and Tu and Tf are


the torques on 2 hydraulic actuators.
The above equation is unfolded as follows
8
H I1 I 2f m1 P 1 P 2m1 m2f P 1 P 22 m2f P 2 P 2m2 2m2f P 1 P 2  P 2 P m2 cos 2  m1 ;
>
>
> 11
>
< H 12 I m P 1 P 2  P 2 P m2 cos 2  m1 ;
2f
2f
>
> H 21 H12 ;
>
>
: H I m P P2 ;
22

2f

2f

2 m2

9
8
>
C  2m2f P 1 P 2  P 2 P m2 _ 2 sin 2  m1 ;
>
> 11
>
< C C =2;
12
11
> C 21 m2f P 1 P 2  P 2 P m2 _ 1 sin 2  m1 ;
>
>
>
:
C 22 0;
(

10

G1  m1 gP 1 P m1 sin 1 m1  m2f gP 1 P 2 sin 1 m1 P 2 P m2 sin 1 2 ;


G2  m2f gP 2 P m2 sin 1 2 ;

11
where m1 is the upper arm mass including cylinder, m2 is the
forearm mass, mf is the load mass, I1 is the inertia of the upper arm
around the shoulder, I2 is the inertia of the forearm around the
elbow, I2f is the inertia of the forearm around the elbow including
load, m2f is the forearm mass with the load mass, m1f is the load
mass of the upper arm, P 1 P 2 is the length of the upper arm, P 1 P m1
is the distance from the upper arm centroid to shoulder, P 2 P m2 is
the distance from the forearm centroid to elbow, and m1 is the
eccentric angle of cylinder to shoulder.
The known mechanical parameters are substituted into (9)
(11), Tu and Tf can be computed. Then the external load FL in (6) is
divided into two items and described as follows
8
T u 1 ; 2
>
>
;
> F Lu 1 ; 2
>
<
l 1 1
12
T f 1 ; 2
>
>
;
F Lf 1 ; 2
>
>
l 2 2
:
where FLu is the load force on shoulder hydraulic actuator, FLf is the
load force on elbow hydraulic actuator, and the dynamic force
arms li i ; i 1; 2 are computed by triangle geometry.
Remark 1. Due to measurement noise and uncertainty disturbance in engineering practice, FLu and FLf are difcult to be
obtained. The computed value in (12) may deviated from the true
value. Thus, these values should be handled by a lter before used
in parametric estimation law.

2.3. External load model


3. Parametric adaptive backstepping controller
The main external load on two hydraulic actuators are described as the driving torques of Two-DOF robotic arm as shown in
(3). In this study, FL can be described as the variable disturbance
different from other unknown constant parameters such as K, b,
e, Ctl, Cd, w, . To signicantly suppress this dynamic disturbance,
FL is estimated in parametric estimation laws, which is used in
backstepping design.
If all the mechanical parameters are known, the external load FL
is constructed by the Lagrange equation as follows
_ q_ Gq T ;
Hqq Cq; q
"
H
"

H 11

H 12

H 21
#

H 22

1
q
;
2

"
;

"

Tu
Tf

C 11

C 12

C 21

C 22

"
;

G1
G2

#
;
8

where H, C, and G are inertia matrix, Coriolis matrix and gravity

In this paper, all the states xi i 1; ; 4 are known for the


controller design. If these unknown parameters are dened as
1 mK ; 2 mb ; 3 FmL , 4 4VetAp ; 5 4VetC tl ; 6 4Vet pC dw, then the
dynamics of EHS can be described as follows [28,31]
8_
x 1 x2 ;
>
>
>
>
Ap
>
>
>
< x_ 2 m x3  1 x1  2 x2  3 ;
p
13
x_ 3  4 x2  5 x3 6 ps  tanhkx4 x3 x4 ;
>
>
>
>
>
1
K sv
>
>
x4
u:
: x_ 4 
T sv
T sv
Remark 2. Generally, i i 2; 4; 5; 6 are often perturbed by different hydraulic physical characteristic. If the load spring constant
K and the load mass m are given, 1 can be considered as a known
constant. The parameter 3 is either an unknown constant or a
dynamic variable due to the characteristic of the external load FL.

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

Without loss of generality, we dene all of them are unknown


parameters in the study.

If the parametric adaptive estimation laws and the virtual


control 2 are designed as follows

3.1. Recursive controller design

^ 1  k1 x1 z2 ;

At rst, the state error zi i 1; ; 4 and the parametric estimated error ~ i i 1; 6 are dened as follows
(
z1 x1  x1d ;
14
zi xi  i  1 i 2; 3; 4;

~ i i  ^ i ;

i 1; ; 6

where x1d is the demand of cylinder displacement,


virtual control, ^ i is the estimation of i.

15

i is the ith

Theorem 1. Assuming that the unknown parameters i i 1; ; 6


are constant, if the virtual control variables i A C 1 i 1; 2; 3, and
their derivatives _ i i 1; 2; 3 are smooth, then there exists a backstepping controller u, which guarantees zi t i 1; ; 4-0,
~ i t i 1; ; 6-0, as t-1.
Proof. If x_ 1d a 0, we see that
z_ 1 z2 1  x_ 1d :

16

2 

the virtual control

18

1 is designed as follows

1 c1 z1 x_ 1d ;

26

where c2 is a positive constant.


Then the derivative of z3 is
z_ 3 x_ 3  _ 2  4 x2  5 x3 6

q
ps  tanhkx4 x3 x4  _ 2 :

27

If the Lyapunov function V3 is dened as


6
2
X
1
1 
i  ^ i ;
V 3 V 2 z23
2
2ki
i4

28

the derivative of V3 is

Ap
V_ 3  c1 z21  c2 z22 z2 z3 z3  4 x2  5 x3
m
6

_
_
 ^_
q
^
^
4~
ps  tanhkx4 x3 z4 3  _ 2 
4  5 ~ 5  6 ~ 6
0

Ap
z2
@m

k4

k5

^ 6 ps  tanhkx4 x3 z4

k6

A
p
ps  tanhkx4 x3 3  ^ 4 x2  ^ 5 x3  _ 2
0
0
1
1
_
_
_
q
^ 4
^ 5
^ 6
 ~ 4 @ x2 z3 A  ~ 5 @ x3 z3 A  ~ 6 @  x4 z3 ps  tanhkx4 x3 A;
k4
k5
k6
 c1 z21  c2 z22 z3

^ 6
0
1

29

19

where c1 is a positive constant.


Therefore, the derivative of V1 is
V_ 1 c1 z21 z1 z2 :


m
c2 m
m ^
z1 
z2
1 x1 ^ 2 x2 ^ 3  c1 x2 c1 x_ 1d x 1d ;
Ap
Ap
Ap

Ap
V_ 2  c1 z21  c2 z22 z2 z3 ;
m

V_ 1 z1 z2 1  x_ 1d ;

24

25

17

Since the derivative of V1 is

^ 3  k3 z2 ;

we see that the derivative of V2 is

Then the Lyapunov function V1 is dened as


V 1 12 z21 :

^ 2  k2 x2 z2 ;

20

where k4, k5, and k6 are positive constants.


If the parametric adaptive estimation laws and the virtual
control 3 are designed as follows
q
_
_
_
^ 4  k4 x2 z3 ; ^ 5  k5 x3 z3 ; ^ 6 k6 x4 z3 ps  tanhkx4 x3 ;
30

Since the derivative of z2 is


z_ 2 x_ 2  _ 1

Ap
x3  1 x1  2 x2  3 c1 x2 c1 x_ 1d  x 1d
m

Ap
Ap
z3 2  1 x1  2 x2  3 c1 x2  c1 x_ 1d  x 1d ;
m
m

21

31

we see that the derivative of V3 is


q
V_ 3  c1 z21  c2 z22  c3 z23 z3 z4 ^ 6 ps  tanhkx4 x3 ;

the Lyapunov function V2 is dened as follows


3
2
X
1
1 
V 2 V 1 z22
i  ^ i ;
2
2k
i
i1

m
z2 c3 z3 ^ 4 x2 ^ 5 x3 _ 2
Ap
;
3 
p
^ p  tanhkx x

22

where k1, k2, and k3 are positive constants.


Therefore, the derivative of V2 is

Ap
Ap
z3 2  1 x1  2 x2  3 c1 x2
V_ 2 c1 z21 z1 z2 z2
m
m
 ^_
_
_
^
1
2
^ 3
 c1 x_ 1d  x 1d  ~ 1  ~ 2  ~ 3
k1
k2
k3

A
A
p
p
 c1 z21 z2 z1 z3 2  ^ 1 x1  ^ 2 x2  ^ 3 c1 x2
m
m

where c3 is a positive constant.


If the Lyapunov function V4 is dened as
V 4 V 3 12 z24 ;

33

then the derivative of V4 is


V_ 4 V_ 3 z4 z_ 4  c1 z21  c2 z22  c3 z23 z3 z4 ^ 6

q
ps  tanhkx4 x3

z4 x_ 4  _ 3  c1 z21  c2 z22  c3 z23




q
x4 K sv
z4 z3 ^ 6 ps  tanhkx4 x3 

u  _ 3 :
T sv T sv

0
1
0
1
0
1

_
_
_
^ 1
^ 2
^ 3
 c1 x_ 1d  x 1d  ~ 1 @ x1 z2 A  ~ 2 @ x2 z2 A  ~ 3 @ z2 A:
k1
k2
k3

23

32

34

If the controller is designed as


u 

T sv
x4 T sv
T
c4 z4

_ 3  sv z3 ^ 6
K sv
K sv K sv
K sv

q
ps tanhkx4 x3 ;

35

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

then the derivative of V4 is


V_ 4  c1 z21  c2 z22  c3 z23  c4 z24 o 0;

36

where c4 is a positive constant.


Thus, by the parametric adaptive estimation laws (24) and (30),
the elements involving ~ i i 1; ; 6 vanish in V_ 4 . Simultaneously, V_ 4 is negative denite by the virtual control (19), (25),
(31) and the backstepping controller (36). This iteration controller
guarantees the parametric estimation errors ~ i i 1; ; 6 and
the system state error zi i 1; ; 4 converging to zero.
From the assumption in Theorem 1, the derivative of 2 and 3
used in (25) and (31) can be simplied as follows:
m
Ap

_ 2  z2 1  x_ 1d  c2 z3  c2 2
c1 x2  c1 x_ 1d  x 1d 

_ 3 z3 
^ 5

c2 m
 ^ 1 x1  ^ 2 x2  ^ 3
Ap

c1 m ^
x_ 2 ;
Ap

37

m
c2 z2 z1 c3 x_^ 3  _ 2 ^ 4 x_^ 2
Ap

x_^ 3 2
p ;
^ 6 ps  tanhkx4 x3

38

where x_^ 2 and x_^ 3 are estimated by the dynamic model of EHS as
follows
8
< x_^ Ap x  ^ x  ^ x  ^ ;
2
3
1 1
2 2
3
m
:
39
p
:^
^
^
^
x_ 3  4 x2  5 x3 6 ps tanhkx4 x3 x4 :
3.2. Decayed memory lter design
As shown in (37) and (38), although the derivatives _ 2 and _ 3
are computed in backstepping design, the algorithm is still very
complicated. If these derivatives directly used in the controller u,
the robustness of this backstepping controller may decline.
Therefore, a decayed memory lter is designed to re-estimate
these derivatives and ltering the variable parameter 3 mentioned in Remark 1. The lter form is described as follows [32]
8

>
>
>
_^ k  1  T
^ k  1
> E k k 
>
c
>
<
2 ; 3 ; 3
40
^
>
^ k ^ k  1 _ k  1  T c G  E k
>
>
>
>
>
_^ k
_^ k  1 H E k;
:
Tc
^ k is the estimation
where k is the kth calculated value of ,
^
_
_ , T is the
of k, k is the kth estimation of the derivative
c
2
interval of controller, the lter parameters are G 1  ,
H 1  2 , and the range of the lter factor is (0, 1).
Since the calculated values of 2 and 3 are obtained from (25)
and (31), the estimation values ^ i ; _^ i i 2; 3 are obtained by (40).
From (12), the calculated value of 3 is described as follows
8
F ; T ;
>
>
> 31 Lu 1 2 u 1 2 ;
>
<
m1f
m1f l1 1
41
F

T
>
1
2
Lf
f 1 ; 2
>
>
;

32
>
:
m2f
m2f l2 2
where 31 and 32 represent the two calculated values of the
parameter 3 for two hydraulic actuators. After (41) is substituted
^
into (40), the estimation values ^ 3i ; _ 3i i 1; 2 are obtained.
Then the convergence of the proposed hlter should be anaT
lyzed. If the vector k is dened as k ^ k; _^ k , then the

lter (40) is described as the following state space equation


#"
"
# "
# " #
1  G 1  GT c
G
^ k 1
^ k
42
H
H k:
1H
_^ k  1
_^ k
Tc
Tc
Theorem 2. Assuming that k is the input of discrete linear system
(42), and ^ k and _^ k are the outputs, then (42) is input-to-state
stable (ISS), and ^ k-k, _^ k-_ k, as k-1.
Proof. (1) If two constants are dened as 1 1  G, 2 1  H,
then we can see that 1 o 1, 2 o 1, 1 o 2 . The characteristic
polynomial of system matrix A in (42) is described as follows
"
#
s  1  1 T c


sI  A
43
s2  1 2 s 1 0:
 2Tc 1 s  2
Substituting 1 , 2 2  into (43), the eigenvalues of
system matrix A are given by
q
1 2 7 1 2 2 41
:
44
s1;2
2
2

Since the lter factor o 1, the eigenvalues norm is strictly less


than 1. According to the stability of linear discrete system, if the
system input k is bounded, then (42) is input-to-state
stable (ISS).
(2) Since the two outputs ^ k and _^ k are ISS, the lter convergence can be analyzed by two steps. Firstly, from (42), we can
see that

^ k 1  G^ k  1 1 GT c _^ k  1 Gk:

45

When ^ k reach its steady state, there exists a sufciently large


integer n0 ; 8 k 4 n0 , _^ k-0 and ^ k ^ k  1. Substituting
these two conditions into (45), we can obtain that ^ k-k.
Secondly, From (42), the dynamics of _^ k is given by
H
Tc

H
Tc

_^ k  ^ k  1 1  H_^ k  1 k:

46

When _^ k reaches its steady state, there exists a sufciently large


integer n1, 8 k 4n1 , ^ k  1-k  1 and _^ k _^ k 1. Simultaneously, we can see that THc k  k  1 H _ k 1. Substituting these three conditions into (46), we can obtain that
_^ k-_ k.
3.3. Revised parametric adaptive control law
In Theorem 1, the unknown parameter 3 is assumed to be
constant. This assumption is not appropriate due to the variable
external loads on two hydraulic actuators. Thus, aforementioned
parametric adaptive backstepping controller should be revised to
guarantee the EHS (13) convergence.
Theorem 3. Assuming that the unknown parameters 3 is variable
but its derivative _ 3 is bounded, and the other unknown parameters
i i 1; 2; 4; 5; 6 are constants, if the virtual control variables
i A C 1 i 1; 2; 3, and their derivatives _ i i 1; 2; 3 are bounded,
then there exists a revised backstepping controller u, such that
zi t i 1; 4-0, ~ i t i 1; 6-0, as t-1 is guaranteed.
_
Proof. The parametric adaptive estimation law of ^ 3 in (24) can
be revised as follows


_
^ 3  k3 z2 _^ 3 _ 3 sgn3  ^ 3 ;
47
max


_
_
is the bound of the lter error 3 (i.e.,
where 3
max

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

_ 3 _ 3  _^ 3 ),

^
_^
3 is the calculated value from (41), 3 and 3 are
the lter outputs.
Due to _ 3 a 0, the Lyapunov function V2 is rewritten as follows

Ap
Ap
0
z3 2  ^ 1 x1  ^ 2 x2  ^ 3 c1 x2
V_ 2 c1 z21 z1 z2 z2
m
m
 ^_
_
_^
1
2
^ 3  _ 3 ~
3
 c1 x_ 1d  x 1d  ~ 1  ~ 2 
k3
k1
k2

Ap
Ap
 c1 z21 z2 z1 z3 2  ^ 1 x1  ^ 2 x2  ^ 3
m
m
0
1
0
1

^_
^_

1
2
c1 x2  c1 x_ 1d  x 1d  ~ 1 @ x1 z2 A  ~ 2 @ x2 z2 A
k1
k2
0
1
_
^ 3  _ 3
48
z 2 A:
 ~ 3 @
k3
_
_
Substituting (47) and ^ 1 and ^ 2 in (24) into (48), we can see
that

Ap
~ 3  _ _
0
3  3 sgn ~ 3 :
V_ 2 c1 z21  c2 z22 z2 z3
m
k3
max

49

According to the convergence of the lter, _ 3 is bounded by




_
. This denotes that the last element in (49) is less than 0.
3
max

Similarly, if the lter errors _ 2 , _ 3 (i.e., _ 2 _ 2  _^ 2 ,






_ 3 _ 3  _^ 3 ) are bounded by _ 2 max , _ 3 max , then the virtual control

d
^
maxt-1 f_ di t  _^ i tg and maxt-1 f_ 3 t  _ 3 tg, where _ di t


d

j_ i jmax sin 2 t for i 2; 3, _ 3 t _ 3
sin 2 t. The constants
max



_
j_ i jmax and 3
can be estimated in the following simplied
max

controller (55).
Therefore, the revised backstepping controller (52) guarantees
zi i 1; 4 and ~ i i 1; 6 converging to zero. This controller
involves the parametric estimation laws (24), (30), and (47), the
state errors (14), the virtual control laws (19), (25), (50), and the
ltering estimation (40) as shown in Fig. 2.

4. Experiment
In our studies, the Two-DOF robotic arm is employed to implement
and test the performance of the proposed control method. Meanwhile,
several working conditions will be considered when both arms of
robotic are driven simultaneously, or either one joint is run instead.
The comparative experiments are demonstrated by which used
parametric adaptive estimation law and the proposed lter or not.
4.1. Experiment setup
The Two-DOF robotic arm is manufactured by Italian Institute of
Technology shown in Fig. 3. The specic parameters and brands of
main components in the experimental architecture are listed in

3 is revised as follows



m
z2 c3 z3 ^ 4 x2 ^ 5 x3 _^ 2  _ 2 max sgn z3
Ap
:
3 
p
^ p tanhkx x
6

50

Substituting (50) and (30) into (29), the Lyapunov function V3 is


rewritten as follows
q
0
V_ 3 c1 z21  c2 z22 c3 z23 z3 z4 ^ 6 ps  tanhkx4 x3

~ 3  _ _
3  3
k3

max






sgn ~ 3  z3 _ 2 _ 2 max sgn z3 ;
51

where the last element is less than 0.


Finally, if the controller u is redesigned as follows

T sv
x4 T sv ^
T
c4 z4

_ 3  sv _ 3 max sgn z4
K sv
K sv K sv
K sv
q
T sv ^

z3 6 ps  tanhkx4 x3 ;
K sv

u 

Fig. 2. Block diagram of the control system.

52

then the derivative of V4 is rewritten as follows




~ 3
0


sgn ~ 3
V_ 4 c1 z21  c2 z22 c3 z23  c4 z24 _ 3  _ 3
k3
max




 z3 _ 2 _ 2 max sgn z3  z4 _ 3 _ 3 max sgn z4 o 0:
53
By the parametric adaptive estimation law (47), the sign of the
0
dynamic element ~ 3 is guaranteed to be negative in V_ 4 . Similarly,
the signs of z3 and z4 become negative by the virtual control (50)
and the revised backstepping controller (53). This iteration controller guarantees ~ i i 1; ; 6 and zi i 1; ; 4 converging
to zero.








Remark 3. The constants _ 3
, _ 2 max and _ 3 max need
max

to be predened before the revised backstepping controller (52).


Form (41) and (46), these three constants can be estimated by

Fig. 3. The experimental equipment (1 robotic arm, 2 hydraulic cylinder, 3


servo valve, 4 pressure gauge, 5 relief valve, 6 xed displacement pump, 7
servo motor, 8 encoder, 9 pressure sensor, 10 tank, 11 IPC).

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

Table 1
The specic parameters and brand of main components.
Element

Type

Marks

Quantity

Servo motor
Fixed displacement pump
Servo valve
Hydraulic cylinder
Relative encoder
Pressure sensor

BSM63N-375
TFH-315
D633-R02K01M0NSM2
LB6-1610-0080-4M
AEDA-3300-BE1
M3041-000006-350BG

BALDOR
Takako
Moog
Hoerbiger
AVAGO
MEAS

1
1
2
2
2
4

Table 2
Hydraulic parameters used in experiments.
Parameter

Value

Cd
xv max
ps
Vt

0.62

0.024 m

7.9 mm
40 bar

Smax
Ap

79 mm
2:01 cm2

1:74  10  5 m3

Ksv

7:9  10  4 m=V
1000 N/m

e
Tsv

2:2  108 Pa
12 ms

2200 Ns/m

K
C tl
m1
mf
I2
m2f
P1 P2
P 2 P m2

2:5  10  11 m3 =s Pa
1.772 kg
1 kg
0:015 kg m2
1.739 kg
0.35 m
0.12 m

Parameter

m2
I1
I2f
m1f
P 1 P m1
m1

Value

800 kg=m3
0.739 kg
0:071 kg m2
0:022 kg m2
3.511 kg
0.16 m
7.9

Table 1. Some hydraulic and mechanical parameters of this EHS are


shown in Table 2. Since the hydraulic parameters Cd, w, , e, K, b, and
Ctl are not obtained exactly, which have some uncertainties in different experimental conditions, it is necessary to estimate the
uncertain parameters i i 1; 2; 4; 5; 6 by the proposed parametric
adaptive estimation law. However, the approximate values C d , w, ,
e , K , b, and C tl can be preset from some hydraulic references as
shown in Table 2. If the proposed parametric adaptive estimation law
is not adopted, the approximate values i i 1; 2; 4; 5; 6 without
considering parametric uncertainties can also be used in the backstepping controller instead of the estimation values. From the parametric denition in (13), the known nominal values are given by
8
>
s
e
K
K
>
>
1
; 1
>
>
>
m
m2f
1f
>
>
>
>
>
e
b
b
> s
>
>
2
; 2
>
>
m1f
m2f
>
>
>
>
<
4 A
;
54
4 e p
>
Vt
>
>
>
>
>
>
> 5 4 e C tl
>
>
>
Vt
>
>
>
>
>
4 C w
>
>
epd
>
>
: 6
Vt
s

where i i 1; 2, i i 1; 2 are two approximate values of the


shoulder and elbow actuators.
Thus, the simplied backstepping controller is redesigned as
follows

Fig. 4. The 4 state responses of this EHS by the proposed controller.

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

Fig. 5. The 6 estimation values by parametric adaptive estimation laws.

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

8
p
T sv
x4
T sv
T
>
>
u 
c4 z4

_ 3  sv z3 6 ps  tanhkx4 x3
>
>
>
K
K
K
K
sv
sv
sv
sv
>
>
>
m
c m
c m
>
>
>
_  z2 1  x_ 1d  c2 z3  c2 2 2  1 x1  2 x2  3 c1 x2 c1 x_ 1d  x 1d  1 x_ 2
>
< 2
Ap
Ap
Ap
>
x_ 3 2
m
>
>
_ 3  z3  c2 z2 z1 c3 x_ 3  _ 2 4 x_ 2 5 p
>
>
A
>
p

p
>
6
s  tanhkx4 x3
>
>
>
p

>
A
p
>
: x_ 2 x3  1 x1  2 x2  3 x_ 3  4 x2  5 x3 6 ps  tanhkx4 x3 x4 :
m

The stroke of cylinder Smax is 79 mm, but in the experiment, the


displacement of cylinder is not more than 58 mm to avoid the
boundary collision of robotic arm. The initial values of estimated
parameters are predened as zero, i.e., ^ i0 0 i 1; ; 6 to verify
the convergence effectiveness of the parametric estimation laws.
Some control parameters are designed as follows:
(1) The constants: k1 10, k2 0:1, k3 10  3 , k4 10  6 ,
k5 10  9 , and k6 10  5 .
3
(2) The constants: c1 10, c2 1, c3 10  5 , c 4 10
.
_
3
_ 2
(3) The
constants:

5

10
,
5  107 ,


3
max


max
_ 3
0:01.
max
(4) The lter parameters: 0:7. T c 10  3 s.
4.2. Result of the proposed controller
In order to verify the proposed parametric adaptive backstepping controller in (52), two sinusoidal demands of the cylinder
displacement are selected as xs1d 29 sin 0:6 t mm, xe1d
29 sin t mm. The four state responses, six parametric estimations, and two control voltages of the proposed controllers are
shown in Figs. 46.
Fig. 4 a shows the good displacement tracking responses of two
sinusoidal demands by the proposed controllers. The load pressures measured by pressure sensors are not more than the supply
pressure 40 bar shown in Fig. 4c. Due to larger dynamic load on
the shoulder actuator, the motion frequency of shoulder should be
lower than the elbow. During time slice (0, 0.8 s), the control
saturation arises since the initial tracking error is signicant. After
0.8 s, the spool positions of two servo valves are not exceeded the
saturation 7 xv max , which are similar to the dynamic characteristic of two control voltages us and ue shown in Fig. 6. In the
experiment, the robotic arm motion will generate resonance effect
when the tracking tendency of two joint angles approaches each

55

other. Especially in some time such as 8 s, 11 s, 18 s, two hydraulic


actuators need simultaneously suffering two maximum external
loads, the instant control supplement is signicant as shown in
Fig. 6a and b.
The six uncertain parameters are estimated respectively by
the parametric adaptive estimation laws (24), (30), and (47)
s
e
shown in Fig. 5. The parametric estimations ^ and ^ respecs

tively approach the nominal values 1 , 1 , which denote that the


load spring constant is almost consistent with the approximate
preset values. The other four parametric estimations ^ 2 , ^ 4 , ^ 5 ,
and ^ 6 also converge to respective steady-state values like ^ 1 .
However, these parametric estimations have obviously deviated
from the approximate preset values respectively. This phenomenon indicates that the permanent parameters 2 , 4 , 5 , and 6
cannot show the actual hydraulic parameters Cd, w, , e, b, Ctl
with some uncertainties in different working conditions. Different from the steady characteristic of the other ve parameters,
s
e
^ , ^ are dynamic estimations, which described two variable
3

external loads of the hydraulic actuators caused by Two-DOF


robotic arm motion. If the parametric estimated errors of ^ i i
2; 4; 5; 6 are signicant, the static control bias emerges in the
long time shown in the initial 1 s in Fig. 7f. If the dynamic estis
e
mations of ^ and ^ are inaccurate, the dynamic control
3

saturation emerges in Fig. 8f.


4.3. Comparison results
4.3.1. Comparison with simplied backstepping controller
To illustrate the problem, the comparison results for two different backstepping controllers are given in two critical conditions
where the external load of the hydraulic actuator is close to the
limitation. The proposed controller (52) involves the parametric
estimation laws and the ltering estimation. The other simplied

Fig. 6. The control voltages of 2 hydraulic actuators by the proposed controller.

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

10

Q. Guo et al. / ISA Transactions ()

Fig. 7. The comparison result for the demand input xs1d 14:5 sin 2t mm.

backstepping controller (55) has no parametric estimation laws


and no ltering estimation, where the approximate preset parameters 1 , 2 , 4 , 5 , 6 are determined by (54), 3 is computed
by (41), not used (40) and (47). Different from the above experimental condition, the elbow joint is xed and the displacement
demands of shoulder are selected as xs1d 14:5 sin 2 t mm,
xs1d 29 sin t mm. The comparison results are shown in
Figs. 7 and 8. The derivatives _ 2 and _ 3 obtained by the proposed

lter are signicantly smaller than the computation values by (55)


as shown in Fig. 7ad. This experimental phenomenon denote that
the proposed lter has more capability to suppress the violent
derivative of virtual control than the conservative approach based
on model computation. Fig. 7e shows that the proposed controller
has higher tracking performance, since the parametric estimation
laws is used to adapt the actual hydraulic parameter with uncertainties. On the contrary, the simplied backstepping controller

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

11

Fig. 8. The comparison result for the demand input xs1d 29 sin t mm.

(55) with certain parameters assumption declines the dynamic


behavior of the closed control loop and results into the sharp
controller as shown in Fig. 7f.
Especially in a large stroke motion experiment of shoulder
actuator as shown in Fig. 8, the simplied backstepping controller
could not obtain the satisfactory tracking performance. However,
to some extent, the proposed lter still guarantees the dynamic
response of shoulder actuator. These results show that the large
stroke motion is easier to cause the saturated and sharp controller

than the high frequency motion due to the dynamic external load
on the shoulder actuator.
4.3.2. Comparison with PI controller
Similarly for the same tracking demands of two conditions shown
in the above section, the comparison results with PI controller are
given in Figs. 9 and 10. Here the PI controller is described as
R
us K p xs1 K i xs1 dt, where xs1 xs1d  xs1 . The two parameters
Kp 100 and Ki 15, which have been well tuned to guarantee fast

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

12

Q. Guo et al. / ISA Transactions ()

Fig. 9. The comparison result for the demand input xs1d 29 sin t mm.

Fig. 10. The comparison result for the demand input xs1d 14:5 sin 2t mm.

Fig. 11. The comparison result for the demand input xs1d 29 sin 2t mm.

response of the shoulder actuator. When the motion frequency is


0.5 Hz, the tracking error of the shoulder cylinder is no more than
3 mm as shown in Fig. 9a, which indicates that both two controllers
can obtain good dynamic response performance under the low frequency demand xs1d 29 sin t mm. Then the motion frequency is

increased to 1 Hz (i.e., xs1d 14:5 sin 2 t mm). The dynamic


behavior of the proposed controller is better than the PI controller as
shown in Fig. 10a. Slight chatters emerge in PI control as shown in
Fig. 10b. If the control parameters are reduced to eliminate the
chatter phenomenon, the tracking accuracy will be declined this

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i

Q. Guo et al. / ISA Transactions ()

moment. This denotes that the proposed controller is more effective


for the parametric uncertainties adaptation and load disturbance
suppression comparison with the PI controller.
4.3.3. Comparison with general parametric adaptive controller
To verify the effectiveness of the proposed controller in more
severe condition, the third tracking demands xs1d 29 sin 2 t
mm and the elbow cylinder is always extended. Both controllers
are compared in this condition. One is the proposed controller, and
the other is the general parametric adaptive controller with no
decay memory lter, which is derived from (24), (30), (19), (25),
(31), (35). When the frequency is 1 Hz and the motion range is
increased to the whole stroke of cylinder, the external load disturbance is signicantly increased. The dynamic performance of
the proposed controller is better than the general parametric
adaptive controller as shown in Fig. 11a. The latter controller
emerges control saturation and slight chatters as shown in Fig. 11b,
which indicates that the proposed decay memory lter can
attenuate the control saturation caused by the derivative explosion
of virtual control in backstepping design.

5. Conclusions
In this paper, a parametric adaptive backstepping controller
based on state feedback is presented to implement the displacement control of electro-hydraulic actuator. The dynamic models of
this EHS and the external loads were rst constructed. Then, the
parametric adaptive estimation law is designed to estimate six
uncertainty parameters, and the recursive backstepping controller
was established by Lyapunov technique. After that, a decayed
memory lter is used to estimate the two virtual control variables
and corresponding derivatives. Simultaneously, the backstepping
controller is also revised to guarantee the convergence of the state
error and the parametric estimated error. The sinusoidal tracking
results verify this proposed method in the motion control of the
robotic arm. Furthermore, the effectiveness of the proposed controller comparison with the simplied backstepping controller is
well demonstrated in critical working conditions.

Acknowledgments
The authors acknowledge Prof. Andrew Plummer, Director of
Power Center for Power Transmission and Motion Control,
Department of Mechanical Engineering, University of Bath for the
funding assistance and previous work. This work was supported
by the National Natural Science Foundation of China (Nos.
61305092 and 51205045), the Postdoctoral Science Foundation of
China (No. 2013M542487), and the Open Foundation of the State
Key Laboratory of Fluid Power Transmission and Control (No.
GZKF-201515).

References
[1] Manring ND. Hydraulic control systems. Hoboken, NJ: John Wiley and Sons;
2005.
[2] Fales R, Kelkar A. A robust control design for a wheel loader using mixed
sensitivity H-innity and feedback linearization based methods. In: Proceedings of the 2005 American control conference, Portland, OR, USA, 2005. p.
43816.

13

[3] Mili V, itum , Essert M. Robust H1 position control synthesis of an electrohydraulic servo system. ISA Trans 2010;49(October (4)):53542.
[4] YU H, Feng ZJ, Wang XY. Nonlinear control for a class of hydraulic servo system. J Zhejiang Univ Sci 2004;5(November (11)):14137.
[5] Yao B, Bu FP, Reedy J, Chiu T-C. Adaptive robust motion control of single-rod
hydraulic actuators theory and experiments. IEEE/ASME Trans Mechatron
2000;5(March (1)):7991.
[6] Bu F, Yao B. Observer based coordinated adaptive robust control of robot
manipulators driven by single-rod hydraulic actuators. In: Proceedings of the
2000 IEEE-ICRA, San Francisco, CA, USA, 2000. p. 30349.
[7] Kim W, Shin D, Won D, Chung CC. Disturbance-observer-based position
tracking controller in the presence of biased sinusoidal disturbance for electrohydraulic actuators. IEEE Trans Control System Technol 2013;21(November
(6)):22908.
[8] Won D, Kim W, Shin D, Chung CC. High-gain disturbance observer-based
backstepping control with output tracking error constraint for electrohydraulic systems. IEEE Trans Control Syst Technol 2015;23(March (2)):787
95.
[9] Guo Q, Yu T, Jiang D. Robust H 1 positional control of 2-DOF robotic arm driven
by electro-hydraulic servo system. ISA Trans 2015;59(November (11)):5564.
[10] Guo Q, Yu T, Jiang D. High-gain observer-based output feedback control of
single-rod electro-hydraulic actuator. IET Control Theory Appl 2015;9
(November (16)):2395404.
[11] He W, Zhang S, Ge SS. Robust adaptive control of a thruster assisted position
mooring system. Automatica 2014;50(July (7)):184351.
[12] He W, Zhang S, Ge SS. Adaptive control of a exible crane system with the
boundary output constraint. IEEE Trans Ind Electron 2014;61(August
(8)):412633.
[13] Niksefat N, Sepehri N. Design and experimental evaluation of a robust force
controller for an electro-hydraulic actuator via quantitative feedback theory.
Control Eng Pract 2000;8(December (12)):133545.
[14] Ursu I, Toader A, Halanay A, Balea S. New stabilization and tracking control
laws for electrohydraulic servomechanisms. Eur J Control 2013;19(January
(1)):6580.
[15] Ursu I, Ursu F, Popescu F. Backstepping design for controlling electrohydraulic.
J Frankl Inst 2006;343(January (1)):94110.
[16] Alleyne A, Liu R. A simplied approach to force control for electro-hydraulic
systems. Control Eng Pract 2000;8(December (12)):134756.
[17] Alleyne A, Liu R. Systematic control of a class of nonlinear systems with
application to electrohydraulic cylinder pressure control. IEEE Trans Control
Syst Technol 2000;8(July (4)):62334.
[18] Krstic M, Kanellakopoulos I, Kokotovic PV. Nonlinear and adaptive control
design. New York: Wiley; 1995.
[19] Khalil HK. Nonlinear systems. 3rd ed. New Jersey: Prentice-Hall; 2001.
[20] Bu FP, Yao B. Nonlinear model based coordinated adaptive robust control of
electro-hydraulic robotic arms via overparametrizing method. In: Proceedings
of the 2001 IEEE-ICRA, Seoul, Korea, 2001. p. 345964.
[21] Guan C, Pan SX. Nonlinear adaptive robust control of single-rod electrohydraulic actuator with unknown nonlinear parameters. IEEE Trans Control
Syst Technol 2008;16(May (3)):43445.
[22] Guan C, Pan SX. Adaptive sliding mode control of electro-hydraulic system
with nonlinear unknown parameters. Control Eng Pract 2008;16(November
(11)):127584.
[23] Kaddissi C, Kenne JP, Saad M. Identication and real-time control of an electrohydraulic servo system based on nonlinear backstepping. IEEE/ASME Trans
Mechatron 2007;12(February (1)):1222.
[24] Ahn K, Nam D, Jin M. Adaptive backstepping control of an electrohydraulic
actuator. IEEE/ASME Trans Mechatron 2014;19(June (3)):98795.
[25] Sun H, Chiu GT-C. Nonlinear observer based force control of electro-hydraulic
actuators. In: Proceedings 1999 American control conference, San Diego, CA,
USA, 1999. p. 7648.
[26] Yao J, Jiao Z, Ma D. Extended-state-observer-based output feedback nonlinear
robust control of hydraulic systems with backstepping. IEEE Trans Ind Electron
2014;61(November (11)):628593.
[27] Pi YJ, Wang XY. Observer-based cascade control of a 6-DOF parallel hydraulic
manipulator in joint space coordinate. Mechatronics 2010;20(September
(6)):64555.
[28] Kim W, Won D, Shin D, Chung CC. Output feedback nonlinear control for
electro-hydraulic systems. Mechatronics 2012;22(September (6)):76677.
[29] Kim W, Won D, Chung CC. High gain observer-based nonlinear position control for electro-hydraulic servo systems. In Proceedings of the 2010 American
control conference, Baltimore, MD, USA, 2010. p. 14406.
[30] Nakkarat P, Kuntanapreeda S. Observer-based backstepping force control of an
electrohydraulic actuator. Control Eng Pract 2009;17(August (8)):895902.
[31] Kim W, Won D, Tomizuka M. Flatness-based nonlinear control for position
tracking of electrohydraulic aystems. IEEE/ASME Trans Mechatron 2015;20
(February (1)):197206.
[32] Zarchan P. Tactical and strategic missile guidance. Reston, VA: AIAA; 1997.

Please cite this article as: Guo Q, et al. Parametric adaptive estimation and backstepping control of electro-hydraulic actuator with
decayed memory lter. ISA Transactions (2016), http://dx.doi.org/10.1016/j.isatra.2016.02.009i