IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO.

3, JUNE 2014 987

Adaptive Backstepping Control of an

Electrohydraulic Actuator
Kyoung Kwan Ahn, Member, IEEE, Doan Ngoc Chi Nam, and Maolin Jin, Member, IEEE

Abstract—This paper presents an adaptive position control for

a pump-controlled electrohydraulic actuator (EHA) based on an
adaptive backstepping control scheme. The core feature of this
paper is the combination of a modified backstepping algorithm
with a special adaptation law to compensate all nonlinearities and
uncertainties in EHA system. First of all, the mathematical model
of the EHA is developed. The position control is then formulated
using a modified backstepping technique and the uncertainties in
hydraulic system are adapted by employing a special Lyapunov
function. The control signal consists of an adaptive control signal
to compensate the uncertainties and a simple robust structure to
ensure the robustness corresponding to a bounded disturbance.
Experimental results proved strongly the ability of the proposed
control method.
Index Terms—Adaptive control, backstepping control, electro-
hydraulic actuator (EHA), pump-controlled hydraulic system.

I. INTRODUCTION Fig. 1. Structure of the pump-controlled EHA.

YDRAULIC piston–cylinders have been the most popu-

H lar actuators in modern industries where high actuating
forces are needed. Applications of hydraulic piston–cylinder directly by operation of the bidirectional pump. Obviously, the
have been studied for years using valve control systems [1]–[3]. pump-controlled EHA acts as a Power-Shift system, which
The configurations of these systems have been generally sim- shifts the high speed from electric motor into the high force
ple, where a pressure compensated or load–sensing pump is at hydraulic cylinder. Thus, comparing with the conventional
employed to supply pressurized fluid for the valve-controlled hydraulic system, the pump-controlled EHA creates a sleeker,
piston–cylinder. Then, the operation is decided via adjusting cleaner way to produce hydraulic power with higher energy effi-
the state of the control valve. However, this method usually re- ciency. Because of its advantages, the pump-controlled EHA has
sults in low-energy efficiency due to leakage via bypass valves been developed as commercial products, e.g., the mini-motion
of hydraulic pump and throttle loss at control valves. Conse- package from Kayaba Industry Company [6], the intelligent
quently, several techniques have been developed to overcome hydraulic servo drive pack from Yuken Kogyo Company [7].
this drawback and to satisfy the developing demands in hy- Most of the previous literatures focused onto force or
draulic engineering. displacement control of a valve-controlled hydraulic piston–
Recently, a novel concept of pump-controlled electrohy- cylinder. The control problems started with linear and linearized
draulic actuator (EHA) has been introduced as an advanced control techniques without concerning the system uncertain-
hydraulic system [4], [5]. The structure of this EHA is shown ties [1], [8], and [9]. Obviously, valve-controlled hydraulic
in Fig. 1. Here, the position of hydraulic cylinder is adjusted piston–cylinder is a highly nonlinear system with many uncer-
tainties due to the nonlinear dynamics of cylinder and uncertain
fluid parameters. Hence, nonlinear adaptive control techniques
have been introduced based on backstepping control [3], [10],
Manuscript received October 1, 2012; revised January 22, 2013 and May 1,
2013; accepted May 8, 2013. Date of publication June 13, 2013; date of current and [11], or on an adaptive friction model [12]. These nonlin-
version April 11, 2014. Recommended by Technical Editor S. C. Mukhopad- ear adaptive control laws have solved the control problem in
hyay. This work was supported by the SW Fusion Component Development of valve-controlled hydraulic piston–cylinder. However, the con-
the KEIT grant funded by the Korea Government Ministry of Trade, Industry
and Energy under Grant 10043810. trol problems are more challenging for pump-controlled EHA
K. K. Ahn is with the School of Mechanical Engineering, University of Ulsan, systems.
Ulsan 680-749, Korea (e-mail: [email protected]). For the pump-controlled EHA system, the dynamics of the
D. N. C. Nam is with the Graduate School of Mechanical Engineering, Uni-
versity of Ulsan, Ulsan 680-749, Korea (e-mail: [email protected]). bidirectional pump has to be considered together with the hy-
M. Jin is with the Research Institute of Industrial Science and Technology, draulic cylinder. Moreover, hydraulic cylinders are generally
Pohang 790-330, Korea (e-mail: [email protected]). asymmetric systems, i.e., the different annulus areas on two
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. sides of actuating piston. Thus, check valves are needed to sup-
Digital Object Identifier 10.1109/TMECH.2013.2265312 ply or discharge the supplement volume of fluid as shown in
1083-4435 © 2013 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.
988 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO. 3, JUNE 2014

Fig. 1. Consequently, the dynamics of pump-controlled EHA Assume there is no external leakage, the dynamics of oil flow
system become more complex. can be written as in [18], [19]
In order to obtain a position tracking for a pump-controlled ⎧
⎪ βe
EHA, fuzzy logic self-tuning PID controllers [13], [14] have ⎪
⎨ ṖH = V + A y (QH − AH ẏ − Ct (PH − PR ))
been developed. However, the design of fuzzy structure depends 01 H
(2)
crucially on expert experience and may not have enough relia- ⎪
⎪ β
⎩ ṖR = e
(QR + AR ẏ + Ct (PH − PR ))
bility in some applications. To deal with the nonlinearities and V02 − AR y
uncertainties of pump-controlled EHA, robust position controls
using sliding mode control scheme [15], [16], and nonlinear where Ct is the coefficient of the internal leakage of the cylinder,
adaptive control scheme based on backstepping controller [17] βe is the effective bulk modulus of the hydraulic fluid in the
have been adopted. Since the variable control structure is ap- container, V01 , V02 are the original total control volumes of the
plied, the sliding mode control methods [15], [16] face chatter- two chambers, respectively (including the volume of pipelines
ing problems, which can damage the experimental instruments. and initial cylinder chambers), QH represents the supply flow
In [17], the nonlinearities of EHA system are considered as rate to the H chamber (cylinder end), and QR represents the
single term, and the developed controller is applicable for a supply flow rate to the R chamber (rod end).
symmetric actuating system only. Based on the configuration of EHA system, flow rates into
In this paper, an adaptive control of a pump-controlled EHA two chambers are calculated as
system is developed based on a novel adaptation law and a mod- 
QH = QP H + Qv 1 − Qv 3
ified backstepping technique. First of all, the dynamics of the (3)
pump-controlled EHA system is investigated. Different from QR = QP R + Qv 2 − Qv 4
the works in [4], [15], and [17], the considered EHA system in where QP H = −QP R = Qpum p is the pump flow rate
this paper employs an asymmetric hydraulic cylinder. Hence,
the flow-rate terms through supplement check valves have to be Qpum p = Dω − kleakage (PH − PR ) . (4)
considered in the dynamic analysis incorporated with the hy-
draulic cylinder and the bidirectional pump. Consequently, the Here, kleakage is the leakage constant, D is displacement of the
uncertainties in this paper include not only the variation of the ef- pump, and ω is the speed of the servo pump system. The terms:
fective bulk modulus but also the flow-rate through supplement Qv 1 , Qv 2 , Qv 3 , and Qv 4 are flow rates through valves v1 , v2 , v3 ,
check valves, and the initial hydraulic volume, i.e., the hydraulic and v4 , respectively.
volume in two chambers of hydraulic cylinder, hoses, and servo
pump; which were not considered in [9]–[12], and [15]–[17].
B. Control Problem Statement
Since these parameters are all treated as uncertainties, a novel
adaptive law is developed based on a special Lyapunov func- In normal operation, the system pressure values: PH , PR
tion. Consequently, a control scheme is constructed based on a should be kept under the setting value Pset of the two relief
modified backstepping algorithm. The control signal consists of valves: v3 and v4 . It means the flow rates through the two relief
an adaptive control signal to compensate the uncertainties and valves: Qv 3 and Qv 4 maintain zero. Consequently, (3) can be
a simple robust structure to ensure the robustness correspond- rewritten as
ing to a bounded disturbance. Finally, an experimental test rig 
is setup and three cases of study are carried out. The experi- QH = QP H + Qv 1
(5)
mental results with high tracking accuracy (up to ±0.25 mm of QR = QP R + Qv 2 .
accuracy) and robustness in various working conditions prove
strongly the applicability of proposed control method in modern For the ease of calculation, let the system states be defined as
hydraulic applications, e.g., press machine, punching machine. Δ
x = (x1 , x2 , x3 , x4 ,)T = (y, ẏ, PH , PR ,)T . (6)

II. MODELING OF EHA SYSTEM Then, the simplified mathematical model of EHA system can
be described by employing (1)–(5)
A. Mathematical Modeling of EHA System ⎧
⎪
⎪ ẋ1 = x2
Considering the EHA system as shown in Fig. 1, the dynamics ⎪
⎪
⎪
⎪ 1 F
of the piston can be described by the second Newton’s Law ⎪ ẋ2 =
⎪ (AH x3 − AR x4 ) −
⎪
⎪ m m
⎪
⎪
⎪
⎨ βe 
mÿ = PH AH − PR AR − F (1) ẋ3 = Dω − (kleakage + Ct ) (x3 − x4 )
V01 + AH x1
⎪
⎪
⎪
⎪ +Qv 1 − AH x2
⎪
⎪
where m is the equivalent mass, y is system displacement; ⎪
⎪ βe 
⎪
⎪ ẋ4 = − Dω + (kleakage + Ct ) (x3 − x4 )
PH , PR are pressure in two chambers, respectively, AH , AR ⎪
⎪ V02 − AR x1
⎪
⎩
are actuating areas, respectively, and F is external loaded force −Qv 2 + AR x2 .
on the cylinder. (7)
AHN et al.: ADAPTIVE BACKSTEPPING CONTROL OF AN ELECTROHYDRAULIC ACTUATOR 989

In order to simplify the system (7), define Then,

⎧ AH x3 − AR x4
⎪
⎪ x34 =
⎪
⎪ m V̇1 = z1 ż1 = z1 z2 − k1 z12 . (16)
⎪
⎪
⎪
⎪ F
⎪
⎪ d=−
⎪
⎪
⎨ m It is obvious that if the error z2 is close to zero, the tracking
kleak = kleakage + Ct (8) error z1 will converge to zero. Thus, the next step is to control
⎪
⎪
⎪
⎪ A β z2 as small as possible.
⎪
⎪ β=
H e
⎪
⎪ m Step 2: Taking the time derivative of (13), one obtains
⎪
⎪
⎪
⎪ A
⎩ kA = R .
AH ż2 = ẋ2 − ẋ2d = x34 + d − ẍ1d + k1 ż1 . (17)
Then, the state space (7) can be rewritten in strictly feedback
form as Virtual control signal for the system (17) is designed as
⎧
⎪
⎪ ẋ1 = x2
⎪
⎪
⎪
⎪ ẋ = x34 + d Δ H
⎪
⎪ 2
x34d = ẍ1d − z1 − k1 ż1 − k2 z2 , k2 = (18)
⎪
⎪ δ
⎪
⎪ β kA β
⎨ ẋ34 = + Dω+
V01 + AH x1 V02 − AR x1
⎪ where H = sup |d|, and δ > 0 is the acceptable bound of |z2 |.
⎪
⎪ β
⎪
⎪ + (Qv 1 − kleak (x3 − x4 ) − AH x2 ) Define the state error
⎪
⎪ V + AH x1
⎪
⎪
01
⎪
⎪ kA β
⎪
⎩ − (−Qv 2 + kleak (x3 − x4 ) + AR x2 ) .
Δ
z34 = x34 − x34d = ż2 + z1 + k2 z2 − d. (19)
V02 − AR x1
(9)
Obviously, system states are adjusted by the speed of the bidi- Consider a Lyapunov function
rectional pump that is driven by an ac servo motor. Given a
bounded desired trajectory: x1d , the objective of this paper is to 1
determine the input speed command for the ac servo motor ω to V2 = V1 + z22 . (20)
2
control the output position x1 track as close as possible to x1d .
The development of position controller will be described in the Then,
following section.
V̇2 = V̇1 + z2 ż2 = −k1 z12 − k2 z22 + z2 (z34 + d) . (21)
III. CONTROL DESIGN
In order to find the correct speed command for EHA system,
It is obvious that if the error z34 is close to a small value, the
a backstepping controller is developed for the state space in (9).
error z2 will converge to around δ. Hence, the next step is to
The procedure for designing the controller is presented in this
control z34 as small as possible.
section as follows.
Step 3: This step aims to synthesize an actual control law for
Step 1: Let the system tracking error be defined as
the input ω of the system. For simplicity, the last equation in (9)
Δ
z1 = x1 − x1d . (10) can be rewritten as

Then, the time derivative term of tracking error is calculated as h1 (x) ω + h2 (x)
follows: ẋ34 = (22)
g (x)
ż1 = x2 − ẋ1d . (11)
where
The virtual control signal for the system in (11) is defined as
⎧
Δ
x2d = ẋ1d − k1 z1 , k1 > 1. ⎪ g (x) = (V01 + AH x1 ) (V02 − AR x1 )
(12) ⎪
⎪
⎪
⎪
⎪
⎪ = V01 V02 + (V02 AH − V01 AR ) x1 − AH AR x21 > 0
Define the state error ⎪
⎪
⎪
⎪
⎪ h1 (x) = (βV02 + kA βV01 ) D
⎪
Δ
z2 = x2 − x2d . (13) ⎪
⎪
⎪
⎨ h2 (x) = β (V02 − AR x1 ) (Qv 1 − kleak (x3 − x4 ) − AH x2 )
It follows ⎪ −βkA (V01 +AH x1 ) (−Qv 2 +kleak (x3 − x4 )+AR x2 )
⎪
⎪
⎪
⎪
z2 = ż1 + k1 z1 . (14) ⎪
⎪
⎪
⎪ = − (V02 + kA V01 ) kleak β (x3 − x4 )
⎪
⎪
⎪
⎪ − (V02 AH +kA V01 AR ) βx2 + (−Qv 1 +Qv 2 ) AR βx1
Consider a Lyapunov function ⎪
⎪
⎪
⎩
1 2 +βAR (AH − AR ) x1 x2 + β (V02 Qv 1 + kA V01 Qv 2 ) .
V1 = z . (15) (23)
2 1
990 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO. 3, JUNE 2014

Here, uncertainties in hydraulic system are treated as an adapted as follows:

vector ⎧ 1
⎧  T ⎪
⎪ ω= T (ω1 + ω2 )
⎪ ⎪
⎪ α̂h1 h̄1 (x)
⎪
⎪ α
Δ
= α 1 α 2 α3 α 4 α5 α6 α7 α 8 ⎨
⎪
⎪
⎪
⎪ ⎪
1
ω1 = α̂gT ḡ(x)ẋ34d − α̂gT ḡ˙ (x)z34 − α̂h2
T (32)
⎪
⎪ Δ ⎪
⎪
h̄2 (x)
⎪
⎪ α1 = V01 V02 ⎪
⎩ 2
⎪
⎪
⎪
⎪ Δ ω2 = −z2 − k3 z34 − k4 Hsign(z34 )
⎪
⎪ α2 = V02 AH − V01 AR
⎪
⎪
⎪
⎪ Δ where ω1 is the adaptive signal to compensate the uncertainties
⎪
⎨ α3 = βV02 + kA βV01 inside the EHA system, ω2 is the robust signal which deals the
Δ (24)
⎪ α4 = (V02 + kA V01 ) kleak β disturbance d (t) , k3 , and k4 > 1 are positive constants. The
⎪
⎪
⎪
⎪ Δ idea to form (32) comes from the inversion of (22). With the
⎪
⎪ α5 = (V02 AH + kA V01 AR ) β
⎪
⎪ above control signal, one obtains
⎪
⎪α =
⎪
⎪
Δ
(−Qv 1 + Qv 2 ) AR β
⎪
⎪
6 1
⎪
⎪ V̇3 = −k3 z34
2
− z2 z34 − k4 Hz34 sign (z34 ) + α̃gT ḡ˙ (x) z34
2
⎪
⎪
Δ
α7 = βAR (AH − AR ) 2
⎪
⎪
⎪
⎩ Δ
α8 = β (V02 Qv 1 + kA V01 Qv 2 ) .
T
+ z34 α̃h1 h̄1 (x) ω − z34 α̃gT ḡ (x) ẋ34d + z34 α̃h2
T
h̄2 (x) .

Then, (23) can be rewritten as (33)

⎧ The adaptation scheme for uncertainties is developed by con-
⎨ g (x) = α1 + α2 x1 − AH AR x1 > 0
2
⎪
sidering the following Lyapunov function:
h1 (x) = α3 D > 0
⎪
⎩ 1  T −1 
h2 (x) = −α4 (x3 − x4 ) − α5 x2 + α6 x1 + α7 x1 x2 + α8 . Va = V2 + V 3 + α̃ T α̃g + α̃h1 T −1
Th1 T
α̃h1 + α̃h2 −1
Th2 α̃h2
(25) 2 g g
(34)
Since g (x) ≥ 0, define the Lyapunov function as
where Tg , Th1 , and Th2 are the positive definite constant diag-
1 2 onal matrices. The time derivative of (34) is calculated
V3 = g(x)z34 . (26) 
2
V̇a = V̇2 + V̇3 − α̃gT Tg−1 α̂˙ g + α̃h1
T −1 ˙
Th1 T
α̂h1 + α̃h2 −1 ˙
Th2 α̂h2
The time derivative of (26) is calculated as follows:
1 2
= −k1 z12 − k2 z22 + z2 d − k3 z34 2
− k4 Hz34 sign (z34 )
V̇3 = ġ(x)z34 + g(x)z34 ż34

2 1
+ α̃gT 2
ḡ˙ (x) z34 − z34 ḡ (x) ẋ34d − Tg−1 α̂˙ g
1 2
2
= ġ(x)z34 + z34 (h1 (x) ω + h2 (x) − g(x)ẋ34d ) . (27) 
2
+ α̃T z34 h̄1 (x) ω − T −1 α̂˙ h1
h1 h1
Here, the time derivative of ẋ34d is calculated as

−1 ˙
ẋ34d = ˙ẍ1d − k1 z̈1 − ż1 − k2 ż2 . (28)
T
+ α̃h2 z34 h̄2 (x) − Th2 α̂h2 . (35)

For simplicity, define The update laws can be chosen so that these inequalities hold
⎧ T ⎧

⎪
⎨ αg = [ α1 α2 1 ] ⎪
⎪ α̃ T 1
˙
ḡ (x) z 2
− z ḡ (x) ẋ − T −1 ˙
α̂ ≤0
⎪
⎪ g 2 34 34 34d g g
αh1 = α3 (29) ⎪
⎨
⎪
⎩ 
−1 ˙ (36)
αh2 = [ α4 α5 α6 α7 α8 ] T
⎪ α̃ T
z 34 h̄ 1 (x) ω − T α̂ h1 ≤0
⎪
⎪
h1 h1
⎪
⎪ 
and ⎩ α̃T z h̄ (x) − T −1 α̂˙ ≤ 0.
⎧ 2 T
h2 34 2 h2 h2

⎨ ḡ (x) = [ 1 x1 −AH AR x1 ]
⎪
In this paper, the adaptation laws for uncertain terms are de-
h̄1 (x) = D (30) signed as follows:
⎪
⎩
h̄2 (x) = [ − (x3 − x4 ) −x2 x1 x1 x2 1 ]T . ⎧ 1
⎪ ˙
⎨ α̂g = 2 Tg z34 (ḡz 34 − 2ḡ ẋ34d )
⎪ ˙
Obviously, the following statement holds:
⎧ α̂˙ h1 = Th1 z34 h̄1 ω (37)
g (x) = αgT ḡ (x) ⎪
⎪
⎪
⎨ ⎩
T α̂˙ h2 = Th2 z34 h̄2 .
h1 (x) = αh1 h̄1 (x) (31)
⎪
⎩ T
h2 (x) = αh2 h̄2 (x) . For system (22), there exist positive boundaries
αg m in , αh1 m in > 0 where
Let α̂g , α̂h1 , and α̂h2 be the estimation of αg , αh1 , and αh2 , 
respectively; α̃g = αg − α̂g , α̃h1 = αh1 − α̂h1 , and α̃h2 = αh1 > αh1 m in
(38)
αh2 − α̂h2 be estimation errors. The control input ω is designed αg > αg m in .
AHN et al.: ADAPTIVE BACKSTEPPING CONTROL OF AN ELECTROHYDRAULIC ACTUATOR 991

Fig. 2. Structure of experiment test rig using two EHAs.

Thus, these boundary conditions are employed TABLE I

⎧  SETTING PARAMETERS FOR THE EHA SYSTEM
⎪ α̂g = αg m in
⎪
⎪ if α̂g < αg m in , then ˙
⎨ α̂g = 0
 (39)
⎪
⎪ α̂h1 = αh1 m in
⎪
⎩ if α̂h1 < αh1 m in , then
α̂˙ h1 = 0.
Obviously, the boundary conditions (39) still satisfy the
inequalities (36).
Theorem 1: Given a nonlinear pump-controlled EHA system
with uncertainties as (7), employing the controller (32) with
the adaptation scheme (37) and boundary condition (39), all
signals z1 , z2 , z34 , α̃g , α̃h1 , α̃h2 , α̂g , α̂h1 , α̂h2 are bounded and
lim z1 ≤ kδ1 .
t→∞
Proof: From (34), (35), (37), and (39), one can obtain
V̇a ≤ −k1 z12 − k2 z22 + z2 d
− k3 z34
2
− k4 Hz34 sign (z34 ) . (40)
Considering the following term:
sup |d| 2
Z = −k2 z22 + z2 d = − z2 + z2 d. (41) Applying Laplace transform into (14), one obtains
δ
z2
Note that z1 = . (45)
⎧ s + k1
⎨ if (|z2 | ≥ δ) then Z ≤ 0
⎪
Thus,
if ( (|z2 | < δ) and (z2 d ≤ 0)) then Z ≤ 0 (42)
⎪
⎩ δ
if ((|z2 | < δ) and (z2 d > 0)) then Z ≤ Hδ. lim z1 ≤ . (46)
t→∞ k1
Then, Consequently, Theorem 1 is proven.
V̇a ≤ −k1 z12 − k3 z34
2
− k4 Hz34 sign (z34 ) + Hδ . (43)
IV. EXPERIMENTS OF AN EHA SYSTEM
Since H and δ are defined positive constants, it can be
concluded that the errors z1 , z2 , z3 , and estimation signals In this section, experimental results for the pump-controlled
α̃g , α̃h1 , α̃h2 , α̂g , α̂h1 , α̂h2 are all bounded. EHA using the designed adaptive controller are carried out. First,
From (35) and (41), it can be concluded that an experimental test rig is setup. The experimental apparatus
include two pump-controlled EHA systems. An EHA system
lim z2 ≤ δ. (44) operates as position generator, while the other plays the role
t→∞
992 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO. 3, JUNE 2014

Fig. 3. Experimental apparatus.

of a load simulator. Applying the designed algorithm onto the

position generator, experiments are carried out for verifying the
proposed controller.

A. Experimental Test Rig

Schematic diagram of the whole pump-controlled EHA sys-
tem is shown as Fig. 2. The system consists of a PC-based control
system, sensing system, and two intelligent hydraulic servo drive
packages from Yuken Kogyo Company, which serve as motion
generator (main EHA) and load simulator (load generator). In
this configuration, the movement of main cylinder is adjusted Fig. 4. System response at 0.05-Hz sinusoidal trajectory.
by the speed of an ac servo motor (SGDM-50 ADA) through
a fix displacement piston pump, reservoir, and hydraulic con-
trol circuit. One linear encoder, two pressure transducers, and  T 
one load cell are installed to the system to measure the cylinder Tg = diag 1 10−2 0
position, the pressure in two chambers of main EHA, and the
loading force, respectively. Th1 = 10−1
The developed controller is implemented on a personal com-  T 
puter (Core2 Duo 1.8 GHz) within Simulink environment com- Th2 = diag 0 10 10−3 10−4 10−3 .
bined and Real-time Windows Target Toolbox of MATLAB. Here, m is the equivalent mass of the cylinder, H is the maxi-
One encoder Quad–04 card from Measure Computing Corpo- mum nominal load, δ is the acceptable bound of |z2 |. Selection
ration and two multifunction data acquisition Advantech cards, of the control gains (k1 , k3 , and k4 ) and adaptation matrices
A/D 1711 and D/A 1720 are installed on the PCI slots of the Tg , Th1 , and Th2 must be done carefully. Small control and
PC to perform the peripheral interfaces. In order to evaluate the adaptation values can make the system fail to track the tra-
main EHA system in various working conditions, the simulated jectory, while big values of these parameters make the system
load is generated out from the load simulator via force control unstable.
applied to the load generator. Setting parameters for the EHA In order to evaluate the designed controller, performances
system are shown in Table I, while the system apparatus is setup of conventional PID and fuzzy PID have been carried out as
as shown in Fig. 3. comparison. Parameters of PID controller are chosen via trial
and error (KP = 0.25, KI = 0.05, and KD = 0.01), while the
B. Experimental Parameters structure of fuzzy PID is developed as shown in [20].
For the designed experimental test rig, the control parameters
of the developed adaptive backstepping controller are selected C. Experimental Results
as Case study 1: In this case, no loading force is applied, that
is, F = 0 kN. The desired trajectory is selected as a sinusoidal
m = 50 Kg signal with 100 mm of amplitude and 0.05 Hz of frequency
H = 2.5
x1d = 100 sin (0.1πt) [mm] .
δ = 0.4
Applying the proposed control method, response of the EHA,
k1 = 10 ; k3 = 4; k4 = 5 control input signal, and system states are shown in Figs. 4–6,
AHN et al.: ADAPTIVE BACKSTEPPING CONTROL OF AN ELECTROHYDRAULIC ACTUATOR 993

Fig. 5. Control input signal at 0.05-Hz sinusoidal trajectory. Fig. 7. Comparison at 0.05-Hz sinusoidal trajectory and no load condition.

Fig. 8. Generated 10 kN load for 0.5-Hz sinusoidal trajectory.

provides a comparable result with the adaptive backstepping

controller because it can deal with the nonlinearities in specific
cases.
Case study 2: In order to investigate a more challenging work-
ing condition; the input frequency is increased up to 0.5 Hz, the
amplitude is chosen as (50 mm), and loading force is decided
as F = 10 kN. Applying the same controller, external loading
force, response of the EHA, control input signal, and system
states, are shown in Figs. 8, 9, 10, and 11, respectively. At
the beginning, the system with adaptive backstepping controller
Fig. 6. System states at 0.05-Hz sinusoidal trajectory. cannot track the reference smoothly. This is originated by the
difference between the system initial state and the states of de-
sired trajectory. However, after 0.5 s the state error of the system
respectively. As seen in Fig. 4, the adaptive backstepping z34 is controlled to be bounded within ±2000. Consequently,
controller ensure the bounded state errors z1 , z2 , and z34 . the position error is bounded within ±2 mm.
Consequently, the position response can be controlled accu- With this case study, the effectiveness of the proposed con-
rately and the position error mostly lied within ±0.02 mm. troller over the conventional PID and fuzzy PID controller is
There are two notch error points in one cycle of the cylin- strongly confirmed. As seen in Fig. 12, while the proposed
der movement. These notch error points are due to the dead adaptive backstepping controller provides robust response for
zone when the cylinder changes it direction. The form of con- a 0.5-Hz sinusoidal input signal, the PID and fuzzy PID show
trol input signal in Fig. 5 is generally smooth which implies insufficient responses. At a steady state, the conventional PID
a good performance for the proposed controller. Fig. 7 shows controller generates largest tracking error (within ±12 mm).
comparison performance between the proposed controller, the Next, the fuzzy PID controller generates smaller error (within
conventional PID controller, and fuzzy PID controller. It can be ±5 mm). By the adaptation of internal parameters, the proposed
seen that the PID provides worst response because it cannot han- backstepping controller can produce good tracking response for
dle the system nonlinearities. Appropriate fuzzy PID controller constant 10 kN loaded condition.
994 IEEE/ASME TRANSACTIONS ON MECHATRONICS, VOL. 19, NO. 3, JUNE 2014

Fig. 9. System response at 0.5-Hz sinusoidal trajectory. Fig. 11. System states at 0.5-Hz sinusoidal trajectory.

Fig. 10. Control input signal at 0.5-Hz sinusoidal trajectory.

Fig. 12. Comparison at 0.5-Hz sinusoidal trajectory and 10 kN load.

Case study 3: In order to investigate the performance of the

proposed controller in a variable load condition, a multistep load
is generated for a sinusoidal response. Here, the input frequency
is 0.05 Hz and amplitude is 50 mm. As seen in Fig. 13, the load
condition varies from 2.5 to 11 kN at 10th s, and from 12
to 8 kN at 25th s. Despite the challenging load, the proposed
backstepping controller provides a robust response with high
tracking accuracy (within ±0.1 mm) comparing with the PID
and fuzzy PID controllers as seen in Fig. 14. The experimental
results prove strongly that the proposed backstepping controller
can maintain robust performance for the EHA system even in
disturbed working conditions. Fig. 13. Variable load for 0.05-Hz sinusoidal trajectory.
AHN et al.: ADAPTIVE BACKSTEPPING CONTROL OF AN ELECTROHYDRAULIC ACTUATOR 995

[12] W. H. Zhu and J. C. Piedboeuf, “Adaptive output force tracking control

of hydraulic cylinders with applications to robot manipulators,” ASME J.
Dyn. Syst., Meas., Control, vol. 127, pp. 206–217, 2005.
[13] L. Jun, F. Yongling, Z. Guiying, G. Bo, and M. Jiming, “Research on fast
response and high accuracy control of an airborne brushless DC motor,”
in Proc. IEEE Conf. Robot. Bio., Shenyang, China, 2004, pp. 807–810.
[14] D. Q. Truong and K.K. Ahn, “Force control for press machines using an
online smart tuning fuzzy PID based on a robust extended Kalman filter,”
Expert Syst. Appl., vol. 38, no. 5, pp. 5879–5894, 2011.
[15] Y. Lin, Y. Shi, and R. Burton, “Modeling and robust discrete-time sliding-
mode control design for a fluid power electrohydraulic actuator (EHA)
system,” IEEE Trans. Mechatronics, vol. 18, no. 1, pp. 1–10, 2013.
[16] M. Perron, J. Lafontaine, and Y. Desjardins, “Sliding-mode control of a
servomotor-pump in a position control application,” in Proc. IEEE Conf.
Elect. Commun. Eng., Saskatoon, SK, Canada, 2005, pp. 1287–1291.
[17] H. M. Kim, S. H. Park, J. H. Song, and J. S. Kim, “Robust position control
of electro-hydraulic actuator systems using the adaptive back-stepping
control scheme,” J. Syst. Control Eng., vol. 224, pp. 737–746, 2010.
[18] H. E. Merritt, Hydraulic Control Systems. New York, NY, USA: Wiley,
Fig. 14. Comparison at 0.05-Hz sinusoidal trajectory and variable load. 1967.
[19] N. D. Manring, Hydraulic Control Systems. New York, NY, USA:
Wiley, 2005.
[20] D. Q. Truong and K. K. Ahn, “Force control for hydraulic load simula-
V. CONCLUSION tor using self-tuning grey predictor—fuzzy PID,” Mechatronics, vol. 19,
In this paper, a robust position control using the modified pp. 233–246, 2009.
backstepping algorithm was developed and successfully applied
Kyoung Kwan Ahn (M’02) received the B.S. de-
to a nonlinear EHA with uncertainties. A practical asymmetric gree from the Department of Mechanical Engineer-
pump controlled EHA system was modeled and a backstepping ing, Seoul National University, Seoul, Korea, in 1990,
control scheme was developed. The control structure is designed the M.Sc. degree in mechanical engineering from the
Korea Advanced Institute of Science and Technology
by selecting a special Lyapunov function, while all uncertain (KAIST), Daejeon, Korea, in 1992, and the Ph.D. de-
terms are adapted by selecting another Lyapunov function. The gree from the Tokyo Institute of Technology, Tokyo,
whole control structure is the combination of adaptive technique Japan, in 1999.
He is currently a Professor in the School of Me-
and a simple robust structure to compensate the nonlinearities chanical Engineering, University of Ulsan, Ulsan,
and bounded disturbance. Experimental results prove that the Korea. His research interests include design and con-
proposed controller can obtain accurate performance even in trol of smart actuators using smart materials, fluid power control, and active
damping control.
the presence of an external disturbance. Prof. Ahn is a member of the American Society of Mechanical Engineers
(ASME), Society of Instrument and Control Engineers (SICE), Robotics So-
ciety of Japan (RSJ), Japan Society of Mechanical Engineers (JSME), Korean
REFERENCES Society of Mechanical Engineers (KSME), Korean Society of Precision Engi-
[1] P. M. FitzSimons and J. J. Palazzolo, “Part I: Modeling of a one-degree-of- neers (KSPE), Korean Society of Automotive Engineers (KSAE), Korea Fluid
freedom active hydraulic mount—Part II: Control,” ASME J. Dyn. Syst., Power Systems Society (KFPS), and Japan Fluid Power System Society (JFPS).
Meas., Control, vol. 118, no. 4, pp. 439–448, 1996.
[2] A. Alleyne and J. K. Hedrick, “Nonlinear adaptive control of active sus- Doan Ngoc Chi Nam received the B.S. degree in
pension,” IEEE Trans. Control Syst. Technol., vol. 3, no. 1, pp. 94–101, mechatronics from the Department of Mechanical
Jan. 1995. Engineering, Hochiminh City University of Tech-
[3] B. Yao, F. Bu, and G. T. C. Chiu, “Adaptive robust motion control of nology, Vietnam, in 2008, and the Ph.D. degree in
single-rod hydraulic actuators: Theory and experiments,” IEEE Trans. mechatronics from the University of Ulsan, Ulsan,
Mechatronics, vol. 5, no. 1, pp. 79–91, Mar. 2000. Korea, in 2013.
[4] S. Habibi and A. Goldenberg, “Design of a new high-performance electro- He is currently a Postdoctoral Researcher in the
hydraulic actuator,” IEEE Trans. Mechatronics, vol. 5, no. 2, pp. 158–164, School of Mechanical Engineering, University of Ul-
Jun. 2000. san. His research interests include smart actuators/
[5] J. Grabbel and M. Ivantysynova, “An investigation of swash plate control materials, robotics, fluid power control, nonlinear
concepts for displacement controlled actuators,” Int. J. Fluid Power, vol. 6, control, and energy harvesting systems.
no. 2, pp. 19–36, 2005.
[6] Mini-Motion Package Catalog, Kayaba Industry Company, Ltd. (2013).
Maolin Jin (S’06–M’08) received the B.S. degree
[Online]. Available: http://www.kybfluidpower.com/Mini_Motion_
in material science and mechanical engineering from
Package.html
Yanbian University of Science and Technology, Jilin,
[7] Intelligent Hydraulic Servo Drive Pack Catalog, Yuken Kogyo
China, in 1999, and the M.S. and Ph.D. degrees in
Company Ltd. (2013). [Online]. Available: http://www.yuken.co.uk/
mechanical engineering from the Korea Advanced
long_cat/%5BK%5D/816-817.pdf
Institute of Science and Technology (KAIST), Dae-
[8] T. C. Tsao and M. Tomizuka, “Robust adaptive and repetitive digital con-
jeon, Korea, in 2004 and 2008, respectively.
trol and application to hydraulic servo for noncircular machining,” Trans.
In 2008, he was a Postdoctoral Researcher in the
ASME J. Dyn. Syst., Meas., Control, vol. 116, pp. 24–32, 1994.
Mechanical Engineering Research Institute, KAIST.
[9] R. Vossoughi and M. Donath, “Dynamic feedback linearization for electro-
He is currently with the System Solution Research
hydraulically actuated control systems,” ASME J. Dyn. Syst., Meas., Con-
Department, Research Institute of Industrial Science
trol, vol. 117, no. 4, pp. 468–477, 1995.
and Technology, Pohang, Korea. His research interests include robust control
[10] A. Alleyne and R. Liu, “A simplified approach to force control for electro-
of nonlinear plants, time-delay control, robot motion control, electrohydraulic
hydraulic systems,” Control Eng. Practice, vol. 8, pp. 1347–1356, 2000.
actuators, and factory automation.
[11] C. Kaddissi, J. P. Kenne, and M. Saad, “Identification and real-time control
Dr. Jin is a member of the Institute of Control, Robotics, and System Engi-
of an electrohydraulic servo system based on nonlinear backstepping,”
neers in Korea.
IEEE Trans. Mechatronics, vol. 12, no. 1, pp. 11–22, 2007.