You are on page 1of 10

ISA Transactions 59 (2015) 5564

Contents lists available at ScienceDirect

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

Research Article

Robust H1 positional control of 2-DOF robotic arm driven


by electro-hydraulic servo system$
Qing Guo a,b,n, Tian Yu b, Dan Jiang c
a
School of Aeronautics and Astronautics, University of Electronic Science and Technology of China, Chengdu 611731, China
b
Center for Power Transmission and Motion Control, Department of Mechanical Engineering, University of Bath, BA2 7AY, Bath, UK
c
School of Mechatronics Engineering, University of Electronic Science and Technology of China, Chengdu 611731, China

art ic l e i nf o a b s t r a c t

Article history: In this paper an H 1 positional feedback controller is developed to improve the robust performance under
Received 7 June 2014 structural and parametric uncertainty disturbance in electro-hydraulic servo system (EHSS). The robust
Received in revised form control model is described as the linear statespace equation by upper linear fractional transformation.
17 March 2015
According to the solution of H 1 sub-optimal control problem, the robust controller is designed and
Accepted 14 September 2015
simplied to lower order linear model which is easily realized in EHSS. The simulation and experimental
Available online 23 October 2015
This paper was recommended for publica- results can validate the robustness of this proposed method. The comparison result with PI control shows
tion by Dr. Q.-G. Wang that the robust controller is suitable for this EHSS under the critical condition where the desired system
bandwidth is higher and the external load of the hydraulic actuator is closed to its limited capability.
Keywords: & 2015 ISA. Published by Elsevier Ltd. All rights reserved.
Robust H1 control
Structural and parametric uncertainty
Electro-hydraulic servo system
2-DOF robotic arm

1. Introduction easy to be used in industry. Variable control parameters of PID


controllers [4,5] are adopted to suit the variable characteristic of
In most hydraulic applications, electro-hydraulic servo system dynamic model. In addition, several types of adaptive sliding mode
(EHSS) is widely used in control practice, because of high load control methods [6,7] are presented in EHSS and get some good
efciency, fast response, and high tracking accuracy. However, performance. The asymptotically stable adaptive controller is
EHSS has some model uncertainties such as parametric dis- constructed and the adaptation law is designed to estimate
turbances, structural uncertainties. The parametric disturbances unknown parameters. In order to improve the dynamic response,
[1] include ow gain of valve, effective bulk modulus, ow- the backstepping control method [8,9] is used to suppress dis-
pressure coefcient, viscous damping coefcient, etc., which are turbance based on Lyapunov stability analysis. The robust con-
consequences of hydraulic physical characteristics, oil leakage, oil troller design [1012] is presented with parameter uncertainty
temperature variations, characteristics of valves near null, and so disturbance, which is converted to H 1 standard problem and
on. In addition, the load variation is the main structural uncer- solved by linear matrix inequality optimization. Afterward the
tainty caused by the force or torque motivated the mechanical mixed sensitivity H 1 control method [13] is presented to improve
motion. the robust stability of EHSS and to satisfy the control constraint.
Refs. [2,3] show the theory of uid power transmission and Then, to improve the robustness in the case of nonlinearities
analysis of the feedback control of position, velocity, and force. The and uncertainties, the two-stage nonlinear robust controller [14] is
classical approach to the control of EHSS is PID control which is used for velocity tracking control. Based on state observer, an
output feedback control and a cascade control algorithm [15,16]
are presented in EHSS. Afterward the variable structure approach

This work was supported by UK Engineering and Physical Sciences Research [17] is presented to solve the friction nonlinearities. In [18,19], a
Council project (No. EP/H024190/1), the National Natural Science Foundation of
China (No. 61305092 and No. 51205045) and the Postdoctoral Science Foundation predictive generic model and a fractional order control method are
of China (No. 2013M542487). presented to improve the dynamic response in the case of the
n
Correspondence to: No. 207#, School of Aeronautics and Astronautics, Uni- robust stability. In addition, to improve the tracking accuracy, an
versity of Electronic Science and Technology of China, No. 2006, Xiyuan Ave, West
adaptive nonlinear optimal compensation controller [20] is
Hi-Tech Zone, ChengDu 611731, China. Tel.: 86 15008496370.
E-mail addresses: guoqinguestc@uestc.edu.cn (Q. Guo), employed for electro-hydraulic load simulator to suppress sudden
ty267@bath.ac.uk (T. Yu), jdan2002@uestc.edu.cn (D. Jiang). disturbance.

http://dx.doi.org/10.1016/j.isatra.2015.09.014
0019-0578/& 2015 ISA. Published by Elsevier Ltd. All rights reserved.
56 Q. Guo et al. / ISA Transactions 59 (2015) 5564

This paper developed an H 1 output feedback controller to Table 1


drive the motion control of 2-DOF robotic arm. It is organized as The specic parameters and brands of main components.
follows. At rst, the experimental architecture is introduced. Sec-
Element Type Marks Quantity
ondly, the dynamic model of EHSS is constructed. Then the typical
parametric and structural uncertainties are quantitatively ana- Servo motor BSM63N-375 BALDOR 1
lyzed to construct the robust control model. According to the Fixed displacement pump TFH-315 Takako 1
Servo valve D633-R02K01M0NSM2 Moog 2
solution of H 1 suboptimal problem, the controller is designed to
Double-rod cylinder LB6-1610-0080-4 M Hoerbiger 2
satisfy certain robust performance. Finally, both the simulation Relative encoder AEDA-3300-BE1 AVAGO 2
and experiment validate the robustness of this controller.

effective bulk modulus, mt is the load mass, K is the load spring


2. Experimental architecture constant, Be is the viscous damping coefcient of oil, FL is the
external load on the hydraulic actuator from the mechanical
The experimental architecture of EHSS in this work is shown in structure of 2-DOF robotic arm.
Fig. 1. The hydraulic part is composed of 2 servo valves, 2 double- The dynamic characteristic of servo valve can be described as a
acting cylinders, a xed displacement pump, a servo motor, a relief second-order linear model given by
valve, etc. The mechanical part is 2-DOF robotic arm. The upper xv K sv 2sv
arm and forearm rotate their joints driven by 2 cylinders. The main 2 2
u s 2 sv sv s 2sv
load is a disc on the top of forearm. The angle of joint is measured
by encoder, which can be transferred into the cylinder displace. where sv is the damping ratio, sv is natural frequency, Ksv is the
Here the hydraulic position feedback control is used to realize the servo valve gain, u is the control voltage.
angular servo control of the robotic arm. The control algorithm is
executed through IPC, equipped with the data acquisition card. 3.2. Analysis of model uncertainty
The 2-DOF robotic arm is manufactured by Italian Institute of
Technology, which refers to the Robotic Bigdog. The specic In (1), Vt, Be, Kq are 3 parametric uncertainties, which are
parameters and brands of main components in the experimental caused by operating conditions, environmental variability,
architecture are listed in Table 1. hydraulic characteristic variations and the linearized model error.
Because of the asymmetry cylinder, the value of Kq is different in
cylinder extension (Kqa) or retraction (Kqb) respectively shown as
3. Dynamic model of EHHS follows:
s
3.1. Linearized hydraulic model P s P l  P vloss  P tube
K q K q0
Ps
The positional control mechanism of EHSS is shown in Fig. 2. r
2
The shoulder and elbow joints are driven by 2 hydraulic actuators. K aq K q
1 3
The joint angles 1 ; 2 are the only output feedback information. s
The linear model of hydraulic cylinder is shown as follows [2]: b 2P s  P vloss  P tube P l
Kq Kq : 3
  P s  P vloss  P tube  P l 1 3
K iq 1 Vt
xv  2 K ic s FL
Aa Aa 21 2 e
Y ! ! where Ps is supply pressure, Pl is load pressure, Kq0 is unloading
mt K ic KK i
V t mt
s3
Be V t
s
B K
2 1 e ce1
VtK
s 2c ow gain of the valve, P vloss is pressure loss of the micro-pipe,
21 2 e A2a A2a 21 2 e A2a A2a 21 2 e A2a Aa
P tube is pressure loss of the pipe from valve to cylinder.
1 The load pressure Pl can be estimated by Lagrange equation. If
where Y is the displacement of piston, xv is the spool position of the parameters shown in Table 2 are substituted into (3), Kqa and
the servo valve, K iq ; K ic i a; b are the ow gain of the servo valve Kqb are bounded by 0:012 r K aq r0:02, 0:019 rK bq r 0:027. So the
and ow-pressure coefcient, Aa and Ab are the annulus areas of ow gain of the servo valve Kq can be represented as follows:
the 2 chambers ( Aa =Ab ), Vt is the cylinder volume, e is the K q K q 1 pK q K q  1 r K q r 1 4
2
9 where the nominal values K q is 0.02 m /s, the maximum relative
1 uncertainty pK q is 0.375 with the relative variation K q .
The actual cylinder volume Vt can be described as follows:

2 V a V a0 Aa y V b V b0 Ab y
V a0 V b0 V t =2: 5
3 8
Although the initial volumes of 2 chambers are unknown, Va0
4 and Vb0 satisfy the following conditions:
5
6 Lc  dr max Aa =2 rV a0 r Lc drmax Aa =2
Lc  drmax Ab =2 rV b0 r Lc drmax Ab =2: 6
7
After the known parameters are substituted into (6), the
cylinder volume Vt can be represented as follows:

V t V t 1 pvt vt 1 r V t r 1 7
Fig. 1. The experimental equipment (1-robotic arm, 2-hydraulic cylinder, 3-servo 3
valves, 4-pressure gauge, 5-relief valve, 6-xed displacement pump, 7-servo motor, where the nominal values V t of is 2.85e  5 m , the maximum
8-tank, 9-IPC). relative uncertainty pV t is 0.663 with the relative variation V t .
Q. Guo et al. / ISA Transactions 59 (2015) 5564 57

Fig. 2. The electro-hydraulic servo control mechanism of 2-DOF robotic arm.

Table 2
Hydraulic parameters and mechanical parameters used in simulation and
experiments. H 11 I 1 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
Parameter Value Parameter Value
H 12 I 2f m2f P 1 P 2  P 2 P m2 cos 2  m1
xvmax 0.79 mm 0.62 H 22 I 2f m2f P 2 P 2m2
Ps 40 bar Pr 2 bar
Aa 2.01 cm2 Kq0 0.015 m2/s C 11 2m2f P 1 P 2  P 2 P m2 _ 2 sin 2  m1
3.95  10  4 m/V e 2  108 Pa
C m P P  P P _ sin 
Ksv
sv 353.6 rad/s sv 0.707 12 2f 1 2 2 m2 2 2 m1
pvloss 1.4 bar ptube 7.5 bar C 21 m2f P 1 P 2  P 2 P m2 _ 1 sin 2  m1
Lc 157 mm drmax 79 mm
m1 1.772 kg m2 0.739 kg C 22 0
mf 1 kg I1 0.071 kg m2 G1  m1 gP 1 P m1 sin 1 m1  m2f gP 1 P 2 sin 1 m1
I2 0.015 kg m2 I2f 0.022 kg m2
P 2 P m2 sin 1 2 
m2f 1.739 kg m1f 3.511 kg
P1P2 0.35 m P1Pm1 0.16 m G2  m2f gP 2 P m2 sin 1 2 : 10
P2Pm2 0.12 m m1 7.9
where H, C and G are inertia matrix, Coriolis matrix and gravity
vector, respectively, Tu and Tf are the torques on upper arm and
Similarly, the viscous damping coefcient Be can be described forearm, m1 is upper arm mass including cylinder, m2 is forearm
as follows: mass, mf is load mass, I1 inertia of upper arm around the shoulder,
I2 inertia of forearm around the elbow, I2f is inertia of forearm
Be B e 1 pBe Be 1 r Be r1 8
around the elbow including load, m1f ; m2f are upper arm and
where the nominal values B e is 3500 Ns/m, the maximum relative forearm mass with load, P 1 P 2 is upper arm length, P 1 P m1 is dis-
uncertainty pBe is 0.428 with the relative variation Be . tance from upper arm centroid to shoulder, P 2 P m2 is distance from
Now the external load FL is the maximum disturbance of forearm centroid to elbow, m1 is eccentric angle of cylinder to
hydraulic model, which is considered as an un-modeled dis- shoulder.
turbance [10]. In this paper, the quantitative relationship between Step 2: Some known parameters shown in Table 2 are sub-
the external load FL and the position of cylinder is analyzed stituted into (10), Tu and Tf can be computed. Then 2 ctitious
quantitatively to describe the structural uncertainty. proportional gains are dened as follows:
Step 1: The dynamic model of 2-DOF robotic arm is constructed F Lu 1 ; 2 T u 1 ; 2
by Lagrange equation as follows: K F Lu 1
c1 1 l1 1 c1 1
Hqq Cq; q
_ q_ Gq T F Lu 1 ; 2 T u 1 ; 2
" # " # K F Lu 2
H 11 H 12 C 11 C 12 c2 2 l1 1 c2 2
H C F Lf 1 ; 2 T f 1 ; 2
H 21 H 22 C 21 C 22 K F Lf 1
" # " # " # c 1 1 l2 2 c1 1
G1 Tu 1 F Lf 1 ; 2 T f 1 ; 2
G T q 9 K F Lf 2 11
G2 Tf 2 c 2 2 l2 2 c2 2
58 Q. Guo et al. / ISA Transactions 59 (2015) 5564

1500
KFLu(1)
KFLu(2)
KFLf(1)
1000
KFLf(2)

500
KFL(N/m)

500

Fig. 4. The robust model with parametric and structural uncertainties.

1000
0 0.5 1 1.5 2
t(s)
2 3
0 1 0 0 0
Fig. 3. The dynamic ranges of K F Lu and K F Lf in one motion duration. 6  2 2 sv sv 0 0 0 7
6 sv 7
6 7
6 0 0 0 1 0 7
A6 7
where the dynamic length ci i and arm li i (i 1; 2) are com- 6 0
6 0 0 0 1 7
7
puted by triangle geometry. 4 4K q Aa 4e K F Lu K c 4 e A2a B K KF  4K c e
5
e
0    e e c  m1fLu  mB1fe
V t m1f V t m1f V t m1f V t m1f Vt
The dynamic ranges of K F Lu and K F Lf in a duration are shown in
Fig. 3. Here K F Lu 1 and K F Lf 2 are larger than the other 2 gains, 2 3
0 0 0 0 0 0
respectively, which means that the external load FLu is inuenced 6 0 0 0 0 0 0 7
6 7
by the shoulder angle 1 more than 2. The external load FLf is 6 7
6 7
similar to FLu. B1 6 0 0 0 0 0 0 7
6 0 7
So the structural uncertainty caused by FL can be described as 6 0 0 0 0 0 7
4 pK q pK pBe 5
   pK F pBe  pvt
F Lu
follows: Vt Vt Vt Lu

F Lu K F Lu 1 y1 F Lf K F Lf 2 y2  T
B2 0 K sv 2sv 0 0 0
K F Lu K F Lu pK FLu K FLu  1 r K FLu r 1
2 3
4K q e Aa
K F Lf K F Lf pK FLf K FLf 1 r K FLf r1 12 0 0 0 0
6 m1f 7
6 7
6 0 0 4 eKc
0 0 7
where the nominal values K F Lu and K F Lf are 0, the maximum 6 m1f 7
6 7
relative uncertainties pK FLu , pK FLf are 1200 N/m, 300 N/m with the 6 b e K c 7
6 0 0 0 0 7
relative variations K FLu , K FLf . 6 m1f 7
C1 6
6 0
7
6 0 0 1
m1f 0 7 7
6 7
3.3. Robust model construction 6 7
6 0 0 0 0 Be 7
6 m1f 7
6 7
The parameters Kq, Vt, Be are represented as an upper LFT: 4 4K q e A 4e K F Lu K c 4 A2 B K  4K c e 5
0   e a e e c
K q F u M K q ; K q , V t F u M V t ; V t , Be F u M Be ; Be , where V t m1f V t m1f V t m1f V t m1f Vt

2 3 " # " # 2 3
0 Kq  pV t 1=V t 0 Be 0 0 0 0 0 0
MK q 4 5 MV M Be : 6
6 0 0 0 0 0 0
7
7
pK q K q t
 pV t 1=V t pBe B e 6 7
6 0 0 0 0 0 0 7
6 7
13 D11 6 7
6 0 0 0 0 0 0 7
6 7
The structural uncertainty parameters K F Lu , K F Lf can be repre- 6 0 0 0 0 0 0 7
6 7
sented as K F Lu F u M K F Lu ; K F Lu , K F Lf F f M K F ; K F , where 4 pK q pK p Be 5
   pvt
F Lu
Lf Lf 0 0
" # " # Vt Vt Vt
0 1 0 1
M K F Lu p K MK F p K : 14 C 2 0 0 1 0 0: 16
K F Lu F Lu Lf K F Lf F Lf

Taking the shoulder hydraulic actuator for example, the robust Here the inuence of load stiffness is neglected, i.e., K 0. The
model with parametric and structural uncertainties is shown in measurement noise Yd(t) can be described in frequency domain as
Fig. 4. follows:
If we dene some vectors as follows: the state variable vector 1 0:12s
X x1 ; x2 ; x3 ; x4 ; x5 T xv ; x_ v ; y; y;
_ y
T , the control inputs u(t) is the Y d s Gd sNs Gd s 0:006  17
1 0:001s
voltage of servo valve, the vector of exogenous inputs
wt w1 ; w2 ; w3 ; w4 ; w5 ; w6 T , the vector of measurements y(t), where N(s) is the noise input. The block diagram of the closed-loop
the regulated output vector zt z1 ; z2 ; z3 ; z4 ; z5 ; z6 T , then the system with robust performance requirements is shown in Fig. 5.
linear uncertain statespace model is described as follows: The linear uncertain statespace model is described as block
Ghyd, the block diagK q ; V t ; Be ; K FLu . The demand displace-
X_ t AXt B1 Wt B2 ut
ment of cylinder r and the measurement noise d are 2 inputs. The
Zt C 1 Xt D11 Wt weighting functions Wp and Wu reect the relative signicance of
Yt C 2 Xt Y d t; 15 the performance requirement.
Q. Guo et al. / ISA Transactions 59 (2015) 5564 59

Fig. 8. Block diagram for the description of robust stability and robust
performance.

Fig. 5. The block diagram of the closed-loop system with robust performance 10
0

requirements.
upper bound
low bound

1
10
0 10
Nominal
Log Magnitude(dB)

Perturbed

2
10
2
10

4

10
0.1 1 6.28 10 100
3
Bode Plots of Perturbed Plants
10
0
Phase (degrees)

50
4
100 10
1 0 1 2
10 10 10 10
150 Frequency/(rad/sec)

Fig. 9. The maximum robust stability bound of 4 uncertainty parameters.


200
0.1 1 6.28 10 100
Frequency (rad/sec)

Fig. 6. The frequency response of the open-loop system with varying uncertainty
parameters Kq, Vt, Be and K F Lu . 0
10 nominal
upper bound
1 low bound
10

0.1
10
0
10

Magnitude

1
10 0.2
10

1
2 (Wp )
10
n
(Gc ) 10
1
10
0
10
1 2
10
Frequency/(rad/sec)

Fig. 10. The maximum robust performance bound of 4 uncertainty parameters.


3
10
4 3 2 1 0 1 2
10 10 10 10 10 10 10
uncertainties. In addition, the controller should reduce the output
Frequency/(rad/sec)
sensitivity due to the presence of measurement noise.
Fig. 7. The singular values of the inverse function Wp and the nominal closed-loop
system Gcn.
4.1. Analysis of open-loop system

4. Robust controller design The frequency response of the open-loop system with uncer-
tainties is shown in Fig. 6. The magnitude is little different between
The goal of the controller design is to satisfy the robust stability perturbed parameter and nominal parameter. But the phase is sig-
and performance in the case of the parametric and structural nicantly different, which is mainly caused by the structural
60 Q. Guo et al. / ISA Transactions 59 (2015) 5564

Fig. 11. The whole simulation structure of this EHSS. (a) The simulation structure of robust H1 positional control. (b) 2-DOF robotic arm simechanics model. (c) The
animation scene of 2-DOF robotic arm.

uncertainty parameter K F Lu . The stable phase of perturbed system designed as follows [21]:
will lead or lag approximately 40 degree comparing to the nominal
system in middle frequency stage. So this robust controller is mainly s=M s c
Wp W u 6  10  4 18
designed to regulate the phase margin of system. s c A e

4.2. Weight function design where the maximum steady accuracy requirement Ae is 0.01, the
peak of sensitivity Ms is 2, the prospective bandwidth of closed-
According to robust performance requirement and control loop c is 2 . The weighting function Wu should satisfy the
constraint, the weighting functions Wp and Wu are iteratively maximum control voltage 710 V of servo valve.
Q. Guo et al. / ISA Transactions 59 (2015) 5564 61

The singular values of the inverse function


 W p and the nominal (1) To nd a stabilizing controller K such that the H 1 norm of

closed-loop system Gcn are described as W p 1 , Gnc shown in the closed-loop system is less than a given positive constant , i.e.,

Fig. 7. So the H 1 norm of the nominal closed-loop system satises J F l Gic ; K k1 J F l F u Ghyd ; ; K k1 o 20
the following rule:


where 4 0 minK J F l Gic ; K J 1 .
W p Gn r W p Gn o 1: 19 stabili sin g
c 1 c
(2) The closed-loop system composed of the inter-connection
system Gic and the controller K should satisfy the following con-
4.3. Robust control problem statement dition:
" #
W p I Gic K  1 Gd
The robust controller will solve 2 problems: robust stability and
o1: 21
robust performance requirement which can be described in Fig. 8. W u KI Gic K  1
1
shoulder angle()

20
0 4.4. Robust controller design
-20
-40 demand
simulation The robustness analysis for different uncertainty parameters
-60
-80
experiment
are given in Figs. 9 and 10. The relative uncertainty varies from
0 5 10 15 20 25
 1 to 1 for uncertainty parameters Kq, Vt, Be and K F Lu .
120 In Fig. 9, the singular value of perturbed closed-loop system is
elbow angle()

100 0.37, which is less than 1. So the closed-loop system Gc is stable


80
and its relative uncertainties can be extended to 1/0.37. In
60
40
Fig. 10, the robust performance requirement is 1.06, which is a
0 5 10 15 20 25 little more than 1 in low frequency stage. It means the tracking
t(s) accuracy may be not excellent. But in middle frequency stage 4
20 rad/s, the dynamic trace performance and robustness margin
Fig. 12. The square response of simulation and experiment.
meet the requirement.
When the iterative design nishes, the robust controller is
seven orders. Then the controller is simplied to 4 orders by
10
Balanced Truncation Method [22] as follows:
shoulder control(V)

5
n4 s4 n3 s3 n2 s2 n1 s n0
K 1 s y~  y 22
0
d4 s4 d3 s3 d2 s2 d1 s d0

simulation
where n4 0.053, n3 1.049, n2 155.8, n1 5376.2, n0 7.6, d4 1,
5
experiment d3 146.2, d2 1938.3, d1 6198.7, d0 77.4, y~ is the demand input
10 and y is the measurement output.
0 5 10 15 20 25

10

5. Simulation result
elbow control(V)

0 The nonlinear simulation can be implemented by Matlab/


Hydraulic and Simechanics modules. The whole simulation
5
structure includes 2 joint demand inputs, 2 robust controllers,
10
2 servo valves, 2 double-acting cylinders, 2-DOF robotic arm
0 5 10 15 20 25 simechanics model, 2 encoders model with noise shown in
t(s) Fig. 11a. The robotic arm is constructed by simechanics modules
Fig. 13. The control voltage of simulation and experiment for square demand. shown in Fig. 11b. The animation scene of 2-DOF robotic arm
corresponding to simechanics model is shown in Fig. 11c. The

40
demand
20
shoulder angle()

simulation
0 experiment

-20
-40
-60
-80
0 1 1.65 2 3 3.3 4 5 6 7 8 9 10

140

120
elbow angle()

100

80

60

40
0 1 2 3 4 5 6 7 8 9 10
t(s)

Fig. 14. The sinusoidal response of simulation and experiment.


62 Q. Guo et al. / ISA Transactions 59 (2015) 5564

shoulder control(V)
0

-2
simulation
-3
experiment
-4
0 1 2 3 4 5 6 7 8 9 10

5
4
elbow control(V)

-5
0 1 2 3 4 5 6 7 8 9 10
t(s)

Fig. 15. The control voltage of simulation and experiment for sinusoidal demand.

hydraulic parameters of this EHSS and the mechanical parameters 30


of 2-DOF robotic arm are shown in Table 2.
20
The maximum stroke of cylinder is 79 mm. So the ranges of the
10
shoulder and elbow joint angles are (  60, 20), (40, 130). In
addition, the relative variations V t , Be , K FLu , K FLf are all chosen as
shoulder angle()

1, which means the maximum uncertainties of perturbed system. -10

The simulation results of square and sinusoidal response are -20


shown in Figs. 1215. -30
The 2 square frequencies are 0.1 Hz, 0.1 Hz and the sinusoidal -40
frequencies are 0.3 Hz, 0.5 Hz, respectively. Because the external -50
load on the shoulder joint is heavier than the elbow joint, the demand
-60 robust control
sinusoidal frequency for the shoulder motion is not allowed too PI control
-70
much. Due to the light external load on the elbow arm, the 0 1 2 3 4 5 6 7 8
tracking accuracy is higher than the shoulder arm. The control t(s)
voltages of 2 servo valves are no more than the saturation 710 V
Fig. 16. The sinusoidal response of experiment result by two control method.
shown in Fig. 13.
Similarly for the sinusoidal response, the maximum dynamic
errors are less than 3, which are no more than the relative error 10
robust control
5% of demand input shown in Fig. 14. The control voltages are also PI control
normal shown in Fig. 15. So these 2 robust controllers can be
validated effectively in simulation. 5
shoulder control(V)

6. Experimental result 0

In this experiment, the supply pressure Ps is 40 bar, and the


motor rev is xed 1000 rpm. From Figs. 1215, the experiment
-5
results are very close to the simulation results. The steady errors of
2 joint angles are less than 2 and the maximum dynamic errors
are less than 4 in sinusoidal response results. So these controllers
can satisfy the robustness and dynamic response performance in -10
0 1 2 3 4 5 6 7 8
the case of model uncertainties. t(s)
The experimental comparison for 2 control methods is shown
Fig. 17. The control voltage of experiment result by two control method.
in Figs. 1617. To illustrate this problem, the sinusoidal frequency
is 1 Hz and the elbow is always extended like the snapshot of
Fig. 18b, which is the worst control condition for the shoulder t0 s. One second later, the elbow cylinder is extended to reach its
hydraulic actuator. In fact, when the frequency and the external maximum stroke. Then the elbow cylinder has turned to retract
load of the shoulder hydraulic actuator are increased simulta- and the shoulder cylinder is extended to reach its maximum
neously, the dynamic performance of PI controller becomes worse stroke at t1.65 s. At t 2 s, the elbow cylinder is retracted
than the robust controller. The control saturation emerges for PI entirely and the shoulder cylinder turns to retract gradually. Then
controller, which would result in clear dynamic error. the elbow cylinder is extended to reach its maximum stroke once
The snapshots of sinusoidal experiment process are shown in again at t 3 s, which means the duration of elbow joint motion
Fig. 18. The 2-DOF robotic arm is controlled by 2 robust controllers has ended. After 3.3 s, the duration of shoulder joint motion has
with 1 kg disk load. The cylinders are all retracted at initial time also ended.
Q. Guo et al. / ISA Transactions 59 (2015) 5564 63

Fig. 18. The snapshots of sinusoidal experiment process. (a) t 0 s. (b) t 1 s. (c) t 1.65 s. (d) t 2 s. (e) t 3 s. (f) t 3.3 s.

7. Conclusions by Lagrange equation since it has the maximum impact to the


robustness of EHSS. After the linear perturbed model is con-
In this paper, a robust H 1 output feedback controller is structed, 2 robust controllers are designed by H 1 control problem
designed to improve the robustness under the structural and optimization. According to the analysis of the maximum robust
parametric uncertainties in EHSS. The parametric uncertainty performance bound with varying uncertainty, this perturbed sys-
includes the ow gain of servo valve, the effective volume of tem is stable despite of the structural and parametric uncertainty
cylinder and the viscous damping coefcient. The structural disturbance. The consistent results about simulation and experi-
uncertainty is the external load, which is quantitatively estimated ment can validate the robustness of this controller. In addition, the
64 Q. Guo et al. / ISA Transactions 59 (2015) 5564

comparison experimental results with PI controller show that the [7] Guan C, Pan SX. Adaptive sliding mode control of electro-hydraulic system
robust controller is suitable for this EHSS in the critical condition with nonlinear unknown parameters. Control Eng Pract 2008;16(11):127584.
[8] Wang CW, Jiao ZX, Wu S, Shang YX. Nonlinear adaptive torque control of
where the control frequency is higher and the external load of the electro-hydraulic load system with external active motion disturbance.
hydraulic actuator is closed to its limited capability. Mechatronics 2014;24(1):3240.
[9] Ursu I, Toader A, Halanay A, Balea S. New stabilization and tracking control
laws for electro hydraulic servomechanisms. Eur J Control 2013;19(1):6580.
[10] Mili V, itum , Essert M. Robust H? Position control synthesis of an electro-
Acknowledgements hydraulic servo system ISA Trans 2010;49(4):53542.
[11] Weng FL, Ding YC, Tang MK. LPV model-based robust controller design of
electro-hydraulic servo systems. Proced Eng 2011;15:4215.
We thank Prof. Andrew Plummer, director of Power Center for [12] Song XY, Wang Y, Sun ZX. Robust stabilizer design for linear time-varying
Power Transmission and Motion Control, Department of Mechanical internal model based output regulation and its application to an electro
Engineering, University of Bath for the project funding assistance hydraulic system. Automatica 2014;50(4):112834.
[13] Fales R, Kelkar A. A robust control design for a wheel loader using mixed
and previous research work. sensitivity H-innity and feedback linearization based methods. In: Proceed-
ings of 2005 American control conference; 2005. p. 43816.
[14] Kim CS, Hong KS, Kim MK. Nonlinear robust control of a hydraulic elevator
experiment-based modeling and two-stage Lyapunov redesign. Control Eng
Appendix A. Supplementary material Pract 2005;13(6):789803.
[15] Kim W, Won D, Shin D, Chung CC. Output feedback nonlinear control for
Supplementary data associated with this paper can be found in electro-hydraulic systems. Mechatronics 2012;22(6):76677.
[16] Pi YJ, Wang XY. Observer-based cascade control of a 6-DOF parallel hydraulic
the online version at http://dx.doi.org/10.1016/j.isatra.2015.09.014. manipulator in joint space coordinate. Mechatronics 2010;20(6):64855.
[17] Bonchis A, Corke PI, Rye DC, Ha QP. Variable structure methods in hydraulic
servo systems control. Automatica 2001;37(4):58995.
[18] Dai YB, Yang WD, Wang SF, Zhang M. Electro-hydraulic servo control system
References based on a novel generic model control method. J Iron Steel Res Int 2010;17
(10):237.
[1] Fales R, Kelkar A. Robust control design for a wheel loader using H1 and [19] Yin C, Chen YQ, Zhong SM. Fractional-order sliding mode based extremum
feedback linearization based methods. ISA Trans 2009;48(3):31320. seeking control of a class of nonlinear system. Automatica 2014;50(12):3173
[2] Manring N. Hydraulic control systems. New York: John Wiley and Sons; 2005. 81.
[3] Akers A, Gassman M, Smith R. Hydraulic power system analysis. Boca Raton: [20] Yao JY, Jiao ZX, Shang YX, Huang C. Adaptive nonlinear optimal compensation
Taylor and Francis Group; 2006. control for electro-hydraulic load simulator. Chin J Aeronaut 2010;23(6):720
[4] Boiko I. Variable-structure P.I.D. controller for level process. Control Eng Pract 33.
2013;21(5):7007. [21] Gu DW, Petkov PH, Konstantinov MM. Robust control design with MATLAB.
[5] Moradi M. Self-tuning PID controller to three-axis stabilization of a satellite London: Springer-Verlag; 2005.
with unknown parameters. Int J Non-Linear Mech 2013;49:506. [22] Gawronski W. Balanced systems and structures: reduction, assignment, and
[6] Cerman O, Huek P. Adaptive fuzzy sliding mode control for electro-hydraulic perturbations. Control Dyn Syst 1992;54:373415.
servo mechanism. Expert Syst Appl 2012;39(11):1026977.

You might also like