automated postoperative blood pressure control
TRANSCRIPT
Journal of Control Theory and Applications 3 (2005) 207- 212
Automated postoperative blood pressure control
Hang Z H E N G , Kuanyi ZHU (School of Electrical and Electronic Engineering, Nanyang Technological University, Nanyang Avenue, 639798 Singapore)
Abstract: It is very important to maintain the level of mean arterial pressure (MAP). The MAP control is appfied in many
clinical situations,including limiting bleeding during cardiac surgery and promoting healing for patient's post-surgery. This paper
presents a fuzzy controller-based multiple-model adaptive control system for postoperative blood pressure management.
Multiple-model adaptive control (MMAC) algorithm is used to identify the patient model, and it is a feasible system identification.
method even in the presence of large noise. Fuzzy control (FC) method is used to design controller bank. Each fuzzy controller in
the controller bank is in fact a nonlinear proportional-integral (PI) controller, whose proportional gain and integral gain are adjusted continuously according to error and rate of change of error of the plant output, resulting in better dynamic and stable control
performance than the regular PI controller, especially when a nonlinear process is involved. For demonstration, a nonlinear,
pulsatile-flow patient model is used for simulation,and the results show that the adaptive control system can effectively handle the
changes in patient' s dynamics and provide satisfactory performance in regulation of blood pressure of hypertension patients.
Keywords: Multiple-model adaptive control; Fuzzy control; Blood pressure control; Cardiovascular modeling
1 Introduction
After completion of open-heart surgery, some patients
develop high mean arterial pressure ( M A P ) . . S u c h
hypertension should be treated timely to prevent severe
complication. Continuous infusion of vasodilator drags,
such as sodium nitropmsside ( S N P ) , w o u l d quickly lower
the blood pressure in most patients. However , a wide range
of patient drag sensitivities to SNP occurs and critical
patient care o~en requires the infusion of blood, fluid, or
drags to regulate cardiovascular system variables. Therefore,
manual control by clinical personnel could be very tedious,
time consuming and inconsistent. Automation of these
therapies by application o f feedback control techniques can
be effective.
To improve the quality o f patient care, automatic
closed-loop control systems for SNP delivery have been
developed. A proportional-integral (PI ) controller-based
multiple-model adaptive control ( M M A C ) system was built
in [ 1 ] , and was improved by Martin [ 2 ~ 5 ] . MMAC is a
feasible system identification method even in the presence
of large noise. However, our simulations have shown that
PI controller cannot work well because the system is
nonlinear. A fuzzy control system was built in [ 6 ,7 ] . Based
on the fuzzy controller, general fuzzy control method is
proposed in [ 8 , 9 ] , and the nonlinear controller can
overcome the shortcomings that PI controller possesses
inherently. However, no identification of the plant model is
considered. Furthermore, the wide range of patient drag
sensitivities to SNP is not considered. In this paper, we use
multiple-model adaptive control (MMAC) algorithm to
identify the patient model, and use fuzzy control algorithm
to design controller bank. Thus, the proposed control
method can provide satisfactory control performance and
can handle changes in the system dynamics.
2 Plant characteristics
A dynamic model o f patient mean arterial pressure
response to SNP infusion rate was developed by Slate
[ 10]. It is obtained by using animals' as well as actual
patients' data together with physiological knowledge about
SNP effect. Improvement is made to State's model
according to our nonlinear, pulsatile-flow patient model,
which will be illustrated in Section 4 , and dog experiment
results. The modified Slate's model is given by (1) and
( 2 ) .
MAP = MAP 0 +AMAP + MAPn, (1)
where MAP is the actual mean arterial pressure; MAP0 is
the initial blood pressure; AMAP is the change in blood
pressure due to infusion of SNP; MAPn is the plant
background noise. MAP0 is usually 115 - 140 mmHg[ 1 ] .
The variance of MAPn is typically 4 m m H g (for normal
noise levels) or 16 ~ 36 m m H g (for high noise levels).
Received 8 June 2005.
208 H. ZHENG et al. I Journal of Control Theory and Applications 3 (2005) 207- 212
Many people might consider MAP0 as a constant.
However, MAP 0 in fact can change with time in many
cases. For example, after the patient is cured, his MAP 0
gready declines, and the infusion of SNP can be stopped.
"the " Therefore, defining MAP 0 as initial blood pressure is
imprecise, we can consider that MAP 0 is just a parameter in
( 1 ) , and that the parameter can change with time.
The transfer function describing the relationship between
the change in the blood pressure AMAP and the drug
infusion rate is given by
AMAP(s ) Ke-r ,k ' (1 + a e - r ~ ) SNP(s ) - (1 + Tids)(1 + r s ) ' (2)
where AMAP(s) is the change in blood pressure, SNP( s )
is the SNP infusion rate, K represents the sensitivity o f
patients to SNP, Tik is the transport delay from the
injection site, a is the recirculation constant, Tc is the
recirculation time delay, Tid results fi'om the gradual
relaxing of the vascular muscle in response to SNP, r is the
lag time constant resulting from the uptake, distribution,
and biotransformation o f the drug. Note that steady-state
gain of dose response is K(1 + a ) .The parameters in (2)
are chosen as: K = - 0 .25 to - 9 (normal = - 1 ) , Tik =
10 ~ 20s ( n o n n a l = 2 0 s ) , To = 30 - 75s ( n o r m a l = 4 5
s ) , a = 0 - 0 . 4 , TAd = 10 - 40s ( n o m a a l = 2 0 s ) , v =
3 0 - 6 0 s (normal = 40 s ) . The unit o f AMAP is m m H g .
The unit o f the infusion rate of SNP is m l / h . The
concentration o f SNP is 200 rng/1.
3 Control system design
The MMAC procedure is based upon the assumption
that the plant can be represented by one of a finite number
of models and that for each such model a controller can be
a priori designed. All o f these controllers constitute a
controller bank. An adaptive mechanism is then needed for
deciding which controller should be dominant for a given
plant. One procedure for solving this problem is to form a
weighted sum of all the controller outputs, where the
weighting factors are determined by the relative residuals
between the plant response and the model responses. Fig. 1
shows the structure of the MMAC system.
Setpoint
+
~1 Controller 1
Controller Bank)
~NP2] I Npi; sN : / Integrate _ _
MAP2
MA ?s
Ws
(Model Bank
[~ RI [ Operation L
of
.- R 8 i Residuals
w1 Operation w2. of [ Weights
t ~
Fig. 1
Since the plant gain is negauve, the system error is
expressed as
e ( k ) = M A P ( k ) - MAPo(k ) ,
where k is the sampling time and MAPo(k) is the
commanded or set-point pressure level. The rate change of
error is defined by
r ( k ) = [ m a P ( k ) - M i P ( k - 1 ) ] / r ,
where T is the sampling interval, and T = l0 s.
MMAC system structure.
For patient safety, two nonlinear units are built into the
system. The nonlinear unit limiting SNP infusion rate
change is given as
f l ( x ) = for - 40 ~< x ~< 7, (3)
f o r x > 7 .
The reason for this constraint is that SNP is metabolized by
the body into cyanide, and hence, too much SNP can be
H. ZHENG a a l . / Jounuzl of Control Theory and Applications 3 (2005) 207- 212 209
toxic to the patient. The nonlinear unit limiting SNP
infusion rate is given as
0, for x < 0,
f i ( x ) = x , for 0 ~ X ~ U m, (4)
Um, for x > Urn,
where Um is the maximum infusion rate. This is to prevent
rapid decreases in pressure which can cause diminished
blood flows or circulatory collapse. We use the equation in
[ 1 ] to compute Urn,
U m --- 60 x Wp x i m X C; I, (5)
where Wp is patient weight (the unit is k g ) , i m is
maximum recommended dose (the value is 10 Pg" kg -1 .
m in -1 ) , C, is drag concentration (the unit is pg /ml ) . For
example,if Wp = 70kg, then
(60min/h) x 70kg x (lOtzg • kg -1 • min -1) U m =
200/~g/ml
= 210ml/h.
With respect to the MMAC development, the design of
the model bank is described in Section 3 .1 ; in Section 3.
2, the design of the controller bank is given; and finally,
the actual control computation is described in Section 3 .3 .
3.1 Model bank design The model bank consists of 8 models with constant
parameters characterizing the individual plant subspace.
are described by the following transfer These models
functions:
AMAPi ( s ) SNP( s )
Kie-2°*(1 + 0 .2e -5°') = (1 + 40s ) (1 + 45s) ' j = 1 , . . . , 8 .
(6)
MAPj = MAPoINIT + AMAPj, j = 1 , " ' , 8 , (7)
where AMAPi(s) is the blood pressure change of the j t h
model due to infusion of SNP, SNP(s ) is the input of
these 8 models, MAPj is the output of the j t h model,
MAPoINrr is the estimated initial value of patient's MAP0.
MAPoINrr is calculated in the first 30 seconds, as the average
of MAP(0) , MAP( 1 ) and MAP(2) .
Tc has trivial effect on the control system because the
value of a is small, so it can be frozen in the model
bank. Tid and r are lag time constants, and they can affect
the dynamic behavior of the system. Smaller Tid and v restflt
in smaller settling time. However, the range of Tid and r is
small, so their effects are reduced. Therefore, Tid and r can
also be frozen. In the model bank, the three parameters are
selected as Tcj = 50, Tidj = 25, and rj = 45 (for j = 1,
• "" ,8) . It is noted that while K reflects patient response
sensitivity, the effective steady state gain is K( 1 + a ). That
is, the recirculation enhances the effect of infusion [ 11 ] .
However, the effect of a can be compensated by different
Kj (for j = 1 , ' " , 8) . Therefore, a can also be frozen. In
the model bank, aj = 0 .2 (for j = 1 , ' " , 8 ) . The plant
parameter Tik significantly affects undershoot. Controlling a
high-infusion-dehy patient with a low-infusion-delay model
will cause large undershoot. For a blood pressure control
system,large undershoot is not allowed, so that Tik j (for j
= 1 , ' " , 8 ) should be chosen as the ~ u m expected
plant delay time. Although controlling a low-infusion-delay
patient with a high-infusion-delay model causes long settling
time, it can be partly compensated by different Kj (for j =
1 , - " , 8) . The plant parameter K has the greatest effect on
controller performance. If the plant has high gain and is
being controlled by a low-gain model, large overshoots
occur; if the plant has low gain and is being controlled by a
high-gain model, long settling time Hill occur. Table 1
shows the eight models and the range of plants they cover.
The plant gain partition factor is about 1 .56,
i . e . , Kjn~,,/Kjmi,~ ~ 1.56. Notice that ~ / 1 . 5 6 ~ 1 .25 ,
therefore Kj,,ax/K j ~ Kj/Kimi,, ~ 1.25. Kjmin is the mini-
mum plant gain that Kj covers, and Kjm~x is the maximum
plat gain that Kj covers.
Table 1 Gains of the eight models (column 2 ) , along
with the range of plants they cover (column 3).
Model Model Gain/ Plant Gain/
Number (mmHg/(ml" h- ' )) (mmHg/(ml-h- ' ) ) 1 -0.3125(K1) ffom-O.25(K~a,)to-O.39(K~,~)
2 - 0.4872 (K2) from- 0.39 (K2~a.) to - 0.61 (K2n~x)
3 -0.7625(/(3) from- 0.61 (K3,~,) to - 0.95 (K3,~)
4 - 1. 1875 (K4) t~om- 0.95 (K4~) to - 1.48 (K4~)
5 -1.8500(/(5) ~om- 1.48 (Ksmm) to - 2.30 (Ks,,~) 6 -2.8750(/(6) from- 2.30 (Kt~) to - 3_60 (Kt,~)
7 -4.5000(/(7) from- 3.60 (K7~an) to - 5.60 (K7,~)
8 - 7.2000 (Ks) fi'om- 5.60 (Ks~) to - 9.00 (gs~)
3.2 Controller bank design The details about the fuzzy controller are described in
[ 6 , 7 ] . It converted the expert-system-shell-based fuzzy
controller to nonfuzzy control algorithms, and a total of ten
different nonfuzzy control algorithms for 20 different input
combinations were obtained. We compare the nonfuzzy
control algorithms, and find that they can be substituted in
fact only by one algorithm
ASNP( k )
f ( GE x e (k ) ) + f ( GR x r (k ) ) = - G I ' L ' T "
3L - max[f( GE x e ( k ) ) , f ( gR x r ( k ) ) ] '
(8)
210 H. ZHENG et al . /Journal of Control Theory and Applications 3 (2005) 207- 212
where e ( k ) is the error, r ( k ) is the rate of change of
error, G1 is the scalar for incremental SNP infusion
rate, GE is the scalar for the error, GR is the scalar for the
rate of change of error, L is a turning point of the fuzzy
sets in the fuzzy controller. T is the sampling interval.
Ftmctionf(x) = sgn(x)" min(L, Ix 1). After simplifying
the fuzzy control algorithms as (8 ) , it is much easier to
design the controller bank.
There are eight controllers in the controller bank. All of
the parameters except GI are the same in the eight
controllers,and they are chosen as L = 16, GE = 0.25,
GR = 15. In [6 ] , GR = 13.5;in this paper, we increase
GR to 15, because simulations show that the controller
works better when GR = 15. Table 2 shows GI values of
the eight models. Note that Ks" GIj ..~ O. 0460 (for j = 1,
...,8). Table 2 GI values of the eight controllers (column 2),and
the gains of the eight models (column 3) .
Model Model Gain/ GI
Number (mmHg/(rrfl" h- 1 ) )
1 - 0. 1843 ( G I l ) -0.3125 (Kl)
2 -0.1183 (GI2) - 0.4875 (K2) 3 -0.0755 (GI3) -0.7625 (K3)
4 -0.0485 (G/a) - 1.1875 (K4)
5 - 0.0311 (CI5) - 1.8500 (Ks)
6 - 0.0200 (g16) - 2.8750 (K6)
7 - 0.0128 (g17) - 4.5000 (K7)
8 - 0.0080 (GIs) -7.2000(/(8)
3.3 Identification a lgor i thm
To achieve desirable system performance and guarantee
patient safety, the identification algorithm should converge
quickly to the desired values and should react to
time-varying plant characteristics, as well as ensure a
reasonable rate of blood pressure change. Thus, the control
output is computed as a weighted sum of controller bank
signals, i. e . , 8
ASNP(k) = ~ , [ W j ( k ) . A S N P j ( k ) ] , (9) j=l
where W~(k) is the weighting factor of the j th
controller, ASNPj(k) is the output of the j th controller.
In the first 30 s, MAPoINr r is calculated. In the meantime,
the SNP infusion rate should be equal to zero. Therefore, in
the first 30s,
Wj(k) = 0, j = 1 , . . . , 8 .
The convergence algorithm of Wj will be described later.
Since the control variable SNP(k) will be in error before
the convergence of Wj, large overshoots could occur for
plants with high gain. Therefore, from 0.5 min mark to 3
rain mark, the infusion rate is based on the assumption that
the most matched model is Model 8, i. e. from 0.5 rain
mark to 3 min mark,
0, f o r j = 1 , " ' , 7 , Wj(k) = 1, f o r j 8.
The convergence algorithm is run at 30 s mark, and the
weighting factors begin to converge. However, their values
won ' t be used until 3 rain mark, because their values
haven't converged in the first 3 minutes. The convergence
algorithm in [ 1 - 4] is improved, so that the computation
precision of the weighting factors can be enhanced. In the
algorithm, the weights are selected as follows:
1) Recursive update
exp[-R~( k )/2V2]Wj( k - 1 ) Wj'(k) = 8 ' J = 1,-. . ,8;
~ ] exp[-R2(k)/2V2]Vc~i(k-l) i=1
2) Bounding from zero
a , for E . ' ( k ) ~< a , Wj(k) = Wj ' (k ) , for Wj '(k) > c~, j = 1,
(10)
. . . ,8 ;
(11)
where MAPj (k ) is the output of the j th model.
V is a parameter controlling the convergence rate of
Wj(k ) with R~(k ) . We compare [1] with [12] , and
find that 2
V 2 = tTn (MAP0 - MAPc) 2' (14)
3) Normalization
[ W i ' ( k ) ] 2 Wj(k) = 8 , J = 1 , - - - ,8 , (12)
2
i=1
where R2(k) is the relative residual which will be defined
in (13) , V is a parameter controlling the convergence rate
of (k) with Eq. (11) is used to prevent 8
___a~exp[- R 2 ( k ) / 2 V 2 ] W i ( k - 1), the denominator on i=1
the fight in (10) , from being zero.
In the algorithm, the initia ! weighting factors Wj(O)( j =
1 , ' " , 8) must be determined a priori. We choose them as
Wj(0) = 1/8 ( f o r j = 1 , - - - ,8 ) .
The relative residual R 2 ( k ) is defined as the normalized
squared error between plant and model, i. e.
[ M A P j ( k ) - MAP(k)]2
R2(k ) = M---~OINIT--- ~ ' j = 1, °°* , 8 ,
(13)
H. ZHENG et al . / Journal o f Control Theory and Applications 3 (2005) 2 0 7 - 212 211
2 where a , is the variance of noise associated with MAP. For
rapid convergence of Wj(k ) , a smaller value of V is
desired. However, if we choose a small value of V when
high noise level is present, the system ~ be unstable. In
this paper, we choose V as V = 0 .1 .
4 Simulation study
Fig. 2 shows the structure of a nonlinear, pulsafile-flow
patient model, and we develop it based on [5,13 - 15]. It
is a much more complex patient model than modified
Slate' s model, and will be used in computer simulation. In
SNP pharmacokinetic model and pharmacod3mmnic model,
the effect of different values of the infusion delay, sensitivity
to SNP, amotmt of recirculation, and speed of action of
SNP, can all be studied. In cardiovascular model, it is
possible to change, independently or in any combination,
the resistances of the arteries and veins, their comphances,
the strength of the heat 's contraction, heart rate and many
other parameters. We only introduce some most important
parameters in the patient model. ;t and T d are two
parameters in SNP pharmacokinetic model. The gradual
relaxing of the vascular muscle in response to SNP is
determined by Td, and 2 represent SNP recirculation
effect. /31 and f12 are two parameters in SNP
pharmacodynamic model, and they represent the sensitivity
of patients to SNP.
Concentration of Arterial resistance
SNP SNP in the arteries SNP
SNP (t) Pharmacokinetic Phamacodynamic model Concentration of model Venous unstressed
SNP in the veins volume change
Cardiovascular I model I
Arterial pressure
Fig. 2 A nonlinear, pulsatile-flow patient model.
The set-point pressure level is initially chosen as MAP~ =
100. Some important parameters of the patient model are
initially chosen as ;t = 0 .4 , T d = 4 0 , fll = 90 and/32 =
120. Then, Td is changed to 25 at the 4 rain mark, /31 and
/32 are increased to 135 and 180 respectively at the 20 rain
mark, MAPc is decreased to 80 at the 35 min mark, 2 is
changed to 0 .6 at the 50 rain mark. White Gaussian noise
(mean 0, variance 9) is included. Figs. 3 and 4 show the
simulation results.
140
130 120 ZZ
E 110 ~- 100 ~ 9o
80
70
L, i0 80 ' I~tl/r ~ ' - ' ~ , l l . Z:0.4~0.6
Td:40--25 fi1:90~135 ~ . ~
, /~2,:120~!80 , , ~ h ~ ' '" 10 20 30 40 50 60 70
t / min
Fig. 3 Mean arterial pressure.
e -
.o
35[ " fl1:90]~135 " ! " ~ " 30 / t2 :120~180 i ,¢ '" i~.
:04_06 - ls[ f MAp0:100-80 l0 }
,
0 10 20 30 40 50 60 70 t / rain
Fig. 4 Mean arterial pressure.
5 Conclusions
A fiazzy controller-based MMAC algorithm is proposed
to control postoperative hypertension. Modification has
been made to Slate's model and a nonlinear, pulsatile-flow
patient model is developed for computer simulation. The
results of simulations indicate that the control system can
automatically control blood pressure over a fairly wide plant
parameter envelope, even in the presence of large
background noise. Further animal experiments and a flail
clinical test should be conducted.
R e f e r e n c e s
[ 1 ] W.G. He, H. Kaufinan, K. Roy. Multiple-model adaptive control proce-
dure for blood pressure control [J ]. IEEE Trans. on Biomedical
Engineering, 1986,33( 1 ) : 10 - 19.
[ 2 ] J .F . Martin, A. M. Schneider, N. T. Smith. Multiple-model adaptive
control of blood pressure using sodium ultroprusside [J]. IEEE Trans.
on Biomedical Engineering, 1987,34(8):603- 611.
[3] J.F. Martin, A.M. Schneider, M.L. Quinn, et aI. Improved safety and
eflacacy in adaptive control of arterial blood pressure through the use of a
supervisor [ J ] . IEEE Trans. on Biomedical Engineering, 1992, 39
(4) :381 - 388.
[4] J.F. Martin, N. T. Smith, M. L. Quinn, et al. Supervisory adaptive con-
trol of arterial pressure during cardiac surgery [ J ]. IEEE Trans. on
Biomedical Engineering, 1992,39(4) :389 - 393.
[5] "J. F. Martin. Multiple-Model Adaptive Control of Blood Pressure Using
Sodium Nitroprnsside [ D] . San Diego: Department of Applied Mechan-
ics and Engineering Science, University of California, 1987.
[6] H.Ying,M.McEachem,D.W.Eddleman, et al. Fuzzy control of mean
2 1 2 H. ZHENG a a l . / Joumal o f Control Theory and Applications 3 (2005) 2 0 7 - 212
arterial pressure in postsurgical patients with sodium nitropmsside infusion
[ J ]. IEEE Trans. on Biomedical Engineering, 1992,39 (10) : 1060 -
1070.
[7 ] H. Ying, L. C. Sheppard. Kegnlation mean arterial pressure in postsurgical
cardiac patients [J ] . IEEE Engineering in Medicine and Biology,
1994,13(5) :671 - 677.
[ 8 ] H. Ying. Explicit smacmre of the typical two-input fuzzy controllers [ C]
/ / Proc. of the Fifth IEEE Int. Conf. on Fuzzy Systems. New Or-
leans, LA, USA, 1996,1:498 - 503.
[9] F. Lin, H. Ying. Modeling and control of fuzzy discrete event systems
[J] . IEEE Trans. on Systems, Man and Cybernetics, Part B,2002,39
(4) :408 - 415.
[10] J" B ' Slate' L" C" Sheppard' V ' C" Ride°ut ' et al" A m°dd f°r design ° f a
blood pressure controller for hypertensive patients [ C ] / / The 5 th
IFAC Syrup. on Identification and System Parameter Estimation.
Darmstadt: Federal Republic of Germany, 1979.
[ 11 ] J .S . Delapasse, K. Behbehani, K. Tsni, et al. Accorranodation of time
delay variations in automatac infusion of sodium nitroprusside [ J ] .
IEEE Trans. on Biomedical Engineering, 1994, 41 ( 11 ): 1083 -
1091.
[12] C . Y u , tL.J .R.oy, H. Kanfrnan, et al. Multiple-modd adaptive predic-
tive control of mean arterial pressure and cardiac output [J ] . IEEE
Trans. on Biomedical Engineering, 1992,39(8 ) : 765 - 778.
[13] J .F .Mar t in ,A.M.Schneider , J .E .Mandel , et al .A new cardiovascular
model for real-time applications [ J ]. Trans. Soc. Comput.
Simulation, 1986,3 ( 1 ) : 31 - 66.
[ 14] Eileen A. Woodruff, James F. Martin, Madonna Omens. A model for
the design and evaluation of algorithms for dosed-loop cardiovascular
therapy [J ] . IEEE Trans. on Biomedical Engineering, 1997,44(8) :
694 - 705.
[ 15 ] Vincent C. Rideout. Mathematical and Computer Modeling of
Physiological Systems [ M ] . Englewood Cl i~ , NJ: Prentice Hall,
1 9 9 1 : 3 - 4 .
[ 16] K. Behbehani, R . P,.. Cross. A controller for regulation of mean arterial
blood pressure using optimum nitroprussde infusion rate [ J ] . IEEE
Trans. on Biomedical Engineering, 1991,38(6) : 5 1 3 - 520.
[ 17 ] G.I. Voss, H.J . Chizeck, P. G. Katona. Self-mmng controller for drug
delivery systems [J]. Int. J. Control, 1998 ,47(5) :1507- 1520.
t l a n ~ ZI-IF_,NG graduated from the University
of Science and Technology of China, in 2001.
Currendy, he is a master student in School of Elec-
trical and Electronic Engineering of Nanyang
Technological University, Singapore. His research
interests include nonlinear time-varying system
control,multiple model adaptive control and fuzzy
control. Email: zhen0008 @ ntu. edu. sg.
K u a n y i Z I / U graduated from No~heastem
University,China,in 1982.He received his M . E .
and Ph. D from University of Louvain-La-Neuve,
m Belgium in 1986 and 1989, respectively. Cur-
rently, he is associate professor with School of
Hectrical and Electronic En~neering of Nanyang
~ I Technological University, Singapore. His research
interests include model-based predictive control,
biomedical system control and process control.
Ernail: ekyzhu @ ntu. edu. sg.