Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
By
Sayyar Ahmad
Supervised by
Dr. Nisar Ahmed
Certified that the research work presented in this thesis, entitled “Nonlinear Control of
Blood Glucose Level in Type 1 Diabetes Patients” was conducted by Mr. Sayyar Ahmad
under the supervision of Dr. Nisar Ahmed and Dr. Nasim Ullah.
i
Dedication
ii
Acknowledgment
First of all, I would like to express my deepest gratitude to Almighty Allah, the Creator, the
most Merciful and the Beneficent, for His countless blessings and to enable me achieve this goal.
I also offer my praises to Holy Prophet MUHAMMAD (S.A.W.) and his companions who laid the
foundations of Islamic civilization and paved the way for social, moral, economic, and cultural
revolution.
I am highly indebted to my advisor Dr. Nisar Ahmed and co-advisor Dr. Nasimullah. Working
with them has been a wonderful experience from many aspects. First, their belief in my abilities
has kindled my interest in research. Second, their encouragement to come up with my own original
ideas has helped me appreciate the value of formulating meaningful research problems and had a
profound effect in shaping my vision for future research. I thank them for their laudable teaching
and guiding me in accomplishing this work.
I would like to acknowledge GIK Institute for sponsoring my MS studies and for providing me
such a memorable environment. I am also grateful to the following former or current staff at Faculty
of Electrical Engineering,GIK Institute for their various forms of support during my graduate study
Dr. Husnul Maab, Dr. Adnan Noor, Dr. Ziaul Haq Abbas, Dr. Khasan Karimov, Ali Ghais, Mazhar
Javed, Mehran Bashir, Zaiwar Ali, Ikram Ullah, Iftikhar Ali Khan and Muhammad Arshad.
To all my friends, thank you for your understanding and encouragement in my many many
moment of crisis. Your friendship makes my life a wonderful experience. I would like to ac-
knowledge Attiq Ahmed, Ejaz Khan, Mussawer Pervez, Muhammad Tariq, Umer Farooq, Imran,
Ammar Ahmed, Salman Khan, Nihad Ali, Waqar Alam, Attiq ur Rehman, Muhammad Ilyas, Awais,
Muhammad Adil, Obaid, Shaukat, and Shakeel.
Most importantly, none of this would have been possible without the love and patience of my
family. My immediate family to whom this dissertation is dedicated to, has been a constant source
of love, concern, support and strength all these years. I would like to express my heart-felt gratitude
to my family. My extended family has aided and encouraged me throughout this endeavor. I warmly
appreciate the generosity and understanding of all of them.
Sayyar Ahmad
April, 2017
iii
iv
Declaration
I declare that this is my own work and has not been submitted in any form for another degree
or diploma at any university or other institution for tertiary education. Information derived from
published or unpublished work of others has been acknowledged in the text and a list of references
is provided at the end of this dissertation.
Sayyar Ahmad
April, 2017
v
Abstract
Diabetes mellitus is one of the most common diseases in the world. Specifically, type 1 diabetes
mellitus in which pancreas of patient does not produce insulin resulting in high concentration of
glucose in blood plasma for longer periods of time. High concentration of glucose can cause many
complications and may lead to death. For this purpose strict control of blood glucose concentration
is required in type 1 diabetes patients. In this research work nonlinear closed loop control algorithms
have been investigated for blood glucose regulation in type 1 diabetes patients, which can be used
in automatic insulin pumps, referred to as artificial pancreas in literature. Bergman’s minimal model
is considered for studying dynamical behavior of glucose and insulin inside human body.As slid-
ing mode control (SMC) is known for its robustness against disturbances and mismatches, therefore
three nonlinear control strategies based on SMC have been developed and analyzed, namely Classi-
cal Sliding Mode Control (CSMC), Super Twisting Sliding Mode Control (STSMC) and Backstep-
ping Integral Sliding Mode Control (BISMC). CSMC based controller shows chattering, which may
be dangerous in practical actuation of controller. In order, to cope with the problem of chattering
STSMC algorithm is developed. For improving the transient response of controller a novel BISMC
is introduced and developed showing satisfactory results. The advantages of backstepping and in-
tegral SMC are combined for achieving the desired performance. Problem of hypoglycemia related
to previous nonlinear control techniques is also removed by using proposed controller. Three cases
have been considered for analysis of the SMC based controllers. Firstly the controllers are subjected
to patient having hyperglycemia whose blood glucose was high. Secondly, controllers are analyzed
for patients having hyperglycemia and meal disturbance. Finally, controllers are studied for patients
subjected to three meal intakes (as disturbances) corresponding to break-fast, lunch and dinner. It
is shown that proposed BISMC based controller shows best performance and successfully stabilize
the glucose level of patient for above mentioned three cases. A comparison of CSMC, STSMC and
BISMC can be found in results. Meal disturbance to diabetes patient is also explicitly given in sim-
ulation results. Simulation results are provided to show superiority and advantages of the proposed
controller such as robustness to external disturbances, no chattering, less control effort and accurate
output tracking, thus proposed controller can be considered a suitable choice for designing automatic
insulin pumps for type 1 diabetes patients.
vi
Contents
Dedication i
Acknowledgment ii
Declaration iv
Abstract v
List of Figures x
Abbreviations xiii
1 Introduction 1
1.1 Diabetes Mellitus . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Type 1 DM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2 Type 2 DM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.3 Pancreas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.3.1 Alpha Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.3.2 Beta Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.4 Regulation of Glucose . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Control Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.1 Linear Control Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.2 Nonlinear Control Review . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
vii
2.2.1 Glucose Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.2.2 Block Diagram for Glucose Dynamics . . . . . . . . . . . . . . . . . . . . 10
2.2.3 Insulin Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2.4 Block Diagram for Insulin Dynamics . . . . . . . . . . . . . . . . . . . . 11
2.3 Complete Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.3.1 Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4 Backstepping Integral Sliding Mode Control for Blood Glucose Regulation in Dia-
betes Patients 32
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2 Backstepping Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
viii
4.3 Backstepping Integral Sliding Mode Control . . . . . . . . . . . . . . . . . . . . . 33
4.3.1 Derivation of Control Law . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.3.2 Simulation Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.3.2.1 Regulation of Glucose during Hyperglycemia . . . . . . . . . . 34
4.3.2.2 Regulation of Glucose during Hyperglycemia and Meal Distur-
bance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.3.2.3 Regulation of Glucose in Presence of 3 Meal Disturbances . . . 37
4.4 Comparative Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
4.4.0.1 Regulation of Glucose during Hyperglycemia . . . . . . . . . . 39
4.4.0.2 Regulation of Glucose during Hyperglycemia and Meal Distur-
bance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.4.0.3 Regulation of Glucose in Presence of 3 Meal Disturbance . . . . 42
References 47
ix
List of Figures
x
3.25 Insulin concentration in presence of 3 meal disturbance using STSMC . . . . . . . . 31
3.26 Control input in presence of 3 meal disturbance using STSMC . . . . . . . . . . . . 31
xi
List of Tables
xii
Abbreviations
xiii
Chapter 1
Introduction
Control systems design has gain a tremendous attention in tackling the challenging control problems
in biomedical control systems (BCS). BCS in nature shows nonlinear behavior, for attaining accu-
rate control performance nonlinear control is the appropriate option in this scenario. Blood glucose
regulation is one of the most common and important problems in BCS. Abnormal behavior of blood
glucose levels causing different long term and lethal diseases is distinguished by a disease called
Diabetes Mellitus. Diabetes mellitus (DM) is a metabolic disease, which is rapidly increasing all
around the world. According to world health organization (WHO), diabetes victims are increased
from 108 million (1980) to 422 million (2014) [1]. Statistical analysis done by WHO reflect that di-
abetes would be number seven leading cause of death in 2030 [2]. A report released by International
Diabetes Federation (IDF) states if no action is taken against diabetes, approximately there will be
642 million diabetes patients in 2040, which is an increase of about 50 percent [3].
1
Chapter 1 Introduction
1.1.1 Type 1 DM
Type 1 DM is a type of DM disease also known as insulin-dependent DM. In this type of DM the
immune system of human attacks its own pancreas and destroy the cells responsible for secretion of
insulin thus destroying the regulation of blood glucose levels at normal value.
1.1.2 Type 2 DM
It is the most common DM disease. In type 2 DM body of patient shows resistance to the secretion
of insulin by pancreas due to which not sufficient amount of insulin is secreted in order to regulate
the blood glucose level at normal value thus leading to hyperglycemia.
1.1.3 Pancreas
Pancreas is an endocrine organ located in abdominal cavity, playing vital role in homeostasis of
blood glucose level. Hormones responsible for regulation of glucose levels are secreted by pancreas
namely insulin and glucagon.
The cells responsible for secretion of Glucagon are called alpha cells. They are located in islet of
langerhans within pancreas.
2
Chapter 1 Introduction
The cells responsible for secretion of Insulin are called beta cells. They are located in islet of
langerhans within pancreas.
Glucose is an important source of energy for body cells, reaching cells through blood stream. In hu-
mans the glucose-insulin regulation system works under stiff homeostatic control condition. When-
ever glucose level in blood plasma exceeds normal value, insulin is secreted for regulation whereas
low levels of glucose in blood plasma are compensated by secretion of glucagon.
The linear control algorithms developed for blood glucose regulation in DM patients are discussed
in this section. Among linear control strategies the Proportional Integral Derivative (PIDs) based
controllers i.e. Expert PID control [8], Robust PID control [9], Digital PID control [10], Genetic al-
gorithm based PID control [11] and switching based PID control startegy [12] were proposed. Other
linear controllers deployed are glucose control based on feed forward-feedback strategy [13], con-
trol algorithm based on pole placement technique [14], controller based on discrete model prediction
algorithm [15], and H∞ based robust control design [16].
Besides linear control algorithms researchers also show keen interest in designing nonlinear con-
trollers for blood glucose regulation in diabetes patients to achieve the required performance.As
blood-glucose regulation is a nonlinear phenomenon in nature, thus nonlinear control can provide
better performance as compared to linear controllers based on linearized models. Some of nonlinear
3
Chapter 1 Introduction
controllers proposed are nonlinear robust controller based on model predictive algorithm [17], Con-
trol based on fuzzy logic [18], high order sliding mode control [19], adaptive gain high order sliding
mode control using fuzzy logic [20], recurrent based neural network control [21], sliding mode con-
troller based on backstepping technique [22], Fractional order PID controller [23], and fractional
sliding mode controller with adaptive gains [24].
1.4 Objectives
Designing a controller for tight control of blood glucose regulation in DM patients is a challenging
task. The control objectives are listed below:
4
Chapter 1 Introduction
Chapter 4 introduces backstepping integral sliding mode control strategy for blood glucose reg-
ulation in type 1 diabetes patients. The performance of proposed controller is validated by including
simulation results. MATLAB software is used for simulation purposes.
Chapter 5 concludes the dissertation with a summary of findings and highlights future research
challenges.
5
Chapter 2
Modeling of Glucose-Insulin
Dynamics
2.1 Introduction
In order to design a model based controller mathematical modelling of the biomedical system de-
scribing the dynamics of insulin and glucose is necessary [26]. A lot of research is also carried
out on modelling of insulin-glucose dynamics inside human body. Researchers have proposed dif-
ferent models including linear models developed by Bolie 1961 and Ackerman et al 1965 [27,28].
Mentioned linear models are so simple with two states. Regarding nonlinear behaviour of glucose-
insulin dynamics physiological models developed are Tiran et al 1979, Sorensen 1985, Hovorka et
al 2004 and Dalla Man et al 2007 [29-31].Compartmental models derived for glucose-insulin dy-
namics consist of Bergman et al 1981, Cobelli et al 1982, Cobelli and Mari 1983 and Lehmann 1992
[32,33]. Among models of glucose-insulin dynamics discussed above Bergman’s minimal model
have been chosen for simulation study of the controller. Linear models stated are too simple and
cannot tackle challenges for appropriate control. Physiological models are very complex, consist-
ing of many state variables and parameters, making designing of a controller difficult and tedious.
While minimal model is one of the best pioneer models and consists of least parameters necessary.
Moreover, minimal model relates the main states in question i.e. glucose and insulin concentrations
[34].
6
Chapter 2 Modeling of Glucose-Insulin Dynamics
clear and simple way. Minimal models consist of least parameters and set of equations necessary
for explaining the process. One such model for explaining the blood glucose-insulin kinetics in man
was proposed by Richard N. Bergman in 1980. This minimal model is one of the pioneers and well-
known model of glucose-insulin regulatory system in man. In minimal model the body is considered
as a compartment composed of basal values of insulin and glucose concentrations therefore; also
known as one compartment model. Minimal model consists of two parts. One explaining dynamics
of glucose, how concentration of glucose is affected by concentration of insulin in blood. Second
part is given to explain the effect of glucose concentration on insulin concentration in blood. The
derivation of glucose-insulin dynamics used in minimal model is based on Intravenous Glucose
Tolerance Test (IVGTT).
There exist two kinds of masses flowing out of compartment, i.e consumption of glucose by liver and
periphery. There exists one in going mass, i.e production of glucose by liver. This in and out flow
of glucose give rise to a basal concentration of glucose, determined by threshold Gb . According to
Steil et al. [35], basal concentration is given by difference of consumption and production of glucose
independent of glucose and insulin uptginind , prodginind respectively. When glucose level exceeds
Gb , glucose level is decreased by consumption of glucose by liver upli and periphery tissues upph .
When glucose level is below Gb liver produces glucose until Gb is reached. Steil et al, referred this
balance as Net Hepatic Glucose Balance (NHGB) [35]. NHGB and upph are given by:
7
Chapter 2 Modeling of Glucose-Insulin Dynamics
For changing L to dl k is used; a constant and is set to 1. Now using mass balance rule:
Gb (to + ∆t)
=
∆t
QGb1 QGb2 QGb1 ωb1 QGb2 ωb2
Gb + Gb − ( Gb (t) + Gb (t)Xb (t) + Gb (t) + Gb (t)Xb (t))
VGb VGb VGb VGb VGb VGb
QGb1 ωb1 QGb2 ωb2
By replacing = k1 , = k4 , = k5 , = k6 and applying the limit ∆t → 0 the
VGb VGb VGb VGb
following differential equation is derived:
dGb (t)
= k1 Gb + k5 Gb − (k1 Gb (t) + k4 Xb (t)Gb (t)) − (k5 Gb (t) + k6 Xb (t)Gb (t)) (2.3)
dt
Differential equation derived above doesn’t show dynamics of active insulin present in remote com-
partments. Mathematical expression can be derived for dynamics of active insulin using mass bal-
ance rule as follows:
Whenever concentration of insulin in blood plasma Ib (t) is above the basal value Ib , insulin starts
flowing into remote compartment. If concentration goes below the basal value, insulin flows out of
the remote compartment, thus maintaining balance referred to as balins . Other outgoing mass from
remote compartment is proportional to the level of active insulin and is referred to as Xclr . Now
using mass balance theory we have:
Dividing both sides by ∆t and Vxb and applying the limit ∆t → 0 the following differential equation
is derived:
8
Chapter 2 Modeling of Glucose-Insulin Dynamics
Now by letting p1 = k1 + k5 , p2 = k3 and p3 = (k4 + k6 ) we get the model equations given in (2.1)
and (2.2).The block diagram of glucose dynamics given in minimal model is given below:
9
Chapter 2 Modeling of Glucose-Insulin Dynamics
Dynamics of insulin in body is described in minimal model by the following differential equation:
dI b (t)
= −n(Ib (t) − Ib ) + γ[Gb (t) − h]+ t (2.5)
dt
Derivation of insulin kinetics is also based on rule of mass balance like glucose kinetics. Assump-
tions used for derivation are described by Gaetano et al, [36]. Accumulated mass of insulin is
actually the difference between final and initial mass which is given by:
The natural reaction of pancreas for maintaining basal value of insulin is mathematically described
as P (t) = γ[Gb (t) − h]+ t whereas the term [Gb (t) − h]+ has some value for [Gb (t) − h] positive
and 0 otherwise. Now using mass balance rule we have:
Now dividing both sides by VIb and ∆t and applying the limit ∆t → 0 we get:
10
Chapter 2 Modeling of Glucose-Insulin Dynamics
11
Chapter 2 Modeling of Glucose-Insulin Dynamics
12
Chapter 3
3.1 Introduction
Dealing any practical control problem, there exists difference between actual plant and its mathemat-
ical modeling used for designing a controller. These differences or mismatches result from external
disturbances, variation in parameters and dynamics which are not modeled. In order to design a
controller for closed-loop systems that provides desired performance in the presence of such mis-
matches is a challenging objective. This conception has led to the discovery of robust controllers.
Among robust controllers one of the most famous is sliding mode control (SMC) technique.
13
Chapter 3 Sliding Mode Control........
In SMC first of all the system trajectories are forced to reach the designed sliding manifold from their
initial points. This phase of reaching system trajectories to sliding manifold is termed as reaching
phase in SMC theory.
Once system trajectories reach sliding manifold, they are forced to stay on sliding surface without
leaving it and are forced to reach equilibrium point. This phase in which system trajectories slide
over surface is termed as sliding phase in SMC theory. Reaching and sliding phases are shown in
figure below [38]:
dX1b (t)
= −(p1 + X2b (t))X1b (t) + p1 Gb + D(t) (3.1)
dt
dX2b (t)
= −p2 X2b (t) + p3 (X3b (t) − Ib ) (3.2)
dt
14
Chapter 3 Sliding Mode Control........
dX3b (t)
= −n(X3b (t) − Ib ) + u(t) (3.3)
dt
Where γ[Gb (t) − h]+ t term is zero in type 1 patients as there is no release of insulin by pancreas,
D(t) is external meal disturbance and u(t) is control input showing infusion of insulin.
d
S=[ + λ]r−1 e (3.4)
dt
Where r is the relative degree and e = X1b (t) − X1bref is output error. In this case r = 3 as control
input explicitly appears in third derivate of output Gb (t), so we have:
d
S=[ + λ]2 e (3.5)
dt
S = ë + λ1 ė + λ2 e (3.6)
Which is the required sliding manifold where λ1 and λ2 are design parameters.
In order to derive equivalent control law for SMC controller the lyapunov candidate function is
selected as:
1 2
V(X ) = S (3.7)
2
The derivative of lyapunov function will be:
V̇(X ) = S Ṡ
...
Ṡ = e + λ1 ë + λ2 ė
... ... ...
e = X 1b (t) − X 1bref
...
=⇒ e = ψ(X, t) − p3 X1b u(t)
ψ(X, t) = [−p1 (p1 2 + 3p3 Ib ) − p3 Ib (p2 + n) − p3 γ[X1b (t) − h]+ t]X1b (t) + [−p1 2 (1 + Gb ) +
p1 p2 (2Gb − 1) + 2D(t)(p1 + p2 )]X2b (t)+
2
[−2p3 (p1 + D(t))]X3b (t) + [−(p1 + p2 ) − 3p3 Ib ]X1b (t)X2b (t)+
2
[p3 (3p1 + p2 + n)]X1b (t)X3b (t) + [−3(p1 + p2 )]X1b (t)X2b (t) +
2 3
(p1 Gb + D(t))X2b (t) + 3p3 X1b (t)X2b (t)X3b (t) − X1b (t)X2b (t) +
D̈(t) + (p1 Gb + D(t))(p1 2 + 2p3 Ib )
=⇒ Ṡ = ψ(X, t) − p3 X1b u(t) + λ1 ë + λ2 ė
15
Chapter 3 Sliding Mode Control........
Now the equivalent control law can be derived and is given by:
1
ueq (t) = (ψ(X, t) + λ1 ë + λ2 ė) (3.8)
p3 X1b
Ṡ = 0
=⇒ V̇(X ) = 0
In order to make V̇(X ) negative definite for achieving asymptotic stability we have switching control
law which is described in next section.
For the sake of achieving asymptotic stability in SMC, based on lyapunov theory, discontinuous
control law is used also known as switching control law. The switching control law used here is
given by:
usw (t) = k1 S + k2 sig(S) (3.9)
Where k1 and k2 are design parameters, which are tuned for achieving desired response. The overall
control law comes out to be:
1
u(t) = (ψ(X, t) + λ1 ë + λ2 ė + k1 S + k2 sig(S)) (3.10)
p3 X1b
Using this control law given in equation (3.10) derivative of lyapunov function becomes:
This section includes the simulation results for analyzing the performance of derived SMC control
law given in equation (3.10).
16
Chapter 3 Sliding Mode Control........
Parameter Value
p1 0
p2 0.02
p3 0.0000053
γ 0
n 0.3
Ib 7
Gb 80
X1b (0) 250
X3b (0) 40
λ1 0.0331
λ2 0.00017225
k1 0.0199
k2 0.00053
Figurer 3.2 shows stabilization of glucose concentration in blood plasma when the patient is in state
of hyperglycemia. The level of glucose is forced to the basal value in about 650 minutes.
17
Chapter 3 Sliding Mode Control........
Figure 3.3 shows the dynamical behavior of error appearing in glucose concentration of patient.
The output error is driven to zero in about 650 minutes thus bringing the concentration of glucose to
the normal value.
Figure 3.4 shows concentration of insulin in blood plasma. The phenomenon of chattering can
be observed, which is the main drawback of SMC algorithm. The concentration of insulin is driven
to the basal value successfully.
Figure 3.5 shows the infusion of insulin required for blood glucose regulation of diabetes patient,
which is the control input generated by SMC controller. There exists chattering in control input
18
Chapter 3 Sliding Mode Control........
which may be dangerous in practical implementation. To cope with the problem of chattering Super
Twisting SMC algorithm is used.
Figure 3.7 shows the stabilization of blood glucose concentration in the presence of meal disturbance
which occurs at t = 500 min. Meal disturbance has an amplitude of 25 mg/dL but lasts for about 25
minutes. It can be observed that concentration of glucose in blood plasma is brought to normal value
i.e. 80 mg/dL.
19
Chapter 3 Sliding Mode Control........
Figure 3.8 depicts kinetics of output error. It is clear that when disturbance is added to the
system, the error is rapidly increased to about 160mg/dL, which is forced back to zero, it can be seen
in given figure.
Figure 3.9 shows the concentration of insulin in blood plasma. Once disturbance appears at t =
500 min, insulin rapidly rises to lower the glucose concentration to normal. The intrinsic property
of classical SMC, chattering can be observed in the given figure.
20
Chapter 3 Sliding Mode Control........
Figure 3.10 is a picture of control input, the infusion of insulin into patient. A peak can be
noticed at the instant where disturbance is added to the system. This rapid increase in insulin infusion
is due to rapid increase in glucose in the form of disturbance. Chattering can be noticed in the given
figure.
21
Chapter 3 Sliding Mode Control........
Figure 3.12 shows the regulation of blood glucose concentration in the presence of three meal intakes
corresponding to break fast, lunch and dinner. It is clear from the figure that the blood glucose
concentration is stabilized successfully.
Figure 3.13 shows the concentration of insulin in blood plasma. Rise in insulin level correspond-
ing to meal intakes can be observed, in order to lower the blood glucose concentration to normal.
Phenomenon of chattering is clear in the figure which is due to discontinuous control law of classical
SMC.
22
Chapter 3 Sliding Mode Control........
Figure 3.14 shows the infusion of insulin, control input generated by SMC controller in the
presence of three meal intakes. The three peaks appearing in control input are due to rapid increase
in glucose level corresponding to meal intake. Chattering is obvious in the given figure.
The main disadvantages of classical SMC algorithm based controller are given, which can also be
observed in simulation results provided.
23
Chapter 3 Sliding Mode Control........
In this section super twisting sliding mode controller is designed for blood glucose regulation in type
1 diabetes patients.
S = ë + λ1 ė + λ2 e (3.11)
It can be derived in the similar way as done for SMC and is given by:
1
ueq (t) = (ψ(X, t) + λ1 ë + λ2 ė) (3.12)
p3 X1b
The switching algorithm used in super twisting sliding mode controller is given by:
1
2
usw (t) = η1 |S| sign(S) + ν (3.13)
Where ν̇ = η2 sign(S)
So the switching law comes out to be:
1 Z
usw (t) = η1 |S| 2 sign(S) + η2 sign(S)dτ (3.14)
24
Chapter 3 Sliding Mode Control........
Parameter Value
p1 0
p2 0.02
p3 0.0000053
γ 0
n 0.3
Ib 7
Gb 80
X1b (0) 250
X3b (0) 40
λ1 0.0199
λ2 0.0003975
η1 0.000795
η2 9.275 × 10−11
In this section simulation results are provided showing the performance of STSMC based controller.
Figure 3.15 shows regulation of blood glucose level using STSMC based controller. It can be seen
that reaching time is improved and basal level is achieved in about 460 minutes.
25
Chapter 3 Sliding Mode Control........
Figure 3.16 is a picture of error in glucose concentration. The error is reduced to zero suc-
cessfully in about 460 minutes, showing better performance as compare to classical SMC based
controller.
Figure 3.17 explains concentration of insulin in blood plasma. It can be noted that level of
insulin is smooth and do not have chattering. So, using STSMC based controller the problem of
chattering is eliminated successfully.
26
Chapter 3 Sliding Mode Control........
Figure 3.18 shows the infusion of insulin by pump into patient. The control input is smooth
having no chattering like classical SMC algorithm.
Figure 3.19 depicts the performance of STSMC algorithm when the subject is exposed to meal
disturbance, which enters the system at time 450 min. It can be seen that the level of glucose is
forced back to basal value successfully.
27
Chapter 3 Sliding Mode Control........
Figure 3.20 shows the dynamical behavior of output error. Error is reduced to zero in the pres-
ence of meal disturbance at time t = 1050 min.
Figure 3.21 shows the concentration of insulin in blood. It can be noted that concentration of
insulin is smooth having no chattering anymore.
28
Chapter 3 Sliding Mode Control........
Figure 3.22 gives the dynamics of control input generated by STSMC based controller. It is
obvious that there is no chattering in control input, the infusion of insulin to patient is smooth.
Figure 3.23 shows the regulation of glucose in the presence of three meal intakes using STSMC
based controller. Finally the blood glucose concentration is stabilized at its basal value.
29
Chapter 3 Sliding Mode Control........
Figure 3.24 is a picture of output error. The behavior of error is consistent with the output.
Figure 3.35 shows concentration of insulin in blood compartment. The three peaks appearing in
graph are due to the disturbance added at these instances. The behavior of insulin concentration is
smooth having no fluctuations of high frequency.
30
Chapter 3 Sliding Mode Control........
Figure 3.26 shows the control input, infusion of insulin by pump into patient. The control input
is smooth and the phenomenon of chattering is eliminated using STSMC algorithm for designing the
controller.
31
Chapter 4
4.1 Introduction
There are two major objectives, while designing a controller for regulation of blood glucose con-
centration in diabetes patients. Firstly, the controller should be robust against external disturbances
and mismatches. As it is concerned to human body so the controller must be accurate and precise in
its operation. Secondly, the blood glucose concentration should be stabilized in reasonable amount
of time. There is no well defined reaching time as different subjects have different glucose-insulin
dynamics, however reaching time should not be less then two hours. In order to obtain the two
objectives mentioned above a backstepping integral sliding mode control (BISMC) algorithm is in-
troduced. SMC is known for its robustness against mismatches and external disturbances where as
backstepping is used for reducing the reaching time of glucose concentration in blood plasma of
patient. The derivation of control law based on BISMC is given and simulation results are presented
for analyzing the performance of proposed controller.
32
Chapter 4 Backstepping Integral Sliding Mode Control........
Where e1 = x1 (t) − Gb ,
Ṡ1 = ė1 + c1 e1
⇒ Ṡ1 = ẋ1 − Ġb + c1 e1
⇒ Ṡ1 = −p1 x1b (t) − z1 (t)x1b (t) + p1 Gb + D(t) + c1 e1
1
z1 (t) = (−p1 x1b (t) + p1 Gb + D(t) + c1 e1 ) (4.4)
x1b (t)
The second equation is given by:
Ṡ2 = ė2 + c2 e2
⇒ Ṡ2 = ẋ2b − ż1 (t) + c2 e2
⇒ Ṡ2 = −p2 x2b (t) + p3 z2 (t) − p3 Ib − ż1 (t) + c2 e2
33
Chapter 4 Backstepping Integral Sliding Mode Control........
Ṡ3 = ė3 + c3 e3
⇒ Ṡ3 = ẋ3b − ż2 (t) + c3 e3
⇒ Ṡ3 = −n(x3b (t) − Ib ) + u(t) − ż2 (t) + c3 e3
The switching control for keeping the dynamics on the proposed sliding surface is given by:
Parameter Value
c1 0.012
c2 0.6
c3 0.05
k1 0.0007
k2 1.6
Figure 4.1 shows regulation of blood glucose level using BISMC based controller. It can be seen
that reaching time is improved and basal level is achieved in about 360 minutes.
34
Chapter 4 Backstepping Integral Sliding Mode Control........
Figure 4.2 explains concentration of insulin in blood plasma. Concentration of insulin in blood
plasma is stabilized within about 350 minutes using BISMC based controller. It can be seen that
reaching time is improved, also no chattering is there.
Figure 4.3 shows the infusion of insulin by pump into patient. The control input is smooth
having no chattering like classical SMC algorithm.
35
Chapter 4 Backstepping Integral Sliding Mode Control........
Figure 4.4 depicts the performance of BISMC algorithm when the subject is exposed to meal dis-
turbance, which enters the system at time 400 min. It can be seen that the level of glucose is forced
back to basal value successfully.
Figure 4.5 shows the concentration of insulin in blood. It can be noted that concentration of
insulin is smooth having no chattering anymore. The reaching time is also reduced.
36
Chapter 4 Backstepping Integral Sliding Mode Control........
Figure 4.6 gives the dynamics of control input generated by BISMC based controller. It is
obvious that there is no chattering in control input, the infusion of insulin to patient is smooth.
Figure 4.7 shows the regulation of glucose in the presence of three meal intakes using BISMC based
controller. Finally the blood glucose concentration is stabilized at its basal value.
37
Chapter 4 Backstepping Integral Sliding Mode Control........
Figure 4.8 shows concentration of insulin in blood compartment. The three peaks appearing in
graph are due to the disturbance added at these instances. The behavior of insulin concentration is
smooth having no fluctuations of high frequency.
Figure 4.9 shows the control input, infusion of insulin by pump into patient. The control input
is smooth. Level of insulin rapidly increases as meal disturbance appears, as a result blood glucose
level is stabilized.
38
Chapter 4 Backstepping Integral Sliding Mode Control........
Figure 4.10 shows the blood glucose regulation of diabetes patient during hyperglycemia using SMC,
STSMC and BISMC based controllers. The basal value is taken as 80 mg/dL. It is clear from the
figure that reaching time of BISMC based controller is less then other two controllers.In case of
BISMC algorithm the blood glucose level is stabilized at about 350 minutes, which is reasonable
reaching time.
39
Chapter 4 Backstepping Integral Sliding Mode Control........
Figure 4.11 depicts the concentration of insulin in blood of patient during hyperglycemia using
SMC, STSMC and BISMC algorithms. SMC based controller shows chattering which is not desir-
able. However, using BISMC algorithm the concentration of insulin varies smoothly, reaching to
basal value.
Figure 4.12 shows the insulin infusion rate for three controllers. Smooth and accurate behavior
of BISMC based controller is obvious.
40
Chapter 4 Backstepping Integral Sliding Mode Control........
Figure 4.13 shows regulation of blood glucose level in the presence of meal disturbance using three
different controllers.The performance of BISMC based controller is clear in the figure.
Figure 4.14 is a picture of insulin concentration in blood plasma. The peaks appearing in BISMC
based controller are due to its quick response and low reaching time.
41
Chapter 4 Backstepping Integral Sliding Mode Control........
Figure 4.15 shows the control input during hyperglycemia and meal disturbance. Initially more
amount of insulin is injected using BISMC algorithm because of its fast transient response.
Figure 4.16 shows stabilization of glucose level in the presence of three meal intakes by the pa-
tient. The performance of BISMC based controller is clear that its response is fast and accurate as
compared to SMC and STSMC based controllers.
42
Chapter 4 Backstepping Integral Sliding Mode Control........
Figure 4.17 shows the dynamics of insulin in blood plasma during three meal intakes by the
patient. BISMC based controller has no chattering and is maintaining basal level successfully.
Figure 4.18 shows the insulin injection rate of pump in the presence of three meal intakes by the
patient.
43
Chapter 4 Backstepping Integral Sliding Mode Control........
44
Chapter 5
5.1 Conclusion
Robust control of blood glucose concentration in type 1 diabetes patients has been an active area
of research for many decades. This thesis is an attempt to design a robust nonlinear controller for
automatic insulin pumps, in order to regulate blood glucose level in type 1 diabetes patients. Sliding
Mode Control (SMC) based three algorithms were developed and investigated. Traditional SMC
algorithm was designed for blood glucose regulation in type 1 diabetes patients. Bergman’s minimal
model was used for controller design. It has been observed that CSMC show chattering (fluctuations
of high frequency) appearing in control action. STSMC was introduced and designed for getting
rid of chattering phenomenon in control action. STSMC showed smooth control action but tran-
sient response was still comparatively slow. For obtaining suitable performance BISMC algorithm
was introduced and designed. BISMC controller showed best results and can be used for blood
glucose regulation of diabetes patients in artificial pancreas. Main disadvantages of previous con-
trollers include risk of hypoglycaemia, robustness, chattering in classical SMC and more control
effort required. A novel BISMC algorithm has been developed for tackling the challenges regard-
ing regulation of blood glucose concentration in type 1 diabetes patients. Proposed controller was
analyzed for three cases. Firstly, hyperglycemia i.e when blood glucose level of patient is danger-
ously high. Secondly, hyperglycemia and meal disturbance. Finally, case of three meal intakes was
investigated. It has been showed that proposed controller show satisfactory results in stabilizing
blood glucose level by reasonable injection rate of insulin to diabetic patient in question. Designed
controller can be a suitable choice for closed loop automatic insulin pumps.
45
Chapter 5 Conclusion and Future Work
• Using advance optimization techniques for tuning of BISMC controller can result in more optimal
control performance
• Using detailed physiological model for deriving controller may result in more accurate perfor-
mance.
• Using adaptive laws for parameters estimation can further increase robustness of the controller.
46
Bibliography
[1] World Health Organization. "Global report on diabetes WHO Geneva." (2016).
[2] Mathers, Colin D., and Dejan Loncar. "Projections of global mortality and burden of disease
from 2002 to 2030." Plos med 3.11 (2006): e442.
[3] IDF Atlas: About 415 Million Adults Worldwide Have Diabetes. Medscape. Dec 02, 2015.
[4] American Diabetes Association. "Diagnosis and classification of diabetes mellitus." Diabetes
care 33.Supplement 1 (2010): S62-S69.
[5] American Diabetes Association. "Summary of revisions for the 2010 clinical practice recom-
mendations." (2010): S3-S3.
[7] Gomes, Marilia de Brito. "Impact of diabetes on cardiovascular disease: an update," Interna-
tional journal of hypertension (2013).
[8] Chee, F., Fernando, T. L., Savkin, A. V., & Van Heeden, V. (2003). "Expert PID control system
for blood glucose control in critically ill patients," IEEE Transactions on Information Technol-
ogy in Biomedicine,7(4), 419-425.
[9] Ramprasad, Y., Rangaiah, G. P., & Lakshminarayanan, S. (2004). "Robust PID controller for
blood glucose regulation in type I diabetics," Industrial & engineering chemistry research,
43(26), 8257-8268.
[10] Geramipour, A., Khazaei, M., Marjaninejad, A., & Khazaei, M. (2013). "Design of FPGA-
based digital PID controller using Xilinx SysGen
R for regulating blood glucose level of type-I
[11] Slavov, T., & Roeva, O. (2012)." Application of genetic algorithm to tuning a PID controller
for glucose concentration control," Electronic journal: WSEAS Transactions on Systems,11(7),
223-233
47
Chapter 5 Conclusion and Future Work
[12] Marchetti, G., Barolo, M., Jovanovic, L., Zisser, H., & Seborg, D. E. (2008). "An improved
PID switching control strategy for type 1 diabetes," IEEE Transactions on Biomedical Engi-
neering,55(3), 857-865
[13] Marchetti, G., Barolo, M., Jovanovic, L., Zisser, H., & Seborg, D. E. (2008). "A feedforward-
feedback glucose control strategy for type 1 diabetes mellitus",Journal of process con-
trol,18(2), 149-162.
[14] Salzsieder, E., Albrecht, G., Fischer, U., & Freyse, E. J. (1985)."Kinetic modeling of the glu-
coregulatory system to improve insulin therapy",IEEE Transactions on biomedical engineer-
ing,(10), 846-855.
[15] Magni, L., Raimondo, D. M., Bossi, L., Dalla Man, C., De Nicolao, G., Kovatchev, B., &
Cobelli, C. (2007). "Model predictive control of type 1 diabetes: an in silico trial".
[16] Parker, R. S., Doyle, F. J., Ward, J. H., & Peppas, N. A. (2000)."Robust H∞ glucose control in
diabetes using a physiological model",AIChE Journal,46(12), 2537-2549.
[17] Hovorka, R., Canonico, V., Chassin, L. J., Haueter, U., Massi-Benedetti, M., Federici, M.
O., & Wilinska, M. E. (2004)."Nonlinear model predictive control of glucose concentration in
subjects with type 1 diabetes",Physiological measurement,25(4), 905.
[18] Ting, C. W., & Quek, C. (2009)."A novel blood glucose regulation using TSK 0 -FCMAC: A
fuzzy CMAC based on the zero-ordered TSK fuzzy inference scheme",IEEE Transactions on
Neural Networks,20(5), 856-871.
[19] Kaveh, P., & Shtessel, Y. B. (2008)."Blood glucose regulation using higher order sliding mode
control",International Journal of Robust and Nonlinear,18(4?5), 557-569.
[20] Dinani, S. T., Zekri, M., & Kamali, M. (2015)."Regulation of Blood Glucose Concentration
in Type 1 Diabetics Using Single Order Sliding Mode Control Combined with Fuzzy On-line
Tunable Gain, a Simulation Study",Journal of medical signals and sensors,5(3), 131.
[21] Allam, F., Nossair, Z., Gomma, H., Ibrahim, I., & Abdelsalam, M. (2012)."Evaluation of using
a recurrent neural network (RNN) and a fuzzy logic controller (FLC) in closed loop system to
regulate blood glucose for type-1 diabetic patients",International Journal of Intelligent Systems
and Applications,4(10), 58.
[22] Parsa, N. T., Vali, A. R., & Ghasemi, R. (2014)."Back Stepping Sliding Mode Control of
Blood Glucose for Type I Diabetes",World Academy of Science, Engineering and Technology,
International Journal of Medical, Health, Biomedical, Bioengineering and Pharmaceutical
Engineering,8(11), 779-783.
48
Chapter 5 Conclusion and Future Work
[23] Goharimanesh, M., Lashkaripour, A., & Abouei Mehrizi, A. (2015)."Fractional order PID
controller for diabetes patients",Journal of Computational Applied Mechanics,46(1), 69-76.
[24] Heydarinejad H., Delavari H. (2017)."Adaptive Fractional Order Sliding Mode Controller De-
sign for Blood Glucose Regulation-4-3. In: Babiarz A., Czornik A., Klamka J., Niezabitowski
M. (eds)",Theory and Applications of Non-integer Order Systems,Lecture Notes in Electrical
Engineering, vol 407. Springer, Cham.
[26] van Heusden, K., Dassau, E., Zisser, H. C., Seborg, D. E., & Doyle III, F. J. (2012)."Control-
relevant models for glucose control using a priori patient characteristics",IEEE transactions on
biomedical engineering,59(7), 1839-1849.
[28] Ackerman, E., Gatewood, L. C., Rosevear, J. W., & Molnar, G. D. (1965)."Model studies of
blood-glucose regulation",The bulletin of mathematical biophysics,27(1), 21-37.
[29] Sorensen JT. A Physiologic Model of Glucose Metabolism in Man and its Use to Design and
Assess Improved Insulin Therapies for Diabetes, Ph.D. Dissertation, Massachusetts Institute
of Technology. Dept of Chemical Engineering; 1985.
[30] Hovorka, R., Canonico, V., Chassin, L. J., Haueter, U., Massi-Benedetti, M., Federici, M.
O., & Wilinska, M. E. (2004)."Nonlinear model predictive control of glucose concentration in
subjects with type 1 diabetes",Physiological measurement,25(4), 905.
[31] Dalla Man, C., Rizza, R. A., & Cobelli, C. (2007)."Meal simulation model of the glucose-
insulin system",IEEE transactions on biomedical engineering,54(10), 1740-1749.
[32] Cobelli, C., & Mari, A. (1983)."Validation of mathematical models of complex endocrine-
metabolic systems. A case study on a model of glucose regulation",Medical and Biological
Engineering and Computing,21(4), 390-399.
[33] Lehmann, E. D., & Deutsch, T. (1992)."A physiological model of glucose-insulin interaction
in type 1 diabetes mellitus",Journal of biomedical engineering,14(3), 235-242.
[34] Bergman, R. N., Phillips, L. S., & Cobelli, C. (1981)."Physiologic evaluation of factors con-
trolling glucose tolerance in man: measurement of insulin sensitivity and beta-cell glucose
sensitivity from the response to intravenous glucose",Journal of clinical investigation,68(6),
1456.
49
Chapter 5 Conclusion and Future Work
[35] Steil, G. M., Clark, B., Kanderian, S., & Rebrin, K. (2005)."Modeling insulin action for devel-
opment of a closed-loop artificial pancreas",Diabetes technology & therapeutics,7(1), 94-108.
[36] De Gaetano, A., & Arino, O. (2000)."Mathematical modelling of the intravenous glucose tol-
erance test",Journal of mathematical biology,40(2), 136-168.
[37] Utkin, V. (1977)."Variable structure systems with sliding modes", IEEE Transactions on Auto-
matic control,22(2), 212-222.
[38] Chen, H., Pallapa, M., Sun, W. J., Sun, Z. D., & Yeow, J. T. W. (2014)."Nonlinear control of
an electromagnetic polymer MEMS hard-magnetic micromirror and its imaging application",
Journal of Micromechanics and Microengineering,24(4), 045004.
[39] Kokotovic, P. V. (1992)."The joy of feedback: nonlinear and adaptive",IEEE Control sys-
tems,12(3), 7-17.
[40] Lozano, R., & Brogliato, B. (1992)."Adaptive control of robot manipulators with flexible
joints",IEEE Transactions on Automatic Control,37(2), 174-181.
[41] Khalil, H.K. (2002). Nonlinear Systems (3rd ed.). Upper Saddle River, NJ: Prentice Hall. ISBN
0-13-067389-7.
50