
Citation: Dominic Bowd, Campbell McKay, Wendy S. Shaw. Urban greening: environmentalism or marketable aesthetics[J]. AIMS Environmental Science, 2015, 2(4): 935-949. doi: 10.3934/environsci.2015.4.935
[1] | Lela Dorel . Glucose level regulation via integral high-order sliding modes. Mathematical Biosciences and Engineering, 2011, 8(2): 549-560. doi: 10.3934/mbe.2011.8.549 |
[2] | Danilo T. Pérez-Rivera, Verónica L. Torres-Torres, Abraham E. Torres-Colón, Mayteé Cruz-Aponte . Development of a computational model of glucose toxicity in the progression of diabetes mellitus. Mathematical Biosciences and Engineering, 2016, 13(5): 1043-1058. doi: 10.3934/mbe.2016029 |
[3] | Lingmin Lin, Kailai Liu, Huan Feng, Jing Li, Hengle Chen, Tao Zhang, Boyun Xue, Jiarui Si . Glucose trajectory prediction by deep learning for personal home care of type 2 diabetes mellitus: modelling and applying. Mathematical Biosciences and Engineering, 2022, 19(10): 10096-10107. doi: 10.3934/mbe.2022472 |
[4] | Xingjia Li, Jinan Gu, Zedong Huang, Chen Ji, Shixi Tang . Hierarchical multiloop MPC scheme for robot manipulators with nonlinear disturbance observer. Mathematical Biosciences and Engineering, 2022, 19(12): 12601-12616. doi: 10.3934/mbe.2022588 |
[5] | Xiaohan Yang, Yinghao Cui, Zhanhang Yuan, Jie Hang . RISE-based adaptive control of electro-hydraulic servo system with uncertain compensation. Mathematical Biosciences and Engineering, 2023, 20(5): 9288-9304. doi: 10.3934/mbe.2023407 |
[6] | Kimberly Fessel, Jeffrey B. Gaither, Julie K. Bower, Trudy Gaillard, Kwame Osei, Grzegorz A. Rempała . Mathematical analysis of a model for glucose regulation. Mathematical Biosciences and Engineering, 2016, 13(1): 83-99. doi: 10.3934/mbe.2016.13.83 |
[7] | Balázs Csutak, Gábor Szederkényi . Robust control and data reconstruction for nonlinear epidemiological models using feedback linearization and state estimation. Mathematical Biosciences and Engineering, 2025, 22(1): 109-137. doi: 10.3934/mbe.2025006 |
[8] | Hugo Flores-Arguedas, Marcos A. Capistrán . Bayesian analysis of Glucose dynamics during the Oral Glucose Tolerance Test (OGTT). Mathematical Biosciences and Engineering, 2021, 18(4): 4628-4647. doi: 10.3934/mbe.2021235 |
[9] | Huazhong Yang, Wang Li, Maojin Tian, Yangfeng Ren . A personalized multitasking framework for real-time prediction of blood glucose levels in type 1 diabetes patients. Mathematical Biosciences and Engineering, 2024, 21(2): 2515-2541. doi: 10.3934/mbe.2024111 |
[10] | Xingjia Li, Jinan Gu, Zedong Huang, Wenbo Wang, Jing Li . Optimal design of model predictive controller based on transient search optimization applied to robotic manipulators. Mathematical Biosciences and Engineering, 2022, 19(9): 9371-9387. doi: 10.3934/mbe.2022436 |
Blood glucose needs to be maintained within a narrow range, for example, from 3.6 to 6.0 mmol/l (65 to 108 mg/dl) for the fasting people without diabetes. This narrow range differs in people and in time. A high blood glucose concentration outside the range is a symptom of diabetes, one of the prevalent diseases affecting millions of people worldwide. A peptide hormone insulin is required to interact with glucose to achieve this goal.
The interaction between blood glucose and insulin involves multiple feedback control mechanisms and is of extremely high complexity. Here we give a very basic description about it. Glucose comes from food and liver, and utilized by brain and nerve cells (insulin-independent) via the glucose transporter 3 (GLUT3) or by tissue cells such as muscle, kidney, and fat cells (insulin-dependent) via the glucose transporter 4 (GLUT4). Glucose is transported into and out of liver cells by the concentration-driven GLUT2, which is insulin-independent[17]. In response to a low blood glucose level, the $ \alpha $ cells of the pancreas produce the hormone glucagon. The glucagon initiates a series of activations of kinases, and finally leads to the activation of the glycogen phosphorylase, which catalyzes the breakdown of glycogen into glucose[14]. In addition, the series of activations of kinases also result in the inhibition of glycogen synthase and then stop the conversion of glucose to glycogen. In response to a high blood glucose level, the $ \beta $ cells of the pancreas secrete insulin[15,35,36]. Insulin triggers a series of reactions to activate glycogen synthase, which catalyzes the conversion of glucose into glycogen[16,21]. Insulin also initiates a series of activations of kinases in tissue cells to lead to the redistribution of the glucose transporter 4 (GLUT4) from intracellular storage sites to the plasma membrane[13]. Once at the cell surface[8,32], GLUT4 transports glucose into the muscle or fat cells.
Mathematical models for the blood glucose regulation system are required for integrating a glucose monitoring system into insulin pump technology to form a closed-loop insulin delivery system, the so-called "artificial pancreas" (see Hovorka[18], Panteleon et al.[33], Steil et al.[38]). To make this artificial pancreas as close as possible to the natural pancreas, many mathematical models about the interaction between blood glucose and insulin have been proposed. In many of these models, the compartmental modeling approach has been used. In the 1960's, Ackerman et al.[1,2] proposed a simplified compartmental model of two linear ordinary differential equations for the glucose-tolerance test. This model lumped the large number of kinetic parameters into a much smaller number which can, at least in part, characterize the human glucose regulatory system. In the 1970's, Bergman et al.[4,5,6,7] proposed a minimal compartmental model of nonlinear ordinary differential equations. It was assumed that the blood glucose and hormone insulin are contained in two different compartments and interact with each other. The plasma insulin is secreted from the endocrine system under the stimulation of the blood glucose and enters a "remote compartment", where it is active in accelerating glucose utilization. Then the dynamics of the blood glucose, plasma insulin, and the insulin in the remote compartment was modeled by a system of three nonlinear ordinary differential equations. In the 1990's, to determine whether the oscillations of insulin and blood glucose could result from the feedback loops between insulin and glucose, Sturis et al.[39,40] developed a parsimonious model including the major mechanisms involved in glucose regulation. This model comprises two major negative feedback loops describing the effects of insulin on glucose utilization and glucose production, respectively, and both loops include the stimulatory effect of glucose on insulin secretion. Then this model consists of six nonlinear ordinary differential equations, three of which are used to model a time delay for the effect of insulin on glucose production and a sluggish effect of insulin on glucose utilization. On the basis of Sturis et al.'s model, Li et al.[23] introduced two explicit time delays and proposed a more robust alternative model for better understanding the glucose-insulin endocrine metabolic regulatory system and the ultradian insulin secretory oscillations for the cases of continuous enteral nutrition and constant glucose infusion. More complicated compartmental models proposed by Hovorka et al.[18], Liu et al.[24,25,27], Man et al.[29,30,31], and Sorensen[37] have considered labeling more compartments for better understanding the behavior of different parts of the body. To assess the liver ability to extract insulin, Toffolo et al.[41,42], proposed a minimal model of insulin delivery and kinetics by combining the classical C-peptide minimal model. In the 2010's, Vahidi et al.[45,46] developed a model for type Ⅱ diabetes, based on a detailed compartmental physiological model proposed by Sorensen[37] for a healthy human body. There have been a huge set of references on the modeling research and it is hopeless to mention all of them.
In all these models, no parameter uncertainties were introduced. The parameter uncertainties here means that a parameter may vary in a neighborhood of its nominal value. This can happen in practice. For instance, an estimation error may be introduced when a model parameter is estimated. In addition, parameter values may vary slightly from person to person. Therefore, it is of practical interest to introduce parameter uncertainties into a model. As far as the robust tracking is concerned, the robustness against the parameter uncertainties means that a feedback and feedforward controller should be designed in a way such that it can tolerate the parameter variation in a small neighborhood of their nominal values. That is, even though the parameters have a small change, the controller can still achieve its tracking goal. Such a robust problem may be solved by only a robust control design method and other methods such as the sensitivity analysis, bootstrapping to quantify parameter ranges, and stochastic modeling might not be relevant.
As a first attempt in this direction, we introduce parameter uncertainties into the model proposed by Bergman et al.[4,5]. Although this model is simple and consists of only three differential equations, one equation for the glucose dynamics, one equation for the effect of the remote insulin on glucose, and the other equation for the hormone insulin dynamics, it can simulate well the major biological dynamics about the interaction between glucose and insulin, as evidenced in the numerous literature (see. e.g., [4,5,6,7,11]). It has been widely used and further updated later (see, e.g., [10,18,20,23,24,39,40,41]).
Furthermore, we also include a time-dependent exogenous glucose input $ J $ from food through the intestine in this model and assume that the desired glucose reference $ r $ to be tracked is time-dependent, as described in Figure 1. Then we design a robust feedback (using an actual glucose concentration $ g $) and feedforward (using the exogenous glucose input $ J $ and the desired glucose reference $ r $) controller $ u(g, r, J) $ such that the glucose concentration asymptotically tracks the time-varying glucose reference for any small uncertainties. Numerical simulations conducted with published experimental blood glucose data show that our controller is effective in achieving the blood glucose tracking and agree with our theoretical results.
In the 1970's, Bergman et al.[4,5] proposed a mathematical model for the blood glucose regulation system. To establish such a model, they made a simplified assumption. The blood glucose is produced from food through the intestine. The blood glucose and hormone insulin are contained in two different compartments and interact with each other. The blood glucose can be converted into the liver glycogen and used by the tissue metabolism with or without the help of insulin. The plasma insulin is secreted from the endocrine system under the stimulation of the blood glucose and enters a "remote compartment", where it is active in accelerating glucose utilization. It is used up by the insulin metabolism. Under this assumption, the dynamics of the blood glucose and insulin was modeled by the following system of nonlinear differential equations [5]:
$ dgdt=−m1g−hag+J, $
|
(2.1) |
$ dhadt=−m2ha+m3h, $
|
(2.2) |
$ dhdt=−m4h+u. $
|
(2.3) |
In the above equations, $ g $ and $ h $ denote concentrations of blood glucose and plasma insulin, respectively, $ h_a $ is the effect of the remote insulin on glucose, $ J $ is a rate of the exogenous glucose input from the intestine, $ u $ is a rate of insulin secreted from the endocrine system or infused externally, and $ m_1, m_2, m_3, m_4 $ are positive rate constants. The biological meaning of these rate constants are explained in the Table 1. Because the glucose needs to be monitored and regulated, we introduce an output equation:
Parameter | Value | Description |
$ m_1 $ | 0.000296 (/min) | Glucose utilization rate constant |
$ m_2 $ | 0.045 (/min) | Insulin effect rate constant |
$ m_3 $ | 0.000012 (/(pmol/l)/min$ ^2 $) | Insulin effect rate constant |
$ m_4 $ | 0.268 (/min) | Insulin metabolism constant |
$ G_b $ | 0.3218 (mM/min) | Basal glucose input |
$ \bar{g} $ | 6 (mM) | Glucose equilibrium |
$ \bar{h} $ | 200 (pmol/l) | Insulin equilibrium |
$ \bar{h}_a $ | 0.0533 (/min) | Insulin effect equilibrium |
$ \bar{u} $ | 53.63 (pmol/l/min) | Insulin input equilibrium |
$ y=g. $
|
(2.4) |
To maintain the blood glucose at a normal level of about 6 mmol/l, a basal exogenous glucose input $ G_b $ is required[24,47]. Thus the blood glucose input $ J $ could be split into two parts:
$ J=Gb+Gd, $
|
(2.5) |
where $ G_d $ can be treated as a glucose disturbance.
In what follows, $ (\bar{g}, \bar{h}_a, \bar{h}, \bar{u}) $ denotes the equilibrium of the system (2.1)-(2.3) without the glucose disturbance. That is, they are solutions of the following steady state system:
$ 0=−m1ˉg−ˉgˉha+Gb, $
|
(2.6) |
$ 0=−m2ˉha+m3ˉh, $
|
(2.7) |
$ 0=−m4ˉh+ˉu. $
|
(2.8) |
In practice, an estimation error may be introduced when a model parameter is estimated. In addition, the values of the parameters $ m_1, m_2, m_3, m_4 $ may vary slightly from person to person. Therefore, it is reasonable to introduce parameter uncertainties into the system (2.1)-(2.3):
$ dgdt=−(m1+w1)g−gha+w1ˉg+Gb+Gd, $
|
(2.9) |
$ dhadt=−(m2+w2)ha+(m3+w3)h+w2ˉha−w3ˉh, $
|
(2.10) |
$ dhdt=−(m4+w4)h+w4ˉh+u. $
|
(2.11) |
In the above equations, $ m_1, m_2, m_3 $ and $ m_4 $ represent the nominal part of the parameters and $ w_1, w_2, w_3, w_4 $ represent the uncertain part. Since the equilibrium $ (\bar{g}, \bar{h}_a, \bar{h}, \bar{u}) $ would not depend on the uncertainties, the terms $ w_1 \bar{g}, w_2 \bar{h}_a, w_3 \bar{h} $ and $ w_4\bar{h} $ are included such that this is guaranteed.
Because the blood glucose needs to be maintained within a narrow range from 3.6 to 6.0 mmol/l (65 to 108 mg/dl), we need to develop an insulin infusion rate $ u $ in response to a glucose surge to achieve this goal. This problem can be mathematically formulated as follows. Given a glucose reference $ r(t) $ (a constant 6 mmol/l or fluctuating around 6 mmol/l), we want to design a controller $ u = u(g, J, r) $ such that for all sufficiently small uncertainties $ w_1, w_2, w_3, w_4 $
$ \lim\limits_{t\to \infty} (g(t)- r(t)) = 0. $ |
To solve this control problem, we need to express $ G_d $ and $ r $ by using solutions of linear differential systems. The glucose produced from food via the intestine is not exactly known and might be random. This glucose disturbance could be of periodic nature and then can be approximated by a Fourier polynomial
$ Gd=a0+a1sin(ωdt)+b1cos(ωdt)+a2sin(2ωdt)+b2cos(2ωdt)+⋯+ansin(nωdt)+bncos(nωdt). $
|
(3.1) |
The number $ 1 $ and functions cosine's and sine's can be generated by the linear differential system
$ \frac{d\mathbf{v}_d}{dt} = \mathbf{A}_{d} \mathbf{v}_d, $ |
where
$ \mathbf{A}_{d} = \left[00000⋯0000ωd00⋯000−ωd000⋯0000002ωd⋯00000−2ωd0⋯00⋮⋮⋮⋮⋮⋱⋮⋮00000⋯0nωd00000⋯−nωd0 \right]. $
|
Then $ G_d $ can be expressed as
$ G_d = \mathbf{C}_d\mathbf{v}_d, $ |
where $ \mathbf{C}_d = [a_0, a_1, b_1, a_2, b_2, \cdots, a_n, b_n] $ is a constant vector.
Since the glucose reference $ r $ is of periodic nature, too, it can be generated in the same way as follows:
$ r = \bar{g}+\mathbf{C}_r\mathbf{v}_r, $ |
where $ \bar{g} $ is the glucose equilibrium, $ \mathbf{C}_r $ is a constant vector and $ \mathbf{v}_r $ is governed by
$ \frac{d\mathbf{v}_r}{dt} = \mathbf{A}_{r} \mathbf{v}_r $ |
with $ \mathbf{A}_{r} $ being similar to $ \mathbf{A}_{d} $:
$ \mathbf{A}_{r} = \left[00000⋯0000ωr00⋯000−ωr000⋯0000002ωr⋯00000−2ωr0⋯00⋮⋮⋮⋮⋮⋱⋮⋮00000⋯0mωr00000⋯−mωr0 \right]. $
|
Let
$ \mathbf{v} = \left[vrvd \right],\quad \mathbf{A}_e = \left[Ar00Ad \right]. $
|
Then glucose reference and disturbance can be lumped together as follows:
$ dvdt=Aev. $
|
(3.2) |
We introduce the tracking error:
$ e = g-r = g-\bar{g}-\mathbf{C}_r\mathbf{v}_r. $ |
Combining the exosystem (3.2) with the model (2.9)-(2.11), we obtain a composite system
$ dgdt=−(m1+w1)g−gha+w1ˉg+Gb+[0,Cd]v, $
|
(3.3) |
$ dhadt=−(m2+w2)ha+(m3+w3)h+w2ˉha−w3ˉh, $
|
(3.4) |
$ dhdt=−(m4+w4)h+w4ˉh+u, $
|
(3.5) |
$ dvdt=Aev, $
|
(3.6) |
$ e=g−ˉg−[Cr,0]v. $
|
(3.7) |
When the disturbance $ G_d = 0 $ and $ r = \bar{g} $, we can readily show that $ g, h_a, h $ and $ u $ converge to their constant equilibrium, respectively. When $ G_d $ and $ r $ are time-dependent, we can expect that $ g, h_a, h $ and $ u $ would converge to their time-varying "equilibrium" that would depend on $ \mathbf{v} $. In fact, such time-varying "equilibrium" called a stable center manifold is guaranteed by the center manifold theory (see, e.g., [9]). Thus we introduce the following $ \mathbf{v} $-dependent variable transform:
$ g=ˆg+ˉg+G(v),ha=ˆha+ˉha+Ha(v),h=ˆh+ˉh+H(v),u=ˆu+ˉu+U(v) $
|
(3.8) |
to translate the time-varying "equilibrium" to the origin. Substituting the above transformation into the equations (3.3)-(3.7), we obtain
$ dˆgdt+∂G∂vAev=−a1(w)(ˆg+G(v))−ˉg(ˆha+Ha(v))−[ˆgˆh+G(v)ˆha+ˆgHa(v)]−G(v)Ha(v)+[0,Cd]v, $
|
(3.9) |
$ dˆhadt+∂Ha∂vAev=−a2(w)(ˆha+Ha(v))+a3(w)(ˆh+H(v)), $
|
(3.10) |
$ dˆhdt+∂H∂vAev=−a4(w)(ˆh+H(v))+ˆu+U(v), $
|
(3.11) |
$ e=ˆg+G(v)−[Cr,0]v, $
|
(3.12) |
where $ \mathbf{w} = [w_1, w_2, w_3, w_4] $ and
$ a1(w)=m1+w1+ˉha, $
|
(3.13) |
$ a2(w)=m2+w2, $
|
(3.14) |
$ a3(w)=m3+w3, $
|
(3.15) |
$ a4(w)=m4+w4. $
|
(3.16) |
In order to eliminate the terms that do not contain either $ \hat{g} $ or $ \hat{h} $, we set
$ ∂G∂vAev=−a1(w)G−ˉgHa−GHa+[0,Cd]v, $
|
(3.17) |
$ ∂Ha∂vAev=−a2(w)Ha+a3(w)H, $
|
(3.18) |
$ ∂H∂vAev=−a4(w)H+U, $
|
(3.19) |
$ 0=G−[Cr,0]v. $
|
(3.20) |
Notice that the equations (3.17)-(3.19) are the result of the center manifold theory (see[9]). In this way, the tracking problem (3.3)-(3.7) is converted to the following stabilization problem:
$ dˆgdt=−a1(w)ˆg−ˉgˆha−(ˆgˆh+G(v)ˆha+ˆgHa(v)), $
|
(3.21) |
$ dˆhadt=−a2(w)ˆha+a3(w)ˆh, $
|
(3.22) |
$ dˆhdt=−a4(w)ˆh+ˆu, $
|
(3.23) |
$ e=ˆg. $
|
(3.24) |
The regulator equations (3.17)-(3.20) can be solved explicitly as follows:
$ G(v)=[Cr,0]v, $
|
(3.25) |
$ F(v)=[0,Cd]v−a1(w)[Cr,0]v−[Cr,0]Aev, $
|
(3.26) |
$ Ha(v)=F(v)ˉg+[Cr,0]v, $
|
(3.27) |
$ H(v)=∂Ha∂vAev+a2(w)Haa3(w) $
|
(3.28) |
$ =(ˉg+[Cr,0]v)∂F∂vAev−F(v)[Cr,0]Aeva3(w)[ˉg+[Cr,0]v]2+a2(w)F(v)a3(w)(ˉg+[Cr,0]v), $
|
(3.29) |
$ U(v)=∂H∂vAev+a4(w)H=([Cr,0]Aev)∂F∂vAev+(ˉg+[Cr,0]v)∂F∂vA2ev−∂F∂vAev[Cr,0]Aev−F(v)[Cr,0]A2eva3(w)[ˉg+[Cr,0]v]2−2[(ˉg+[Cr,0]v)∂F∂vAev−F(v)[Cr,0]Aev][ˉg+[Cr,0]v]a3(w)[Cr,0]Aeva23(w)[ˉg+[Cr,0]v]4+a2(w)∂F∂vAeva3(w)(ˉg+[Cr,0]v)−a2(w)F(v)a3(w)[Cr,0]Aev)a23(w)(ˉg+[Cr,0]v)2+a4(w)(ˉg+[Cr,0]v)∂F∂vAev−F(v)[Cr,0]Aeva3(w)[ˉg+[Cr,0]v]2+a2(w)a4(w)F(v)a3(w)(ˉg+[Cr,0]v). $
|
(3.30) |
It is easy to see that the equilibrium $ (0, 0, 0) $ of the system (3.21)-(3.24) at $ \mathbf{v} = \mathbf{0} $ can be stabilized exponentially and locally by a static output feedback
$ \hat{u} = k_g\hat{g}. $ |
Therefore, we have constructed a static output feedback
$ u=ˉu+ˆu+U(v)=ˉu+kgˆg+U(v)=ˉu+kg(g−ˉg−[Cr,0]v))+U(v) $
|
(3.31) |
to regulate the blood glucose to its desired reference $ \bar{g}+ [\mathbf{C}_r, \; \mathbf{0}]\mathbf{v} $. Incorporating this feedback controller into the system (3.3)-(3.7), we obtain a closed-loop system:
$ dgdt=−(m1+w1)g−gha+w1ˉg+Gb+[0,Cd]v, $
|
(3.32) |
$ dhadt=−(m2+w2)ha+(m3+w3)h+w2ˉha−w3ˉh, $
|
(3.33) |
$ dhdt=−(m4+w4)h+w4ˉh+ˉu+kg(g−ˉg−[Cr,0]v))+U(v), $
|
(3.34) |
$ dvdt=Aev, $
|
(3.35) |
$ e=g−ˉg−[Cr,0]v. $
|
(3.36) |
We summarize this result in the following theorem.
Theorem 3.1. Suppose that $ k_g $ is a constant to exponentially stabilize the equilibrium $ (0, 0, 0) $ of the system (3.21)-(3.24). Then the closed-loop nonlinear system (3.32)-(3.36) asymptotically tracks the blood glucose reference $ \bar{g} +[\mathbf{C}_r, \; \mathbf{0}] \mathbf{v}(t) $, that is,
$ \lim\limits_{t\to \infty} [g(t)-\bar{g}-[\mathbf{C}_r, \; \mathbf{0}] \mathbf{v}(t)] = 0. $ |
In numerical simulations below, we use the software Matlab and the experimental data from Sturis et al.'s [39]. We use the software Engauge Digitizer 4.1 to read the data.
To conduct a numerical simulation, we use the experimental data from Sturis et al.'s [39] to determine $ \mathbf{C}_d $, $ \mathbf{C}_r $, and $ \mathbf{A}_e $. These data were obtained from normal men without a history of diabetes and the detailed experimental methods and procedures are referred to[39]. Since the data contain 24 hours profiles of glucose and plasma insulin, they are appropriate for testing whether the closed-loop nonlinear system (3.32)-(3.36) asymptotically tracks the blood glucose reference. The data were read from [39] by using the software Engauge Digitizer 4.1. According to the second panel of Figure 2 of Sturis et al.'s [39], the exogenous glucose infusion shown in Figure 2 is given by
$ J = 125[1+0.33\cdot\sin(\pi t/72 )]\; (\mbox{cc/hr}) . $ |
Because the glucose concentration unit is mM, we need to covert the unit of cc per hour of the infused solution to the unit of mM per minute of the infusion rate inside the body. According to Sturis et al.'s paper[39], the infused glucose solution is 20% solution. This could mean that one cc solution contains 0.2 grams of glucose. The blood volume of a person with the weight of 70 kg is about 6 liter. Using the conversion 1 mM = 18 mg/dl, we obtain
$ J = \frac{125\cdot 0.2\cdot 1000}{60\cdot 18\cdot 60} [1+0.33\cdot\sin(\pi t/72 )] = 0.3858 [1+0.33\cdot\sin(\pi t/72 )]\; (\mbox{mM/min}) . $ |
In this simulation, we find that $ G_b = 0.3218 $ mM/min. Thus we obtain
$ \mathbf{C}_d = [0.3858 - G_b, 0.1273, 0 ], $ |
and
$ \mathbf{A}_{d} = \left[00000π/720−π/720 \right] . $
|
Fitting the experimental glucose data shown in the second panel of Figure 2 of Sturis et al.'s [39] into the Fourier polynomial:
$ g = a_0+\sum\limits_{i = 1}^{17}[a_i\sin(i\pi t/750 )+b_i\cos(i\pi t/750)], $ |
we obtain
$ a0=7.599,a1=−0.2394,b1=−1.024,a2=0.328,b2=−0.2217,a3=−0.2314,b3=−0.1684,a4=−0.1663,b4=−0.05297,a5=−0.1167,b5=0.1937,a6=0.07061,b6=−0.1443,a7=0.3512,b7=−0.08337,a8=0.02829,b8=−0.0705,a9=0.2508,b9=−0.259,a10=0.6841,b10=−0.8653,a11=−0.3087,b11=0.5024,a12=−0.2117,b12=0.2057,a13=−0.07391,b13=0.02245,a14=−0.01738,b14=−0.1047,a15=−0.1031,b15=−0.05237,a16=−0.001116,b16=0.004296,a17=−0.04128,b17=0.003007. $
|
The unit of the coefficients $ a_0 $, $ a_i $ and $ b_i $ is mM, where $ i = 1, 2, \cdots, 17 $. The glucose equilibrium is set to $ \bar{g} = 6 $ mM. Thus we obtain
$ \mathbf{C}_r = [a_0 - \bar{g}, a_1, b_1 , a_2, b_2 , \cdots a_{17}, b_{17} ], $ |
and
$ \mathbf{A}_{r} = \left[00000⋯0000π/75000⋯000−π/750000⋯0000002π/750⋯00000−2π/7500⋯00⋮⋮⋮⋮⋮⋱⋮⋮00000⋯017π/75000000⋯−17π/7500 \right] . $
|
Figure 3 shows that the data can be fitted into this Fourier polynomial very well.
The estimation of the parameters $ m_1, m_2, m_3 $, and $ m_4 $ is challenging. Because the expression of the controller $ u $ is too big to be expressed in Matlab, we failed to use the Matlab function "sbioparamestim" to estimate them. Thus, starting with the values from Bergman et al.'s paper[5]: $ m_1 = 0.0296 $ (/min), $ m_2 = 0.0186 $ (/min), $ m_3 = 0.0000065 $ (/($ \mu $U/ml)/min$ ^2 $), $ m_4 = 0.238 $ (/min), we manually adjust them to achieve as a best result as we can. These estimated values are presented in Table 1. Compared with Bergman et al.'s original values, only $ m_1 $ has a large change. The uncertainties $ w_1, w_2, w_3, $ and $ w_4 $ are taken to be 10% of their nominal values. The proportional constant $ k_g $ is taken to be $ 0.01 $ (/min).
The model is solved numerically with Matlab. We first run the model 1500 minutes to achieve a stable state and then set that time as the initial time. Figure 4 shows that the blood glucose $ g $ is asymptotically tracking the desired time-varying reference determined by the experimental data from Sturis et al.'s [39]. Although the simulated insulin profile does not exactly match with the experimental data, the simulated and experimental pulses occur simultaneously. The dynamics of the controller $ u $ and insulin effect $ h_a $ follows the dynamics of the glucose infusion.
Because the static output feedback (3.31) contains the uncertainties $ \mathbf{w} $, it is not robust. If the uncertainties $ \mathbf{w} $ are set to 0 in the static output feedback controller (3.31), Figure 5 shows that the glucose $ g $ fails to track its reference. In fact, it has been shown that there are no static state feedback controls that can solve the robust tracking problem (see[19]).
We now design a robust dynamical feedback and feedforward controller via linearization. The linearization of the system (3.3)-(3.7) at its equilibrium is given by
$ d˜gdt=−a1(w)˜g−ˉg˜ha+[0,Cd]v, $
|
(4.1) |
$ d˜hadt=−a2(w)˜ha+a3(w)˜h, $
|
(4.2) |
$ d˜hdt=−a4(w)˜h+˜u, $
|
(4.3) |
$ dvdt=Aev, $
|
(4.4) |
$ e=˜g−[Cr,0]v, $
|
(4.5) |
where
$ ˜g=g−ˉg,˜ha=ha−ˉha,˜h=h−ˉh,˜u=u−ˉu. $
|
(4.6) |
To construct a robust control law, we introduce the following dynamic compensator:
$ dzdt=G1z+G2(˜g−[Cr,0]v), $
|
(4.7) |
where the matrix pair $ (\mathbf{G}_1, \mathbf{G}_2) $ is determined below. Since the state variables $ \tilde{h}_a $ and $ \tilde{h} $ are not available for feedback, we also need to use the Luenburger observer:
$ dodt=Ao+B˜u+L(˜g−Co), $
|
(4.8) |
where the vector $ \mathbf{L} $ is to be designed and
$ A=[−a1(0)−ˉg00−a2(0)a3(0)00−a4(0)],B=[001],C=[1,0,0]. $
|
(4.9) |
We then introduce the following dynamic output feedback and feedforward controller:
$ ˜u=−Kzz−Koo, $
|
(4.10) |
where the vectors $ \mathbf{K}_o $ and $ \mathbf{K}_z $ are to be designed.
A minimal polynomial of a matrix $ \mathbf{A} $ is a polynomial $ P(\lambda) $ with the least degree such that $ P(\mathbf{A}) = \mathbf{0} $. If $ \omega_r = \omega_d $, then the minimal polynomial of $ \mathbf{A}_{e} $ is given by
$ p(λ)=λ(λ2+ω2d)(λ2+(2ωd)2)⋯(λ2+(lωd)2)=λ2l+1+c1λ2l+⋯+c2lλ+c2l+1, $
|
where $ l = \max(m, n) $. If $ \omega_r\neq\omega_d $, then the minimal polynomial of $ \mathbf{A}_{e} $ is given by
$ p(λ)=λ(λ2+ω2d)(λ2+(2ωd)2)⋯(λ2+(nωd)2)(λ2+ω2r)(λ2+(2ωr)2)⋯(λ2+(mωr)2)=λ2l+1+c1λ2l+⋯+c2lλ+c2l+1, $
|
where $ l = m+n $. Then $ \mathbf{G}_1 $ is defined to be a companion matrix of the minimal polynomial and $ \mathbf{G}_2 $ is set to be a matrix such that the pair $ (\mathbf{G}_1, \mathbf{G}_2) $ is controllable as follows:
$ G1=[010⋯0001⋯0⋮⋮⋱⋮⋮000⋯1−c2l+1−c2l−c2l−1⋯−c1],G2=[00⋮01]. $
|
(4.11) |
In control theory, such a matrix pair $ (\mathbf{G}_1, \mathbf{G}_2) $ is called to incorporate a 1-copy internal model of the matrix $ \mathbf{A}_{e} $. In the following proof, we can see the reason why the pair $ (\mathbf{G}_1, \mathbf{G}_2) $ is set up this way.
To convert this tracking problem to a stabilization problem, we introduce the coordinate transformation
$ ˜g=ˆg+Gv,˜ha=ˆha+Hav,˜h=ˆh+Hv,z=ˆz+Zv,o=ˆo+Ov. $
|
(4.12) |
Substituting this transformation (4.12) and the feedback (4.10) into the composite system (4.1)-(4.5) and the equations (4.7) and (4.8), we obtain
$ dˆgdt+GAev=−a1(w)(ˆg+Gv)−ˉg(ˆha+Hav)+[0,Cd]v,dˆhadt+HaAev=−a2(w)(ˆha+Hav)+a3(w)(ˆh+Hv),dˆhdt+HAev=−a4(w)(ˆh+Hv)−Kz(ˆz+Zv)−Ko(ˆo+Ov),dˆzdt+ZAev=G1(ˆz+Zv)+G2(ˆg+Gv−[Cr,0]v),dˆodt+OAev=A(ˆo+Ov)−B[Kz(ˆz+Zv)+Ko(ˆo+Ov)]+L(ˆg+Gv−C(ˆo+Ov)),e=ˆg+Gv−[Cr,0]v. $
|
In order to eliminate $ \mathbf{v} $ from the above system, we set
$ GAe=−a1(w)G−ˉgHa+[0,Cd], $
|
(4.13) |
$ HaAe=−a2(w)Ha+a3(w)H, $
|
(4.14) |
$ HAe=−a4(w)H−KzZ−KoO, $
|
(4.15) |
$ ZAe=G1Z+G2(G−[Cr,0]), $
|
(4.16) |
$ OAe=AO−BKzZ−BKoO+L(G−CO), $
|
(4.17) |
$ 0=G−[Cr,0] $
|
(4.18) |
and then
$ dˆgdt=−a1(w)ˆg−ˉgˆha, $
|
(4.19) |
$ dˆhadt=−a2(w)ˆha+a3(w)ˆh, $
|
(4.20) |
$ dˆhdt=−a4(w)ˆh−Kzˆz−Koˆo, $
|
(4.21) |
$ dˆzdt=G1ˆz+G2ˆg, $
|
(4.22) |
$ dˆodt=Aˆo−B[Kzˆz+Koˆo]+L(ˆg−Cˆo), $
|
(4.23) |
$ e=ˆg. $
|
(4.24) |
If the regulator equations (4.13)-(4.18) have a solution, then the tracking problem is converted to the problem of stabilization of the system (4.19)-(4.24).
We first show that the equilibrium $ (0, 0, 0, \mathbf{0}, \mathbf{0}) $ of the system (4.19)-(4.24) can be exponentially stabilized at $ \mathbf{w} = \mathbf{0} $ and then robustly stabilized for sufficiently small $ \mathbf{w} $. For this, we set $ \hat{\mathbf{x}} = [\hat{g}, \hat{h}_a, \hat{h}]^T $ and rewrite the system (4.19)-(4.24) at $ \mathbf{w} = \mathbf{0} $ as follows:
$ dˆxdt=Aˆx−B[Kzˆz+Koˆo], $
|
(4.25) |
$ dˆzdt=G1ˆz+G2Cˆx, $
|
(4.26) |
$ dˆodt=Aˆo−B[Kzˆz+Koˆo]+L(Cˆx−Cˆo). $
|
(1.27) |
Setting $ \hat{\mathbf{q}} = \hat{\mathbf{x}}-\hat{\mathbf{o}} $, we derive from the above system that
$ dˆxdt=(A−BKo)ˆx−B[Kzˆz−Koˆq], $
|
(4.28) |
$ dˆzdt=G1ˆz+G2Cˆx, $
|
(4.29) |
$ dˆqdt=(A−LC)ˆq, $
|
(4.30) |
and then
$ ddt[ˆxˆzˆq]=[A−BKo−BKzBKoG2CG1000A−LC][ˆxˆzˆq]. $
|
(4.31) |
It is easy to see that $ (\mathbf{A}, \mathbf{C}) $ is observable and the pair
$ Φ=[A0G2CG1],Ψ=[B0] $
|
(4.32) |
is stabilizable (see, e.g., [19]). So there exist constant vectors $ \mathbf{K}_o, \mathbf{K}_z, \mathbf{L} $ such that all eigenvalues of the coefficient matrix of the system
$ Ω=[A−BKo−BKzBKoG2CG1000A−LC] $
|
(4.33) |
have negative real parts. That is, the equilibrium $ (0, 0, 0, \mathbf{0}, \mathbf{0}) $ of the system (4.19)-(4.24) can be exponentially stabilized at $ \mathbf{w} = \mathbf{0} $.
So it remains to show that the regulator equations (4.13)-(4.18) have a solution. To this end, we set $ \mathbf{X} = [\mathbf{G}, \mathbf{H}_a, \mathbf{H}]^T $ and rewrite the equations (4.13)-(4.17) as one matrix equation:
$ [XZZ]Ae=Q[XZO]+[CT[0,Cd]G2[Cr,0]0], $
|
(4.34) |
where
$ Q=[A−BKz−BKoG2CG10LC−BKzA−LC−BKo]. $
|
(4.35) |
Since $ \mathbf{Q} $ and $ \mathbf{\Omega} $ are equivalent, all eigenvalues of $ \mathbf{Q} $ also have negative real parts. Because all eigenvalues of $ \mathbf{A}_e $ have zero real parts, the above matrix equation has a unique solution (see[19]). To show the solution also satisfies the equation (4.18), we let $ \mathbf{Z} = [\mathbf{Z}_{1}, \mathbf{Z}_{2}, \cdots, ,\mathbf{Z}_{2l+1}]^T $. It then follows from the equation (4.16) that
$ Z2=Z1Ae, $
|
(4.36) |
$ Z3=Z2Ae=Z1A2e, $
|
(4.37) |
$ ⋮Z2l+1=Z2lAe=Z1A2le, $
|
(4.38) |
$ G=[Cr,0]+Z2l+1Ae+c2l+1Z1+c2lZ2+⋯+c1Z2l+1=[Cr,0]+Z1A2l+1e+c2l+1Z1+c2lZ1Ae+⋯+c1Z1A2le=[Cr,0]+Z1(c2l+1I+c2lAe+⋯+c1A2le+A2l+1e)=[Cr,0]+Z10=[Cr,0]. $
|
(4.39) |
Here we have used the property that $ p(\lambda) $ is the minimal polynomial of $ \mathbf{A}_e $, that is, $ p(\mathbf{A}_e) = \mathbf{0} $. The above proof of the equation (4.18) gives the reason why the pair $ (\mathbf{G}_1, \mathbf{G}_2) $ is constructed as in (4.11).
Using the translation (4.6), we obtain a dynamic output feedback and feedforward controller for the original nonlinear system (3.3)-(3.7):
$ u=ˉu+˜u=ˉu−Kzz−Koo, $
|
(4.40) |
$ dzdt=G1z+G2(g−ˉg−[Cr,0]v), $
|
(4.41) |
$ dodt=Ao−B(Kzz+Koo)+L(g−ˉg−Co). $
|
(4.42) |
For the convenience of reference, we collect all relevant equations together to form a closed-loop system:
$ dgdt=−(m1+w1)g−gha+w1ˉg+Gb+[0,Cd]v, $
|
(4.43) |
$ dhadt=−(m2+w2)ha+(m3+w3)h+w2ˉha−w3ˉh, $
|
(4.44) |
$ dhdt=−(m4+w4)h+w4ˉh+ˉu−Kzz−Koo, $
|
(4.45) |
$ dzdt=G1z+G2(g−ˉg−[Cr,0]v), $
|
(4.46) |
$ dodt=Ao−B(Kzz+Koo)+L(g−ˉg−Co), $
|
(4.47) |
$ dvdt=Aev, $
|
(4.48) |
$ e=g−ˉg−[Cr,0]v. $
|
(4.49) |
Theorem 4.1. Suppose that $ \mathbf{K}_o, \mathbf{K}_z, \mathbf{L} $ are constant vectors such that the matrix
$ Ω=[A−BKo−BKzBKoG2CG1000A−LC] $
|
is exponentially stable. Then the blood glucose of the closed-loop nonlinear system (4.43)-(4.49) asymptotically tracks its reference $ \bar{g} + [\mathbf{C}_r, \; \mathbf{0}] \mathbf{v}(t) $ for any sufficiently small uncertainty $ \mathbf{w} $, that is,
$ limt→∞[g(t)−ˉg−[Cr,0]v(t)]=0. $
|
We use the same data as used in the above section to conduct a numerical simulation for this robust approximate controller (4.40). In this simulation, we first used an internal model 1-copy $ \mathbf{G}_1 $ of the matrix
$ Ae=[000⋯0000000π/750⋯000000−π/7500⋯00000000⋯00000000⋯00000⋮⋮⋮⋱⋮⋮⋮⋮⋮000⋯017π/750000000⋯−17π/7500000000⋯00000000⋯0000π/72000⋯000−π/720]. $
|
Unfortunately, Matlab failed to solve the closed-loop control system (4.43)-(4.49) with this high dimensional $ \mathbf{G}_1 $. Thus we have to reduce the dimension to obtain an internal model 1-copy $ \mathbf{G}_1 $ of the matrix
$ ˜Ae=[000000000000π/75000000000−π/7500000000000002π/75000000000−2π/7500000000000003π/75000000000−3π/75000000000000000000000000π/7200000000−π/720]. $
|
In this case, the matrices $ \mathbf{G}_1 $ and $ \mathbf{G}_2 $ are given by
$ G1=[0100000000010000000001000000000100000000010000000001000000000100000000010−3.70⋅10−160−2.89⋅10−110−4.83⋅10−70−0.00214950],G2=[000000001]. $
|
The vectors $ \mathbf{K}_o $ and $ \mathbf{K}_z $ are determined in the way such that the eigenvalues of the matrix $ \mathbf{\Phi}- \mathbf{\Psi}\mathbf{K} $ are: $ -0.0500, -0.0909, -0.1318, -0.1727, -0.2136, -0.2545, -0.2955, -0.3364, $ $ -0.3773, -0.4182, -0.4591, -0.5000, $ where $ \mathbf{K} = [\mathbf{K}_o, \mathbf{K}_z] $. Using the algorithm for computing the feedback gain matrix from Section 3.4 of the book[26], we obtain
$ Ko=[−55379.84,379273.25,2.93],Kz=[−0.00028,−0.019,−0.56,−9.41,−100.48,−728.94,−3701.22,−13180,−33587.83]. $
|
The vector $ \mathbf{L} $ is determined in the way such that the eigenvalues of the matrix $ \mathbf{A}- \mathbf{L}\mathbf{C} $ are: $ -0.01, -0.02, -0.03 $. Using the Matlab function place, we obtain
$ L=[−0.30679,−0.011386,211.93]T. $
|
According to our simulation studies, the eigenvalues should be between $ -0.5 $ and $ -0.01 $ to achieve a good performance of the controller. All other parameter values are the same as in the previous simulations.
Comparing Figure 6 with Figure 4, we found it interesting that the dynamics of all state variables under the robust controller (4.40) is identical to the one under the static controller (3.31).
The static controller (3.31) and the robust dynamic controller (4.40) have their own advantages and disadvantages. The static controller (3.31) is easy to implement, but is not flexible, no parameters being used to adjust to improve its performance. Since it is not robust, a small change in the model parameters results in a deviation of blood glucose from its target reference. On the other hand, the robust dynamical controller (4.40) can tolerate certain small changes in the model parameters and its performance can be improved by adjusting certain controller parameters such as $ \mathbf{K}_o, \mathbf{K}_z $ and $ \mathbf{L} $. However, it is more difficult to implement because the dimension of the closed-loop control system is greatly increased through the dynamic compensator (4.7). As discussed above, Matlab failed to solve the closed-loop control system numerically if the original $ \mathbf{G}_1 $ for the matrix $ \mathbf{A}_e $ is used. Furthermore, its performance critically depends on the gain vectors $ \mathbf{K}_o, \mathbf{K}_z $ and $ \mathbf{L} $. It is a great challenging problem to determine them to achieve the best performance of the controller in a way such that the simulated both glucose and insulin dynamics agrees with the real experimental data. Since the controller performance issue itself is an independent research area[12,22,28], this problem is beyond the scope of this work.
In this work, we have solved a robust glucose tracking problem. Parameter uncertainties were introduced into the mathematical model proposed by Bergman et al.[5]. Using an actual blood glucose level as feedback and an exogenous glucose input and a desired glucose reference as feedforward, we have designed a robust feedback and feedforward controller such that the blood glucose tracks the desired time-varying glucose reference for any small uncertainties. Numerical simulations conducted with published experimental blood glucose data show that our controller is effective in achieving the blood glucose tracking and agree with our theoretical results.
Limitations exist in this work. Although our robust feedback and feedforward controllers might have a potential application in the insulin pump technology, more comprehensive tests and improvements with experimental and clinic data are needed to be done before they might be implemented in an insulin pump. The interaction between glucose and insulin is complex. Our current simple model missed considering many other important control mechanisms such as the glucose regulation by the hormone glucagon and the organ liver. Therefore we may need to consider more complex mathematical models, such as the model proposed by Sturis et al.[40] and the model proposed by Liu et al.[24], and introduce parameter uncertainties into them. In all these cases, the regulator equations are more complex and it might be impossible to solve them analytically. Then a different approach such as used in[3] is needed to design a controller.
The author heartily appreciates the reviewers' constructive comments, which greatly enhance the quality of the paper to meet the journal standard.
The author declares no conflicts of interest in this paper.
[1] |
Lankao PR (2007) Are we missing the point? Particularities of urbanization, sustainability and carbon emissions in Latin American cities. Environ Urban 19: 159-175. doi: 10.1177/0956247807076915
![]() |
[2] | Betsill MM, Bulkeley H (2006) Cities and the multilevel governance of global climate change. Glob Gov 12: 141-159. |
[3] | Heidegger M (2001) The Fundamental Concepts of Metaphysics. Indiana: Indiana University Press. |
[4] |
Gibbs DC (2000) Ecological modernization, regional economic development and Regional Development Agences. Geoforum 31: 9-19. doi: 10.1016/S0016-7185(99)00040-8
![]() |
[5] |
Myllyla S, Kuvaja K (2005) Societal premises for sustainable development in large southern cities. Glob Environ Change 15: 224-237. doi: 10.1016/j.gloenvcha.2005.01.001
![]() |
[6] |
Deonie A, Olive V, Arthur S, et al. (2015). Urban Sediment Transport through an Established Vegetated Swale: Long Term Treatment Efficiencies and Deposition. Water 7: 1046-1067 doi: 10.3390/w7031046
![]() |
[7] |
Hamilton C (2010) Consumerism, self-creation and prospects for a new ecological consciousness. J Clean Prod 18: 571-575. doi: 10.1016/j.jclepro.2009.09.013
![]() |
[8] |
Li J, Wai OWH, Li YS, et al. (2010) Effect of green roof on ambient CO2 concentration. Build Environ 45: 2644-2651. doi: 10.1016/j.buildenv.2010.05.025
![]() |
[9] |
Susca T, Gaffin SR, Dell'Osso GR (2011) Positive effects of vegetation: Urban heat island and green roofs. Environ Pollut 159: 2119-2126. doi: 10.1016/j.envpol.2011.03.007
![]() |
[10] | Habermas J (1987) The Theory of Communicative Action, Volume 2: Life-World and System: A Critique of Functionalist Reason. Boston: Beacon Press. |
[11] | Foucault M (1980) Power-knowledge: selected interviews and other writings, 1972-1977. New York: Pantheon Books. |
[12] | Rabinow P (1991) The Foucault Reader. London: Penguin. |
[13] | Chichilnisky G (2011) What is sustainability? Int J Sust Econ 3: 125-140. |
[14] |
Corburn J (2009) Cities, Climate Change and Urban Heat Island Mitigation: Localising Global Environmental Science. Urban Stud 46: 413-427. doi: 10.1177/0042098008099361
![]() |
[15] | Jasanoff S (2004) States of knowledge: the co-production of science and the social order. London: Routledge. |
[16] |
Bai X (2007) Integrating global environmental concerns into urban management: the scale and readiness arguments. J Ind Ecol 11: 15-29. doi: 10.1162/jie.2007.1202
![]() |
[17] | Miller C (2004) Resisting empire; globalism, relocalization, and the politics of knowledge. In: S Jasanoff, M Martello (eds.) Earthly Politics: Local and Global in Environmental Governance, 81-102. Cambridge, MA: MIT Press. |
[18] | Ingold T (2000) The perception of the environment: essays on livelihood, dwelling and skill. Melbourne: Psychology Press. |
[19] | Meinig DW, Jackson JB, Lewis PF, et al. (1979) The Interpretation of Ordinary Landscapes: Geographical Essays. New York: Oxford University Press. |
[20] |
Logan MI (1964) Manufacturing Decentralization in the Sydney Metropolitan Area. Econ Geogr 40: 151-162. doi: 10.2307/142194
![]() |
[21] |
Duncan J (1995) Landscape Geography, 1993-94. Prog Human Geogr 19: 414-422. doi: 10.1177/030913259501900308
![]() |
[22] | Cosgrove D, Jackson P (1987) New directions in cultural geography. Area 19: 95-101. |
[23] |
Muir R (1998) Landscape: a wasted legacy. Area 30: 263-271. doi: 10.1111/j.1475-4762.1998.tb00071.x
![]() |
[24] |
Seddon G (1986) Landscape Planning: A Conceptual Perspective. Landsc Urban Plan 13: 335-347. doi: 10.1016/0169-2046(86)90051-4
![]() |
[25] | Copley S (1994) The Politics of the Picturesque: Literature, Landscape, and Aesthetics since 1770. Cambridge: Cambridge University Press. |
[26] | Kant I (1952) The science of right (W. Hastie, trans.). In: R Hutchins (ed.) Great books of the Western world, Vol. 42: Kant, 397-446. Chicago: Encyclopaedia Britannica. |
[27] | Sauer C (1925) The Morphology of Landscape. Los Angeles: Univ Calif Publ Geogr 22: 19-53. |
[28] |
Gulinck H (1986) Landscape ecological aspects of agro-ecosystems. Agr Ecosyst Environ 16: 79-86. doi: 10.1016/0167-8809(86)90095-2
![]() |
[29] |
Olwig KR (1996) Recovering the substantive nature of landscape. Ann Assoc Am Geogr 86: 630-653. doi: 10.1111/j.1467-8306.1996.tb01770.x
![]() |
[30] |
Zhang LW, Jianping ZY, Jiong S (2004) A GIS-based gradient analysis of urban landscape pattern of Shanghai metropolitan area, China. Landsc Urban Plan 69: 1-16. doi: 10.1016/j.landurbplan.2003.08.006
![]() |
[31] | Dickinson RE, Crowe PR (1939) Landscape and society. Scott Geogr Mag 55: 1-15. |
[32] |
Lewis P (1983) Learning from looking: geographic and other writing about the American cultural landscape. Am Quart 35: 242-261. doi: 10.2307/2712650
![]() |
[33] | Jackson JB (1984) Discovering the vernacular landscape. New York: Yale University Press. |
[34] | Cosgrove D, Daniels S (1988) The iconography of landscape: essays on the symbolic representation, design and use of past environments. London: Cambridge University Press. |
[35] | Anderson K, Gale F (1992) Inventing places: Studies in cultural geography. Melbourne: Longman Cheshire. |
[36] |
Zukin S (1998). Urban lifestyles: diversity and standardisation in spaces of consumption. Urban Stud 35: 825-839. doi: 10.1080/0042098984574
![]() |
[37] |
Gobster P, Nassauer J, Daniel TC, et al. (2007) The shared landscape: what does aesthetics have to do with ecology? Landsc Ecol 22: 959-972 doi: 10.1007/s10980-007-9110-x
![]() |
[38] |
Johnston R (1998) Approaches to the perception of landscape. Archeo Dialog 5: 54-68. doi: 10.1017/S1380203800001161
![]() |
[39] |
Seamans S (2013) Mainstreaming the environmental benefits of street trees. Urban For Urban Gree 12: 2-11. doi: 10.1016/j.ufug.2012.08.004
![]() |
[40] | Baker AR, Biger G (2006) Ideology and landscape in historical perspective: essays on the meanings of some places in the past, Vol. 18. London: Cambridge University Press. |
[41] | Thompson JB (1984) Studies in the Theory of Ideology. Los Angeles: University of California Press. |
[42] | Žižek S (1989) The sublime object of ideology. London: Verso. |
[43] | Žižek S, Daly G (2004) Conversations with Zizek. Cambridge: Polity. |
[44] | Cosgrove DE (1984) Social Formation and Symbolic Landscape. Totowa, NJ: Barnes and Noble. |
[45] |
Robinson GM, Liu Z (2015) Greening and “un”greening Adelaide, South Australia. AIMS Enviro Sci 2: 511-532. doi: 10.3934/environsci.2015.3.511
![]() |
[46] |
van Der Heijden H (1999) Environmental movements, ecological modernisation and political opportunity structures. Environ Polit 8: 199-221. doi: 10.1080/09644019908414444
![]() |
[47] | Baker A (1992) Introduction: on ideology and landscape. In: Baker A, and Biger G (eds.) Ideology and landscape in historical perspective, 1-14. London: Cambridge University Press. |
[48] |
Bailey I, Gouldson A, Newell P (2011) Ecological modernisation and the governance of carbon: a critical analysis. Antipode 43: 682-703. doi: 10.1111/j.1467-8330.2011.00880.x
![]() |
[49] |
Blühdorn I, Welsh I (2007) Eco-politics beyond the paradigm of sustainability: a conceptual framework and research agenda. Environ Polit 16: 185-205. doi: 10.1080/09644010701211650
![]() |
[50] | Derrida J (1990) Force of Law: The Mystical Foundation of Authority. Cardozo L Rev 11: 919. |
[51] |
Checker M (2011) Wiped out by the ‘Greenwave’: environmental gentrification and the paradoxical politics of urban sustainability. City Soc 23: 210-229. doi: 10.1111/j.1548-744X.2011.01063.x
![]() |
[52] |
Shaw W, Menday (2013) Fibro Dreaming: Greenwashed Beach-house Development on Australia's Coasts. Urban Stud 50: 2940-2958. doi: 10.1177/0042098013482507
![]() |
[53] | Morris ES (2011) Down with ECO-towns! Up with ECO-communities. Or Is There a Need for Model Eco-towns? A Review of the 2009-2010 Eco-town Proposals in Britain. In: T Wong, B Yuen (eds.) Eco-city Planning-Policies, Pract and Design, 113-130. Dordrecht: Springer. |
[54] |
Meppem T, Bourke S (1999) Different ways of knowing: a communicative turn toward sustainability. Ecol Econ 30: 389-404. doi: 10.1016/S0921-8009(99)00053-1
![]() |
[55] |
Lee KN (2006) Urban sustainability and the limits of classical environmentalism. Environ Urban 18: 9-22. doi: 10.1177/0956247806063940
![]() |
[56] | Birch E, Wachter S (2008) Growing greener cities: Urban sustainability in the twenty-first century. Philadelphia: University of Pennsylvania Press. |
[57] |
Leff E (1993) Marxism and the environmental question: From the critical theory of production to an environmental rationality for sustainable development. Capital Nat Social 4: 44-66. doi: 10.1080/10455759309358531
![]() |
[58] | Scott AJ (1997) The cultural economy of cities. London: Sage. |
[59] | Friedmann H (2005) From colonialism to green capitalism: Social movements and the emergence of food regimes. Res Rural Sociol Dev 11: 227-264. |
[60] |
Bakker K (2005) Market Environmentalism in Water Supply in England and Wales. Ann Assoc Am Geogr 95: 542-565. doi: 10.1111/j.1467-8306.2005.00474.x
![]() |
[61] |
Gibbs D, Jonas, A (2000) Governance and regulation in local environmental policy: The utility of a regime approach. Geoforum 34: 299-313. doi: 10.1016/S0016-7185(99)00052-4
![]() |
[62] |
Kovacs ZI, LeRoy CJ, Fischer DG, et al. (2006) How do Aesthetics Affect our Ecology. J Ecol Anthropol 10: 61-65. doi: 10.5038/2162-4593.10.1.5
![]() |
[63] |
Svensson K, Eliasson I (2002) Diurnal air temperatures in built-up areas in relation to urban planning. Landsc Urban Plan 61: 37-54. doi: 10.1016/S0169-2046(02)00076-2
![]() |
[64] |
Arnfield AJ (2003) Two decades of urban climate research: a review of turbulence, exchanges of energy and water, and the urban heat island. Int J Climatol 23: 1-26. doi: 10.1002/joc.859
![]() |
[65] | Oke TR (1987) Boundary Layer Climates, 2nd edition. New York: Methuen. |
[66] |
Tan J, Zheng Y, Tang X, et al. (2010) The urban heat island and its impact on heat waves and human health in Shanghai. Int J Biometeorol 54: 75-84. doi: 10.1007/s00484-009-0256-x
![]() |
[67] |
Newman PWG (1999) Sustainability and cities: extending the metabolism model. Landsc Urban Plan 44: 219-226. doi: 10.1016/S0169-2046(99)00009-2
![]() |
[68] | Durkheim E (1972) Emile Durkheim: selected writings. Sydney: Cambridge University Press. |
[69] | United Nations, City planning will determine pace of global warming. United Nations, 2007. Available from: http://www.un.org/News/Press/docs/2007/gaef3190.doc.htm. |
[70] | Parker JH (1983) Landscaping to reduce the energy used in cooling buildings. J Forest 81: 82-105. |
[71] |
Nowak DJ, Crane DE (2002) Carbon storage and sequestration by urban trees in the USA. Enviro Pollut 116: 381-389. doi: 10.1016/S0269-7491(01)00214-7
![]() |
[72] |
Chua BH (1998) World cities, globalisation and the spread of consumerism: a view from Singapore. Urban Stud 35: 981-1000. doi: 10.1080/0042098984646
![]() |
[73] |
Hurrell A, Sengupta S (2012) Emerging powers, North-South relations and global climate politics. Int Affair 88: 463-484. doi: 10.1111/j.1468-2346.2012.01084.x
![]() |
[74] |
Norman J, MacLean HL, Kennedy CA (2006) Comparing high and low residential density: life-cycle analysis of energy use and greenhouse gas emissions. J Urban Plan Dev 132: 10-21. doi: 10.1061/(ASCE)0733-9488(2006)132:1(10)
![]() |
[75] |
Jaffe AB, Newell RG, Stavins RN (2005) A tale of two market failures: Technology and environmental policy. Ecol Econ 54: 164-174. doi: 10.1016/j.ecolecon.2004.12.027
![]() |
[76] | Thorpe A, Graham K (2009) Green buildings-are codes, standards and targets sufficient drivers of sustainability in New South Wales? Environ Plan Law J 26: 486-497. |
[77] |
Wolf D, Lundholm JT (2008). Water uptake in green roof microcosms: Effects of plant species and water availability. Ecol Eng 33: 179-186. doi: 10.1016/j.ecoleng.2008.02.008
![]() |
[78] | Barnes K, Morgan J, Roberge M (2001) Impervious surfaces and the quality of natural built environments. Baltimore: Department of Geography and Environmental Planning. |
[79] |
Bowler DE, Buyung-Ali L, Knight TM, et al. (2010) Urban greening to cool towns and cities: A systematic review of the empirical evidence. Landsc Urban Plan 97: 147-155. doi: 10.1016/j.landurbplan.2010.05.006
![]() |
[80] | Dunnett N, Kingsbury N (2004). Planting green roofs and living walls, Vol. 254. Portland, OR: Timber Press. |
[81] |
Francis RA, Lorimer J (2011) Urban reconciliation ecology: the potential of living roofs and walls. J Environ Manag 92: 1429-1437. doi: 10.1016/j.jenvman.2011.01.012
![]() |
[82] |
Saiz S, Kennedy C, Bass B, et al. (2006) Comparative life cycle assessment of standard and green roofs. Environ Sci Technol 40: 4312-4316. doi: 10.1021/es0517522
![]() |
[83] |
Ottelé M, van Bohemen HD, Fraaij AL (2010) Quantifying the deposition of particulate matter on climber vegetation on living walls. Ecol Eng 36: 154-162. doi: 10.1016/j.ecoleng.2009.02.007
![]() |
[84] |
Ford LR (1992) Reading the skylines of American cities. Geogr Rev 82: 180-200. doi: 10.2307/215431
![]() |
[85] |
Sudhira HS, Ramachandra TV, Jagadish KS (2004). Urban sprawl: metrics, dynamics and modelling using GIS. Int Appl Earth Obs Geoinf 5: 29-39. doi: 10.1016/j.jag.2003.08.002
![]() |
[86] |
Pincetl S (2010) Implementing municipal tree planting: Los Angeles million-tree initiative. Environ Manag 45: 227-238. doi: 10.1007/s00267-009-9412-7
![]() |
[87] | Pacey A (1983) The Cult of Tech. Cambridge, MA: The MIT Press. |
[88] |
Sklair L (2010) Iconic Architecture and the Culture-ideology of Consumerism. Theor Cult Soc 27: 135-159. doi: 10.1177/0263276410374634
![]() |
[89] |
Jackson T (2005) Live Better by Consuming Less? Is There a ‘Double Dividend’ in Sustainable Consumption. J Ind Ecol 9: 19-36. doi: 10.1162/108819805775247873
![]() |
[90] |
Cluley R, Dunne S (2012) From commodity fetishism to commodity narcissism. Mark Theor 12: 251-265. doi: 10.1177/1470593112451395
![]() |
[91] |
Kumar M, Kumar P (2008) Valuation of the ecosystem services: A psycho-cultural perspective. Ecol Econ 64: 808-819. doi: 10.1016/j.ecolecon.2007.05.008
![]() |
[92] | Gupta S, Ogden D (2006). The attitude-behavior gap in environmental consumerism. APUBEF Proc 3: 199-206. |
[93] | Sanne C (2002) Willing consumers-of locked in? Policies for a sustainable consumption. Ecol Econ 42: 273-287. |
[94] |
Skollerhorn E (1998) Habermas and Nature: The Theory of Communicative Action for Studying Environmental Policy. J Environ Plan Manag 41: 555-573. doi: 10.1080/09640569811452
![]() |
[95] |
Hackworth J (2002) Postrecession gentrification in New York City. Urban Affair Rev 37: 815-843. doi: 10.1177/107874037006003
![]() |
[96] | LeGates RT, Hartman C (1986) The anatomy of displacement in the United States. In: N Smith, P Williams (eds.) Gentrification of the City, 178-199. Boston: Allen & Unwin. |
[97] | Zizek S (2010) Living in the End Times. London: Verso Books. |
[98] |
Mapes J, Wolch J (2011) ‘Living Green’: The promise and pitfalls of new sustainable communities. J Urban Design 16: 105-126. doi: 10.1080/13574809.2011.521012
![]() |
[99] | Gray M (2002) Urban Surveillance and Panopticism: will we recognize the facial recognition society. Surveill Soc 1: 314-330. |
1. | J. Sooknanan, D. M. G. Comissiong, Trending on Social Media: Integrating Social Media into Infectious Disease Dynamics, 2020, 82, 0092-8240, 10.1007/s11538-020-00757-4 | |
2. | Aníbal Coronel, Luis Friz, Ian Hess, María Zegarra, On the existence and uniqueness of an inverse problem in epidemiology, 2021, 100, 0003-6811, 513, 10.1080/00036811.2019.1608964 | |
3. | Winfried Just, Joan Saldaña, Ying Xin, Oscillations in epidemic models with spread of awareness, 2018, 76, 0303-6812, 1027, 10.1007/s00285-017-1166-x | |
4. | Fahad Al Basir, Santanu Ray, Ezio Venturino, Role of media coverage and delay in controlling infectious diseases: A mathematical model, 2018, 337, 00963003, 372, 10.1016/j.amc.2018.05.042 | |
5. | Chao Liu, Jane Heffernan, Stochastic dynamics in a delayed epidemic system with Markovian switching and media coverage, 2020, 2020, 1687-1847, 10.1186/s13662-020-02894-5 | |
6. | Fulgensia Kamugisha Mbabazi, J. Y. T. Mugisha, Mark Kimathi, Global Stability of Pneumococcal Pneumonia with Awareness and Saturated Treatment, 2020, 2020, 1110-757X, 1, 10.1155/2020/3243957 | |
7. | M. A. Aziz-Alaoui, Jean M.-S. Lubuma, Berge Tsanou, Prevalence-based modeling approach of schistosomiasis: global stability analysis and integrated control assessment, 2021, 40, 2238-3603, 10.1007/s40314-021-01414-9 | |
8. | David Juher, David Rojas, Joan Saldaña, Robustness of behaviorally induced oscillations in epidemic models under a low rate of imported cases, 2020, 102, 2470-0045, 10.1103/PhysRevE.102.052301 | |
9. | Tanuja Das, Prashant K. Srivastava, Anuj Kumar, Nonlinear dynamical behavior of an SEIR mathematical model: Effect of information and saturated treatment, 2021, 31, 1054-1500, 043104, 10.1063/5.0039048 | |
10. | Shikha Jain, Sachin Kumar, Dynamic analysis of the role of innate immunity in SEIS epidemic model, 2021, 136, 2190-5444, 10.1140/epjp/s13360-021-01390-3 | |
11. | Ellen P. Carlin, Koya C. Allen, Jeffrey J. Morgan, Jean-Paul Chretien, Suzan Murray, Deborah Winslow, Dawn Zimmerman, Behavioral Risk Modeling for Pandemics: Overcoming Challenges and Advancing the Science, 2021, 19, 2326-5094, 447, 10.1089/hs.2020.0209 | |
12. | Khadija Akdim, Adil Ez-Zetouni, Mehdi Zahid, A stochastic vaccinated epidemic model incorporating Lévy processes with a general awareness-induced incidence, 2021, 14, 1793-5245, 2150044, 10.1142/S1793524521500443 | |
13. | Talal Daghriri, Michael Proctor, Sarah Matthews, Evolution of Select Epidemiological Modeling and the Rise of Population Sentiment Analysis: A Literature Review and COVID-19 Sentiment Illustration, 2022, 19, 1660-4601, 3230, 10.3390/ijerph19063230 | |
14. | Ze Chen, Yuan Wang, Yanjun Guan, Michael Jie Guo, Rong Xu, Long‐term effect of childhood pandemic experience on medical major choice: Evidence from the 2003 severe acute respiratory syndrome outbreak in China, 2023, 1057-9230, 10.1002/hec.4659 | |
15. | Tanuja Das, Prashant K. Srivastava, Effect of a novel generalized incidence rate function in SIR model: Stability switches and bifurcations, 2023, 166, 09600779, 112967, 10.1016/j.chaos.2022.112967 | |
16. | Rabindra Kumar Gupta, Rajanish Kumar Rai, Pankaj Kumar Tiwari, Arvind Kumar Misra, Maia Martcheva, A mathematical model for the impact of disinfectants on the control of bacterial diseases, 2023, 17, 1751-3758, 10.1080/17513758.2023.2206859 | |
17. | A. K. Misra, Kuldeep Kumar Tripathi, Maitri Verma, Modeling the effects of psychological fear and media-induced awareness on the dynamics of infectious diseases, 2023, 0, 1531-3492, 0, 10.3934/dcdsb.2023118 | |
18. | Jin Yang, Likun Guan, Zhuo Chen, Yuanshun Tan, Zijian Liu, Robert A. Cheke, Bifurcation analysis of a nonlinear pulse SIR model with media coverage, 2023, 111, 0924-090X, 19543, 10.1007/s11071-023-08869-x | |
19. | Aníbal Coronel, Fernando Huancas, Ian Hess, Alex Tello, The diffusion identification in a SIS reaction-diffusion system, 2023, 21, 1551-0018, 562, 10.3934/mbe.2024024 | |
20. | Makayla Preston, Alexandria Carter, Eric Numfor, Modeling the Effects of Media Awareness on SARS-CoV-2 Transmission in Georgia, 2024, 10, 2349-5103, 10.1007/s40819-024-01759-9 | |
21. | Arjun Kumar, Uma S. Dubey, Balram Dubey, Modeling the effect of vaccinations, hospital beds, and treatments on the dynamics of infectious disease with delayed optimal control and sensitivity analysis, 2024, 139, 2190-5444, 10.1140/epjp/s13360-024-05508-1 | |
22. | Saduri Das, Prashant K. Srivastava, Pankaj Biswas, The role of social media in a tuberculosis compartmental model: Exploring Hopf-bifurcation and nonlinear oscillations, 2025, 230, 03784754, 323, 10.1016/j.matcom.2024.11.015 | |
23. | Pritam Saha, Kalyan Kumar Pal, Uttam Ghosh, Pankaj Kumar Tiwari, Dynamic analysis of deterministic and stochastic SEIR models incorporating the Ornstein–Uhlenbeck process, 2025, 35, 1054-1500, 10.1063/5.0243656 | |
24. | MAITRI VERMA, KULDEEP KUMAR TRIPATHI, ARVIND KUMAR MISRA, MODELING THE EFFECTS OF PSYCHOLOGICAL FEAR, TV AND SOCIAL MEDIA ADVERTISEMENTS ON THE CONTROL OF INFECTIOUS DISEASES, 2025, 33, 0218-3390, 553, 10.1142/S0218339025500147 | |
25. | Xinghua Chang, Jianrong Wang, Maoxing Liu, Xue Yan, An SIS infectious disease model with nonlinear incidence and disease awareness on complex networks, 2025, 196, 09600779, 116349, 10.1016/j.chaos.2025.116349 |
Parameter | Value | Description |
$ m_1 $ | 0.000296 (/min) | Glucose utilization rate constant |
$ m_2 $ | 0.045 (/min) | Insulin effect rate constant |
$ m_3 $ | 0.000012 (/(pmol/l)/min$ ^2 $) | Insulin effect rate constant |
$ m_4 $ | 0.268 (/min) | Insulin metabolism constant |
$ G_b $ | 0.3218 (mM/min) | Basal glucose input |
$ \bar{g} $ | 6 (mM) | Glucose equilibrium |
$ \bar{h} $ | 200 (pmol/l) | Insulin equilibrium |
$ \bar{h}_a $ | 0.0533 (/min) | Insulin effect equilibrium |
$ \bar{u} $ | 53.63 (pmol/l/min) | Insulin input equilibrium |
Parameter | Value | Description |
$ m_1 $ | 0.000296 (/min) | Glucose utilization rate constant |
$ m_2 $ | 0.045 (/min) | Insulin effect rate constant |
$ m_3 $ | 0.000012 (/(pmol/l)/min$ ^2 $) | Insulin effect rate constant |
$ m_4 $ | 0.268 (/min) | Insulin metabolism constant |
$ G_b $ | 0.3218 (mM/min) | Basal glucose input |
$ \bar{g} $ | 6 (mM) | Glucose equilibrium |
$ \bar{h} $ | 200 (pmol/l) | Insulin equilibrium |
$ \bar{h}_a $ | 0.0533 (/min) | Insulin effect equilibrium |
$ \bar{u} $ | 53.63 (pmol/l/min) | Insulin input equilibrium |