Content uploaded by Mahmoud Elbayoumi
Author content
All content in this area was uploaded by Mahmoud Elbayoumi on Dec 23, 2021
Content may be subject to copyright.
A Novel Power-Aware Task Scheduling for Energy
Harvesting-Based Wearable Biomedical Devices
Using FPA
Retaj Yousri1, Mahmoud Elbayoumi2, Ahmed Moawad1, M. Saeed Darweesh1, and Ahmed Soltan3
1Wireless Intelligent Networks Center (WINC), Nile University, Giza 12677, Egypt
2Electrical Engineering Department, Faculty of Engineering, Fayoum University, Fayoum 63514, Egypt
3Nanoelectronics Integrated Systems Center (NISC), Nile University, Giza 12677, Egypt
Abstract—Power management and saving in energy harvesting-
based biomedical wearable devices are mandatory to ensure
prolonged and stable operation under a stringent power budget.
Thus, power-aware task scheduling can play a key role in min-
imizing energy consumption to improve system durability while
maintaining device functionality. This paper proposes a novel
biosensor task scheduling for optimizing energy consumption
through wearable biomedical devices. The proposed approach is
based on Flower Pollination Algorithm (FPA). The biomedical
functionality constraints are enforced with a Hamming-based
Tikhonov regularization. We proposed a greedy approach to com-
pute the Tikhonov regularization term efficiently. The algorithm
has been tested for scheduling the tasks of two biosensors: a heart
rate sensor and a temperature sensor on a lab-based biomedical
device.
Index Terms—Energy harvesting, biosensors, optimization,
FPA, wearable devices, tasks scheduling, Tikhonov regulariza-
tion, telemedicine.
I. INTRODUCTION
In healthcare systems, wearable biomedical devices have
attracted significant interest [1]. This is due to their usage
for monitoring various bio-signals, early detection of medical
illnesses, and emergency notification [2], [3]. Embedding
the biosensors into wearable biomedical devices is essential
as they are responsible for acquiring different bio-signals
(e.g., heart rate, temperature, and oxygen saturation of the
blood) [4], [5]. Those sensing signals are measured based on
large-scale sensing techniques such as the optical, thermal,
and electrochemical ones [6]. Wearable devices enable more
mobility to their users. This adds crucial factors to consider in
its design (e.g., size, power supply, energy consumption, and
electrodeposition). In order to prevent the interruption of an
operating wearable device, providing a continuous source of
power is indispensable.
One of the major challenges of those devices is the inter-
mittent nature of the Energy-harvesting sources [3]. In order
to prevent the interruption of an operating wearable device,
providing a continuous source of power is indispensable
[7], [8]. Battery-based wearable devices have proven their
inefficient and constrained operation [3]. This is because of
the long-term charging and discharging cycle, which leads to
operational interruption and human intervention. In spite that
the battery-based wearable devices can be designed to be with
large capacity, the increase in their weight and size is still
considered a limitation [9].
Another approach utilizes both the source capability and
power management to ensure stable, long-lasting operation
[10], [11]. There have been many researches that adopted the
concept of optimization either to minimize the power con-
sumption or to maximize the energy throughout the energy har-
vesting devices [12]. Different approaches employed the opti-
mization in various ways, including circuit design, operation,
communication, and components [12]. The optimization in the
context of the operation has been adopted in the literature as
the main concern of this paper is power saving. Authors in
[13] considered minimizing the sensors’ average contribution
to power consumption by introducing pulsed sensor excitation.
Kansal et al. [14] employed a dynamically-controlled duty
cycle of sampling in order to maximize the harvested power
utilization. They considered duty cycle reduction whenever the
scavenged energy is low as at night and vice versa. Also,
this research group introduced the concept of energy-neural
operation, which means that, on average, energy is consumed
only as much as it is harvested. In [15], the limitation of
minimally adaptive duty-cycling mechanism was addressed by
introducing a new technique that ensures duty cycle stability
without assuming prior knowledge about the incoming energy.
A framework for energy management in energy harvesting em-
bedded systems was presented in [16]. The authors provided
a real-time scheduling approach and reward maximization
for optimizing the system performance given certain energy
constraints. Luo et al. [17] considered time scheduling and
power allocation optimization by investigating how to apply
to relay to improve the short-term performance of energy
harvesting communication systems. Their study employed a
relaxed energy profile and the directional water-filling (DWF)
algorithm. In [18], a reinforcement learning-based throughput
on-demand provisioning dynamic power management method
was proposed for sustainable operation of energy harvesting
wireless sensor node. Authors in [19] introduced a fuzzy-
based novel for power management in energy harvesting-based
devices. Because of the unpredictability of the harvested en-
ergy, fuzzy control theory was employed, showing promising
results. Qiu et al. [20] introduced Lyapunov optimization for
energy harvesting wireless sensor communications.
By addressing the need to managing and saving power
throughout the energy harvesting-based wearable biomedical
devices, the contributions of this work can be listed as follows:
•Proposing biosensors task scheduling for the first time in
the energy harvesting-based wearable biomedical devices.
•Formulating, mathematically, multi-device task schedul-
ing problem.
•Proposing a task scheduling technique based on Flower
Pollination Algorithm (FPA) that considers low power
operation constraint. In addition, we proposed a greedy
approach to compute Hamming-based Tikhonov regular-
ization to mandate the feasibility of the obtained solution.
The paper is structured as follows. The system overview is
presented in Section 2. The problem formulation, including
the objective function, operation constraints, Tikhonov regu-
larization used for the optimization approach, and FPA, are
found in Section 3. The results are presented and discussed in
Section 4, while the conclusion is summarized in Section 5.
II. SY ST EM OV ERVIEW
This section presents our wearable biomedical system that
will be used as a case study for the proposed optimization
approach. This is followed by the description of the energy
consumption data profiling utilized to guide the proposed FPA-
based technique to obtain an optimal schedule.
A. Wearable Biomedical System
The wearable biomedical system consists of a piezoelectric
harvester, a bridge rectifier, a supercapacitor, a processing unit,
a temperature sensor, and a heart rate sensor as depicted in
Fig. 1. The piezoelectric harvester is an energy scavenger that
converts ambient energy into electrical energy. The electrical
energy then passes through a bridge rectifier followed by a
supercapacitor. The supercapacitor is a high-capacity capacitor
that is capable of storing energy for a relatively long time; thus
it is considered the energy reservoir of the circuit. By focusing
on the monitoring function of the wearable biomedical device,
two essential sensors were chosen for this circuit: a heart
rate sensor and a temperature sensor. Fig. 1 shows the two
biosensors connected to the circuit micro-controller.
B. Energy Consumption Profiling
As the energy consumption depends on a nonlinear load
(e.g., processing units, sensors..etc.), the consumption will
vary depending on the current/voltage passing through/across
the supercapacitor. Accordingly, we calculated a dataset to
profile the energy consumption and its corresponding voltage
drop in the capacitor from the datasheet of the different IC
components of our system [21], [22], [23]. Fig. 2 illustrates a
sample for the calculated dataset. The first column represents
the Analog-to-Digital conversation (ADC)-value. Which is
corresponds to the voltage value across the supercapacitor.
Piezoelectric harvester
Bridge rectifier Supercapacitor
Temperature sensor
Heart rate sensor
I2CADC
Fig. 1: Wearable system powered by supercapacitor.
The ADC-value is used instead of the voltage value because
those values are measured with ADC converters (24-bit in
our situation). The second column represents the combina-
tion (ON/OFF status of the two sensors) packed in integer
representation. In other words, the combination column has
repetitive decimal values ranging from 0to 3, where these
values represent the biosensors status, whether ON (1) or OFF
(0). As there are two biosensors in our system, they can be
represented by two bits: the first bit denotes the status of the
heart rate sensor, and the second bit denotes the status of the
temperature sensor. By decoding the decimal values to their
corresponding binary ones, we can deduce that 2, for example,
is equivalent to 1 0 in binary, which means that the heart rate
sensor is ON while the temperature sensor is OFF, this applies
as well to the other three decimal values. The third column
depicts the resultant voltage drop (∆V) shown in the ADC
value.
𝑽
𝒔𝒕𝒂𝒓𝒕
(adc) Combination
∆
𝑽
167772150316710106111281664299724482165758883563514696840041462973111289
.
..
.
..
.
..
Fig. 2: Sample from the generated dataset.
III. PROB LE M FOR MU LATI ON
In this part, we present the problem formulation that governs
the energy consumption (and its corresponding voltage drop
across the supercapacitor) through an energy harvesting-based
wearable biomedical device in the presence of two biosensors:
a heart rate sensor and temperature sensor. This formulation
will be utilized in the proposed optimization algorithm that
robustly manages the biosensors tasks while considering the
system’s constraints (which is achieved by Tikhonov’s regu-
larization).
A. Objective Function
In our approach, the biomedical device is intended to
operate for a certain period of time. This period will be
divided into Nslots time-slots. Our approach aims to obtain
an optimized schedule for the operation of the temperature
and heart rate sensors to save power throughout the device
while maintaining the device’s functionality constraints. For
each time-slot i∈ {1,2, ..., Nslots}there is an associated
decision variable xi∈ {00,01,10,11}. Where xidetermines
whether two sensors are turned ON(1)/OFF(0). The LSB bit
corresponds to the temperature sensor, while the MSB bit
corresponds to the heart rate sensor. Our objective function
is defined as follows:
f(X) = Vfinal (X) + TT ikhonov(X)(1)
where X={x1, x2, ..., xNslots }represent a schedule over
Nslots time-slots, Vf inal is the final ADC voltage value (which
corresponds to analog voltage) across the supercapacitor, and
TT ikhonov is a regularization term that will be described in the
next subsection.
Accordingly, the optimum solution is defined as:
ˆ
X= argmax
∀X
(f(X)) (2)
where ˆ
Xis the optimum schedule.
B. Tikhonov Regularization and Operation Constraints
Tikhonov regularization is a commonly used regularization
technique [24]. Its strategy is based on adding a regularization
term to the objective function in order to approach a particular
solution with desirable properties [25]. Tikhonov’s regulariza-
tion term generally aims to enforce the objective function to
exclude infeasible solutions. The optimization algorithm aims
to find or approach the optimum solution ˆ
Xthat satisfies the
adopted objective function. In this study, feasible solutions
can be approached relying on the objective function found in
Eq. (1) that contains regularization term TT ikhonov defined as
follows:
TT ikhonov (X) = −λ× ||X−XNF | |H(3)
Where XNF is the nearest feasible solution to Xwhile
||X−XNF | |His the Hamming distance between Xand
XNF . Hamming distance between two vectors of equal length
is the number of positions at which the corresponding elements
of the vector differ. In the context of dealing with binary
values (0or 1), Hamming distance represents the L2(2-norm)
of X⊕XNF . For example, if we have two vectors of the
elements shown in Fig. 3, the Hamming distance between them
is equal to √2.
The nearest feasible solution (XNF ) can be easily derived
by specifying the operation constraints. In this study, the
sensors are supposed to acquire new measurements on a
periodic basis as follows:
for today’s energy harvesting wireless sensor node (EHWSN)
8.An approximate multi-parametric programming algorithm was proposed
in [8] for the adap- tive power management of an energy harvesting embed- ded
system
In this paper a framework for energy management in energy harvesting em-
bedded systems is presented. As a possible example scenario, we focus on wire-
less sensor nodes which are powered by solar cells. We demonstrate that classical
power management solutions have to be reconceived and/or new problems arise
if perpetual operation of the system is required. In particular, we provide a set
of algorithms and methods for different application scenarios, including real-
time scheduling, application rate control as well as reward maximization. The
goal is to optimize the performance of the application subject to given energy
constraints. Our methods optimize the system performance which allows the
usage of, e.g., smaller solar cells and smaller batteries. Our theoretical results
are supported
V1:0 0 1010
V2:1010 0 0
3
Fig. 3: Visualizing the binary elements of two vectors.
1) The temperature sensor acquires body temperature one
and only one with period PT.
2) The heart rate sensor acquires body heart rate one and
only one with period PHR .
Accordingly, by setting the values of PTand PH S , the
nearest feasible solution (XNF ) can be easily estimated to
be used in Tikhonov approach. For more elaboration, let us
consider the following:
•Nslots =5time slots.
•X={00,10,11,11,10}.
•PHR =2and PT=3.
By assigning the first bit to the heart rate sensor and the
second bit to the temperature sensor, we can generate (XNF )
for the solutions of both sensors as shown in Fig. 4
for today’s energy harvesting wireless sensor node (EHWSN)
8.An approximate multi-parametric programming algorithm was proposed
in [8] for the adap- tive power management of an energy harvesting embed- ded
system
In this paper a framework for energy management in energy harvesting em-
bedded systems is presented. As a possible example scenario, we focus on wire-
less sensor nodes which are powered by solar cells. We demonstrate that classical
power management solutions have to be reconceived and/or new problems arise
if perpetual operation of the system is required. In particular, we provide a set
of algorithms and methods for different application scenarios, including real-
time scheduling, application rate control as well as reward maximization. The
goal is to optimize the performance of the application subject to given energy
constraints. Our methods optimize the system performance which allows the
usage of, e.g., smaller solar cells and smaller batteries. Our theoretical results
are supported
X:01 1 1 1
XNF :01010
X:00110
XNF :00100
3
Fig. 4: Visualizing the solution and the nearest feasible solution for both
sensors.
Algorithm 1 Tikhonov Regularization Term
Inputs: SolH R,S olT,PHR,PT,Nslots ,SHR ,ST
Outputs: TT ikhonov
1: Initialization
2: HRF easibleSol ←zeros
3: TF easibleSol ←zeros
4: λ←1011
5: Finding the nearest feasible solution for each sensor
6: for j:=SHR to Nslots step PH R do
7: HRF easibleSol (j)←1
8: end for
9: for j:=STto Nslots step PTdo
10: TF easibleSol (j)←1
11: end for
12: for i:Nslots do
13: HDHR (i)←(HRF easibleS ol(i)−S olHR (i))2
14: HDT(i)←(TF easibleSol (i)−SolT(i))2
15: end for
16: S1←Sum(H DH R)
17: S2←Sum(H DT)
18: Return Tikhonov Term
19: TT ikhonov ← − λ×(√S1+√S2)
In order to force the adopted optimization technique to
alleviate the infeasible solutions as much as possible, Tikhonov
is used to penalizing the infeasible solution depending on how
much the solution Xviolates the above-mentioned constraints.
This is achieved by computing the Hamming distance between
Xand XNF . This is because the more violation between X
and XNF , the more the objective function defined in Eq. (1)
will be penalized. Algorithms 1 and 2 depict the computation
of TT ikhonov and Vf inal , respectively.
Algorithm 2 Finding the Final Voltage
Inputs: Sol, tabularData, voltage (V)
Outputs: Vfinal
Initialization
H eightOf T able ←height(tabularData)
∆V←zeros
Finding ∆V
for j:LengthOf Sol do
minIndex ←(Heig htOf T able −4 + sol(j))
maxIndex ←sol(j)
maxV ←V olt at maxIndex
minV ←V olt at minI ndex
if V < minV then
CurrentIndex ←minIndex
P recedingI ndex ←(minIndex −4)
Extrapolation⇒∆V
V←V−∆V
else if V > maxV then
CurrentIndex ←maxIndex
NextIndex ←(maxIndex + 4)
Extrapolation⇒∆V
V←V−∆V
else
Current ←sol(i)
Next ←(sol(i) + 4)
i←4
for i: (Heig htOf T able −4) do
if (Vequals V olt at C urrent orN ext)
then
Found at table⇒∆V
V←V−∆V
Break
else if (Vbetween V olt at Current&Next)
then
Intrapolation⇒∆V
V←V−∆V
Break
else
i←i+ 4
Current ←(sol(i) + i)
Next ←(sol(i) + 4 + i)
end if
end for
Return Final Voltage
Vfinal ←V
In Algorithm 1, HR refers to the heart rate sensor while
Trefers to the temperature sensor. This algorithm takes the
generated solutions for both sensors along with slot numbers at
which each sensor starts to operate, SHR and ST. Depending
on these inputs, the algorithm generates the nearest feasible
solutions for the two sensors then calculates Hamming dis-
tance HD, found in lines 13 and 14, to eventually give out
TT ikhonov as illustrated in Algorithm 1. By iterating through
the generated data set and accumulating the voltage drop (in
ADC values) at the end of each time slot, Algorithm 2 can find
out the final Voltage Vf inal. The returns of both algorithms
formulate the objective function as found in Eq. (1).
C. Flower Pollination Optimization Algorithm
FPA is a nature-inspired optimization technique [26]. Flower
pollination is mainly about the pollen transfer for reproduction.
In nature, pollination can be either abiotic or biotic. Biotic
pollination is the predominant form where pollen is transferred
by a pollinator (insects or animals) through a short distance.
In contrast, the abiotic form occurs with the aid of wind and
water diffusion by which pollen can travel a long distance
[27]. In the context of optimization, biotic pollination is
recognized as global pollination (exploration) process with
pollen-carrying pollinators performing L´
evy flight. On the
other hand, the abiotic pollination form is considered local
pollination (exploitation). The algorithm is illustrated in the
flow chart found in Fig. 5.
Start
Setting FPA parameters: population size n, number of
iterations and value of p
Initializing the population for n
flowers with random solutions
Selecting the best solution according
to a predefined objective function
For iteration t
Rand > p
Exploiting (local pollination) Exploring (global pollination)
Calculating the objective function corresponding to new
solutions
Updating the current best solution
t > number of iterationst = t+1
Getting the best (optimal) solution
End
Yes
No
No Yes
Fig. 5: FPA flowchart.
TABLE I: Biosensor tasks scheduling for different time slots (minutes) with initial capacitor voltage Vadc=16710110 (≈5v olt).
Slots Number Solution Final Voltage (×107)
5 3,0,3,0,1 1.6698 (≈4.9764 volt)
10 3,0,3,0,3,2,3,1,0,1 1.6681 (≈4.9713 volt)
15 3,2,2,1,2,0,3,2,3,1,3,1,3,1,1 1.6659 (≈4.9648 volt)
20 3,1,1,0,0,0,2,1,3,0,3,1,2,0,0,1,1,2,3,2 1.6663 (≈4.9660 volt)
IV. RESULTS AND DISCUSSIONS
The experiment are conducted with the following parameter
setting; Nslots =5,PHR =2,PT=2,n=20,p=0.8,λ
=1011, and 1000 FPA iterations. The initial capacitor ADC
voltage Vadc =16710110 (corresponding to 5Volt), the best
solution obtained is {3,0,3,0,1}Dec. which is equivalent to
{11,00,11,00,01}Bin. . For the specified PHR and PT, the
optimum solution is expected to be {11,00,11,00,11}Bin.
which means that the generated solution violates the optimal
one in only one value. This gives intuition that the FPA along
with the tikhonov regularization-based objective function are
capable of generating optimal solutions.
time (slots)
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
capacitor ADC voltage
×107
1.6696
1.6698
1.67
1.6702
1.6704
1.6706
1.6708
1.671
1.6712
FPA solution
(a)
time (slots)
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
capacitor ADC voltage
×107
1.6692
1.6694
1.6696
1.6698
1.67
1.6702
1.6704
1.6706
1.6708
1.671
1.6712
Optimal solution
(b)
Fig. 6: The voltage drop across the supercapacitor throughout the five slots.
Fig. 6 depicted the voltage drop across the supercapacitor
due to the schedule obtained from the FPA-based approach
and the optimum solution. It is obvious that both solutions
cause the same voltage drop across the supercapacitor except
in the 5th slot. According to the operation constraints, the
temperature sensor was supposed to operate during the 5th
slot, making an additional voltage drop across the superca-
pacitor. However, the FPA solution turns it off, causing a
relatively flattened voltage drop in the 5th time slot. For
visualization, Fig. 7 shows the tasks schedule of the two bio-
sensors throughout five time-slots according to the solution
generated using FPA. Table 1 listed the generated solutions
and the final ADC voltage across the superconductor for 5,
10,15, and 20 slots. Fig. 8 tackles the voltage drop across the
superconductor for the different time-slots.
12 3 4 5
Time Slots
Scheduling Solution
Heart Rate Sensor Temperature Sensor
ON
OFF
ON
OFF OFF
ON
OFF
ON
OFF
ON
Fig. 7: The scheduling solution of FPA for five slots.
time (slots)
0 2 4 6 8 10 12 14 16 18 20
Final ADC voltage
×107
1.665
1.666
1.667
1.668
1.669
1.67
1.671
1.672
5 slots
10 slots
15 slots
20 slots
Fig. 8: Task Scheduling for different slots.
Fig. 9 depicts the effect of the violations between the
generated and the expected solutions on the final ADC voltage
value, which was constructed. It shows the final ADC voltage
across the supercapacitor in the case of the optimal solution
and the FPA solution for several experiments with a different
number of slots. In the context of power saving, the solution
that results in a higher final voltage across the superconductor
is supposed to be the preferable one. However, this is true if
there are no operation constraints; thus, the value of the final
voltage across the capacitor along with satisfying the operation
constraints are the concerns of this approach.
No. of slots
5 10 15 20
Final ADC voltage
×107
1.663
1.664
1.665
1.666
1.667
1.668
1.669
1.67
FPA
Optimal
Fig. 9: This bar chart shows the final ADC voltage across the supercapacitor
by experimenting different time slots.
V. CONCLUSION
Energy harvesting-based wearable biomedical devices have
a stringent power budget. Accordingly, power-saving and
management are crucial in such devices to maintain a stable
and continuous operation. In this paper, we have introduced
a novel power-aware task scheduling for energy harvesting-
based wearable biomedical devices. The system of this ap-
proach contains two biosensors: heart rate and temperature
sensors. By formulating a multi-device problem, we were able
to optimize the energy consumption throughout the device.
A proposed task scheduling technique based on FPA was
proposed taking into account low power operation constraints.
The experiments showed promising output results, indicating
that task scheduling can be extended further in future work
with different optimization techniques and more sensors.
ACKNOWLEDGMENT
This research was supported by Nile University.
REFERENCES
[1] C. Glaros and D. I. Fotiadis, “Wearable devices in healthcare,” in
Intelligent paradigms for healthcare enterprises. Springer, 2005, pp.
237–264.
[2] M. Ghamari, B. Janko, R. S. Sherratt, W. Harwin, R. Piechockic, and
C. Soltanpur, “A survey on wireless body area networks for ehealthcare
systems in residential environments,” Sensors, vol. 16, no. 6, p. 831,
2016.
[3] Y.-W. Chong, W. Ismail, K. Ko, and C.-Y. Lee, “Energy harvesting for
wearable devices: A review,” IEEE Sensors Journal, vol. 19, no. 20, pp.
9047–9062, 2019.
[4] A. P. Turner, “Biosensors: sense and sensibility,” Chemical Society
Reviews, vol. 42, no. 8, pp. 3184–3196, 2013.
[5] F. Scheller and F. Schubert, Biosensors. Elsevier, 1991.
[6] T. Cass, Biosensors. Oxford University Press, 2004, no. 268.
[7] R. Hesham, A. Soltan, and A. Madian, “Energy harvesting schemes
for wearable devices,” AEU-International Journal of Electronics and
Communications, p. 153888, 2021.
[8] S. Khalifa, G. Lan, M. Hassan, A. Seneviratne, and S. K. Das, “Harke:
Human activity recognition from kinetic energy harvesting data in
wearable devices,” IEEE Transactions on Mobile Computing, vol. 17,
no. 6, pp. 1353–1368, 2017.
[9] H. Li, C. Han, Y. Huang, Y. Huang, M. Zhu, Z. Pei, Q. Xue, Z. Wang,
Z. Liu, Z. Tang et al., “An extremely safe and wearable solid-state
zinc ion battery based on a hierarchical structured polymer electrolyte,”
Energy & Environmental Science, vol. 11, no. 4, pp. 941–951, 2018.
[10] J. Park, G. Bhat, A. Nk, C. S. Geyik, U. Y. Ogras, and H. G. Lee,
“Energy per operation optimization for energy-harvesting wearable iot
devices,” Sensors, vol. 20, no. 3, p. 764, 2020.
[11] M. M. Sandhu, S. Khalifa, R. Jurdak, and M. Portmann, “Task schedul-
ing for energy harvesting-based iot: A survey and critical analysis,” IEEE
Internet of Things Journal, 2021.
[12] M. R. Sarker, S. Julai, M. F. M. Sabri, S. M. Said, M. M. Islam,
and M. Tahir, “Review of piezoelectric energy harvesting system and
application of optimization techniques to enhance the performance of
the harvesting system,” Sensors and Actuators A: Physical, vol. 300, p.
111634, 2019.
[13] S. W. Arms, C. Townsend, D. Churchill, J. Galbreath, and S. Mundell,
“Power management for energy harvesting wireless sensors,” in Smart
Structures and Materials 2005: Smart Electronics, MEMS, BioMEMS,
and Nanotechnology, vol. 5763. International Society for Optics and
Photonics, 2005, pp. 267–275.
[14] A. Kansal, J. Hsu, S. Zahedi, and M. B. Srivastava, “Power management
in energy harvesting sensor networks,” ACM Transactions on Embedded
Computing Systems (TECS), vol. 6, no. 4, pp. 32–es, 2007.
[15] C. M. Vigorito, D. Ganesan, and A. G. Barto, “Adaptive control of
duty cycling in energy-harvesting wireless sensor networks,” in 2007
4th Annual IEEE communications society conference on sensor, mesh
and ad hoc communications and networks. IEEE, 2007, pp. 21–30.
[16] C. Moser, J.-J. Chen, and L. Thiele, “Dynamic power management in
environmentally powered systems,” in 2010 15th Asia and South Pacific
Design Automation Conference (ASP-DAC). IEEE, 2010, pp. 81–88.
[17] Y. Luo, J. Zhang, and K. B. Letaief, “Optimal scheduling and power
allocation for two-hop energy harvesting communication systems,” IEEE
Transactions on Wireless Communications, vol. 12, no. 9, pp. 4729–
4741, 2013.
[18] R. C. Hsu, C.-T. Liu, and H.-L. Wang, “A reinforcement learning-
based tod provisioning dynamic power management for sustainable
operation of energy harvesting wireless sensor node,” IEEE Transactions
on Emerging Topics in Computing, vol. 2, no. 2, pp. 181–191, 2014.
[19] F. A. Aoudia, M. Gautier, and O. Berder, “Fuzzy power management for
energy harvesting wireless sensor nodes,” in 2016 IEEE International
Conference on Communications (ICC). IEEE, 2016, pp. 1–6.
[20] C. Qiu, Y. Hu, Y. Chen, and B. Zeng, “Lyapunov optimization for energy
harvesting wireless sensor communications,” IEEE Internet of Things
Journal, vol. 5, no. 3, pp. 1947–1956, 2018.
[21] CC2640R2F SimpleLink™ Arm® Cortex®-M3 Bluetooth®
low energy Wireless MCU datasheet. Texas Instru-
ments. [Online]. Available: https://www.ti.com/product/CC2640R2F?
keyMatch=CC2640R2Ftisearch=search-everythingusecase=GPN (Last
accessed: Sept. 3rd, 2021)
[22] Infra Red Thermometer datasheet. Melexis. [Online]. Avail-
able: https://www.melexis.com/en/product/MLX90614/Digital-Plug-
Play-Infrared-Thermometer-TO-Can (Last accessed: Sept. 3rd, 2021)
[23] Heart rate sensor datasheet. World Famous Electronics llc. [Online].
Available: https://pulsesensor.com (Last accessed: Sept. 3rd, 2021)
[24] H.-X. Dou, H.-B. Li, Q.-Y. Fan, and Y.-C. Chen, “Signal restoration
combining modified tikhonov regularization and preconditioning tech-
nology,” IEEE Access, vol. 5, pp. 24 275–24 283, 2017.
[25] C. Rao and V. N. Gudivada, Computational analysis and understanding
of natural languages: principles, methods and applications. Elsevier,
2018.
[26] X.-S. Yang, “Flower pollination algorithm for global optimization,”
in International conference on unconventional computing and natural
computation. Springer, 2012, pp. 240–249.
[27] B. J. Glover, Understanding flowers and flowering. Oxford University
Press, 2007.
