See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/370939627

A Systematic Literature Review of TinyML for Environmental Radiation

Monitoring System

Conference Paper · April 2023

DOI: 10.2991/978-94-6463-134-0_44

CITATION READS

1 113

3 authors:

Istofa Istofa Prawito Prajitno

National Research and Innovation Agency of Indonesia University of Indonesia
6 PUBLICATIONS 22 CITATIONS 108 PUBLICATIONS 294 CITATIONS

SEE PROFILE SEE PROFILE

I Putu Susila
Badan Riset dan Inovasi Nasional
43 PUBLICATIONS 144 CITATIONS

SEE PROFILE

All content following this page was uploaded by I Putu Susila on 22 May 2023.

The user has requested enhancement of the downloaded file.

 A Systematic Literature Review of TinyML
for Environmental Radiation Monitoring System

Istofa1,2(B) , Prawito Prajitno1 , and I. Putu Susila2

1 Department of Physics, Faculty of Mathematics and Natural Sciences, Universitas Indonesia,
Depok 16424, West Java, Indonesia
[email protected], [email protected]
2 Research Center for Radiation Detection and Nuclear Analysis Technology, Research

Organization for Nuclear Energy, National Research and Innovation Agency, Tangerang Selatan,
Banten, Indonesia

Abstract. Tiny machine learning (TinyML) is an important and growing field, but
academic research related to the term is still at a very early stage at this time. As
a result, efforts to synthesize TinyML research into broad knowledge integration
are relatively limited. Therefore, the development of tinyML in the environmen-
tal radiation monitoring system is carried out in stages and relies on the latest
developments in tinyML technology. The application of TinyML for environmen-
tal radiation monitoring systems has not been used or has not been discovered
in a published paper. To meet the needs of developing systems using tinyML,
we started by conducting a systematic literature review (SLR). Until now there
is no direct literature related to tinyML and environmental radiation monitoring
systems, so SLR focuses on the literature on developing tinyML technology. The
review results obtained are expected to meet the requirements and specifications of
the developed system. In particular, this article contributes to the TinyML literature
by synthesizing current research on the following aspects: scientific publication
trends, hardware, and use case experiment sets. This review is carried out by select-
ing papers from journals and proceedings in scientific databases, namely Scopus,
ScienceDirect, IEEE Xplore, and Web of Science. The process of selecting papers
using special keyword strings for each database, using inclusion and exclusion
criteria, and extracting data using Mendeley and Microsoft Excel. Paper selection
resulted in 34 selected papers.

Keywords: TinyML · machine learning · systematic review · radiation

monitoring

1 Introduction
Environmental radiation monitoring systems, especially around nuclear facilities, are
needed to ensure the protection and safety of the public. An online radiation monitoring
system that continuously detects the presence of radioactive material in the environment
is an important system for the rapid detection of radioactive releases into the air. An
IoT-Based Meteorological and Environmental Gamma Radiation Monitoring System

© The Author(s) 2023

T. P. Syawitri et al. (Eds.): MEST 2022, AER 222, pp. 461–473, 2023.
https://doi.org/10.2991/978-94-6463-134-0_44
462 Istofa et al.

have been developed and tested around the Nuclear Facility Engineering Center at the
Serpong Nuclear Complex [1]. Furthermore, the system development uses machine
learning to detect nuclear pollution in an open environment using the Support Vector
Machine algorithm and Linear Discriminant Analysis [2].
Machine learning has become an indispensable part of the existing technology
domain. Edge computing and the Internet of Things (IoT) together present new opportu-
nities to imply machine learning techniques on resource-constrained embedded devices
at the network edge. Conventional machine learning requires enormous power to pre-
dict scenarios. The TinyML (Tiny Machine Learning) paradigm aims to transform the
majority from high-end traditional systems to low-end clients.
TinyML’s goal is to apply machine learning inference to very low power (under one
milliwatt) and low-cost microcontrollers (MCUs). Through the deployment of various
battery-powered MCUs and streaming applications, TinyML facilitates real-time on-site
data collection, processing, analysis, and interpretation [3]. TinyML is a fast-growing
research area committed to democratizing deep learning for all-pervasive microcon-
trollers (MCUs). Challenged by the constraints on power, memory, and computation,
TinyML has achieved significant advancement in the last few years [4]. The main obsta-
cles currently hindering the growth of the TinyML paradigm are estimated to have four
main aspects, minimum battery energy of 10–100 mAh, a processor clock speed of
10–1000 MHz, a memory of less than 1 MB, and cost [5].

2 Methodology
Literature studies related to TinyML for environmental radiation monitoring systems
have used a systematic literature review method. A systematic literature review guide was
first proposed by Kitchenham and Charters which identifies, assesses, and interprets all
literature findings on a research topic to answer research questions (RQ) [6]. To evaluate
and interpret all available research relevant to a particular research question, topic area,
or phenomenon of interest requires a means of a systematic review. Systematic reviews
aim to provide a fair evaluation of research topics using a reliable, rigorous, and auditable
methodology. The need for systematic reviews arises from the need for researchers to
summarize all available information about some phenomena in a comprehensive and
unbiased manner. It is possible to draw general conclusions about some phenomena
than is possible from individual studies, obtain information on the latest research status,
or as a prelude to further research activities.
The purpose of the literature review paper related to tinyML is to get the latest sta-
tus related to the development of tinyML applications to be applied in environmental
radiation monitoring systems. For that, we need data related to the hardware used, avail-
able memory, power consumption, framework, and algorithms used. The results of this
review will be used for the development of a radiation monitoring system using tinyML.

2.1 Research Questions

We identified three main research questions (RQ) about tinyML as follows: [7].
 A Systematic Literature Review of TinyML 463

1. RQ 1: How many studies related to this research theme, publication trends, citation
analysis, paper distribution, and geographic distribution were last published?
2. RQ 2: What are the applications of tinyML and the main components of microcon-
troller devices among published papers?
3. RQ 3: What is the scope of tinyML’s future research for the development of
environmental radiation monitoring systems?

2.2 Scientific Databases for Papers Searching (Search Process)

A literature review was conducted by collecting relevant papers on four scientific

databases, which are Scopus, ScienceDirect, IEEE Xplore, and Web of Science. We
selected only journals and proceeding conference was published during last 10 years
and written in English.

2.3 Study Selection of Papers

At the time this paper was written, there was only one research paper we could find
on the application of tinyML in radiation monitoring systems. This means that tinyML
technology has not been widely used or published, so the search topic is focused on
keywords: tinyML or tiny-ML or tiny machine learning. To search for papers in each
database, a specific keyword string is used, as described below.

4. Scopus: https://www.scopus.com/search/form.uri#basic

Keywords: (TITLE (tinyml) OR TITLE (“tiny-ml”) OR TITLE (“tiny ml”) OR

TITLE (“tiny machine learning”)

5. ScienceDirect: https://www.sciencedirect.com/

Keywords: tinyml OR “tiny-ml” OR “tiny ml” OR “tiny machine learning”

6. IEEE Xplore: https://ieeexplore.ieee.org/Xplore/home.jsp

Keywords: tinyml OR “tiny-ml” OR “tiny ml” OR “tiny machine learning”

7. Web of Science: https://www.webofscience.com/wos/woscc/basic-search

Keywords: (tinyml OR “tiny-ml” OR “tiny ml” OR “tiny machine learning”).

2.4 Inclusion and Exclusion Criteria

Inclusion Criteria:

8. Topic research should focus on TinyML, environmental monitoring, and low power
9. Research tinyML experimental applications or use cases
464 Istofa et al.

Selecon of papers Merger Removing Data Extracon

• Scopus : 76 •Total : 335 • Selected by inclu- • Paper dis-

• ScinceDirect : 52 papers sion and exclusion tribuon
• IEEEXplore : 123 criteria :34 paper • Topic
• WOS : 84 paper

Fig. 1. Papers selection process

10. Uses marketed hardware and open source software

11. Papers published in four database journal (Scopus, ScienctDirect, IEEE Xplore and
Web Of Science)
12. Papers published from 2012 until 2022 and written in English Exclusion criteria:

a. Paper not related this topic research

b. Duplicate papers
c. Other article types such as books, encyclopedias, mini-reviews, and papers
without full text (not open access)

2.5 Data Extraction

Extract data by identifying the occurrence of certain words, topics, or concepts from
the text in the paper. The aim is to integrate various studies into a conceptual map that
describes the development of published papers (publishing trends, distribution, coun-
tries, citations) and the key elements of TinyML (hardware, use cases, frameworks, data
sets, and algorithms/models) and directions for future development. We use Mendeley
reference manager to manage and save selected documents. Data and information were
recorded in Microsoft Excel for analysis.

3 Results and Discussion

To provide a more accountable insight, the selected studies only contained journal liter-
ature and proceedings, other publications were not included in the review. Publications
are selected in two steps: first by reviewing the title, abstract, and keywords, then with a
full-text review. The choice is not limited by the year of publication but papers must be
written in English. All web-based resources can be accessed in print or download form.
Figure 1 shows the results of the initial literature search yielding 335 results. After
reviewing titles, abstracts, and keywords, and deleting duplicates, we obtained 81 papers
for further evaluation. After applying the inclusion and exclusion criteria while reading
full-text, we obtained a final total of 34 papers (15 journals and 19 proceedings) used
for data extraction.
 A Systematic Literature Review of TinyML 465

Fig. 2. Publication trend line (a) and author’s country distribution of the paper publications (b)

3.1 Publication Trends and Author’s Country Distribution of the Paper

Publications

Figure 2 (a) shows the distribution of papers from 2019 to 2022. During the last 10 years,
no paper related to tinyML was found before 2019. This is natural because tinyML tech-
nology was introduced in 2019 and is developing very quickly. Based on the publication
trend, shows a very significant increase as technology is increasingly developed and
many researchers are interested. Figure 2 (b) shows a graph of the author’s country dis-
tribution of the paper publications. The data show that researchers are evenly distributed
around the world. This shows that this technology was not developed by one country
and shows the enthusiasm of many researchers around the world. Italy is the country
with the most published papers. Almost all continents are also involved and actively
contribute to the development of this tinyML technology.

3.2 Publication Papers Distribution Among Journals and Conferences

Studies on TinyML have been published in various journals and conferences. From
34 papers selected consisted of 15 papers were published in journals and 19 papers
were published in conferences. Figure 3 (a) is the number of papers published at each
conference. The research paper was published in 16 proceedings. Figure 3 (b) is the
number of papers published in each journal. The research paper was published in 13
journals.

3.3 Citation Analysis

One method to assess the level of expertise of a researcher is by the number of papers pro-
duced and cited by other researchers. The more papers that are cited by other researchers,
the more information from those papers is needed by others to conduct research. Sco-
pus platform is used for citation analysis, used to analyze the number of citations in
this review. Table 1 shows the top 5 papers cited for research on this topic. Miguel de
Prado’s paper entitled Robustifying the Deployment of tinyML Models for Autonomous
mini-vehicles was cited by most of the other researchers with a total of 8 citations.
466 Istofa et al.

Fig. 3. Number of papers published in each proceeding (a) and number of papers published in
each journal (b)

Table 1. Citation papers

Title papers Year Citations

Robustifying the Deployment of tinyML Models for Autonomous 2021 8
mini-vehicles
An Unsupervised TinyML Approach Applied for Pavement Anomalies 2021 7
Detection Under the Internet of Intelligent Vehicles
A TinyML Platform for On-Device Continual Learning with Quantized 2021 6
Latent Replays
Adaptive Traffic Control with TinyML 2021 4
An SRAM Optimized Approach for Constant Memory Consumption and 2021 4
Ultra-fast Execution of ML Classifiers on TinyML Hardware

3.4 Main Component of TinyML Among Papers Publication

TinyML is a combination of hardware and software, with very stringent hardware and
software support. TinyML is a rapidly growing research area committed to democratiz-
ing deep learning for all-pervasive microcontrollers (MCUs). Challenged by power con-
straints, memory limitations, and computations, TinyML has made significant progress
in recent years [5]. Currently, for tinyML hardware requirements, several microcon-
trollers are available in the market such as those used for experiments in several papers
that have been reviewed. Table 2 shows some of the hardware and its specifications.

3.5 Use Case Themes in Research Publication

Although TinyML is still in its infancy, there are many use cases that apply TinyML
to solving real-life problems. In this review, we identify use cases, such as anomaly
detection in food supply chain security, fall detection for the elderly using sound tech-
nology, greenhouses temperature forecasting, support elderly care, autonomous driving
 A Systematic Literature Review of TinyML 467

Table 2. Hardware platform to support tinyML

Hardware Processor CPU Flash SRAM Power/Voltage Connectivity Sensor/Connector

Clock (MB) (kB)
(MHz)
Arduino Nano 33 Cortex M4 64 1 256 3,3V, 15 mA/pin UART, SPI, I2C, IMU, microphone,
[8–17] – nRF52840 USB, BLE gesture, light,
proximity,
barometer,
temperature,
humidity
STM32F 32-bit ARM 48 MHz 1 MB 192 KB 3–5 V LQFP100 I/O, Accelerometer,
Discovery Cortex-M4 USB microphone,
[18–23] FPU Core
Arduino Portenta ARM 480 MHz, 16 MB 8 MB 3.7–5 V, WiFi, BLE, Temperature,
H7 Cortex- 240 MHz SDRAM Li-Po cell, 10/100 camera extension
[24] M7, ARM 700mAh Ethernet Phy,
Cortex-M4 USB, MIPI DSI,
GPU MPI D-PHY
Nordic Semi ARM Cortex 64 MHz 192 KB 24 KB 1.7–5 V BLE5,Bluetooth –
nRF52840 DK M4 Li-Po mesh, Thread,
[18, 23, 25] Zigbee,802.15.4,
ANT, 2.4 GHz,
NFC,UART
Raspberry Pi 4B 64-bit ARM 1.5 GHz – 256 KB 3.8–4 W, WiFi, BLE, CSI, Temperature
[26–28] Cortex-A72 3.3–5 V DSI, HDMI,
quad core, USB, Ethernet
Broadcom
BCM2711
ESP-32 32-bit 240 MHz 4 MB 8 MB 3.3 V WiFi, USB, SPI, 2MP camera
[18, 23, 29, 29] ESP32 PSRAM I2C,
UART, BLE
ATSAMD51G19A 32-bit ARM 120 MHz 1 MB 256 kB UART, SPI, I2C, SiPM micro
[30] Cortex-M4F USB, radiation sensor

mini-vehicles, environmental Predictions, and more. We added the use case “Technol-
ogy Upgrade” as another type of use case. Table 3. Shows tinyML use case among paper
publications.

Gamma Radiation Classifier

TinyML is implemented for automatic radiation detection and identification using a
small low power gamma sensor as SiPM (Silicon Photo Multiplier). The TinyML model
is designed using the Edge Impulse cloud platform, adopting a Convolutional Neural
Network (CNN). Gamma source dataset provided by the International Atomic Energy
Agency (IAEA), augmented using MATLAB software. To deploy the model on embed-
ded devices, Edge Impulse offers deployment systems for multiple platforms. In our
case, the ATSAMD51P microcontroller embedded in the WIO terminal was used. The
deployment process includes an optimizer to reduce the amount of memory used by the
program. All the libraries required by the model are also included [30].

Image Classification
The most popular and largely deployed case of TinyML is image classification.
468 Istofa et al.

Table 3. Use case TinyML among papers publications

Use Case References

Gamma radiation classifier [30]
Image classification [24, 27, 31–34]
Gas detection [13, 28, 29, 35]
Voice recognition [10, 12, 20]
Activity detection [26, 37–39]
Anomaly detection [8, 11, 30]
Forecasting [16, 22]
Motor control [21, 40]
Threat prediction [9]
Fingerprint classification [40]
Technology upgrade [15, 17, 18, 23, 37]

Researchers bring Smart farming to the agricultural ecosystem to increase its productiv-
ity. TinyML and LoRaWAN were used to propose an energy-efficient model capable of
detecting fruit to demonstrate the technological capabilities in the agricultural domain
to perform TinyML, a microcontroller equipped with a camera and a LoRaWan commu-
nication module [24]. A new non-destructive system for detecting ripeness and disease
defects in tomatoes, using digital image processing techniques with a minimal num-
ber of features compared to existing systems [33]. In the case of farmland, an animal
disturbance is a major threat to crop productivity, reducing profits for farmers and affect-
ing food security. Raspberry cameras are used to capture images of intruders entering
farmland. The images are analyzed with a machine learning algorithm running on a
Raspberry pi board. Furthermore, conclusions are drawn as the output of the machine
learning algorithm [27].

Gas Detection
Vehicles are the main source of air pollution in big cities. Improper vehicle condi-
tions can result in excessive CO2 levels and produce harmful gases. Modern cars are
equipped with multiple sensors, which offer the opportunity to develop algorithms that
can monitor and diagnose vehicle performance more efficiently. TinyML is embedded
into typical OBD-II automotive scanners to function as a soft sensor and estimate carbon
dioxide emissions [35]. The adoption of embedded algorithms for mounting vehicles is a
promising approach, which can significantly benefit smart city macro systems. As smart
vehicles become interconnected, we can better understand how air pollution is gener-
ated and how public actions can be adopted to reduce its negative impact [29]. Another
TinyML application is to detect hazardous gas leaks. The system can be trained to detect
irregularities and notify occupants using BLE technology via messages sent to smart-
phones. This device can be installed in a household environment to warn occupants of
fumes or gases such as Liquefied Petroleum Gas (LPG), when the system is placed in a
 A Systematic Literature Review of TinyML 469

garage. The system is based on TinyML BLE, creating an autonomous system that does
not require an internet connection, access to the cloud for data processing and alerting,
or communication with other devices [13].

Voice Recognition
TinyML can be implemented to classify words in audio data. To find words in raw audio
data continuously requires pre-processing, unlike image classification. Audio recogni-
tion is an application that requires a lot of raw data pre-processing, but the network
model used for classification cannot accept such large input data. Therefore, the fea-
tures are extracted using several complex preprocessing steps [20]. The human voice
can be recognized with this tinyML technology. Everyone has a different sound fre-
quency, amplitude, and pronunciation time. Therefore, it is necessary to look for the
unique characteristics of the target word in the voice data series. TinyML technology
can be implemented to determine coarse age with speech recognition. Research has been
carried out by building a TinyML dataset and a model for rough age classification based
on voice commands. Data points undergo pre-processing by the Audacity app before
being uploaded to the Edge Impulse platform for labeling [12].

Activity Detection
TinyML based optimized real-life solution applications are extensively developed.
Human Activity Recognition (HAR) and fall detection have been hot research topics
since the advent of IoT along with the integration of various types of low-cost sensors
in smartphones and other everyday gadgets. The older we get, the weaker our physical
strength becomes. The need for fall detection devices for the elderly is becoming increas-
ingly important. A convolutional neural network model was developed and trained to
study fall detection using voice technology so that the model can have autonomic recog-
nition capabilities for elderly falls from other sounds [38]. Another paper proposes the
use of accelerometer sensors in a non-invasive way to detect falls to support elderly care.
Using deep learning edge, machine learning based detection system using BLE sensor
network. Machine learning models can run in real-time on edge devices using Edge
Impulse [25]. In the case of electric vehicles, by utilizing a connected personal mobility
vehicle, a Machine Learning (ML)-based fall detection system is introduced on a new
device that analyzes data taken from various sensors integrated into the On Board Unit
(OBU) prototype [26].

Anomaly Detection
Performance Extreme industrial environments present harsh operating conditions and
limited resources, which are further exacerbated by the reduced computing capabilities
of most IoT devices. An ML-based anomaly detection system has been developed that
uses retrofit kits based on limited and cost- effective IoT devices [29]. In the area of
food supply chain security, anomaly detection systems can be developed to detect and
report when abnormalities are found throughout the supply chain. This abnormality can
be caused by a defective food product, incorrect operation of the machine, or an attempt
to tamper with the monitoring device [8].

Forecasting
A statistical approach can be designed to perform weather forecasting activities using
470 Istofa et al.

machine learning (ML) techniques. The recent growing interest in the Edge Computing
Tiny Intelligent architecture suggests a shift towards implementing ML algorithms on
Tiny Embedded Systems. Deep but Tiny Neural Networks (DTNN) can be designed
to be cost-effective and can be automatically converted into STM32 microcontroller-
optimized C libraries via the X-CUBE-AI toolchain [22]. Another approach using
tinyML focuses on using machine learning (ML) algorithms to forecast greenhouse
temperatures [16].

Motor Control
It has been developed utilizing the compact, high-throughput tinyCNN family to con-
trol the mini-vehicle, which learns in the target environment by imitating the expert’s
computer vision algorithm [21].

Threat Prediction
The increasing use of tinyML also poses a major challenge, namely security. To keep
the security aspect guaranteed, it is necessary to implement a Cyber Threat Intelligence
Platform on the Internet of Things (IoT) using the TinyML approach. In addition, if
we design a threat prediction model using the TinyML platform, the integration in the
microcontroller and subsequently in the smart device will be smoother and more energy
efficient. A model has been designed based on a machine learning approach that uses a
Naive Bayes classifier to extract potential threat intelligence from heterogeneous data
sources and the model can predict threat incidents accurately [15].

Fingerprint Classification
Electrical loads are monitored by analyzing changes in voltage and current measured
at the connection points of household power plugs. The idea is to use their unique
fingerprint to identify and classify active equipment. This helps to separate the total
power consumption and the individual power consumption [40].

Technology Upgrade
The topic of the paper is more towards improving tinyML technology than using tinyML
technology in real life.

3.6 Future Research on the Development TinyML for Environmental Radiation

Monitoring System
TinyML broadly covers the area of machine learning technologies that can be used to
analyze sensor data on devices with very low power consumption. Between hardware
advancements and the recent innovations of the TinyML community in the machine
learning field, it is now possible to execute increasingly complex deep learning mod-
els (the basis for most modern applications of artificial intelligence) directly on the
microcontroller.
The results of the review obtained from this SLR, there are several possibilities for
the application of tinyML in the environmental radiation monitoring system. Aspects
that can be developed are related to radiation counting systems, data transmission, and
device management (anomaly detection).
 A Systematic Literature Review of TinyML 471

The next study is the application of TinyML for environmental radiation monitoring
systems, the system created requires handling with machine learning embedded in the
microcontroller. Parts that require handling machine learning include radiation detection
systems and device management systems. The machine learning method was chosen as
the element analysis method because according to some literature it has a fast element
detection rate with high accuracy. The monitoring station is installed independently and
uses a solar panel power supply, so a device management system is needed that can detect
any anomalies or malfunctions during the operation of the device. With the system to
be developed, it is hoped that from the detection results, the system can report or repair
itself on anomalies that occur.

4 Conclusions

TinyML is an important and rapidly growing field that requires interchange between
various integral components (hardware, software, machine learning algorithms). In this
paper, we contribute a systematic literature review that draws on the results of synthe-
sizing data from 34 publications on TinyML since 2019. We focus on five elements:
hardware, frameworks, data sets, use cases, and algorithms/models. From the results
of this literature review, there is a possibility that tinyML can be used to be applied to
the environmental radiation monitoring system that will be developed. Furthermore, the
selection of hardware and algorithms to be used can be done.

References
1. I.P. Susila, Istofa, G. Kusuma, Sukandar, and I. Isnaini, in AIP Conf. Proc. (American Institute
of Physics Inc., 2018).
2. G. Kusuma, R.M. Saryadi, S.K. Wijaya, S. Soekirno, P. Prajitno, and I.P. Susila, AIP Conf.
Proc. 2381, (2021).
3. G. Signoretti, M. Silva, P. Andrade, I. Silva, E. Sisinni, and P. Ferrari, Sensors 21, (2021).
4. H. Ren, D. Anicic, and T.A. Runkler, Proc. Int. Jt. Conf. Neural Networks 2021-July, (2021).
5. P.P. Ray, J. King Saud Univ. - Comput. Inf. Sci. 34, 1595 (2022).
6. B. Kitchenham, O. Pearl Brereton, D. Budgen, M. Turner, J. Bailey, and S. Linkman, Inf.
Softw. Technol. 51, 7 (2009).
7. A. Tris Susanto, P. Prajitno, S. Kusuma Wijaya, S. Soekirno, and I. Putu Susila, J. Phys. Conf.
Ser. 1528, (2020).
8. V. Tsoukas, A. Gkogkidis, A. Kampa, G. Spathoulas, and A. Kakarountas, Inf. 13, (2022).
9. R.R. Prasanna, P. Chowdary Kakarla, V.P. Simha, and N. Mohan, J. Xi’an Univ. Archit.
Technol. XIV, 160 (2022).
10. N.L. Giménez, M.M. Grau, R.P. Centelles, and F. Freitag, Electron. 11, (2022).
11. M. Silva, G. Signoretti, T. Flores, P. Andrade, J. Silva, I. Silva, E. Sisinni, and P. Ferrari,
in 2022 IEEE Int. Work. Metrol. Ind. 4.0 IoT, MetroInd 4.0 IoT 2022 - Proc. (Institute of
Electrical and Electronics Engineers Inc., 2022), pp. 408–413.
12. A.D.I.A. Kadir, A. Al-Haiqi, and N.M. Din, in 15th IEEE Malaysia Int. Conf. Commun.
Emerg. Technol. IoT 5G, MICC 2021 - Proc. (Institute of Electrical and Electronics Engineers
Inc., 2021), pp. 75–80.
472 Istofa et al.

13. A. Gkogkidis, V. Tsoukas, S. Papafotikas, E. Boumpa, and A. Kakarountas, in 2022 11th Int.
Conf. Mod. Circuits Syst. Technol. MOCAST 2022 (Institute of Electrical and Electronics
Engineers Inc., 2022).
14. P. Andrade, I. Silva, G. Signoretti, M. Silva, J. Dias, L. Marques, and D.G. Costa, in 2021
IEEE Int. Work. Metrol. Ind. 4.0 IoT, MetroInd 4.0 IoT 2021 - Proc. (Institute of Electrical
and Electronics Engineers Inc., 2021), pp. 642–647.
15. A. Dutta and S. Kant, in Int. Conf. Electr. Comput. Commun. Mechatronics Eng. ICECCME
2021 (Institute of Electrical and Electronics Engineers Inc., 2021).
16. M.F. Alati, G. Fortino, J. Morales, J.M. Cecilia, and P. Manzoni, in Proc. - IEEE Consum.
Commun. Netw. Conf. CCNC (Institute of Electrical and Electronics Engineers Inc., 2022),
pp. 691–694.
17. K. Kopparapu, E. Lin, J.G. Breslin, and B. Sudharsan, in 2022 IEEE Int. Conf. Pervasive
Comput. Commun. Work. Other Affil. Events, PerCom Work. 2022 (Institute of Electrical
and Electronics Engineers Inc., 2022), pp. 79–81.
18. B. Sudharsan, J.G. Breslin, and M.I. Ali, IEEE Internet Things J. 9, 15007 (2022).
19. D. Pi˛atkowski and K. Walkowiak, Appl. Sci. 12, 1 (2022).
20. J. Kwon and D. Park (2021).
21. M. de Prado, M. Rusci, R. Donze, A. Capotondi, S. Monnerat, L. Benini, and N. Pazos, in
Proc. - IEEE Int. Symp. Circuits Syst. (Institute of Electrical and Electronics Engineers Inc.,
2021).
22. F. Alongi, N. Ghielmetti, D. Pau, F. Terraneo, and W. Fornaciari, in Proc. - 2020 IEEE Int.
Conf. Smart Comput. SMARTCOMP 2020 (Institute of Electrical and Electronics Engineers
Inc., 2020), pp. 350–355.
23. B. Sudharsan, P. Yadav, J.G. Breslin, and M. Intizar Ali, in Proc. - 2021 IEEE Int. Conf. Serv.
Comput. SCC 2021 (Institute of Electrical and Electronics Engineers Inc., 2021), pp. 319–328.
24. C. Nicolas, B. Naila, and R.C. Amar, in Int. Conf. Ubiquitous Futur. Networks, ICUFN (IEEE
Computer Society, 2022), pp. 256–259.
25. K. Tjonck, C.R. Kancharla, J. Vankeirsbilck, H. Hallez, J. Boydens, and B. Pang, 2021 30th
Int. Sci. Conf. Electron. 2021 - Proc. (2021).
26. R. Sanchez-Iborra, L. Bernal-Escobedo, J. Santa, and A. Skarmeta, Comput. Mater. Contin.
71, 3869 (2022).
27. K. Balakrishna, F. Mohammed, C.R. Ullas, C.M. Hema, and S.K. Sonakshi, Glob. Transitions
Proc. 2, 169 (2021).
28. M.F. Yousuf, T. Siddique, and M.S. Mahmud, in 2022 IEEE Microelectron. Des. Test Symp.
MDTS 2022 (Institute of Electrical and Electronics Engineers Inc., 2022).
29. P. Andrade, I. Silva, M. Silva, T. Flores, J. Cassiano, and D.G. Costa, Sensors 22, (2022).
30. M. Antonini, M. Pincheira, M. Vecchio, and F. Antonelli, in 2022 IEEE Int. Work. Metrol.
Ind. 4.0 IoT, MetroInd 4.0 IoT 2022 - Proc. (Institute of Electrical and Electronics Engineers
Inc., 2022), pp. 397–402.
31. M. Altayeb, M. Zennaro, and E. Pietrosemoli, Nucl. Eng. Technol. (2022).
32. G. Datta, S. Kundu, Z. Yin, R.T. Lakkireddy, J. Mathai, A.P. Jacob, P.A. Beerel, and A.R.
Jaiswal, Sci. Rep. 12, 14396 (2022).
33. L. Ravaglia, M. Rusci, D. Nadalini, A. Capotondi, F. Conti, and L. Benini, IEEE J. Emerg.
Sel. Top. Circuits Syst. 11, 789 (2021).
34. S. Dhakshina Kumar, S. Esakkirajan, S. Bama, and B. Keerthiveena, Microprocess. Microsyst.
76, (2020).
35. A. Raha, S. Ghosh, D. Mohapatra, D.A. Mathaikutty, R. Sung, C. Brick, and V. Raghunathan,
in Proc. - IEEE Int. Conf. Comput. Des. VLSI Comput. Process. (Institute of Electrical and
Electronics Engineers Inc., 2021), pp. 13–16.
36. T. Flores, M. Silva, P. Andrade, J. Silva, I. Silva, E. Sisinni, P. Ferrari, and S. Rinaldi, in
(Institute of Electrical and Electronics Engineers (IEEE), 2022), pp. 18–23.
 A Systematic Literature Review of TinyML 473

37. M. Giordano, N. Baumann, M. Crabolu, R. Fischer, G. Bellusci, and M. Magno, IEEE Trans.
Instrum. Meas. 71, (2022).
38. K. Tjonck, C.R. Kancharla, J. Vankeirsbilck, H. Hallez, J. Boydens, and B. Pang, in 2021
30th Int. Sci. Conf. Electron. 2021 - Proc. (Institute of Electrical and Electronics Engineers
Inc., 2021).
39. K. Fang, Z. Xu, Y. Li, and J. Pan, in Proc. - 11th Int. Conf. Inf. Technol. Med. Educ. ITME
2021 (Institute of Electrical and Electronics Engineers Inc., 2021), pp. 222–225.
40. A. Navaas Roshan, B. Gokulapriyan, C. Siddarth, and P. Kokil, in 2021 Int. Conf. Wirel.
Commun. Signal Process. Networking, WiSPNET 2021 (Institute of Electrical and Electronics
Engineers Inc., 2021), pp. 451– 455.
41. D. Pau, R.D. Marc, and D. Denaro, in 2022 IEEE Int. Work. Metrol. Ind. 4.0 IoT, MetroInd 4.0
IoT 2022 - Proc. (Institute of Electrical and Electronics Engineers Inc., 2022), pp. 149–154.

Open Access This chapter is licensed under the terms of the Creative Commons Attribution-
NonCommercial 4.0 International License (http://creativecommons.org/licenses/by-nc/4.0/),
which permits any noncommercial use, sharing, adaptation, distribution and reproduction in any
medium or format, as long as you give appropriate credit to the original author(s) and the source,
provide a link to the Creative Commons license and indicate if changes were made.
The images or other third party material in this chapter are included in the chapter’s Creative
Commons license, unless indicated otherwise in a credit line to the material. If material is not
included in the chapter’s Creative Commons license and your intended use is not permitted by
statutory regulation or exceeds the permitted use, you will need to obtain permission directly from
the copyright holder.

View publication stats