0% found this document useful (0 votes)
50 views27 pages

Artificial Intelligence of Things (AIoT) for Smart Agriculture

Uploaded by

juan corrales
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
50 views27 pages

Artificial Intelligence of Things (AIoT) for Smart Agriculture

Uploaded by

juan corrales
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 27

Journal of Network and Computer Applications 228 (2024) 103905

Contents lists available at ScienceDirect

Journal of Network and Computer Applications


journal homepage: www.elsevier.com/locate/jnca

Review article

Artificial Intelligence of Things (AIoT) for smart agriculture: A review of


architectures, technologies and solutions
Dalhatu Muhammed a ,∗, Ehsan Ahvar b , Shohreh Ahvar b , Maria Trocan a , Marie-José Montpetit c ,
Reza Ehsani d
a
Institut Supérieur d’Electronique de Paris (ISEP), 10 Rue de Vanves, Issy-les-Moulineaux, 92130, France
b Nokia, 91300, Massy, France
c Iowa State University, Ames, IA, USA
d University of California, Merced, CA, USA

ARTICLE INFO ABSTRACT

Keywords: The Artificial Intelligence of Things (AIoT), a combination of the Internet of Things (IoT) and Artificial
Artificial Intelligence of Things Intelligence (AI), plays an increasingly important role in smart agriculture (SA). AIoT has been adopted in
Smart agriculture many applications including agriculture, such as crop yield estimation, soil and water conservation, pest and
Internet of Things
disease detection and supply chain management. While there are plenty of studies on AIoT applications in
Machine learning
healthcare, smart cities, manufacturing, and transportation, SA still has a small share of the reported research.
Sensors
This paper presents a comprehensive review of the existing literature in AIoT and Federated Learning (FL) for
SA. It identifies current and potential challenges and provides research direction for the future investment in
both academia and industry.

1. Introduction Internet of Things (IoT) and Artificial Intelligence (AI), is providing so-
lutions in SA that combine automatic decision-making with the classic
One of the main goals of the United Nations (UN) is to eradicate IoT sensing and system controls.
hunger by 2030 (Pathan et al., 2020). The growth of the global pop- Agriculture benefits from IoT devices’ ability to sense, process, and
ulation, estimated to reach 9 billion by 2050 and 11 billion by 2100, transmit environmental data, such as vision, acoustic, and ambient
emphasizes the critical need for a more sustainable and efficient food data, including temperature, soil humidity, and nutrition, from usually
production system. At the same time, many challenges, such as climate
vast distributed lands. AI, on the other hand, can be trained to analyze
change leading to high temperatures, water scarcity, soil degradation,
sensor data, improving crop and livestock wellness and management
land use change and environmental pollution (Durai and Shamili, 2022)
by making informed decisions about irrigation, fertilization, and pest
will have a negative impact on the way food is produced worldwide.
Global reduction in agricultural land mass, for cultivation and animal and animal control (Alzuhair and Alghaihab, 2023).
husbandry, due to rising water and desertification are already the cause While storing and analyzing data on a centralized cloud brings many
of food shortages. As a result, one of the major issues facing the world benefits, including cost efficiency and high computing and storage ca-
today and in the future is a looming food crisis (Hu et al., 2022). pabilities, it also is ineffective or even infeasible for some applications
To meet the next decade’s production challenges, Smart Agriculture that generate high volumes of data, require a high level of scalability,
(SA) has been proposed as a promising approach to improve agri- constant network connectivity, and, increasingly, data governance.
cultural yields and increase food production. SA solutions have been Hence as will be seen in the paper, AIoT uses different combinations
shown to improve farm management with soil, weather, crop, and of edge, fog, and cloud computing to achieve its goals.
temperature monitoring leading to water, harvest, energy and supply In particular, because of its requirements, SA may need dedicated
chain management (Katiyar and Farhana, 2021). AIoT architectures and technologies. SA requires a high number of
IoT and AI have already been successfully applied in healthcare, devices to interconnect in real-time for decision-making. While con-
transport and energy management over the past years. And, increas-
necting to the AI training is feasible, relying on a remote facility
ingly, Artificial Intelligence of Things (AIoT), a combination of the

∗ Corresponding author.
E-mail addresses: [email protected] (D. Muhammed), [email protected] (E. Ahvar), [email protected] (S. Ahvar),
[email protected] (M. Trocan), [email protected] (M.-J. Montpetit), [email protected] (R. Ehsani).

https://doi.org/10.1016/j.jnca.2024.103905
Received 8 July 2023; Received in revised form 2 May 2024; Accepted 25 May 2024
Available online 1 June 2024
1084-8045/© 2024 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Fig. 2. Paper general structure.

present some of the existing studies on FL as they pertain to SA. Finally,


we present the existing surveys in SA, especially those related to AIoT
in their corresponding subsections.

2.1. Surveys on AIoT


Fig. 1. Relation (overlap) between AIoT and SA.
There have been recent reviews of AIoT applications by researchers
in the field. Table 2 presents the previous work from three perspectives:
AI, IoT and communications. It can be noted that most of these do
for timely decision-making could be challenging outside of a main not consider all three aspects of AIoT technologies combined. Table 1
metropolitan area. In this case, using edge computing can overcome also shows that there are only a few that mention SA as a potential
the limitations of the centralized cloud when it comes to the distributed application.
data analytics and decision-making that are essential for SA.
In addition, because SA may need different farms to cooperate, 2.2. Surveys on FL in AIoT
emerging AI approaches such as Federated Learning (FL) are required.
FL enables distributed devices to train a shared AI model collabora- Looking at recent research, it can be seen that FL plays an increas-
tively while keeping the training data stored locally for privacy and ingly important role in AIoT architectures and systems. A summary of
providing local automatic decision support (Ahvar et al., 2022). It is some of the existing available reviews on FL is presented in Table 3.
now widely believed that when included in the AIoT toolkit, FL can While there is some research on the role of FL in AIoT solutions in
help to develop more efficient solutions for SA. SA (Yin et al., 2022; Saha et al., 2020), at the time of writing, no survey
FL’s benefits have been proved already in other domains such as paper focused on investigating this topic.
healthcare (e.g., a study by Wassan et al. (2022)). Apart from FL, some
new techniques and technologies such as Digital Twins and Extended 2.3. Survey on AIoT in SA
Reality (XR) showed their future potential usage in SA.
In view of these recent developments, our paper main contributions As mentioned above, after a comprehensive review of the literature,
are as follows: we found only a few AIoT surveys focusing on SA (i.e., Katiyar and
Farhana (2021), Gupta et al. (2020), Mitra et al. (2022) and Yang et al.
• We first review AIoT applications, architectures and technologies. (2021)). The work reported in Katiyar and Farhana (2021) includes
• We collate, review and analyze the existing AIoT-based solutions solutions and technologies, while (Mitra et al., 2022) briefly discussed
for SA. architecture and networks as well as solutions and technologies. Au-
• We collate, review and analyze the existing architectures using thors in Gupta et al. (2020), Yang et al. (2021) focused on SA security
AIoT for SA. and privacy. Adli et al. (2023) focused on a Systematic Literature
• We collate, review and analyze the AIoT technologies have been Review (SLR) for highlighting the increasing trends of AIoT publication
used for SA. in SA. They summarized AIoT application and other AI/IoT enabling
• We finally identify challenges of using AIoT for SA and provide techniques and challenges of AIoT adoption. See Tables 1 and 4 for
research direction for the future. summarizing surveys in AIoT for all domains and specially in SA.

To illustrate our contributions, the intersection of SA and AIoT is 2.4. Research gaps
presented in Fig. 1.
The remainder of the paper is organized as follows: Section 2 The results of our literature review can be summarized in Tables 1
reviews namely previous related or similar surveys. Section 3 presents and 4. It showed that there are only a few surveys on AIoT for SA. Non
the methodology, Section 4 presents some AIoT background, Section 5 of them considered the whole spectrum of AIoT in SA (i.e., architecture,
surveys the AIoT landscape for SA and Section 6 defines challenges and solutions and technologies). In addition, they did not consider details
future directions. Finally, Section 7 summarizes the salient contribu- of networking and communication. Unlike the existing surveys on AIoT
tions of the paper. The complete structure of the paper is available in for SA, we provide a comprehensive survey on AIoT architectures,
Fig. 2. solutions and technologies for SA and dwell on the details of net-
working and communication technology aspects essential to create the
2. Related work underlying infrastructure. Also, we reviewed the related surveys in FL
as it is considered as the technology used in training the distributed ML
In this section, we first review the existing surveys on AIoT. Because models in AIoT and similarly for SA which was not provided in related
of the increasingly close relationship between AIoT and FL, we also surveys.

2
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 1
Existing AIoT surveys articles.
Refs. Application No. of Surveys
Pise et al. (2022), Chen et al. (2021), Pappakrishnan et al. (2021), Smart Healthcare 14
Jain et al. (2021), Qian et al. (2021), Alshehri and Muhammad
(2020), Alshamrani (2021), Amin and Hossain (2020), Chamola
et al. (2020), Durga et al. (2019), Tunc et al. (2021), Ali et al.
(2022) and Kakhi et al. (2022)
Nozari et al. (2022), Salih et al. (2022), Pan and Zhang (2021) and Smart Industries 4
Barton et al. (2022)
Molokomme et al. (2022) Smart Grids 1
Mitra et al. (2022), Katiyar and Farhana (2021), Gupta et al. Smart Agriculture 5
(2020), Yang et al. (2021) and Adli et al. (2023)
Huang et al. (2021), Wazid et al. (2021), Patel et al. (2022) and Smart (security) Management 4
Sigov et al. (2022)
Ghoreishi et al. (2022), Bronner et al. (2021) Smart Business 2
Guo et al. (2022) Smart Transportation 1
Seng et al. (2022), Kuguoglu et al. (2021), Wu et al. (2019), Dong Smart Cities 5
et al. (2021) and Zhang et al. (2023)

Table 2 3.1. Research objectives


Summary of AIoT surveys on technologies.
Ref. AI IoT Communications
Wu et al. (2019) ✓ ✓ The major objective of this research is to investigate solutions and
Zhao et al. (2020) ✓ architectures as well as technologies and challenges of AIoT and FL
Hao et al. (2021) ✓ for SA. To achieve this objective, we start by reviewing the AIoT and
Lin (2021) ✓
FL related surveys as presented in Section 2 to show the trends of the
Pan and Zhang (2021) ✓ ✓
Altalak et al. (2022) ✓ ✓ available surveys across AIoT application domains and briefly introduce
Qazi et al. (2022) ✓ ✓ the available existing literature. As we moved to SA in Section 5 which
Vyas et al. (2022) ✓ ✓
is the main targeted application of this research, we dwell into the
Hashni et al. (2022) ✓ ✓
Guo et al. (2022) ✓ existing AIoT solutions for SA both the academia and industry solutions,
Ahmed et al. (2022) ✓ we considered architectures and technologies specialized for SA.
Esenogho et al. (2022) ✓ ✓ ✓
Our survey ✓ ✓ ✓

AI: Artificial Intelligence; IoT: Internet of things. 3.2. Research questions

This section explains the research as follows:

• Q1. What are the available AIoT-based solutions for SA (academia


and industry)?
• Q2. What are the specialized AIoT-based architectures used in SA?
• Q3. What are the AIoT technologies used in SA?
• Q4. What are the current AIoT challenges in SA and what are the
directions for the future?

3.3. Articles collection procedure

The articles used in this research were collected mostly from rel-
evant and reputable databases such as Elsevier (ScienceDirect (SD)),
IEEE Explorer, ACM Digital Library, MDPI, Springer published from
2018 to 2023. For finding the industrial solution, we relied on the list of
Fig. 3. Summary of the reviewed articles.
companies presented.1 To exhaust all the available search avenues, the
following keywords were used ‘‘AIoT Surveys’’ ‘‘Artificial Intelligence
of Things (AIoT) Applications’’, ‘‘AIoT-based Solutions for Smart Agri-
3. Methodology
culture’’, ‘‘AIoT-based Architecture’’ ‘‘AIoT-based Architecture for SA’’,
‘‘AIoT Technologies’’, ‘‘AIoT Technologies for SA’’, ‘‘Surveys on FL in
This section explains the methodology used in this paper for inves- AIoT ’’ and ‘‘FL for Smart Agriculture’’. We used inclusion and exclusion
tigating the related work including the used keywords, our research criteria for article selection as presented in Table 5. Fig. 3 presents the
objective and questions and the selection and quality assessment cri- statistics related to filtered articles. Similarly, a quality assessment is
teria. This process was carried out in three major steps of planning,
presented in Table 6.
reviewing and reporting. In planning, research objectives and research
questions were formulated, in the reviewing stage articles were selected
based on our selection and quality assessment criteria. Lastly, we
1
reported the findings of our research. https://ausagritech.org/about/what-is-ausagritech/

3
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 3
Existing surveys articles on FL applications.
Ref. Application No. of Surveys
Nguyen et al. (2022), Mahlool and Abed (2022), Pfitzner et al. Smart Healthcare 5
(2021), Antunes et al. (2022) and Ali et al. (2022)
Pham et al. (2021), Zhou et al. (2021), Aledhari et al. (2020) Smart Industries 3
Briggs et al. (2021), Mothukuri et al. (2021), Campos et al. (2021), Smart (security) Management 11
Li et al. (2021), Ghimire and Rawat (2022), Hou et al. (2021), Li
et al. (2022a), Nguyen et al. (2021), Ali et al. (2021), Billah et al.
(2022) and Campos et al. (2021)
Du et al. (2020) Smart Transport 1
Jiang et al. (2020), Zheng et al. (2022, 2021), Pandya et al. (2023) Smart Cities 4
and Ramu et al. (2022)

Table 4
Existing surveys articles on AIoT for SA.
Ref. Title. Year Objective
Gupta et al. Security and Privacy in Smart Farming: 2020 Studied the security and privacy in
(2020) Challenges and Opportunities smart farming ecosystems and outline
the multilayered architecture for
precision farming and present the
security and privacy issues in a
dynamic/distributed cyber–physical
environment. They also highlight
potential cyberattack scenarios.
Yang et al. A survey on SA: Development modes, 2021 This paper surveyed the state-of-the-art
(2021) technologies, and security and privacy work related to smart development
challenges modes, technologies, applications and
privacy and security.
Katiyar and SA: The Future of Agriculture using AI 2021 Present the research work for agriculture
Farhana (2021) and IoT automation using sensors, agricultural
robots and drones as well as AI-driven
technologies to improve productivity.
Mitra et al. Everything you wanted to know about 2022 Presents the solutions, technologies
(2022) SA trends, available datasets, network
options and deployment challenges.
Adli et al. Recent Advancements and Challenges of 2023 Presents AIoT concepts, IoT smart
(2023) AIoT Application in Smart Agriculture: A devices and AI techniques adoption, the
Review trends in increasing publication in AIoT
applications using SLR. Highlighted
application of AIoT and other AI/IoT
enabling techniques and challenges of
AIoT adoption.
Our survey Artificial Intelligence of Things (AIoT) 2023 A comprehensive survey of AIoT for SA
for Smart Agriculture: A Review of considering AIoT architecture, solutions
Architectures, Technologies and and technologies. Also, reviewed the
Solutions related surveys in FL as it is considered
as the technology used in training
distributed ML models in AIoT and
similarly highlighted the challenges and
future directions of AIoT for SA.

Table 5
Inclusion/Exclusion method.
Inclusion Criteria
1 Title with all keywords in search string published between 2018 to 2023
2 Title with AIoT plus any of its application domains in the search string
3 Focusing on AIoT-based architecture
4 Focusing on AIoT technologies
5 Focusing on AIoT for SA and any of its 6 solutions
Exclusion Criteria
1 Title that does not include any of the AIoT application domains
2 Articles that are not within the context of the study domain
3 Articles published before 2018 or not peer review
4 Short articles or duplicated articles published by different journals

4
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 6
Quality assessment.
ID Question
Q1 Is the research objective defined clearly in line with the research
problem?
Q2 Are the research questions answered accordingly?
Q3 Does the research focus on AIoT and any of its applications
domains, architectures or technologies?
Q4 Does the research scope include FL for AIoT applications or SA
solutions?
Q5 Is the research report well elaborated and the experiment is clearly
explained?
Q6 Is the research findings valid and relevant to the study domain?

Fig. 5. General AIoT applications.

Fig. 4. AIoT background.

4.1. AIoT applications

4. AIoT background To start our wide-ranging survey, we review existing AIoT applica-
tions and propose a taxonomy where AIoT applications are classified
The combination of AI technologies with the IoT infrastructure into two broad categories: city applications (i.e., smart cities) and
refers to AIoT (Ahvar et al., 2022; Dia et al., 2022). We can consider non-city specific as presented in Fig. 5. In cities, AIoT applications
two main forms for collaborating AI and IoT in AIoT systems: loosely target healthcare, buildings, offices, businesses, transportation, and
coupled (i.e., AI for IoT) and tightly coupled (i.e., AI on IoT). In government (safety and security). More broadly defined AIoT applica-
loosely coupled AIoT (e.g., a traditional combination/collaboration of tions include smart manufacturing, smart grids, and SA. To mention
IoT and AI), the data is generated in IoT devices and forwarded to examples, of each domain, a healthcare AIoT solution for identifying
a local or remote location (e.g., cloud facility) to be analyzed by an persons of interest (POI) in COVID-19 analysis was implemented based
AI-based method such as neural networks. In tightly coupled AIoT on far distance camera with low-resolution handling in Istiklal street
(e.g., a modern combination/collaboration of IoT and AI), IoT devices of Istanbul in Turkey.2 For transport example, we can refer to images
are equipped with AI processing capabilities and AI algorithms are from the autopilot camera which with the help of AI are used to
run on IoT devices partially or even completely. In other words, the automatically convert the wiper speed to the intensity of rain or snow
IoT devices have AI processing capabilities in tightly coupled AIoT. in Tesla (Slama et al., 2023).
The systems that use the loosely coupled form of collaborating AI and
IoT are called loosely coupled AIoT systems and the ones that use the 4.2. AIoT architectures
tightly coupled form are called tightly coupled AIoT systems (Ahvar
et al., 2022). To functionally distribute the needs of AIoT, several layered ar-
Loosely coupled AIoT already is improving traditional farmers prac- chitectures have been recently proposed that include some or all the
tices (e.g., smart dynamic irrigation patterns instead of fixed scheduling following features: devices for data acquisition, an edge/fog layer that
irrigation (Amatya et al., 2016)). Tightly coupled AIoT systems can contains computing devices, for aggregation and pre-processing, and
bring some benefits in comparison to the loosely coupled AIoT systems. gateways for connectivity to the cloud layer for storage and analysis.
In traditional loosely coupled AIoT systems, data are generated by IoT In each architecture, the number of layers and the nature of the
devices and should be transmitted to a remote location (e.g., a cloud operations performed at each layer can vary based on the specific
data center) for AI processing. In modern AIoT applications (e.g., many requirements of the targeted system. Because common need for data
smart agriculture applications), because of a high volume of generated gathering and processing, three-layer and four-layer architectures are
data, high scalability of networks and data privacy issues, it may not becoming almost standard.
be possible to transfer data to a remote location.
A broad perspective of AIoT architectures, technologies, and appli-
2
cations is presented in this section and summarized in Fig. 4. https://github.com/vahit19/4DeepAnalytics.com

5
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

4.2.1. Layered models


Generally, the device layer (also called the data acquisition layer)
is where the sensing and measurement systems are located. This layer
may host numerous devices responsible for the generation and collec-
tion of data based on the target application. Different devices including
but not limited to sensors, actuators, and other embedded devices as
well as drones, cameras, smartphones, and radio frequency identifiers
(RFID) may be used as data sources in that layer. They are deployed
in specific areas based on the application requirements for sensing,
processing, and communicating the data amongst themselves and to the
other layers.
The edge/fog layer refers to the next layer that receives the raw or
processed data from the device layer performs aggregation and further
processing, including security and then sends the results to the cloud
layer using some form of communication interface. Although some of
the existing architectures consider the edge/fog layer as one, others
follow the OpenFog Consortium recommendations and make the dis-
tinction between fog and edge. While fog computing provides resources
Fig. 6. AIoT common architecture.
(e.g., computing, storage and networking) anywhere from the cloud
to the end devices (the edge cloud continuum), edge computing may
be limited to computing locally (OpenFogConsortium, 2017; Ramya,
2021). As was stated in Digiteum (2022), ‘‘As a layer in between clouds Depending on the second layer choice, we define 4-layered ar-
and edges, fog offers the advantages of both. It utilizes the cloud and chitectures Fog–Cloud (F–C), Edge–Cloud (EC) and Edge/Fog–Cloud
communicates directly with it to distribute data that does not require (E/F–C).
immediate processing. Fog is also positioned nearer the edge at the In addition, the AI mechanisms can be centralized or distributed in
same moment. It uses local processing and storage resources for real- the network across every layer, as illustrated in Fig. 7. In a centralized
time analytics and prompt event reaction. Fog is decentralized, with deployment, AI processing occurs only in one of the layers of the
numerous nodes, just like the edge. Fog, in contrast to the edge, has architecture or even in only one device in the edge cloud continuum, as
a network architecture. Fog nodes are interconnected and can disperse shown in Fig. 8. Whereas in a distributed deployment, AI inference can
computing and storage to complete specific tasks more effectively. Edge occur across the devices themselves, with support at the edge/fog and
provides the lowest latency and fastest reaction to data because it is cloud layers in a vertically, horizontally or hybrid manner as shown in
the closest to end devices. The structure in edge computing is typically Fig. 9 (Ahvar et al., 2022).
more loosely coupled, with edge nodes handling data on their own’’. It is in the distributed deployment that FL becomes an asset. In FL,
Either part of the edge or fog layer or as a layer of its own, the ML algorithm training is performed across multiple decentralized edge
connectivity layer relays the data between layers. As an independent devices or servers each holding local data samples. FL trains the model
layer, in some literature, it offers added services such as encryption, from the local data generated and sends the local model update to the
and service and data discovery. server for global aggregation. This improves the model performance
Finally, the cloud layer receives the data from all preceding layers, because the client nodes can use the global modeling results from
performs data analysis and stores the data for future use. The cloud many nodes received from the server to train their local model while
layer enables end-users to access the relevant data virtually for their reducing the network traffic sent to the central server. FL also provides
specific tasks. Some architectures use the cloud as the uppermost layer, improvements in data privacy and security by keeping the data at the
whereas others further separately use the application layer as an upper source.
layer. Similar, to FL, distributed learning is the process of training ML
Examples of specific architecture illustrate the concept. The au- models using multiple computing resources that are interconnected.
thors in Wazid et al. (2021) presented the architecture of AIoT as Rather than relying on a single machine, distributed learning har-
a blockchain-based secure framework consisting of three major com- nesses the collective computational power of a network of machines
ponents: IoT devices/users, gateways and cloud. In this study, in the or nodes. By dividing the workload and data across multiple nodes,
second layer, the gateway nodes convert the huge amount of received distributed learning enables parallel processing, leading to faster and
data from the smart IoT devices into partial blocks and forward them more efficient training of machine learning models (Gülen, 2023).
to the cloud server. The cloud server converts the partial block into However, in centralized learning (CL), environmental row data is re-
the full block and forwards it to the Peer-to-Peer Cloud Server (P2PCS) quired by the clients which are transmitted to the central server after
network for mining and incorporates it with the blockchain to secure initial pre-processing to perform respective model training tasks (heavy
the data. Another layered architecture with a different nomenclature computation) (Drainakis et al., 2020a). There are some studies in
is proposed in Yu et al. (2021). This architecture consists of the data different domains comparing central learning and FL from different
layer, the security layer, the processing layer (with AI modules) and the aspects (Drainakis et al., 2020a,b).
application layer which focuses on securing the data from attacks and Getting help from distributed learning and FL, and depending on
threats by malicious agents preventing their access. At the data layer, applications and required performance, different layered AIoT archi-
data is generated and transmitted to the processing layer through the tectures have been used in related work. A summary of the common
security layer to prevent and recover from attacks. AIoT architectures is presented in Table 7 with their corresponding
Hence, after review, a common architecture emerges as is shown in application areas. Moreover, we identified the level of AI and IoT
Fig. 6. It consists of four layers. In the device layer, end devices are coupling in these architectures. The most common layered architecture
deployed for gathering data related to a particular application domain; studied in our survey is E–C architecture. This choice can be related
the generated data is then sent to the edge/fog layer through the to a lack of standard definition and support for the fog layer. Tightly
connectivity layer for further processing. Finally, the data is transmitted coupled have been already proposed in most of the domains. SA AIoT
to the cloud for processing and storage. architectures will be addressed separately in Section 5.

6
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Fig. 7. AI deployment in AIoT architectures.

Fig. 8. Centralized AI deployments (Ahvar et al., 2022).

Fig. 9. Distributed AI deployment (Ahvar et al., 2022).

7
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 7
General AIoT architectures and systems.
Ref. Architecture NL AIoT Systems Application
E/F–C E–C F–C C-B Loosely Tightly
Zhang and Tao (2020) ✓ 3 ✓ General
Wazid et al. (2021) ✓ 3 ✓ Security
Liu et al. (2021) ✓ 4 ✓ Industry
Yu et al. (2021) ✓ 4 ✓ Security
Ning (2021) ✓ 4 ✓ Smart grid
Seng et al. (2022) ✓ 3 ✓ Smart Homes
Mitra et al. (2022) ✓ 3 ✓ Smart Agriculture

NL: Number of Layers; E/F–C: Edge/Fog–Cloud; E–C: Edge–Cloud; F–C: Fog–Cloud; C-B: Cloud-Based; Loosely: Loosely coupled; Tightly: Tightly
coupled.

Fig. 10. Classification of AI/ML technologies (Wazid et al., 2021).

4.3. AIoT technologies 4.3.2. IoT technologies


IoT technologies are the backbone of AIoT applications as they
In this section, we divide AIoT technologies into AI, IoT and com- enable data acquisition across different types of hardware devices
munications to present a comprehensive overview. (e.g., sensors, smart cameras, RFID, embedded systems, drones, smart-
phones, etc.) and provide the connectivity to enable the devices to com-
4.3.1. AI/ML technologies municate with each other and forward the data to a target destination
AIoT uses extensive ML to implement model-based and data-centric for processing and storage (Chang et al., 2021).
decision-making. There are many classifications of AI/ML encompass-
An IoT devices classification model presented by Cvitić et al. (2021)
ing supervised, unsupervised and re-reinforcement learning. They in-
for classifying traffic features generated by IoT devices and analyzed
clude a wide variety of intelligent applications such as natural lan-
the possibilities of applying those features for classifying IoT devices.
guage processing, speech recognition, virtual agents, decision manage-
ment, bio-metrics, robotic process automation technologies, etc. Fig. 10 IoT technologies can be further divided into three: hardware, soft-
proposed by Wazid et al. (2021) summarizes the different approaches. ware and networking technologies. IoT hardware collects the data.
As an application of distributed ML and as mentioned before, FL IoT software supports data normalization, analysis, manipulation and
enables distributed devices to collaboratively train a shared AI model security as well as, when needed, AI model deployment. Various IoT
while keeping all the training data locally. With the development of software solutions are readily available (Knud, 2019; Satyajit, 2022).
FL and the increasing computing power of edge devices, FL is rapidly IoT connectivity includes the actual networking technologies to connect
becoming an effective solution for data privacy-preserving in the AIoT the devices to the gateways and the cloud. Different IoT hardware and
domain. By locating the computation in edge devices and ensuring that software are used in AIoT and are presented in Fig. 11. Connectivity in
the user’s data does not leave the local area, FL protects user privacy IoT will be explained in the next section (Note that connectivity and
against leakage (Zhang et al., 2020; Yin et al., 2022). communication are used in this paper interchangeably).

8
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Fig. 11. Classification of IoT technologies.

Fig. 12. Communications technologies.

4.3.3. Communications technologies For communications between the end-devices layer and edge/fog
Different communications technologies are used to connect AIoT de- layer, medium-range solutions such as Narrowband Internet of Things
vices to each other, and to edge/fog or cloud nodes. These technologies (NB-IoT), Long Range Wide Area Networks (LoRaWAN) and SigFox,
can be further divided into three main short, medium and long-range and short-range solutions such as Near Field Communication (NFC),
groups as shown in Fig. 12. Table 8 also compares the communications RFID, Bluetooth, Z-Wave, Wireless Fidelity (WIFI), Building Automa-
technologies based on technology standards, data rate, frequency and tion and Control networks (BACnet) and ZigBee are used.
range.
For the literature, the most commonly used communications tech-
Long-range communications technologies including both wired and
nologies well-adapted for SA at the end-devices layer of SA are Lo-
wireless, such as General Packet Radio Service (GPRS), Long-Term
Evolution (LTE), Third Generation (3G), Fourth Generation (4G), Fifth RaWAN, ZigBee and BACnet due to their low power consumption, small
Generation (5G) and soon Six Generation (6G) are mainly used for size, ease of implementation, simplicity, scalability, range (usually
communications between the edge/fog and the cloud. For SA, 5G and medium) and in the case of BACnet heritage from industrial automa-
6G technologies will most likely predominate in the future because tion. In addition, there are some studies to adapt WIFI for SA, such as
of high reliability, low latency, wide communications ranges and high ‘‘WIFI-based long distance (WiLD)’’ proposed by Ahmed et al. (2018)
data transmission capabilities (Qazi et al., 2022). However, deploying for monitoring and controlling smart farming in rural areas because of
these technologies in rural areas is a major challenge due to instal- its low cost.
lation costs (with small cells) and the lack of access to high-speed It is worth mentioning that an investigation and comparison of the
backhauls. In this case, private networks (e.g., private 5G) can already performance of different wireless communications technologies (WIFI,
be used.3 Experimental wireless and long-range optical platforms are LoRaWAN and ZigBee), in terms of energy consumption in agriculture
also considered.4
monitoring, showed that LoRaWAN works well for SA when network
lifetime and energy consumption are priorities (Ray, 2018; Sadowski
3
https://www.uctel.co.uk/blog/private-5g-and-agriculture-the-digital-age and Spachos, 2020). There is also an increasing number of sensors that
4
https://arawireless.org support LoRa and that should enable its penetration in SA.

9
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 8
Comparison of communications technologies.
Tech Stand Range Freq Max.DR
ZigBee ZigBee alliance, IEEE less than 1 km 902–928 MHz, 2.4 GHz 250 Kbps
802.15.4
Z-wave Z-wave 100 m 868 MHz 100 Kbps
WiFi IEEE 802.11 100 m 2.4–60 GHz 10 Mbps
Bluetooth IEEE 802.15.1 100 m 2.45 GHz 1–3 Mbps
RFID Many standards 1 m 13.56 MHz 423 Mbps
NFC ISO /IEC 13157 0.1 m 13.56 MHz 106 kbps–424
Kbps
BACnet ANSI/ASHRAE Standard 1.2 km (RS-485) Depends on physical layer 10 Mbps to 100
135 used (Ethernet, RS-485) Gbps (Ethernet),
9.6 kbps to
115.2 kbps
(RS-485)
LoRa LoRa alliance 15 km 915–928 MHz 50 Kbps
SigFox SigFox Collaboration of 20+km 868 and 915–928 MHz 100 Kbps
ETSI
NB-IoT 3GPP 1 km (urban), 700, 800, 900 MHz 200 Kbps
10 km (rural)
GPRS 3GPP 25 km/10 km GSM 850, 1900 MHz 171 Kbps
LTE 3GPP 28 km/10 km 700–2600 MHz 0.1–1 Gbps
3G/4G UMTS/LTE 26 km/28 km GSM 850, 1900 400.73
MHz/700–2600 MHz Mbps/0.1–1
Gbps
5G IEEE 802.11ac/ITU 28 km 700 MHz–72 GHz 20 Gbps
IMT-2020
6G IEEE 802.11ax 10 THz 1 Tbps

Tech: Technologies; Stand: Standards; Freq: Frequency; Max.DR: Maximum data rate.

5. Smart agriculture (SA) to show the scale of the problem in SA. This farm is using ambient
power-enabled IoT devices, which are either battery-less or with limited
The agricultural evolution started with Agriculture 1.0, with in- energy storage capability (e.g. several hundred micro-watts) getting
digenous tools for farming, then continued to Agriculture 2.0, with energy through the harvesting of radio waves, light, motion, heat
fertilizers and tractors then to Agriculture 3.0, where monitoring sys- or similar sources. The optimal scale considered for this greenhouse
tems and decision systems were introduced, and finally to Agriculture construction is 8 10 m span, 80 100 m length. A single greenhouse
4.0, where smart farming (SA) is currently in use (Liu et al., 2020) area reaches nearly 70,000 square meters, equivalent to ten standard
with automation and embedded-decision systems. As mentioned before football field sizes. A team of eight people controls this greenhouse
AI and IoT are inherently linked to SA. Specific technologies that in front of the computer and knows everything that is going on in
can be included in agriculture 4.0 include robotics, blockchain and the greenhouse. ‘‘For example, the temperature and humidity in each
other cyber-security technologies, drones (Unmanned Aerial Vehicles area of the 70,000 square meters greenhouse, the temperature of the
or UAV), satellites and of course a wide variety of sensors including underground heating tube, the concentration of carbon dioxide are how
novel devices using nanotechnology (Mitra et al., 2022). much, whether the fan is opened, and whether the nutrition is enough
Traditional (conventional) farming of levels 2 and 3 in the evolu- for each tomato’’ (AmbientIoT3GPP, 2023).
tion, in which traditional techniques are used for crop cultivation (Du- We can classify AIoT-based proposed SA solutions into academic
rai and Shamili, 2022) is still dominant in the industry. SA has special and industrial and further categorize the academia solutions into six
requirements such as sensing, processing and connectivity that are still sub-classes as shown in Fig. 13 (Katiyar and Farhana, 2021; Mitra et al.,
not available everywhere. In addition, AIoT solutions designed for other 2022), crop, soil, water and disease management, enhanced harvesting
non-SA use cases may not comply with SA requirements. However, we techniques and supply chain management. These classes can inter-work
have seen that the technologies presented in Section 4.3 can be adapted to provide special use cases. In this section, we analyze each solution
to SA when they meet a wide range of requirements since SA targets in terms of: type, features, ML model if appropriate and data sources.
open-field solutions (with drones for example) as well as greenhouses This is summarized in the form of specific tables in their corresponding
and vertical farms (Controlled Environment Agriculture (CEA)). subsections.
Some studies in SA compared the result of AIoT-based solutions
with traditional solutions called farmer treatment (FMR), which is the 5.1.1. Academic solutions
normal practice of the local farmers (e.g., smart irrigation comparison The details of SA academic research are presented below.
in Amatya et al. (2016)). Crop Management: Crop management is a set of agricultural prac-
While SA may or may not include AI, in this section, we focus on tices to improve growth, development and yield. It begins with seedbed
the AIoT architectures and solutions that have been adapted to SA. preparations, sowing and crop maintenance, including pest and disease
detection, and it ends with harvest, processing, storage and distribu-
5.1. AIoT solutions for smart agriculture (SA) tion. Hence it includes end-to-end crop production activities that are
under the responsibility of the producer.
AIoT solutions in SA may need to support quite large-scale envi- Several crop management systems consider growth monitoring and
ronments. A tomato farm which was referred in the study on Ambient yield prediction (Reddy and Kumar, 2021; Schwalbert et al., 2020;
IoT in 3GPP release 19 (AmbientIoT3GPP, 2023), is a good real case Kumar et al., 2019a) for different types of production such as wheat,

10
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Fig. 13. AIoT solutions for SA.

maize, rice, coffee, tea, cherries, etc. Crop disease detection (scouting) encouraging carbon farming, more and more farmers are engaging in
and remediation is the focus of Jiang et al. (2019), Guo et al. (2020). using this practice (Moinet et al., 2023; Feng et al., 2023; Payen et al.,
Numerous crop recommendation approaches using different models 2023) as presented in Table 10.
are available in Kulkarni et al. (2018), Durai and Shamili (2022). Water Management: Water management in agriculture includes
Crop recommendations are also the focus of many other papers (Doshi practices that cover both the quality and the quantity of water used.
et al., 2018; Kumar et al., 2019b; Suchithra and Pai, 2020; Pawar and As an important component of SA, water management uses smart irri-
Chillarge, 2018; Chougule et al., 2019; Patel and Patel, 2020; Kedlaya gation systems to properly control the use of water (Tomaszewski and
et al., 2021; Patil et al., 2021; Pande et al., 2021; Sadia et al., 2021). Kołakowski, 2023). There are several known use cases in smart water
More recently, a SA soybean crop yield prediction using FL was the management such as AIoT solutions for real-time water quality moni-
topic of Manoj et al. (2022). In this paper, the researchers compared toring (Wang et al., 2021; Chiu et al., 2022; Alahi et al., 2018). In Abi-
the performance of FL, Deep Learning (DL) and other more generic naya et al. (2019), Miao et al. (2022), authors used water analysis to
ML for the prediction of soybean yield. The result showed the overall evaluate fish well-being in a fish-farm operation. Guillén-Navarro et al.
performance of the comparison for the validation set metrics: FL using (2020), focused on optimization of water consumption, using anti-frost
ResNet-16 Regression outperformed ML and DL but at the expense sprinkler irrigation technique. A real-time soil wetness monitoring for
of added complexity. In addition, Durrant et al. (2022) proposed a water management in irrigation is the target of Nursyahid et al. (2019).
data-sharing approach for improving production optimization through A use case for managing hydroponic farming water management is
soybean yield prediction using a variety of ML approaches including developed in the paper of Mehra et al. (2018). The research presented
FL, Convolutional Neural Networks (CNNs), rectified linear activation in Mohammed et al. (2019), Dahane et al. (2020) focused on real-time
function (ReLU), recursive neural networks (RNN), long short-term smart irrigation, while Usmonov and Gregoretti (2017) proposes a drip
memory (LSTM) and multi-layer-perceptron (MLP) on remote sensing irrigation control system for irrigation control. A flood observation and
images, weather and soil data. The results indicated that even simple early warning system using AIoT (Sung et al., 2022) was presented
approaches can give improved measurements compared to manual to effectively forecast floods using a fuzzy algorithm. A special use
evaluation. case, presented in Manikandan et al. (2022), addressed paddy rice
While AI shows promise across all research papers, factors that irrigation by considering the most important growth parameters in rice
have been rarely considered in academic research are the costs of including the crop growth level. Evapotranspiration (ETo) evaluation
investments, the price of the targeted crops and the added revenue was available in the papers of Antonopoulos and Antonopoulos (2017),
generated by the SA operation. Authors of Priyadharshini et al. (2021) Nema et al. (2017), Elbeltagi et al. (2020) and Elbeltagi et al. (2022)
used traditional linear regression (LR) and Neural Networks (NNs) to for requirements, estimation, calculation and prediction. ETo is an
consider the combined effects of the cost of cultivation, crop price mod- important derived parameter needed to evaluate how plants react to
eling, nutrient contents, rainfall and temperature data. They showed the irrigation and environmental humidity. The research focused on
that data-driven and AI approaches in SA can improve revenues by predicting ETo using different ML models. As presented in Table 11,
better planning and forecasting. most of the papers on water and irrigation considered soil moisture,
Soil Management: There are different types of soil and many humidity, temperature, oxygen, water level and pH as important pa-
crops require specific soil types, fertilization and nutrients for optimum rameters. However, Wang et al. (2021) also considered nitrates and
growth and production. There are several published use cases in soil phosphates concentration (Nitrate-Nitrogen (NO3-N), Phosphate (PO4),
management including soil type classification (Vincent et al., 2019), Nitrite-Nitrogen (NO2-N), Ammonia-Nitrogen (NH3-N)) and other pa-
land recommendation focus (Patel and Patel, 2020), soil quality mon- rameters. For open field SA, Antonopoulos and Antonopoulos (2017),
itoring (Murugamani et al., 2022), moisture monitoring (Araya et al., Nema et al. (2017), Elbeltagi et al. (2020) and Elbeltagi et al. (2022)
2020; Bhattacherjee et al., 2020), overall soil health monitoring (Jain considered weather, solar radiation, rain forecast, rain probability and
et al., 2020), salinity estimation (Klibi et al., 2020), nutrient content wind speed also to have accurate models of water consumption.
analysis (Dong et al., 2018), and soil manure composition (Ather et al., In the AI realm, Fog-assisted FL (FogFL) (Saha et al., 2020) proposed
2022; Chang et al., 2019). As part of soil management, carbon farming a smart irrigation scheduling application where sensor nodes monitor
is a new agricultural method by which sequestering atmospheric carbon weather and soil data and communicate the data to the edge devices for
in crop vegetation, soil and biomass reduces atmospheric greenhouse the local model update. The local model is forwarded to the fog node
gas emissions. At the same time, the carbon sequestered in plant ma- for local aggregation and after certain rounds, the cloud application
terial and soil can improve farm productivity and increase soil health selects an optimal fog node, based on a greedy heuristic approach, to
including aiding plant growth, building drought and flood resistance, do the global aggregation. The FogFL framework reduced the global
increasing soil water retention capacity and decreasing the amount of aggregation rounds, communication latency and energy consumption
fertilizers used in the farm. Due to some governmental policies for of the resource-constrained edge devices and increased the reliability of

11
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 9
Crop management solution.
Subsolution Ref. Features ML model Data source Comments
Kumar et al. (2019a) images OpenCV private growth prediction
crop monitoring
Torres-Sanchez et al. temperature MLR, MNLR private predict shelf life
(2020) quality losses
Mehra et al. (2018), Jung water level, pH, ANN, NARX and private action control
et al. (2020) temperature, RNN-LSTM prediction
humidity, and light
intensity
Reddy et al. (2019) depth, texture, pH, CHAID, KNN, Naive private predicting crop to
soil color, water Bayes the farmers
holding
Setiadi et al. (2020) weather, yields, Naive Bayes public (BMKG) suggesting suitable
selling prices crop to farmers
Doshi et al. (2018) temperature and Naive Bayes, SVM, public selecting crop to
crop rainfall RF farmers
recommendation
Kumar et al. (2019b) soil color, pH, SVM,DT, Logistic R public crop
rainfall, temperature recommendation
Patel and Patel (2020) soil types, quality, KNN, SVM private predicting crop to
crop, climate, water the farmer
Pande et al. (2021) crop, year, season, SVM, KNN, ANN, public crop
soil type, area, RF, MLR recommendation
region
Durai and Shamili (2022) pH, rainfall Naıve-B, Logistic-R, public (Kaggle) predict crop,
SVM, DT, RF, KNN, pesticides, weed and
cost
Priyadharshini et al. rainfall, NN and Linear-R public (Kaggle) profit analysis of
(2021) temperature, pH, crops based on the
soil type, NPK and previous data
location
fertilizer recommendation Haban et al. (2020) NPK, season Fuzzy public predict fertilizer
Manoj et al. (2022) crop, soil,weather FL,ResNet- public soybean yield
16,ResNet-28 prediction
yield prediction Durrant et al. (2022) images, weather, FL,CNN, ReLU, private data sharing for
soil RNN, LSTM MLP soybean yield
prediction
iron deficiency prediction Yu et al. (2022) images FL public soybean iron
deficiency chlorosis
prediction

Table 10
Soil management solution.
Subsolution Ref. Features ML model Data source Comments
soil type Vincent et al. soil texture, granular, ANN, MLP private land suitability
classification (2019) water content, degree assessment
saturation, pH, salinity
land Patel and soil types, soil quality, SVM, KNN public land
recommendation Patel (2020) crop, climate, water recommendation
demand
soil quality Murugamani soil moisture, soil pH, SVM private assessment of soil
monitoring et al. (2022) humidity temperature quality
soil moisture Araya et al. images SVR, RF, ANN, private soil moisture
monitoring (2020), RVR, BRT prediction
Bhattacherjee
et al. (2020)
soil health Jain et al. images SVR, RF private soil health
monitoring (2020) assessment
soil salinity Klibi et al. images SVM,AE, KNN,DT private soil salinity
monitoring (2020) prediction
phosphorous Dong et al. images CNN private soil phosphorus
monitoring (2018) prediction
soil manure Ather et al. soil pH, temperature, and ANN public manure prediction
composition (2022) NPK
carbon sequestering Moinet et al. soil nutrient, retention Nill Nill reduce atmospheric
(2023), Feng carbon
et al. (2023)
and Payen
et al. (2023)

12
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 11
Water management solution.
Subsolution Ref. Features ML model Data source Comments
Wang et al. NO3-N, PO4, NO2-N, GRNN, MPR laboratory data water quality
(2021) NH3-N
Abinaya et al. temperature, pH, Naive Bayes private aquaculture water
water quality (2019) dissolved oxygen, water quality monitoring and
monitoring level, foul smell, control
ammonia
Miao et al. dissolved oxygen, pH, DL private water quality
(2022) temperature
Chiu et al. pH, temperature, ANN private fish fond water quality
(2022) dissolved oxygen, monitoring
turbidity
anti-frost sprinkler Guillén-Navarro temperature, humidity, LSTM public (SIAM) accuracy to predict low
et al. (2020) wind speed temperature
soil wetness Nursyahid et al. soil moisture Linear regression private soil moisture
monitoring. (2019) monitoring
Dahane et al. soil moisture, air LSTM, GRU private optimizing water
(2020) temperature, air resources
irrigation humidity
monitoring
Rahmouni et al. soil moisture, not specify private Irrigation prediction
(2022a) temperature, humidity,
barometric pressure
flood monitoring Sung et al. water level, rainfall Fuzzy private flood prediction
(2022), intensity, water speed accuracy
Manikandan land slop
et al. (2022)
Alves et al. weather, soil moisture, Penman Monteith, private water saving
(2023) DAP, rain forecast Fuzzy
Souza et al. soil moisture, weather, fuzzy private water saving
(2020) rain forecast, ET
irrigation Granata (2019) temperature, solar ANN public evapotranspiration
scheduling radiation, wind speed, estimation
humidity
Elbeltagi et al. weather data ANN public evapotranspiration
(2020, 2022) prediction
Başağaoğlu et al. weather data NGBoost-XGBoost, public evapotranspiration
(2021) probability prediction
Saha et al. soil moisture, FL, MLP, ReLU public watering planning
(2020) temperature, humidity

the system by reducing dependency on a centralized entity. The water solutions are facilitated by the emergence of high-quality hyperspectral
management solutions are summarized in Table 11. cameras and libraries of image processing algorithms.
Disease Management: Pests and diseases are major challenges that One approach of particular interest is an FL-based method using
affect the quality and quantity of crop production. Crop growth is an UAVs imaging for disease identification and classification is presented
important element in the optimization of adequate food production and in Khan et al. (2022). In this paper, UAVs on four different farms
is affected by disease and stress including, as seen previously, improper locations are used to detect pest occurrences. The proposed pest classi-
irrigation, biotic stress, diseases and soil salinization (sodium chloride fication solution accurately classified the nine available pests in Kaggle
(NaCl) which occur naturally or due to improper anthropogenic ac- pest datasets using FedAvg. In research conducted in Patros et al.
tivities). Disease prediction, detection and remediation are thus major (2022), another FL framework was developed for rural weed detection
research areas in SA. Disease and stress identification and prediction using hyper-spectral pasture images captured from three different sites.
Smart Harvesting: Smart harvesting can be performed with the
can help in providing more food productivity (Udutalapally et al.,
help of autonomous harvesters and robots. Smart harvesting has used
2020) by allowing us to determine the best approaches to remediation.
different types of sensors and imaging systems such as ultrasonic sen-
Plant stresses are either biotic (infectious) or abiotic (non-infect-
sors, global navigation satellite systems (GNSS), depth cameras, single-
ious). Biotic stresses are usually caused by infection causal agents
shot multi-box detectors, RFID, Three-dimensional (3D) sensors and
such as fungi, bacteria, parasitic plants, viruses and nematodes. The
red, green and blue (RGB) cameras to guide the harvester and define
non-infectious stress (abiotic) is caused by nutrient deficiencies, poor optimal times for collecting the crops.
farm management, or unfavorable environmental conditions. These in- Several solutions have been published in smart harvesting. Object
clude too-low or too-high temperatures, inappropriate moisture levels, detection is the focus of Chen et al. (2020), Cheng and Zhang (2020),
high winds or uneven wind speed, drought or flood, soil compaction, whereas the paper in Hsu et al. (2022) targeted fruit classification.
frequent and heavy rain, improper water management, deficiency or Finally, the authors in Paul et al. (2021) used color recognition to
excess of nutrients and chemical injury caused by pesticides or salt. determine the ripeness of crops and decide on the best harvesting times
There are several use cases in crop disease management (scouting) as presented in Table 13.
in the literature. They include disease detection and prediction, weed Supply Chain Management: Supply Chain Management (SCM) is
monitoring, disease classification, and pest monitoring, detection and the process of planning, implementing and controlling the operations
classification. They are listed in Table 12. Disease management in these of the supply chain to meet customer requirements as efficiently as

13
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 12
Disease management solution.
Subsolution Ref. Features ML model Data source Comments
Jiang et al. (2019) images CNN GoogleNet public apple leaf disease
inception, rainbow detection
Pallagani et al. images CNN public crop disease
(2019), Udutalapally prediction
et al. (2020)
Bhatia et al. (2021) images Tensoflow public crop disease
disease monitoring prediction
Li et al. (2022b) images InceptionV3, ViT, public plant disease
MobileNet detection
Chatterjee et al. diseases symptoms, Fully-Connected private cow disease
(2021) behavior changes NN(FCNN) prediction
Chen et al. (2019) temperature, CNN private rice blast disease
humidity, rainfall, detection
barometric pressure
Antico et al. (2022) images FL, CNN, public maize disease
prediction
Saberi Anari (2022) images Multiple SVM public classifying images of
crop disease
Murugamani et al. images SVM private detect and control
(2022) cotton leaf diseases
disease detection
and control Li et al. (2022d) images ViT public detect and control
cotton leaf diseases
Liu et al. (2019) images CNN public pests detection and
classification
Chen et al. (2020) images YOLOv3, CNN, private pest identification
LSTM
pests management
Guo et al. (2020) images Chan–Vese, RPN public pest identification
Khan et al. (2022) images FL, CNN,ResNet- public pests management
101,ResNet50
Patel and Patel soil types, pH, KNN, SVM private pesticide
(2020) electric conductivity recommendation
Partel et al. (2019) images YOLOv3 with CNN private weed detection,
mapping, spraying
weeds
management Patros et al. (2022) images FL,ANN, ReLU, private unwanted weeds
management

Table 13
Smart harvesting solution.
Subsolution Ref. Features ML model Data source Comments
object detection Cheng and images YOLOv4, CNN public flower detection
Zhang (2020) ResNet and classification
fruit Hsu et al. (2022) images CNN, Private dragonfruit
classification YOLOv3-tiny ripeness
prediction
color recognition Paul et al. images Naive Bayes private cultivation
(2021) prediction

possible. The use of AI in SCM is currently a major research area. For example, there are SA solutions delivered by Nokia in several
AIoT can have an important impact on SCM in SA: SCM defines the countries in the series of projects called real action (Nokia, 2023; Aero-
requirements at each layer of food production that AIoT can meet. Farms, 2021). In India, Nokia in collaboration with Vodafone started
Supply chain management contributions in SA include (Ahamed and the pilot project across Indian states in 100 locations in which 50,000
Vignesh, 2022; Nahr et al., 2021; Nozari et al., 2022; Liu et al., 2016) farmers are expected to benefit from the proposed solution, SmartAgri,
dedicated to ensuring a secure food supply chain and safe delivery of in improving their productivity and income. More than 400 sensors
the food items (see Table 14). In addition, a technical solution based on were installed as part of the project across over 100,000 hectares
FL was presented by Durrant et al. (2022) that used decentralized data of farmland. This solution uses Nokia’s Worldwide IoT Network Grid
to develop a cross-silo ML model. This solution facilitates data sharing (WING). For crop management and includes smart pesticide control,
across the supply chain for improving production optimization through smart irrigation, a platform for commodity exchange and proactive
soybean yield prediction. The authors also provided potential use cases information sharing frameworks on weather and crops which offers
in which such methods can assist in other problem settings. weather forecasts and information on irrigation management and sup-
ports local languages (NS Agriculture Staff Writer, 2020). Also in 2019,
5.1.2. Industrial solutions Nokia piloted a Smart Agriculture-as-a-Service in Algeria.
There are an increasing number of industrial solutions presented Another existing solution is from Metos (by Pessl) in Austria for
by different companies helping farmers to make proper decisions in disease management which provides more than 80 disease management
managing their farms. Some are presented in this section. models for more than 40 different crops. The solutions are distributed

14
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 14
Supply chain management solution.
Ref. Features ML model Comment
Ahamed and autonomous vehicle, DLT food supply chain
Vignesh robot
(2022)
Nahr et al. food materials Not specify green supply chain
(2021)
Nozari et al. interviews, Fuzzy DEMATEL supply chain
(2022) questionnaires
Aliahmadi consumers, producers, Not specify sustainable supply
et al. (2022) suppliers chain
Durrant et al. images FL soft fruit production
(2022) optimization and fresh
food distribution

to Europe, Asia-Middle East, Asia, North America, Central America, LEMKEN provides SA solutions such as soil cultivation (i.e., plough-
South America, Africa, Australia and New Zealand (Metos, 2022). ing, reconsolidation, seedbed preparation and stubble cultivation), sow-
MyEasyFarm and its global partners work with farmers every day ing (drill seeding, precision seeding and intercrop seeding) and crop
to help them make the best choices and be more effective in max- care (fertilization technology) across Europe, Asia, North America and
imizing competitiveness and profitability while protecting the envi- Africa (LEMKEN, 2023).
ronment (MyEasyFarm, 2023). MyEasyCarbon engaged in low-carbon Microsoft developed the FarmBeats system, recently transitioned to
projects by increasing carbon storage in the soil (carbon sequester- a product called Azure FarmBeats, the Microsoft cloud-based agricul-
ing) to reduce carbon emissions using digital tools that support the ture offering.6 It is a complete digital agriculture solution. Microsoft
agroecological transition. works with its SA partners to prototype agricultural services for farm-
Solutions provided by startups include Ferme d’Hiver (FH) in Can- ers. The system gathers a lot of information, spatial, historical and
ada.5 FH is specialized in vertical agriculture. In FH’s farm, strawberries temporal, about every farm, from several data sources, such as drones,
are grown vertically in a closed controlled environment using LED illu- satellites, sensors, cameras, tractors and weather stations. The technol-
mination, advanced HVAC and automatic irrigation without chemicals. ogy employs AI to combine this data and infer likely occurrences on
Intelligent controllers and AIoT are planned to be added to the product
the farm. Partners have access to the information via APIs and they
offering. FH intends to sell its technology to provide local production
may subsequently create agricultural insights for growers using their
and reduce the need for foreign imports.
in-depth agricultural knowledge (Chandra et al., 2022; Ye, 2021).
ISAGRI is a solution for grain growers that helps farmers to make
Agro Smart lab is another solution that provides mathematical mod-
the best choices for their farming operations. It includes regulatory
els for scouting using the weather data obtained from the weather sta-
safety as well as means to meet environmental challenges for better
tion to predict diseases and monitor the appearance of pests (Smartlab,
optimization of pig farming, dairy farm management, sucker cattle
farming and small ruminant farming (ISAGRI, 2023). 2023).
ATIM monitors temperature in greenhouses and allows fast action Finally, Farms.io (farmsio, 2023) is an application for farm man-
via SMS alerts and for example ensures mist spraying in the vineyard to agement, crop monitoring, carbon analysis, link to markets, production
prevent damage due to frost (ATIM, 2023) until the temperature rises traceability, land use and land cover, post-harvest climate change mon-
again. It has been deployed in the Burgundy wine-producing region in itoring and agriculture advisory. This is dedicated to helping a farmer
France. in decision-making as presented in Table 15.
Another solution for crop monitoring, cultivation and production
was developed by Elzeard. It enables farmers to schedule their pro- 5.1.3. Discussion
duction activity for the entire season(s) using a planning module. It
To summarize the entire review, we conclude that research on
handles the end-to-end agricultural plans from the determination of
crop and disease management has the highest number of published
the marketing requirements, the setting and production goals and the
articles, followed by smart monitoring and water management, then
optimization of crop rotations. The cultivation module consists of the
soil management, smart harvesting and finally, the least cited topic,
features required for crop monitoring, farm activity organization, and
supply chain management. This shows where research has focused and
data collecting for operation traceability (Elzeard, 2023).
opens ways for innovation.
The Agribot Platform, a United Kingdom smart agriculture solution
Finally, regarding the datasets, there are some available datasets for
platform (Agribot Platform, 2023; Rai et al., 2022), enables farmers to
make disease treatment decisions that are both environmentally and individual plants but we could not find an open or standard dataset for
economically advantageous by giving them an understanding of crop agricultural and environmental parameters across different crops. One
health which can be received on their smartphone. Agribot can perform of the common websites providing public datasets is Kaggle.com used
several tasks including AI-based disease diagnosis, crop health status, by several papers. However, the lack of data is a major challenge to the
the detection of soil stress and the measurement of localized weather. implementation of AIoT in SA.
The company also proposes an Agribot Edge solution to capture pic-
tures that are then processed using image processing software and AI 5.2. AIoT infrastructure for SA
decision-making. When the application identifies an unhealthy crop or
stress problem, it will deliver a message outlining the problem and
This section explains AIoT architecture and frameworks used in SA.
can automatically perform the appropriate action depending on the
situation.

6
https://www.microsoft.com/en-us/research/project/farmbeats-iot-
5
www.fermedhiver.com agriculture/

15
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 15
Industrial solutions.
Ref. Company Solutions
Crop-M Soil-M Water-M Disease-M Smart-H Supply chain-M
Nokia (2023), AeroFarms (2021) Nokia ✓ ✓ ✓ ✓
Metos (2022) Metos (by Pessl) ✓ ✓ ✓ ✓
ISAGRI (2023) ISAGRI ✓ ✓ ✓
ATIM (2023) ATIM ✓
Elzeard (2023) Elzeard ✓ ✓ ✓
Agribot Platform (2023), Rai et al. (2022) Agribot ✓ ✓
LEMKEN (2023) LEMKEN ✓ ✓
MyEasyFarm (2023) MyEasyFarm ✓ ✓ ✓
Chandra et al. (2022), Ye (2021) FarmBeats ✓ ✓ ✓
Smartlab (2023) AGRO SMART LAB ✓ ✓ ✓
farmsio (2023) Farmsio ✓ ✓ ✓ ✓ ✓

Crop-M: Crop management; Soil-M: Soil Management; Water-M: Water management; Disease-M: Disease Management; Smart-H: Smart Harvesting;
Supply chain-M: Supply chain Management.

for the crop growth analysis. This solution uses devices (i.e., smart-
phones, drones, and sensors) without AI capabilities and the AI is solely
cloud-based.
The architecture presented by Tomar and Kaur (2021) is also a
two-layer C-B architecture consisting of a data collection layer and
a cloud computing layer (cloud platform). The data collection layer
includes many different sensors (temperature, humidity, and moisture)
responsible for generating environmental data. The sensing devices
used in this study are intelligent devices that can perform some basic
AI-related inference tasks. The raw data can be sent to a gateway
through short and medium-range communications technologies for ex-
ample Bluetooth, WIFI, LORA, and SigFox and then to the cloud layer.
Data analysis, processing and storage are performed in the cloud. Users
can get access to the data and the analyzed results through the internet.
Preliminary implementation results validated the approach.
In the architecture presented in Chen et al. (2020), Tomar and
Fig. 14. Percentage of the 5 most popular ML models used in the articles reviewed
Kaur (2021), there is no specified layered architecture and each device
in this survey. The percentage of every other remaining model (i.e., not mentioned in is vertically integrated into the cloud. This can bring some concerns
this figure) is under 5%. such as network traffic management and its impact on AI applications
efficiency and real-time decision-making.
The architecture described in Mitra et al. (2022) is a good exam-
5.2.1. SA architectures ple of the three-layer Edge–Cloud (E–C) architecture for SA with an
While there is no standard AIoT infrastructure for SA, many similar agriculture devices layer, an edge layer, and a cloud layer. In the
architectures have been proposed. A common SA architecture com- device layer, sensors are deployed in various locations on the farm
prises at the minimum an agricultural device layer, an edge/fog layer to acquire physical and environmental parameters: in the field, in the
or data gateway and a cloud layer as shown in Fig. 6. Different IoT greenhouses, in animal paddocks, and on tractors. These end devices
and AIoT devices can be deployed in the device layer for sensing, are not intelligent as they cannot perform in-device processing and
would not be able to participate in AI decision-making: the intelligence
monitoring and tracking different activities of SA. Connectivity sub-
is potentially in the edge and the cloud. As in other solutions, the raw
layers can be added consisting of the short-range and medium-range
data is transferred to the edge layer through WIFI, Bluetooth, and Z-
to long-range communications technologies. A connectivity layer 1
Wave gateways. Edge devices are responsible for processing, filtering
connects the AIoT devices layer with the edge/fog layer with short to
and encrypting the data before transmitting it to the cloud through
medium-range technologies and connectivity layer 2 links the edge/fog
high-speed cellular technologies. The cloud layer can process, analyze
layer with the cloud layer with long-range technologies. The edge/fog
and save the massive data gathered on the farms.
devices are responsible for the processing of the data generated at
The authors in Islam et al. (2021) also proposed a three-layer archi-
the device layer before transferring it to the cloud where it can be
tecture consisting of perception, network and application layers. The
further processed and stored as illustrated in Fig. 6. As was mentioned perception layer consists of sensing devices, such as RFID tags, terminal
previously, AI can be implemented horizontally and vertically at each devices and readers, used to collect data about pests, plant diseases,
layer. nutrient levels, humidity, wind speed and temperature. The collected
A more succinct cloud-based (C-B) two-layer infrastructure was data is transferred to the application layer via network connectivity.
presented by Chen et al. (2020) where mobile devices and drones in the Users can remotely access the cloud applications for monitoring and
first layer were used to capture pest images in fruit trees. The collected controlling their farms.
data was then uploaded to a database in the cloud (cloud layer) for Another three-layer C-B AIoT infrastructure is proposed in Guillén-
processing (e.g., recognition of the pests images and identification of Navarro et al. (2020) with the now familiar devices, connectivity and
the pests locations using deep learning algorithms). In addition, in cloud layer. The devices layer of this system measures the environ-
the 1st layer, environmental sensors collect data to provide meta-data ment and includes wind speed, humidity and temperature sensors. The

16
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

sensed data is transmitted to the cloud layer where the LSTM analysis 5.2.2. Frameworks/practical implementation of AIoT for SA
capabilities are located. The data and results are accessible from the This section presents the available frameworks and practical imple-
cloud layer for monitoring and decision-making. mentation of AIoT for SA. Device management, model management,
A more distributed infrastructure is available in Gupta et al. (2020). support of various AI techniques and edge device integration are the
It introduces a four-layer E–C AIoT architecture with a physical layer, four main features of AIoT platforms mentioned by Panduman et al.
a network layer, edge layer and a cloud layer. Here again as in all (2024). These frameworks need a data model to be built on top of it.
SA infrastructures, the physical layer, the ‘‘IoT devices layer’’ includes Most of these frameworks follow the FIWARE data model framework.7
sensors installed in a greenhouse or on farmlands, embedded in live- Considering AIoT frameworks, the research presented by Rahmouni
stock, and added to autonomous tractors and drones (UAVs). Use cases et al. (2022b) illustrated an example of AIoT framework for precision
for these devices include sensing and monitoring the environment and agriculture which showcases how AIoT can impact modern agriculture
application-specific data and transmitting it to the edge layer through by implementing data-driven solutions based on low-cost devices and
the communications layer. These are traditional sensors in the sense open source technologies, empowered by Edge Intelligence.8 Also, the
that they do not process the data or make decisions. The edge nodes research of Li et al. (2022c) designed an AIoT system for SA based on
are used to complement the sensing by processing the received data: front and rear end separation and a Model View View Model (MVVM)
real-time monitoring, visualization, and online ML model for detection, framework for handling complex business logics and easy integration
prediction, and diagnosis. The edge also transmits the data to the cloud of AI algorithms. They constructed a web page for the front end using
for further processing: AI/ML, database clusters, data visualization, big
Vue.js and Element and the rear end business logic using the Python
data analytics and storage.
Django framework. The data interaction between the front and rear
A four-layer Fog–cloud (F–C) infrastructure is presented by
end was via Axios. They integrated basic application functions such as
Muhammed et al. (2022). In addition, it introduced a user-friendly
historical data query, real-time data monitoring, abnormal data alerting
component for obtaining the data automatically without human inter-
and data visualization. Similarly, integrated deep learning plant disease
vention directly from devices using APIs. This system has the standard
and pest detection algorithms. Another example for helping farmer in
end-device layer with the edge/fog, connection and cloud layers. The
mitigating the effects of climate change and improving their farming
end-devices layer provides the necessary information to provide the
practices to increase the quality of their crop yield and food production
data and create user requests. The edge/fog layer includes a user
was presented.9 In Amazon Web Services (AWS), a smart agriculture
request creator (URC), an inference and decision-making tools. The
startup based in France called Sencrop created a solution that enhances
model training is done in the cloud to benefit from more powerful
prediction accuracy and gives farmers access to precise climate con-
computing resources. Two connection layers are considered: the first
dition insights to support data-driven decision-making. They build a
is located between the end-devices and edge/fog layers to transfer raw
microclimate application on AWS using Amazon EMR (a cloud big
data from end-devices to the edge/fog layer and the second is located
data solution for ML, interactive analytics and petabyte-scale data pro-
between the edge/fog and cloud layers to transfer data between them.
cessing). Almost 30,000 farmers improved the sustainability of SA by
Continuing the four-layer concepts, Ferrández-Pastor et al. (2016)
reducing the use of spray treatment chemicals, water use and the trips
also presented a four-layer E–C architecture consisting of a things
layer, an edge layer, a communications layer and a cloud layer. An tractors and harvesters using Sencrop’s across Europe.10 The research
extension of this infrastructure, which is an E/F–C type is proposed in Slama et al. (2023) presented AIoT Solution design, which is a simple
in Ferrández-Pastor et al. (2018). This architecture added a fog layer canvas to visualize the key functional elements of a solution, and AIoT
between the edge and the cloud. The fog node aggregates the received use case mapping, which clarifies how AIoT can best support typical
data from the edge nodes and transfers them to the cloud through use cases. Additionally, it introduced the AIoT framework, which in-
the communications layer. The cloud processes, analyzes and stores cludes an overview of AIoT, technical execution, development life-cycle
the data. Here, as in most currently proposed architectures, there is perspective, data strategy, and design viewpoints and templates. The
no AI processing at the IoT devices layer (things): all analysis and AIoT framework addresses aspects such as the agile approach, DevOps,
decision-making are done at the edge, fog, and cloud layer. trust and security, reliability and resilience, functional safety, and
Finally, targeted system performance and type of application is an quality management. There are some solutions in the AWS marketplace
important factor to select among the mentioned architectures. Utiliza- examples can be EOSDA Crop Monitoring an online satellite-based
tion of edge intelligence for lightweight applications (e.g., acoustic and precision agriculture platform for field monitoring created by EOS
ambient sensing in SA), will lead to an efficient system. However, more Data Analytics (EOSDA). It is a global provider of AI-powered satellite
compute-intensive applications may need more edge–cloud collabo- imagery analytics. The platform is a one-stop solution that integrates
ration (Alzuhair and Alghaihab, 2023) or edge-to-edge collaboration multiple types of data (crop health, weather conditions, crop rotation,
to provide accurate results and systems with high energy efficiency field activities, elevation, soil moisture, and a host of other types) all
and optimized performance. Distributed learning and FL are two tools in one place. Cropin Apps is another example of a solution in the AWS
that make these collaborations possible. A comparison of central ML marketplace that enables the digitization of complex processes and
and FL performed for crop classification in a smart farm decentralized workflows that span the complete agri-food value chain, a seed from
network was reported by Idoje et al. (2023). The research utilized the farm to a warehouse, all the way up to the fork. These solutions help
climate data consisting of temperature, humidity, rainfall and pH for streamline farm data capture and management and other complex field
crop classification using a federated averaging model. The analysis of operations involved during seed production, seeds strain tracking &
their result using the farm dataset showed that decentralized models trialling across generations, crop protection and nutrition development
achieved a faster convergence and higher accuracy than the centralized both on & off the field for row crops, horticulture and plantations.11
models (binary relevance Gaussian NB, Classifier chain Gaussian NB,
and Label Powerset Gaussian NB).
A summary of the existing SA AIoT infrastructure with their stren- 7
https://www.fiware.org/
gths and weaknesses are presented in Table 16. As Table 16 shows 8
https://github.com/nabs13/Smart-Farming-though-AI-and-IoT
tightly coupled architecture is rarely used in the related work. While 9
https://github.com/IoT-Communications/Smart-Farming
the other domains as seen in Table 7 already started using tightly 10
https://aws.amazon.com/fr/solutions/case-studies/sencrop-case-study/
11
coupled architectures. https://aws.amazon.com/fr/solutions/agriculture/data-analytics/

17
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Table 16
AIoT architectures for SA.
Ref. Architecture NL AIoT Systems Limitation
E/F–C E–C F–C C-B Loosely Tightly NAIPED NEFN
(Khattab et al., 2016) ✓ 3 ✓ ✓
Ferrández-Pastor et al. (2016) ✓ 4 ✓ ✓
Ferrández-Pastor et al. (2018) ✓ 5 ✓ ✓
Chen et al. (2020) ✓ 2 ✓ ✓ ✓
Gupta et al. (2020) ✓ 4 ✓ ✓
Guillén-Navarro et al. (2020) ✓ 3 ✓ ✓ ✓
Tomar and Kaur (2021) ✓ 2 ✓ ✓
Islam et al. (2021) ✓ 3 ✓ ✓ ✓
Mitra et al. (2022) ✓ 3 ✓ ✓
Muhammed et al. (2022) ✓ 4 ✓ ✓

NL: Number of Layers; E/F–C: Edge/Fog–Cloud; E–C: Edge–Cloud; F–C: Fog–Cloud; C-B: Cloud-Based; Loosely: Loosely coupled; Tightly: Tightly
coupled; NAIPED: No AI Processing at the End Devices; NEFN: No Edge or Fog Nodes to process the data generated by end-devices.

5.2.3. Discussion MLP, Linear Reg., Logistic Reg., MLR, OpenCV, MNLR, NARX, RNN-
To summarize this section, we analyzed layering patterns of the LSTM, CHAID, ResNet 16, ResNet 28, RNN, RVR, BRT, AE, GRNN,
available architectures and it was gathered that three layers are the MPR, GRU, DLT, NGBoost, XGBoost, GoogleNet, Inception, Rainbow,
most commonly adopted architectures followed by four layers. How- TensorFlow, Inception V3, MobileNet, FCNN, Chain Vese, RPN, ResNet
ever, the two layers and five layers are found in rear cases. Similarly, 101, ResNet 50, ResNet, YOLOv4 and YOLOv3-Tiny.
Most of the existing architectures are loosely AIoT systems and Cloud-
We then calculate the percentage of using each ML model in the
based architectures as they do not perform AI processing at the end
articles under review. Fig. 14 shows the percentage of the 5 most
devices layer but send their data to the Cloud layer for AI processing.
popular ML models. The percentage of every other remaining model
For AIoT frameworks in SA, the AWS marketplace provides several
frameworks such as Sencrop, EOS Data Analytics (EOSDA) and Cropin (i.e., not mentioned in Fig. 14) is under 5%.
Apps for practical implementation of AIoT for SA.

5.3.2. IoT technologies


5.3. AIoT technologies for SA
IoT technologies collect data from the location or place where they
This section presents the existing technologies in SA in terms of AI, are deployed and use communications technologies to transfer the data
IoT and communications technologies. to the target layer for processing. These devices include but are not lim-
ited to sensors used for environmental monitoring such as temperature
5.3.1. AI technologies/techniques in SA sensors, soil moisture sensors, humidity sensors and so on. Similarly,
In this section, we summarize the findings on the use of AI in SA. drones, satellites and other cameras are used for capturing images of
Based on our reviews, we have noted that in smart monitoring solu- soil, pests and weeds. In this case, lightweight thermal and RGB (red–
tions, Convolutional Neural Networks (CNNs) are widely used because green–blue) cameras are mounted on a UAV to capture images. One of
of their higher accuracy in object detection and image classification. the other types of sensors used is for examining the audio signals. One
For environmental monitoring such as temperature and humidity,
example can be the study which used this information for identifying
CNNs are widely complemented or replaced by more feed-forward arti-
Queenlessness in Honeybee Hives using ML. This identification can
ficial NN (ANN) as well as Recursive NN (RNN). For crop management,
be used by beekeepers as a means of promptly alerting them if the
supervised learning, including k-nearest neighbors (KNN), Naive Bayes,
support vector machines (SVM), Random Forest (RF), Decision Tree queen had died in any of their hives (Ruvinga et al., 2023). Some
(DT) and logistic regression are the most adopted models. Crop rec- other source of data is generated from agricultural machinery. Today’s
ommendation also relies on ensemble learning as shown in Table 9. In agricultural equipment can collect a lot of information about soil and
soil management solutions, it is the SVM model that is mostly adopted, plants and machine performance as they operate in the field. For ex-
ANN, KNN, FR and SVR were also frequently mentioned as presented ample, nowadays, grain harvesting combines are equipped with a yield
in Table 10. Similarly, for water management solutions, LSTM was the monitoring system. Yield monitors can quantify and generate a map
most often cited model with ANN, Naive Bayes and linear regression that shows the yield variability throughout the field. At the same time,
(Table 11). the information generated by the yield monitor can be used to calculate
For disease management solutions, our investigation reveals that position, velocity, and time spent during the harvesting operations. This
CNN is again most often used for disease detection and prediction, and information provides the field manager with information about the field
for pest detection and classification because of the robustness of CNN efficiency, lost time, and operator performance. It can also be used to
and good accuracy in image classification. However, SVM, YOLOv3 and determine the optimum size of the combine for the given field size. The
ViT were also frequently used as they are widely available as presented
gap in the current system includes the lack of sensors that collect all the
in Table 12.
needed information. For example, no suitable sensor is commercially
CNN is still the most commonly adopted model for smart harvesting
available to measure and map the variability of soil nitrate in the field.
which deals with object detection and classification properly. Other
models that were used include You Only Look Once (YOLO) YOLOv3, The data collected manually through field sampling often does not have
YOLOv4 YOLO-tiny as presented in Table 13. the resolution needed for site-specific data analysis. The sheer volume
For supply chain management Fuzzy logic, blockchain and other of data needed for the decision support system and developing accurate
decision algorithms derived from big data were used (Table 14). management decisions means that excellent metadata and good data
Furthermore, we listed all ML models used in the articles reviewed management are needed. Currently, some of the metadata collected
in this survey. These models are RF, LSTM, ReLU, YOLOv3, ViT, SVR, manually is not digitized or has data entry errors and is not dependable.

18
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

5.3.3. Communications technologies 6.1. Energy


In SA, machine-to-machine (M2M) communications using many
network technologies and protocols are needed to collaborate and Access to clean and affordable energy is a major challenge in SA,
share data. In addition, user applications and dashboards can be used especially in greenhouses and CEA. In addition, the need for long-
to communicate requirements and retrieve and illustrate analysis for duration and sustainable power sources for sensors is needed to reduce
decision-making. Since in SA, sensors may be of different types and, the Operational Expenses (OPEX) related to batteries and maintenance.
This is especially challenging for sensors that need to be placed in the
as was seen, many wireless and wired standards can be used, there is
soil at different depths, making them very hard to access individually.
a need for many technologies to provide SA connectivity. As a result,
There are some efforts at the research stage to develop biodegradable
it can be expensive to establish and maintain an SA network (Sahitya
batteries that will resolve some of the sustainability of sensing and
et al., 2016; Chandra et al., 2022). Connectivity is also a major issue
hence SA which can be seen as one of the future directions related to
in large rural areas due to the lack of access to reliable broadband
the energy challenge in SA.
access. This is cited as one challenge in the deployment of SA solu- In addition, there are several efforts to address power consumption
tions. The lack of network infrastructure will affect SA operations. For in SA suggesting using renewable energy solutions (Liu et al., 2018;
example, disruptions in wireless communication channels will prevent Ram et al., 2020; Huang et al., 2020). Also, an efficient micro-grid
SA devices from communicating to the cloud for data processing and architecture with renewable energy was presented by Ebrahimi et al.
decision making (Mitra et al., 2022; Tao et al., 2021). This provides an (2019) but it is a preliminary design. Hydroelectric power is currently
incentive to move a lot of the SA data management to the edge and fog used in countries where it is readily available.
outside well-deserved areas. Given these challenges, there is a need to Hydroelectric power can be part of a sustainable future. As many
provide secure and robust communication for SA and look at functional sensor nodes used in AIoT are equipped with small batteries, balancing
decomposition to use local and edge resources, not just the cloud. The and saving of energy consumption in the network of sensors is another
stakes are high but investments are flowing in. The United States (US) energy-related issue. As a solution, Yu et al. (2022) proposed an energy-
announced an initiative to bring broadband everywhere and compared aware device scheduling solution for optimizing the device selection
it to the electrification efforts of 100 years ago. Initiatives like ARA and assigning communication resources to the optimal edge node to
Wireless12 in the US also plan to bring broadband wireless (including reduce global loss. As a future direction, it is required to offer more
free space optics) to rural areas. The 5GS Ambient IoT service study intelligent solutions for scheduling and management of sensor motes
item in 3GPP release 19 is another initiative that includes agriculture (e.g., combining ML and heuristics methods).
use cases also from communication and architecture point of view Putting all the pieces together, in a nutshell, it is required to design
more energy-efficient sensor motes, find cheaper and simpler solutions
reported in AmbientIoT3GPP (2023).
than existing solutions to recharge the batteries (e.g., new methods of
using solar energy for recharging the batteries) and use more intelligent
5.3.4. Discussion solutions for scheduling and management of sensor motes.
In this section, we presented a summary of our findings in AIoT
technologies based on our reviews. We have noted that the three 6.2. Hardware availability
most widely adopted AI technologies are CNN because of their higher
Maintenance and replacement of equipment such as sensor motes,
accuracy in object detection and image classification, SVM and ANN.
gateways, processing devices, and lights in greenhouses and CEA is
Considering IoT technologies, the commonly adopted IoT devices are
another challenge. As a result of hardware CAPEX, many small farmers
sensors that are used for monitoring activities. Similarly, drones, smart
will avoid deploying SA or will only use simple solutions which in
cameras, embedded systems, smartphones and RFID are also used
the end could affect their revenues. Good quality SA sensors can cost
for capturing field data. For communications technologies, the most
up to hundreds of dollars and drones with cameras cost thousands of
commonly used communications technologies well-adapted for SA at dollars (Chandra et al., 2022).
the end-devices layer are LoRaWAN, ZigBee and BACnet due to their Deciding the number of sensors needed to capture the soil and plant
low power consumption, small size, ease of implementation, simplicity, data is another CAPEX decision. For example, monitoring soil nitrate
scalability, range (usually medium) and in the case of BACnet heritage variations may require many sensors that could be very costly. In this
from industrial automation. In addition, there are some studies to adapt regard, sharing data and virtual devices (e.g., virtual weather station)
WIFI for SA, such as ‘‘WIFI-based long distance (WiLD)’’ proposed usage is going to be one of the future directions.
by Ahmed et al. (2018) for monitoring and controlling smart farming
in rural areas because of its low cost. 6.3. Privacy, security and ethical issues

Like all cloud-based infrastructures, SA is also exposed to server


6. Challenges and future directions attacks (Rettore de Araujo Zanella et al., 2020). Securing the equipment
as well as protecting the data from attacks and unauthorized access
Recent advances in AIoT in SA including AI-based decision systems, need to be considered when engineering an AIoT-based SA system.
robotics, sensors and their applications will positively impact food Some solutions have been proposed to tackle this latter issue (Gupta
production and offer avenues for innovation. In order to address how et al., 2020; Chakraborty and Bhunia, 2009; Bathalapalli et al., 2021).
to achieve the goals of the implementation of AIoT in SA, we present a However, today’s existing SA technologies are resource-limited, making
list of the existing and potential challenges and offer future directions. security measures practice difficult. Thus, data security and privacy
It is important to highlight that these challenges are not only for SA but remain a serious challenge in SA as the security and privacy of sen-
also for many other applications. However, the level of importance and sitive spatial, temporal and spectral information of crops are a serious
concern for farmers (Pham et al., 2021).
priority of a challenge can be different from one application (e.g., SA)
For this reason, blockchain tracing is starting to be used in SA.
to another (e.g., healthcare).
Authors in Jadav et al. (2023) proposed a framework that used AI and
blockchain tracing to minimize the use of pesticides predicting crops
with pesticides above the threshold. Research reported in Zheng et al.
12
arawireless.org (2023), also presented an integration of blockchain in the SA for the

19
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

analysis of the optimal traceability strategies for agricultural products processing systems, they are often too expensive for small-scale farms.
to guarantee fraud-free and sustainable agricultural supply chains. Also, many farmers will want to improve on their existing capabilities
Finally, access to the facilities themselves needs protection. The not just replace all equipment: hence multiple data acquisition systems
equipment security breach can be caused by the farm animals de- need to inter-operate, something that they were not designed for. And
stroying access systems, and by farm workers compromising the SA while they could be co-located often one needs to get to the cloud to
environment. Hence, cyber–physical security is needed across all layers retrieve the data as was previously mentioned. As a result, the data
of the SA infrastructure. will be heterogeneous in nature (different formats, time granularity and
The ethical issues are another concern in using AIoT for SA. Intellec- precision). Delays incurred in cloud access also with impact real-time
tual property (including intellectual data and algorithms/ML models) is operations.
one of the main ethical concerns where SA data owners (e.g., farmers Data normalization is another issue. Sensor nodes, including cam-
and stakeholders) should have control over the data generated by AIoT eras, collect a lot of data, and correlation and bias need to be addressed
devices on the farms and the algorithm and ML model applied to them. when using that data for decision-making. Hence a challenge is to nor-
The data owners should know (and permit) how and where their data is malize and transform the raw data into usable information (a process
being used. They should have the right to consent or opt out. FL is con- called ETL or extract transform load) that can then be used for training,
sidered as a promising technique that can also solve or, at least, reduce inference, decision-making or digital twinning. This normalization is
several ethical and privacy concerns in the future. Developing robust even more important if federation across locations is wished for. In
encryption, authentication and access control mechanisms to safeguard addition, for supervised learning, data labeling is essential. Services like
agricultural data can be future research direction for addressing privacy Sagemaker from AWS can be used.13 However, their use in SA has not
and Security SA. been proven and data labeling may remain a time-intensive human task
for the near future.
6.4. Scalability and reliability But data transformation is not all. The datasets need to be open
for testing and training agriculture systems but also for validation.
In SA, farm sizes vary from small individual farms to large com- To complicate things, equipment suppliers and farm operations alike
mercial operations. We can see an example of a single greenhouse area may be reluctant to open their data as it could reveal trade secrets.
with nearly 70,000 square meters, equivalent to ten standard football Hence a big challenge for SA system developers is creating their own
field sizes in AmbientIoT3GPP (2023). Hence, different quantities of datasets to gather enough data to be able to implement and test both
sensing equipment (sensors and associated computing elements) are new equipment and algorithms. Digital Twins can be used to generate
needed. Consequently, variable data traffic loads will be generated. synthetic data but they themselves need access to some real-time data
And to allow SA growth, this technology needs to be scalable. The so their use can be limited.
layered approaches described previously may provide this scalability Hence, it is a fact that SA development and implementation outside
and it explains that they are widely considered for SA deployment. the confines of a laboratory is limited by the access to open and
SA also needs to provide reliable solutions. Sensors in open fields accessible data. As this is now a problem identified across the industry,
can be exposed to harsh environments, humidity, extreme temperatures there will be more and more initiatives to standardize and open sensor
and heavy rainfall. In greenhouses and CEA they are also exposed and system data in the future. Developing techniques to effectively fuse
to environmental variations. Malfunctions due to sensor damage will and analyze diverse data types for more accurate and holistic insights
disrupt the operations. Inaccurate decisions due to a defective sensor can be a future research opportunity for data access.
also will have impacts on revenues. For example, in paddy rice farming
if damaged sensors cannot report the correct soil water content, it will 6.6. Artificial Intelligence for AIoT
lead to serious consequences: damaging the crop, reducing yield, im-
pacting the food supply chain and eventually causing a price increase. As seen in previous sections, over the last decade, new AIoT tech-
Hence, to reduce OPEX and minimize downtime in the SA operation, nologies/methods such as CNN and FL have been widely used and
there is a need for reliable and fault-tolerant devices to reduce the deployed in a variety of different use cases (e.g., in smart cities and
need for or the number of redundant devices to respond to faults and smart healthcare). These technologies are in the early stage of de-
equipment breakage (Chandra et al., 2022). In addition to reliable and ployment in SA. While AI is now considered essential in achieving
fault-tolerant devices design, providing a mechanism that checks the cost-effective and efficient farming, there are many challenges before
functionality and remaining energy level (residual energy) of the sensor AI can become an agricultural mainstay:
nodes with timely replacement of the batteries or other power sources
will be an essential future work for SA in this type of production. • There is still very little interaction between AI research and the
Reliability is a key element. agricultural industry. As a consequence, the AI researchers are not
well aware of the farming requirements and OPEX and CAPEX
6.5. Data access challenges, and the farmers often do not understand AI tech-
nologies and what they can bring to their operations. Therefore,
The different systems that are part of the sensing layer need integra- there is a need for more interdisciplinary collaborations among
tion to implement AI decision-making. Data science and the advances all stakeholders and a better technology transfer of academic
in ‘‘big data’’ allow the creation and management of the large datasets research into commercial deployment.
needed for SA. As was discussed earlier in this survey, the supply chain • There is a lack of well-established legal framework, policies and
end-to-end efficiency of SA systems can be improved and food security regulations, for the implementation of AI in SA. In academic
issues can be mitigated with predictive analysis and real-time decisions settings, many legal aspects of operations are not addressed.
on large datasets (as an example, Kempenaar et al. (2016) uses the For example, a majority of proposed AIoT solutions in SA are
datasets in the milk industry). The SA data workflow starting with the vertically integrated and cloud-based. Without security audits,
collection to the analysis is presented in Wolfert et al. (2017), Bhat and this may make them vulnerable to data breaches, cyber-attacks
Huang (2021). In addition, the use of big data analysis and data science and compromised privacy. This is one concern that is making
in SA enables new business models (Wolfert et al., 2017). farmers wary of AI technologies. The emergence of edge AI seems
However, a major challenge facing the real deployment of SA is
the lack of integrated and open datasets for research and deployment.
13
While there are existing commercially available integrated sensing and https://aws.amazon.com/sagemaker/data-labeling/

20
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

to slowly mitigate the problem as it provides higher data privacy 6.9. User-centric design
and security, as well as lower latency and cost by processing the
sensor data at the local farm premises even if it adds complexity The number of SA farms is still very low compared to the number
and cost (Bhat and Huang, 2021). of farmers engaged in traditional farming. For example, in Sub-Sahara
• AI can only be a game changer in remote rural areas with the Africa, only 13 percent of small farm operations register for digital
availability of broadband wireless networks. And because of the services and not all of them are active (Chandra et al., 2022). Price is
data volume, it may also need novel data, image and video of course a major driver but user interaction with the technology also
compression (Gia et al., 2019) that sends the compressed data to needs to be considered.
the fog layer and then to the cloud using the broadband network User acceptability will be essential for SA. User-centric design is
now part of any innovation and a major element of system design. For
but at lower rates.
users to accept innovation, they need to be directly involved in the
• AI needs data. Many important data for SA cannot be easily
design process and give their input in the requirements gathering and
collected and it was discussed previously that datasets are hard
the end product design. In addition, designing the system is different
to come by.
from understanding it hence farmers and developers need to interact.
FogFL for training the smart irrigation scheduling (Saha et al., In SA, some of the existing technologies are not user-friendly:
2020) was discussed in the previous section. But while FL can be a they were developed for large-scale operations and required specific
candidate to provide AI in SA using local datasets and information, FL knowledge and expertise. Most farmers in rural areas do not possess
the operational and technical knowledge to integrate SA in their pro-
for multi-farm scenarios is not yet an active area of research. One of the
duction. Providing user-friendly systems and educating the farmers can
challenges may be a lack of trust in data: farmers may manipulate the
address this problem. For example, testbeds and ‘‘Living Lab’’ systems
sensors and the sensors may not provide correct information regarding
can directly show the farmers the advantages (e.g., increase in product,
their farms. However, the complexity of FL (and impacts on OPEX
revenue, minimizing loss, etc.) the systems will provide.
and CAPEX due to the local equipment need and its maintenance)
The support of multiple languages is also important. There are now
versus centralized NN in the cloud has also an impact on its wider automatic translation modules in a large number of languages that
deployment. Therefore, FL applications in multi-farm scenarios can be can be integrated into an application to help the end-users to get an
explored more in the future. In addition, developing more sophisticated understanding of the proposed system (Muhammed et al., 2022) and
AI algorithms for robotic perception, control and decision-making in to use it appropriately.
dynamic agricultural environments and also Optimizing the distri- AI can help in this acceptability. It provides future directions on
bution of computational tasks between edge devices and the cloud, how to design more automatic and robust systems that need less input
ensuring seamless communication and developing efficient edge-based from farmers. The rise of Extended Reality (XR) in agriculture may
AI models can be the focus for the future research in AI for AIoT. also support workers and operators in their daily activities (Anastasiou
et al., 2023). Designing cost-effective, user-friendly solutions that lever-
age local knowledge and resources to enhance agricultural productivity
6.7. Capital investment is a future research opportunity for user-centric design.

In most rural areas and in particular in developing countries, agri- 6.10. Simulators and digital twins
culture earns a meager profit margin. Because of the initial investment
necessary in SA for acquiring, installing and testing advanced tech- One way to test AIoT-based SA systems is to use simulators and
nology SA faces an obstacle before mass-scale deployment can be increasingly Digital Twins. One of the successful simulators in the
considered. (Chandra et al., 2022). Traditional farm loans and other domain is the Agricultural Production Systems Simulator (APSIM) was
proposed in 1990 and provides deterministic modeling of cropping and
agricultural financial instruments are ill-suited for SA and AIoT. New
pasture systems (Holzworth et al., 2018, 2014). APSIM performance has
funding policies are needed that recognize the importance of SA so that
been improved by the Agricultural Production Systems Research Unit
farmers can buy AIoT equipment along with traditional machinery and
in Australia recently (Holzworth et al., 2006; Vogeler et al., 2023).
materials (seeds, fertilizers, pesticides, herbicides, etc.).
In addition to the simulators, some platforms allow testing new
agriculture-related solutions such as Fed4FIRE14 and ARA Wireless.15
6.8. Lack of common standards Moreover, Digital Twins are emerging in SA. A Digital Twin is
the interaction between a physical and digital object which allows
for real-time, two-way communications between digital and physical
There are plenty of wired and wireless standards for communica- items (Alves et al., 2023; Verdouw et al., 2021; Alves et al., 2023)
tions in IoT and AIoT-based SA uses a large number of them. However, proposed a Digital Twin for smart irrigation where farmers can evaluate
customized solutions lack common standards and interoperability is the behavior of an automated system before implementing it. Others are
still provided by gateways with added costs, complexity and reliability in development and part of large research programs such as the one
impacts. The lack of standards also increases the price of the SA in the Institute for Resilient Agriculture (AIIRA) based at Iowa State
products. And as it was discussed before, there are no standards in University.16 Wageningen University and Research also has launched 3
data collection and datasets, which creates delays in SA design and Digital Twin projects in January 2020 including virtual tomato crops.17
implementation. Global uniform and harmonized standards in SA, at Ferme d’Hiver and the University of Montreal also worked on a Digital
the device, network and data levels are needed to reduce the time to Twin prototype for strawberry growth in CEA (Istvan et al., 2023).
market and reduce the price of the products (Mitra et al., 2022). From However, there are not many (if any) mature Digital Twin models for
the communications and architecture point of view, the efforts in 5GS
Ambient IoT service study item in 3GPP including agriculture use cases 14
https://www.fed4fire.eu/demo-stories/oc2/farmsens/
also can be mentioned as 2024 activities (AmbientIoT3GPP, 2023). 15
https://arawireless.org/
Exploring frameworks and policies for responsible AI deployment in 16
https://aiira.iastate.edu/
agriculture, including issues related to data ownership, transparency 17
https://www.wur.nl/en/research-results/research-programmes/research-
and fairness can be a new research opportunity in standards. investment-programmes/digital-twins.htm

21
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

any crops yet and this is a future direction for SA. In addition, effective Aledhari, M., Razzak, R., Parizi, R.M., Saeed, F., 2020. Federated learning: A survey on
decision-making requires consideration of the context features to which enabling technologies, protocols, and applications. IEEE Access 8, 140699–140725.
Ali, M., Karimipour, H., Tariq, M., 2021. Integration of blockchain and federated
Digital Twins belongs. In order to facilitate data exchange, the reuse
learning for internet of things: Recent advances and future challenges. Comput.
of ontologies and improve semantic interoperability, contextualized Secur. 108, 102355.
ontologies (Rico et al., 2023) and ontology quality evaluation design Ali, M., Naeem, F., Tariq, M., Kaddoum, G., 2022. Federated learning for privacy
are needed (Tiwari and Garg, 2022). preservation in smart healthcare systems: A comprehensive survey. arXiv preprint
arXiv:2203.09702.
Aliahmadi, A., Nozari, H., Ghahremani-Nahr, J., 2022. AIoT-based sustainable smart
7. Conclusion supply chain framework. Int. J. Innovat. Manag. Econ. Soc. Sci. 2 (2), 28–38.
Alshamrani, M., 2021. IoT and artificial intelligence implementations for remote
Recently, there has been an increasing amount of carried research, healthcare monitoring systems: A survey. J. King Saud Univ.-Comput. Inf. Sci..
development, and implementation of AIoT in SA. However, there is Alshehri, F., Muhammad, G., 2020. A comprehensive survey of the internet of things
(IoT) and AI-based smart healthcare. IEEE Access 9, 3660–3678.
still lack of a comprehensive survey of AIoT from a smart agriculture
Altalak, M., Alajmi, A., Rizg, A., et al., 2022. Smart agriculture applications using deep
perspective covering all important aspects. It was the goal of this paper learning technologies: A survey. Appl. Sci. 12 (12), 5919.
to survey SA landscape covering AIoT architectures, solutions, and tech- Alves, R.G., Maia, R.F., Lima, F., 2023. Development of a digital twin for smart farming:
nologies as they apply to SA. In addition, the paper wanted to highlight Irrigation management system for water saving. J. Clean. Prod. 135920.
Alzuhair, A., Alghaihab, A., 2023. The design and optimization of an acoustic and
the current challenges and future research directions including data,
ambient sensing aIoT platform for agricultural applications. Sensors 23 (14), 6262.
connectivity, DT and CNN, XR, and FL to trigger further work in this Amatya, S., Karkee, M., Gongal, A., Zhang, Q., Whiting, M.D., 2016. Detection of cherry
emerging field. SA will be essential for the sustainability of agricultural tree branches with full foliage in planar architecture for automated sweet-cherry
production and this paper showed the progress to date and opened harvesting. Biosyst. Eng. 146, 3–15.
paths for the future. In the future work, we will study different key AmbientIoT3GPP, 2023. 3Rd generation partnership project; technical specification
group services and system aspects; study on ambient power-enabled internet of
values that SA solutions should satisfy such as social, economic and things (release 19), 3GPP TR 22.840 V2.0.0 (2023-09). (Accessed on 19 September
exclusiveness aspects as the right technology might be different for 2023).
different countries and cultures. Amin, S.U., Hossain, M.S., 2020. Edge intelligence and internet of things in healthcare:
a survey. IEEE Access 9, 45–59.
Anastasiou, E., Balafoutis, A.T., Fountas, S., 2023. Applications of extended reality (XR)
CRediT authorship contribution statement
in agriculture, livestock farming, and aquaculture: A review. Smart Agric. Technol.
3, 100105. http://dx.doi.org/10.1016/j.atech.2022.100105.
Dalhatu Muhammed: Investigation, Resources, Writing – original Antico, T.M., Moreira, L.F.R., Moreira, R., 2022. Evaluating the potential of federated
draft, Methodology. Ehsan Ahvar: Supervision, Writing – review & learning for maize leaf disease prediction. In: Anais do XIX Encontro Nacional de
editing. Shohreh Ahvar: Supervision, Writing – review & editing. Inteligência Artificial e Computacional. SBC, pp. 282–293.
Antonopoulos, V.Z., Antonopoulos, A.V., 2017. Daily reference evapotranspiration
Maria Trocan: Supervision, Writing – review & editing. Marie-José estimates by artificial neural networks technique and empirical equations using
Montpetit: Writing – review & editing. Reza Ehsani: Writing – review limited input climate variables. Comput. Electron. Agric. 132, 86–96.
& editing. Antunes, R.S., André da Costa, C., Küderle, A., Yari, I.A., Eskofier, B., 2022. Federated
learning for healthcare: Systematic review and architecture proposal. ACM Trans.
Intell. Syst. Technol. 13 (4), 1–23.
Declaration of competing interest
Araya, S.N., Fryjoff-Hung, A., Anderson, A., Viers, J.H., Ghezzehei, T.A., 2020. Machine
learning based soil moisture retrieval from unmanned aircraft system multispectral
The authors declare that they have no known competing finan- remote sensing. In: IGARSS 2020-2020 IEEE International Geoscience and Remote
cial interests or personal relationships that could have appeared to Sensing Symposium. IEEE, pp. 4598–4601.
influence the work reported in this paper. Ather, D., Madan, S., Nayak, M., Tripathi, R., Kant, R., Kshatri, S.S., Jain, R.,
2022. Selection of smart manure composition for smart farming using artificial
intelligence technique. J. Food Qual. 2022.
Data availability ATIM, 2023. Monitor the temperature of the vines. Available online: https://www.atim.
com/en/monitore-vineyards-temperature/ (Accessed on 23 January 2023).
No data was used for the research described in the article. Barton, M., Budjac, R., Tanuska, P., Gaspar, G., Schreiber, P., 2022. Identification
overview of industry 4.0 essential attributes and resource-limited embedded
artificial-intelligence-of-things devices for small and medium-sized enterprises.
References Appl. Sci. 12 (11), 5672.
Başağaoğlu, H., Chakraborty, D., Winterle, J., 2021. Reliable evapotranspiration
Abinaya, T., Ishwarya, J., Maheswari, M., 2019. A novel methodology for monitoring predictions with a probabilistic machine learning framework. Water 13 (4), 557.
and controlling of water quality in aquaculture using internet of things (IoT). Bathalapalli, V.K., Mohanty, S.P., Kougianos, E., Yanambaka, V.P., Baniya, B.K.,
In: 2019 International Conference on Computer Communication and Informatics. Rout, B., 2021. A PUF-based approach for sustainable cybersecurity in smart agri-
ICCCI, IEEE, pp. 1–4. culture. In: 2021 19th OITS International Conference on Information Technology.
Adli, H.K., Remli, M.A., Wan Salihin Wong, K.N.S., Ismail, N.A., González-Briones, A., OCIT, IEEE, pp. 375–380.
Corchado, J.M., Mohamad, M.S., 2023. Recent advancements and challenges of Bhat, S.A., Huang, N.-F., 2021. Big data and ai revolution in precision agriculture:
aIoT application in smart agriculture: a review. Sensors 23 (7), 3752. Survey and challenges. IEEE Access 9, 110209–110222.
AeroFarms, 2021. AeroFarms and nokia unveil partnership for next generation Bhatia, G., Joshi, N., Iyengar, S., Rajpal, S., Mahadevan, K., 2021. Crop prediction
AI-enabled plant vision technology. Available online: https://www.aerofarms. based on environmental conditions and disease prediction. In: Information and
com/2021/08/05/aerofarms-and-nokia-unveil-partnership-for-next-generation-ai- Communication Technology for Intelligent Systems: Proceedings of ICTIS 2020,
enabled-plant-vision-technology/. (Accessed on 24 January 2023). Volume 1. Springer, pp. 335–344.
Agribot Platform, 2023. Precision agriculture should be available for the masses - Bhattacherjee, S.S., Shreeshan, S., Priyanka, G., Jadhav, A.R., Rajalakshmi, P.,
we want to help with that. Available online: https://crop.ai/agribot-platform/. Kholova, J., 2020. Cloud based low-power long-range iot network for soil moisture
(Accessed on 23 January 2023). monitoring in agriculture. In: 2020 IEEE Sensors Applications Symposium. SAS,
Ahamed, N.N., Vignesh, R., 2022. Smart agriculture and food industry with blockchain IEEE, pp. 1–5.
and artificial intelligence. J. Computer Sci.. Billah, M., Mehedi, S., Anwar, A., Rahman, Z., Islam, R., et al., 2022. A systematic
Ahmed, N., De, D., Hussain, I., 2018. Internet of things (IoT) for smart precision literature review on blockchain enabled federated learning framework for internet
agriculture and farming in rural areas. IEEE Internet Things J. 5 (6), 4890–4899. of vehicles. arXiv preprint arXiv:2203.05192.
Ahmed, U., Lin, J.C.-W., Srivastava, G., 2022. Privacy-preserving active learning on the Briggs, C., Fan, Z., Andras, P., 2021. A review of privacy-preserving federated learning
internet of 5G connected artificial intelligence of things. IEEE Internet Things Mag. for the internet-of-things. Federated Learn. Syst. 21–50.
5 (1), 126–129. Bronner, W., Gebauer, H., Lamprecht, C., Wortmann, F., 2021. Sustainable aIoT: How
Ahvar, E., Ahvar, S., Lee, G.M., 2022. Artificial intelligence of things: Architectures, artificial intelligence and the internet of things affect profit, people, and planet.
applications and challenges. In: Springer Handbook of Internet of Things. Springer. In: Connected Business. Springer, pp. 137–154.
Alahi, M.E.E., Pereira-Ishak, N., Mukhopadhyay, S.C., Burkitt, L., 2018. An internet-of- Campos, E.M., Saura, P.F., González-Vidal, A., Hernández-Ramos, J.L., Bernabe, J.B.,
things enabled smart sensing system for nitrate monitoring. IEEE Internet Things Baldini, G., Skarmeta, A., 2021. Evaluating federated learning for intrusion
J. 5 (6), 4409–4417. detection in internet of things: Review and challenges. Comput. Netw. 108661.

22
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Chakraborty, R.S., Bhunia, S., 2009. Security against hardware trojan through a novel Elbeltagi, A., Nagy, A., Mohammed, S., Pande, C.B., Kumar, M., Bhat, S.A., Zsembeli, J.,
application of design obfuscation. In: 2009 IEEE/ACM International Conference on Huzsvai, L., Tamás, J., Kovács, E., et al., 2022. Combination of limited meteoro-
Computer-Aided Design-Digest of Technical Papers. IEEE, pp. 113–116. logical data for predicting reference crop evapotranspiration using artificial neural
Chamola, V., Hassija, V., Gupta, V., Guizani, M., 2020. A comprehensive review of network method. Agronomy 12 (2), 516.
the COVID-19 pandemic and the role of IoT, drones, AI, blockchain, and 5G in Elzeard, 2023. The production management application dedicated to fruit and vegetable
managing its impact. Ieee access 8, 90225–90265. producers. Available online: https://elzeard.co/.(accessed on 24 January 2023).
Chandra, R., Swaminathan, M., Chakraborty, T., Ding, J., Kapetanovic, Z., Kumar, P., Esenogho, E., Djouani, K., Kurien, A., 2022. Integrating artificial intelligence internet
Vasisht, D., 2022. Democratizing data-driven agriculture using affordable hardware. of things and 5G for next-generation smartgrid: A survey of trends challenges and
IEEE Micro 42 (1), 69–77. prospect. IEEE Access.
Chang, Z., Liu, S., Xiong, X., Cai, Z., Tu, G., 2021. A survey of recent advances in farmsio, 2023. Agriculture and climate data solutions combined for people, planet,
edge-computing-powered artificial intelligence of things. IEEE Internet Things J.. profit and purpose. Available online: https://farms.io (Accessed on 06 March 2023).
Chang, W.-J., Su, J.-P., Hsu, C.-H., Chen, L.-B., Chen, M.-C., Chen, H.-C., Lin, C.-F., Feng, J.-C., Sun, L., Yan, J., 2023. Carbon sequestration via shellfish farming: A
2019. iCAP: An IoT-based intelligent liquid waste barrels monitoring system. In: potential negative emissions technology. Renew. Sustain. Energy Rev. 171, 113018.
2019 11th Computer Science and Electronic Engineering. CEEC, IEEE, pp. 156–159. Ferrández-Pastor, F.J., García-Chamizo, J.M., Nieto-Hidalgo, M., Mora-Martínez, J.,
Chatterjee, P.S., Ray, N.K., Mohanty, S.P., 2021. LiveCare: An IoT-based healthcare 2018. Precision agriculture design method using a distributed computing
framework for livestock in smart agriculture. IEEE Trans. Consum. Electron. 67 architecture on internet of things context. Sensors 18 (6), 1731.
(4), 257–265. Ferrández-Pastor, F.J., García-Chamizo, J.M., Nieto-Hidalgo, M., Mora-Pascual, J., Mora-
Chen, S.-W., Gu, X.-W., Wang, J.-J., Zhu, H.-S., 2021. AIoT used for COVID-19 pandemic Martínez, J., 2016. Developing ubiquitous sensor network platform using internet
prevention and control. Contrast Media Molecular Imag. 2021. of things: Application in precision agriculture. Sensors 16 (7), 1141.
Chen, C.-J., Huang, Y.-Y., Li, Y.-S., Chang, C.-Y., Huang, Y.-M., 2020. An aIoT based Ghimire, B., Rawat, D.B., 2022. Recent advances on federated learning for cybersecurity
smart agricultural system for pests detection. IEEE Access 8, 180750–180761. and cybersecurity for federated learning for internet of things. IEEE Internet Things
Chen, W.-L., Lin, Y.-B., Ng, F.-L., Liu, C.-Y., Lin, Y.-W., 2019. RiceTalk: Rice blast J..
detection using internet of things and artificial intelligence technologies. IEEE Ghoreishi, M., Treves, L., Kuivalainen, O., 2022. Artificial intelligence of things as
Internet Things J. 7 (2), 1001–1010. an accelerator of circular economy in international business. In: Megatrends in
Cheng, Z., Zhang, F., 2020. Flower end-to-end detection based on YOLOv4 using a International Business. Springer, pp. 83–104.
mobile device. Wirel. Commun. Mob. Comput. 2020. Gia, T.N., Qingqing, L., Queralta, J.P., Zou, Z., Tenhunen, H., Westerlund, T., 2019.
Chiu, M.-C., Yan, W.-M., Bhat, S.A., Huang, N.-F., 2022. Development of smart Edge AI in smart farming IoT: CNNs at the edge and fog computing with lora. In:
aquaculture farm management system using IoT and AI-based surrogate models. 2019 IEEE AFRICON. IEEE, pp. 1–6.
J. Agric. Food Res. 9, 100357.
Granata, F., 2019. Evapotranspiration evaluation models based on machine learning
Chougule, A., Jha, V.K., Mukhopadhyay, D., 2019. Crop suitability and fertilizers rec-
algorithms—A comparative study. Agricult. Water Manag. 217, 303–315.
ommendation using data mining techniques. In: Progress in Advanced Computing
Guillén-Navarro, M.A., Martínez-España, R., Bueno-Crespo, A., Morales-García, J.,
and Intelligent Engineering. Springer, pp. 205–213.
Ayuso, B., Cecilia, J.M., 2020. A decision support system for water optimization in
Cvitić, I., Peraković, D., Periša, M., Gupta, B., 2021. Ensemble machine learning
anti-frost techniques by sprinklers. Sensors 20 (24), 7129.
approach for classification of IoT devices in smart home. Int. J. Mach. Learn.
Gülen, K., 2023. Elevating ML to new heights with distributed learning. Avail-
Cybern. 12 (11), 3179–3202.
able online: https://dataconomy.com/2023/05/22/what-is-distributed-learning-in-
Dahane, A., Benameur, R., Kechar, B., Benyamina, A., 2020. An IoT based smart farming
ml/. (Accessed on 11 October 2023).
system using machine learning. In: 2020 International Symposium on Networks,
Guo, Y., Zhang, J., Yin, C., Hu, X., Zou, Y., Xue, Z., Wang, W., 2020. Plant disease
Computers and Communications. ISNCC, IEEE, pp. 1–6.
identification based on deep learning algorithm in smart farming. Discrete Dyn.
Dia, I., Ahvar, E., Lee, G.M., 2022. Performance evaluation of machine learning and
Nat. Soc. 2020.
neural network-based algorithms for predicting segment availability in aIoT-based
Guo, H., Zhou, X., Liu, J., Zhang, Y., 2022. Vehicular intelligence in 6G: Networking,
smart parking. Network 2 (2), 225–238.
communications, and computing. Veh. Commun. 33, 100399.
Digiteum, 2022. Difference between cloud, fog and edge computing in IoT.. Available
Gupta, M., Abdelsalam, M., Khorsandroo, S., Mittal, S., 2020. Security and privacy in
online: https://www.digiteum.com/cloud-fog-edge-computing-iot/. (Accessed on
smart farming: Challenges and opportunities. IEEE Access 8, 34564–34584.
22 January 2023).
Haban, J.J.I., Puno, J.C.V., Bandala, A.A., Billones, R.K., Dadios, E.P., Sybingco, E.,
Dong, B., Shi, Q., Yang, Y., Wen, F., Zhang, Z., Lee, C., 2021. Technology evolution
2020. Soil fertilizer recommendation system using fuzzy logic. In: 2020 IEEE
from self-powered sensors to aIoT enabled smart homes. Nano Energy 79, 105414.
REGION 10 CONFERENCE. TENCON, IEEE, pp. 1171–1175.
Dong, W., Wu, T., Sun, Y., Luo, J., 2018. Digital mapping of soil available phosphorus
supported by AI technology for precision agriculture. In: 2018 7th International Hao, Y., Miao, Y., Chen, M., Gharavi, H., Leung, V.C., 2021. 6G cognitive information
Conference on Agro-Geoinformatics (Agro-Geoinformatics). IEEE, pp. 1–5. theory: A mailbox perspective. Big Data Cogn. Comput. 5 (4), 56.
Doshi, Z., Nadkarni, S., Agrawal, R., Shah, N., 2018. AgroConsultant: intelligent Hashni, T., Amudha, T., Ramakrishnan, S., 2022. Smart farming approaches towards
crop recommendation system using machine learning algorithms. In: 2018 Fourth sustainable agriculture—A survey. In: Proceedings of Third International Conference
International Conference on Computing Communication Control and Automation. on Intelligent Computing, Information and Control Systems. Springer, pp. 695–714.
ICCUBEA, IEEE, pp. 1–6. Holzworth, D.P., Huth, N.I., deVoil, P.G., Zurcher, E.J., Herrmann, N.I., McLean, G.,
Drainakis, G., Katsaros, K.V., Pantazopoulos, P., Sourlas, V., Amditis, A., 2020a. Fed- Chenu, K., van Oosterom, E.J., Snow, V., Murphy, C., et al., 2014. APSIM–evolution
erated vs. centralized machine learning under privacy-elastic users: A comparative towards a new generation of agricultural systems simulation. Environ. Model.
analysis. In: 2020 IEEE 19th International Symposium on Network Computing and Softw. 62, 327–350.
Applications. NCA, IEEE, pp. 1–8. Holzworth, D., Huth, N.I., Fainges, J., Brown, H., Zurcher, E., Cichota, R., Verrall, S.,
Drainakis, G., Katsaros, K.V., Pantazopoulos, P., Sourlas, V., Amditis, A., 2020b. Fed- Herrmann, N.I., Zheng, B., Snow, V., 2018. APSIM next generation: Overcoming
erated vs. Centralized machine learning under privacy-elastic users: A comparative challenges in modernising a farming systems model. Environ. Model. Softw. 103,
analysis. In: 2020 IEEE 19th International Symposium on Network Computing and 43–51.
Applications. NCA, pp. 1–8. http://dx.doi.org/10.1109/NCA51143.2020.9306745. Holzworth, D., Meinke, H., DeVoil, P., Wegener, M., Huth, N., Hammer, G., How-
Du, Z., Wu, C., Yoshinaga, T., Yau, K.-L.A., Ji, Y., Li, J., 2020. Federated learning den, M., Robertson, M., Carberry, P., Freebairn, D., et al., 2006. The development
for vehicular internet of things: Recent advances and open issues. IEEE Open J. of a farming systems model (APSIM)-a disciplined approach. In: Proceedings of the
Comput. Soc. 1, 45–61. IEMSs 3rd Biennial Meeting," Summit on Environmental Modelling and Software".
Durai, S.K.S., Shamili, M.D., 2022. Smart farming using machine learning and deep Hou, D., Zhang, J., Man, K.L., Ma, J., Peng, Z., 2021. A systematic literature review
learning techniques. Decis. Anal. J. 3, 100041. of blockchain-based federated learning: Architectures, applications and issues. In:
Durga, S., Nag, R., Daniel, E., 2019. Survey on machine learning and deep learning 2021 2nd Information Communication Technologies Conference. ICTC, IEEE, pp.
algorithms used in internet of things (IoT) healthcare. In: 2019 3rd International 302–307.
Conference on Computing Methodologies and Communication. ICCMC, IEEE, pp. Hsu, C.-W., Huang, Y.-H., Huang, N.-F., 2022. Real-time dragonfruit’s ripeness classifi-
1018–1022. cation system with edge computing based on convolution neural network. In: 2022
Durrant, A., Markovic, M., Matthews, D., May, D., Enright, J., Leontidis, G., 2022. The International Conference on Information Networking. ICOIN, IEEE, pp. 177–182.
role of cross-silo federated learning in facilitating data sharing in the agri-food Hu, W.-C., Chen, L.-B., Huang, B.-K., Lin, H.-M., 2022. A computer vision-based
sector. Comput. Electron. Agric. 193, 106648. intelligent fish feeding system using deep learning techniques for aquaculture. IEEE
Ebrahimi, M., Kelati, A., Nkonoki, E., Kondoro, A., Rwegasira, D., Dhaou, I.B., Sens. J. 22 (7), 7185–7194.
Taajamaa, V., Tenhunen, H., 2019. Creation of CERID: Challenge, education, Huang, C.-H., Chou, T.-C., Wu, S.-H., 2021. Towards convergence of ai and IoT for
research, innovation, and deployment ‘‘in the context of smart MicroGrid’’. In: 2019 smart policing: a case of a mobile edge computing-based context-aware system. J.
IST-Africa Week Conference (IST-Africa). IEEE, pp. 1–8. Global Inf. Manag. (JGIM) 29 (6), 1–21.
Elbeltagi, A., Deng, J., Wang, K., Hong, Y., 2020. Crop water footprint estimation and Huang, K., Shu, L., Li, K., Yang, F., Han, G., Wang, X., Pearson, S., 2020. Photovoltaic
modeling using an artificial neural network approach in the Nile Delta, Egypt. agricultural internet of things towards realizing the next generation of smart
Agricult. Water Manag. 235, 106080. farming. IEEE Access 8, 76300–76312.

23
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Idoje, G., Dagiuklas, T., Iqbal, M., 2023. Federated learning: Crop classification in a Li, H., Li, S., Yu, J., Han, Y., Dong, A., 2022c. AIoT platform design based on front
smart farm decentralised network. Smart Agric. Technol. 5, 100277. and rear end separation architecture for smart agricultural. In: Proceedings of the
ISAGRI, 2023. Agricultural software for grain growers. Available online: 2022 4th Asia Pacific Information Technology Conference. APIT ’22, Association
https://www.isagri.fr/solutions/isagri-prendre-les-meilleures-decisions-sur-votre- for Computing Machinery, New York, NY, USA, pp. 208–214. http://dx.doi.org/10.
exploitation. (Accessed on 23 January 2023). 1145/3512353.3512384.
Islam, N., Rashid, M.M., Pasandideh, F., Ray, B., Moore, S., Kadel, R., 2021. A review Li, H., Li, S., Yu, J., Han, Y., Dong, A., 2022d. Plant disease and insect pest
of applications and communication technologies for internet of things (iot) and identification based on vision transformer. In: International Conference on Internet
unmanned aerial vehicle (uav) based sustainable smart farming. Sustainability 13 of Things and Machine Learning, Vol, 12174. IoTML 2021, SPIE, pp. 194–201.
(4), 1821. Li, Q., Wen, Z., Wu, Z., Hu, S., Wang, N., Li, Y., Liu, X., He, B., 2021. A survey on
Istvan, D., Archambault, P., Quentin Wolak, Q., Vu, C.V., Lalonde, T., Riaz, R., Syri- federated learning systems: vision, hype and reality for data privacy and protection.
ani, E., Sahraoui, H., 2023. Digital twins for cyber-biophysical systems: Challenges IEEE Trans. Knowl. Data Eng..
and lessons learned. Lin, B.-S., 2021. Toward an AI-enabled O-RAN-based and SDN/NFV-driven 5G& IoT
Jadav, N.K., Rathod, T., Gupta, R., Tanwar, S., Kumar, N., Alkhayyat, A., 2023. network era. Netw. Commun. Technol. 6 (1), 6–15.
Blockchain and artificial intelligence-empowered smart agriculture framework for Liu, J., Chai, Y., Xiang, Y., Zhang, X., Gou, S., Liu, Y., 2018. Clean energy consumption
maximizing human life expectancy. Comput. Electr. Eng. 105, 108486. of power systems towards smart agriculture: roadmap, bottlenecks and technologies.
Jain, P., Choudhury, S.B., Bhatt, P., Sarangi, S., Pappula, S., 2020. Maximising value CSEE J. Power Energy Syst. 4 (3), 273–282.
of frugal soil moisture sensors for precision agriculture applications. In: 2020 Liu, Y., Han, W., Zhang, Y., Li, L., Wang, J., Zheng, L., 2016. An internet-of-things
IEEE/ITU International Conference on Artificial Intelligence for Good. AI4G, IEEE, solution for food safety and quality control: A pilot project in China. J. Ind. Inf.
pp. 63–70. Integr. 3, 1–7.
Jain, A., Kushwah, R., Swaroop, A., Yadav, A., 2021. Role of artificial intelligence Liu, W., Lin, H., Wang, X., Hu, J., Kaddoum, G., Piran, M.J., Alamri, A., 2021. D2MIF: A
of things (aIoT) to combat pandemic COVID-19. In: Handbook of Research on malicious model detection mechanism for federated learning empowered artificial
Innovations and Applications of AI, IoT, and Cognitive Technologies. IGI Global, intelligence of things. IEEE Internet Things J..
pp. 117–128. Liu, Y., Ma, X., Shu, L., Hancke, G.P., Abu-Mahfouz, A.M., 2020. From industry 4.0
Jiang, P., Chen, Y., Liu, B., He, D., Liang, C., 2019. Real-time detection of apple leaf to agriculture 4.0: Current status, enabling technologies, and research challenges.
diseases using deep learning approach based on improved convolutional neural IEEE Trans. Ind. Inform. 17 (6), 4322–4334.
networks. IEEE Access 7, 59069–59080. Liu, L., Wang, R., Xie, C., Yang, P., Wang, F., Sudirman, S., Liu, W., 2019. PestNet:
Jiang, J.C., Kantarci, B., Oktug, S., Soyata, T., 2020. Federated learning in smart city An end-to-end deep learning approach for large-scale multi-class pest detection and
sensing: Challenges and opportunities. Sensors 20 (21), 6230. classification. IEEE Access 7, 45301–45312.
Jung, D.-H., Kim, H.S., Jhin, C., Kim, H.-J., Park, S.H., 2020. Time-serial analysis Mahlool, D.H., Abed, M.H., 2022. A comprehensive survey on federated learning:
of deep neural network models for prediction of climatic conditions inside a Concept and applications. arXiv preprint arXiv:2201.09384.
greenhouse. Comput. Electron. Agric. 173, 105402.
Manikandan, R., Ranganathan, G., Bindhu, V., 2022. Deep learning based IoT module
Kakhi, K., Alizadehsani, R., Kabir, H.D., Khosravi, A., Nahavandi, S., Acharya, U.R.,
for smart farming in different environmental conditions. Wirel. Pers. Commun.
2022. The internet of medical things and artificial intelligence: trends, challenges,
1–18.
and opportunities. Biocybern. Biomed. Eng..
Manoj, T., Makkithaya, K., Narendra, V., 2022. A federated learning-based crop yield
Katiyar, S., Farhana, A., 2021. Smart agriculture: The future of agriculture using AI
prediction for agricultural production risk management. In: 2022 IEEE Delhi Section
and IoT. J. Comput. Sci. 17 (10), 984–999.
Conference. DELCON, IEEE, pp. 1–7.
Kedlaya, A., Sana, A., Bhat, B.A., Kumar, S., Bhat, N., et al., 2021. An efficient algorithm
Mehra, M., Saxena, S., Sankaranarayanan, S., Tom, R.J., Veeramanikandan, M., 2018.
for predicting crop using historical data and pattern matching technique. Global
IoT based hydroponics system using deep neural networks. Comput. Electron. Agric.
Trans. Proc. 2 (2), 294–298.
155, 473–486.
Kempenaar, C., Lokhorst, C., Bleumer, E., Veerkamp, R., Been, T., van Evert, F.,
Metos, 2022. Pessl instrument launches new module within FieldClimate’s FarmView
Boogaardt, M., Ge, L., Wolfert, J., Verdouw, C., et al., 2016. Big data analysis
– yield prediction. Available online: https://metos.at/it/pessl-instrument-launches-
for smart farming: Results of TO2 project in theme food security. Wageningen
new-module-within-fieldclimates-farmview-yield-prediction/. (Accessed on 23
University & Research.
January 2023).
Khan, F.S., Khan, S., Mohd, M.N.H., Waseem, A., Khan, M.N.A., Ali, S., Ahmed, R.,
Miao, H.-Y., Yang, C.-T., Kristiani, E., Fathoni, H., Lin, Y.-S., Chen, C.-Y., 2022. On
2022. Federated learning-based UAVs for the diagnosis of plant diseases. In: 2022
construction of a campus outdoor air and water quality monitoring system using
International Conference on Engineering and Emerging Technologies. ICEET, IEEE,
LoRaWAN. Appl. Sci. 12 (10), 5018.
pp. 1–6.
Mitra, A., Vangipuram, S.L., Bapatla, A.K., Bathalapalli, V.K., Mohanty, S.P.,
Khattab, A., Abdelgawad, A., Yelmarthi, K., 2016. Design and implementation of a
Kougianos, E., Ray, C., 2022. Everything you wanted to know about smart
cloud-based IoT scheme for precision agriculture. In: 2016 28th International
agriculture. arXiv preprint arXiv:2201.04754.
Conference on Microelectronics. ICM, IEEE, pp. 201–204.
Klibi, S., Tounsi, K., Rebah, Z.B., Solaiman, B., Farah, I.R., 2020. Soil salinity prediction Mohammed, T.S., Khan, O.F., Al-Bazi, A., 2019. A novel algorithm based on LoRa tech-
using a machine learning approach through hyperspectral satellite image. In: 2020 nology for open-field and protected agriculture smart irrigation system. In: 2019
5th International Conference on Advanced Technologies for Signal and Image 2nd IEEE Middle East and North Africa COMMunications Conference. MENACOMM,
Processing. ATSIP, IEEE, pp. 1–6. IEEE, pp. 1–6.
Knud, L.L., 2019. 40+ Emerging IoT technologies you should have on your radar. Moinet, G.Y., Hijbeek, R., van Vuuren, D.P., Giller, K.E., 2023. Carbon for soils, not
https://iot-analytics.com/40-emerging-iot-technologies-you-should-have-on-your- soils for carbon. Global Change Biol..
radar/. Molokomme, D.N., Onumanyi, A.J., Abu-Mahfouz, A.M., 2022. Edge intelligence in
Kuguoglu, B.K., van der Voort, H., Janssen, M., 2021. The giant leap for smart cities: smart grids: A survey on architectures, offloading models, cyber security measures,
Scaling up smart city artificial intelligence of things (aIoT) initiatives. Sustainability and challenges. J. Sensor Actuator Netw. 11 (3), 47.
13 (21), 12295. Mothukuri, V., Parizi, R.M., Pouriyeh, S., Huang, Y., Dehghantanha, A., Srivastava, G.,
Kulkarni, N.H., Srinivasan, G., Sagar, B., Cauvery, N., 2018. Improving crop productivity 2021. A survey on security and privacy of federated learning. Future Gener.
through a crop recommendation system using ensembling technique. In: 2018 3rd Comput. Syst. 115, 619–640.
International Conference on Computational Systems and Information Technology Muhammed, D., Ahvar, E., Ahvar, S., Trocan, M., 2022. A user-friendly aIoT-based
for Sustainable Solutions. CSITSS, IEEE, pp. 114–119. crop recommendation system (UACR): concept and architecture. In: 2022 16th
Kumar, S., Chowdhary, G., Udutalapally, V., Das, D., Mohanty, S.P., 2019a. Gcrop: International Conference on Signal-Image Technology & Internet-Based Systems.
Internet-of-leaf-things (IoLT) for monitoring of the growth of crops in smart SITIS, IEEE, pp. 569–576.
agriculture. In: 2019 IEEE International Symposium on Smart Electronic Systems. Murugamani, C., Shitharth, S., Hemalatha, S., Kshirsagar, P.R., Riyazuddin, K.,
ISES, Formerly INiS, IEEE, pp. 53–56. Naveed, Q.N., Islam, S., Mazher Ali, S.P., Batu, A., 2022. Machine learning
Kumar, A., Sarkar, S., Pradhan, C., 2019b. Recommendation system for crop identifi- technique for precision agriculture applications in 5G-based internet of things.
cation and pest control technique in agriculture. In: 2019 International Conference Wirel. Commun. Mob. Comput. 2022.
on Communication and Signal Processing. ICCSP, IEEE, pp. 0185–0189. MyEasyFarm, 2023. Precision agriculture easy and connected. Available online: https:
LEMKEN, 2023. Innovative farming technology for high-yield agriculture. Available //www.myeasyfarm.com/solutions/myeasyfarm/. (Accessed on 23 January 2023).
online: https://lemken.com/en-en/agricultural-machines/soil-cultivation. (Accessed Nahr, J.G., Nozari, H., Sadeghi, M.E., 2021. Green supply chain based on artificial
on 23 January 2023). intelligence of things (aIoT). Int. J. Innov. Manag. Econ. Soc. Sci. 1 (2), 56–63.
Li, D., Han, D., Weng, T.-H., Zheng, Z., Li, H., Liu, H., Castiglione, A., Li, K.- Nema, M.K., Khare, D., Chandniha, S.K., 2017. Application of artificial intelligence to
C., 2022a. Blockchain for federated learning toward secure distributed machine estimate the reference evapotranspiration in sub-humid doon valley. Appl. Water
learning systems: a systemic survey. Soft Comput. 26 (9), 4423–4440. Sci. 7 (7), 3903–3910.
Li, H., Li, S., Yu, J., Han, Y., Dong, A., 2022b. AIoT platform design based on front and Nguyen, D.C., Ding, M., Pathirana, P.N., Seneviratne, A., Li, J., Poor, H.V., 2021.
rear end separation architecture for smart agricultural. In: 2022 4th Asia Pacific Federated learning for internet of things: A comprehensive survey. IEEE Commun.
Information Technology Conference. pp. 208–214. Surv. Tutor. 23 (3), 1622–1658.

24
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Nguyen, D.C., Pham, Q.-V., Pathirana, P.N., Ding, M., Seneviratne, A., Lin, Z., Dobre, O., Qian, K., Zhang, Z., Yamamoto, Y., Schuller, B.W., 2021. Artificial intelligence internet
Hwang, W.-J., 2022. Federated learning for smart healthcare: A survey. ACM of things for the elderly: From assisted living to health-care monitoring. IEEE Signal
Comput. Surv. 55 (3), 1–37. Process. Mag. 38 (4), 78–88.
Ning, K., 2021. Data driven artificial intelligence techniques in renewable energy system Rahmouni, M., Hanifi, M., Savaglio, C., Fortino, G., Ghogho, M., 2022a. An aIoT frame-
(Ph.D. thesis). Massachusetts Institute of Technology. work for precision agriculture. In: 2022 IEEE Intl Conf on Dependable, Autonomic
Nokia, 2023. Real action: Smart agriculture. and Secure Computing, Intl Conf on Pervasive Intelligence and Computing, Intl
Nozari, H., Szmelter-Jarosz, A., Ghahremani-Nahr, J., 2022. Analysis of the challenges Conf on Cloud and Big Data Computing, Intl Conf on Cyber Science and Technology
of artificial intelligence of things (aIoT) for the smart supply chain (case study: Congress. DASC/PiCom/CBDCom/CyberSciTech, IEEE, pp. 1–6.
FMCG industries). Sensors 22 (8), 2931. Rahmouni, M., Hanifi, M., Savaglio, C., Fortino, G., Ghogho, M., 2022b. An aIoT
NS Agriculture Staff Writer, 2020. Nokia, vodafone India foundation to im- framework for precision agriculture. In: 2022 IEEE Intl Conf on Dependable, Auto-
prove farmers’ productivity in India. A pilot project is being implemented nomic and Secure Computing, Intl Conf on Pervasive Intelligence and Computing,
across 100 locations in the Indian states of Madhya Pradesh and Maharash- Intl Conf on Cloud and Big Data Computing, Intl Conf on Cyber Science and
tra Available online: https://www.nsagriculture.com/news/nokia-vodafone-india- Technology Congress. DASC/PiCom/CBDCom/CyberSciTech, pp. 1–6. http://dx.doi.
foundation-productivity/. (Accessed on 27 January 2023). org/10.1109/DASC/PiCom/CBDCom/Cy55231.2022.9927989.
Nursyahid, A., Aprilian, T., Setyawan, T.A., Nugroho, A.S., Susilo, D., et al., 2019. Rai, H.M., Chauhan, M., Sharma, H., Bhardwaj, N., Kumar, L., 2022. AgriBot: Smart
Automatic sprinkler system for water efficiency based on lora network. In: 2019 autonomous agriculture robot for multipurpose farming application using IOT.
6th International Conference on Information Technology, Computer and Electrical In: Emerging Technologies for Computing, Communication and Smart Cities:
Engineering. ICITACEE, IEEE, pp. 1–6. Proceedings of ETCCS 2021. Springer, pp. 491–503.
OpenFogConsortium, 2017. OpenFog reference architecture for fog computing.. Ram, S.K., Sahoo, S.R., Das, B.B., Mahapatra, K., Mohanty, S.P., 2020. Eternal-thing:
Available online: https://www.openFogconsortium.org/ra/. (Accessed on 30 May A secure aging-aware solar-energy harvester thing for sustainable IoT. IEEE Trans.
2021)s. Sustain. Comput. 6 (2), 320–333.
Pallagani, V., Khandelwal, V., Chandra, B., Udutalapally, V., Das, D., Mohanty, S.P., Ramu, S.P., Boopalan, P., Pham, Q.-V., Maddikunta, P.K.R., Huynh-The, T., Alazab, M.,
2019. Dcrop: A deep-learning based framework for accurate prediction of diseases Nguyen, T.T., Gadekallu, T.R., 2022. Federated learning enabled digital twins for
of crops in smart agriculture. In: 2019 IEEE International Symposium on Smart smart cities: Concepts, recent advances, and future directions. Sustainable Cities
Electronic Systems (ISES)(Formerly INiS). IEEE, pp. 29–33. Soc. 79, 103663.
Pan, Y., Zhang, L., 2021. Roles of artificial intelligence in construction engineering and Ramya, M., 2021. What is fog computing? Components, examples, and best
management: A critical review and future trends. Autom. Constr. 122, 103517. practices. Available online: https://www.spiceworks.com/tech/edge-computing/
Pande, S.M., Ramesh, P.K., Anmol, A., Aishwarya, B., Rohilla, K., Shaurya, K., 2021. articles/what-is-fog-computing/. (Accessed on 27 January 2023).
Crop recommender system using machine learning approach. In: 2021 5th Interna- Ray, B., 2018. SigFox vs. LoRa: A comparison between technologies & business models.
tional Conference on Computing Methodologies and Communication. ICCMC, IEEE, Link Labs (mayo de 2018). https://www.link-labs.com/blog/sigfox-vs-lora.
pp. 1066–1071. Reddy, D.A., Dadore, B., Watekar, A., 2019. Crop recommendation system to maximize
Panduman, Y.Y.F., Funabiki, N., Fajrianti, E.D., Fang, S., Sukaridhoto, S., 2024. A
crop yield in ramtek region using machine learning. Int. J. Sci. Res. Sci. Technol.
survey of AI techniques in IoT applications with use case investigations in the smart
6 (1), 485–489.
environmental monitoring and analytics in real-time IoT platform. Information 15
Reddy, D.J., Kumar, M.R., 2021. Crop yield prediction using machine learning algo-
(3), 153.
rithm. In: 2021 5th International Conference on Intelligent Computing and Control
Pandya, S., Srivastava, G., Jhaveri, R., Babu, M.R., Bhattacharya, S., Maddikunta, P.K.R.,
Systems. ICICCS, IEEE, pp. 1466–1470.
Mastorakis, S., Piran, M.J., Gadekallu, T.R., 2023. Federated learning for smart
Rettore de Araujo Zanella, A., da Silva, E., Albini, L.C.P., 2020. Security challenges to
cities: A comprehensive survey. Sustain. Energy Technol. Assess. 55, 102987.
smart agriculture: Current state, key issues, and future directions. Array 8, 100048.
Pappakrishnan, V.K., Mythili, R., Kavitha, V., Parthiban, N., 2021. Role of artificial
Rico, M., Taverna, M.L., Galli, M.R., Caliusco, M.L., 2023. Context-aware representation
intelligence of things (aIoT) in Covid-19 pandemic: A brief survey. IoTBDS 229–236.
of digital twins’ data: The ontology network role. Comput. Ind. 146, 103856.
Partel, V., Kakarla, S.C., Ampatzidis, Y., 2019. Development and evaluation of a
http://dx.doi.org/10.1016/j.compind.2023.103856.
low-cost and smart technology for precision weed management utilizing artificial
Ruvinga, S., Hunter, G., Duran, O., Nebel, J.-C., 2023. Identifying queenlessness
intelligence. Comput. Electron. Agric. 157, 339–350.
in honeybee hives from audio signals using machine learning. Electronics 12
Patel, K., Patel, H.B., 2020. A state-of-the-art survey on recommendation system and
(7), http://dx.doi.org/10.3390/electronics12071627, URL https://www.mdpi.com/
prospective extensions. Comput. Electron. Agric. 178, 105779.
2079-9292/12/7/1627.
Patel, C., Vyas, S., Saikia, P., et al., 2022. A futuristic survey on learning techniques
Saberi Anari, M., 2022. A hybrid model for leaf diseases classification based on the
for internet of things (IoT) security: Developments, applications, and challenges.
modified deep transfer learning and ensemble approach for agricultural AIoT-based
Comput. Secur. J..
monitoring. Comput. Intell. Neurosci. 2022.
Pathan, M., Patel, N., Yagnik, H., Shah, M., 2020. Artificial cognition for applications
Sadia, S., Propa, M.B., Al Mamun, K.S., Kaiser, M.S., 2021. A fruit cultivation
in smart agriculture: A comprehensive review. Artif. Intell. Agric. 4, 81–95.
Patil, N., Kelkar, S., Ranawat, M., Vijayalakshmi, M., 2021. Krushi sahyog: Plant disease recommendation system based on pearson’s correlation co-efficient. In: 2021 Inter-
identification and crop recommendation using artificial intelligence. In: 2021 2nd national Conference on Information and Communication Technology for Sustainable
International Conference for Emerging Technology. INCET, IEEE, pp. 1–6. Development. ICICT4SD, IEEE, pp. 361–365.
Patros, P., Ooi, M., Huang, V., Mayo, M., Anderson, C., Burroughs, S., Baughman, M., Sadowski, S., Spachos, P., 2020. Wireless technologies for smart agricultural monitoring
Almurshed, O., Rana, O., Chard, R., et al., 2022. Rural ai: Serverless-powered using internet of things devices with energy harvesting capabilities. Comput.
federated learning for remote applications. IEEE Internet Comput.. Electron. Agric. 172, 105338.
Paul, P.B., Biswas, S., Bairagi, A.K., Masud, M., 2021. Data-driven decision making Saha, R., Misra, S., Deb, P.K., 2020. FogFL: Fog-assisted federated learning for
for smart cultivation. In: 2021 IEEE International Symposium on Smart Electronic resource-constrained IoT devices. IEEE Internet Things J. 8 (10), 8456–8463.
Systems (Formerly INiS). ISES, IEEE, pp. 249–254. Sahitya, G., Balaji, N., Naidu, C., 2016. Wireless sensor network for smart agriculture.
Pawar, M., Chillarge, G., 2018. Soil toxicity prediction and recommendation system In: 2016 2nd International Conference on Applied and Theoretical Computing and
using data mining in precision agriculture. In: 2018 3rd International Conference Communication Technology. ICATccT, IEEE, pp. 488–493.
for Convergence in Technology. I2CT, IEEE, pp. 1–5. Salih, K.O.M., Rashid, T.A., Radovanovic, D., Bacanin, N., 2022. A comprehensive
Payen, F.T., Moran, D., Cahurel, J.-Y., Aitkenhead, M., Alexander, P., MacLeod, M., survey on the internet of things with the industrial marketplace. Sensors 22 (3),
2023. Why do french winegrowers adopt soil organic carbon sequestration 730.
practices? Understanding motivations and barriers. Front. Sustain. Food Syst. 6, Satyajit, S., 2022. T55+ emerging IoT technologies you should have on your radar
994364. (2022 update). https://iot-analytics.com/iot-technologies/.
Pfitzner, B., Steckhan, N., Arnrich, B., 2021. Federated learning in a medical context: Schwalbert, R.A., Amado, T., Corassa, G., Pott, L.P., Prasad, P.V., Ciampitti, I.A.,
a systematic literature review. ACM Trans. Internet Technol. (TOIT) 21 (2), 1–31. 2020. Satellite-based soybean yield forecast: Integrating machine learning and
Pham, Q.-V., Dev, K., Maddikunta, P.K.R., Gadekallu, T.R., Huynh-The, T., et al., 2021. weather data for improving crop yield prediction in southern Brazil. Agricult. Forest
Fusion of federated learning and industrial internet of things: A survey. arXiv Meteorol. 284, 107886.
preprint arXiv:2101.00798. Seng, K.P., Ang, L.M., Ngharamike, E., 2022. Artificial intelligence internet of things:
Pise, A.A., Almusaini, K.K., Ahanger, T.A., Farouk, A., Pareek, P.K., Nuagah, S.J., et A new paradigm of distributed sensor networks. Int. J. Distrib. Sens. Netw. 18 (3),
al., 2022. Enabling artificial intelligence of things (aIoT) healthcare architectures 15501477211062835.
and listing security issues. Comput. Intell. Neurosci. 2022. Setiadi, T., Noviyanto, F., Hardianto, H., Tarmuji, A., Fadlil, A., Wibowo, M., 2020.
Priyadharshini, A., Chakraborty, S., Kumar, A., Pooniwala, O.R., 2021. Intelligent Implementation of naïve bayes method in food crops planting recommendation.
crop recommendation system using machine learning. In: 2021 5th International Int. J. Sci. Technol. Res 9 (02), 4750–4755.
Conference on Computing Methodologies and Communication. ICCMC, IEEE, pp. Sigov, A., Ratkin, L., Ivanov, L.A., Xu, L.D., 2022. Emerging enabling technologies for
843–848. industry 4.0 and beyond. Inf. Syst. Front. 1–11.
Qazi, S., Khawaja, B.A., Farooq, Q.U., 2022. IoT-equipped and AI-enabled next Slama, D., Rückert, T., Thrun, S., Homann, U., Lasi, H., 2023. The digital playbook: A
generation smart agriculture: a critical review, current challenges and future trends. practitioner’s guide to smart, connected products and solutions with AIoT. Springer
IEEE Access. Nature.

25
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Smartlab, 2023. Advice on plant protection against diseases and pests based on Zhang, F., Pan, Z., Lu, Y., 2023. AIoT-enabled smart surveillance for personal data
modern monitoring and signalling system. Advisory regarding application of new digitalization: Contextual personalization-privacy paradox in smart home. Inf.
post-harvest technologies. Available online: https://agrosmartlab.com/en/home/. Manag. 60 (2), 103736.
(Accessed on 06 March 2023). Zhang, J., Tao, D., 2020. Empowering things with intelligence: a survey of the progress,
Souza, G., Aquino, P.T., Maia, R.F., Kamienski, C., Soininen, J.-P., 2020. A fuzzy irri- challenges, and opportunities in artificial intelligence of things. IEEE Internet
gation control system. In: 2020 IEEE Global Humanitarian Technology Conference. Things J. 8 (10), 7789–7817.
GHTC, IEEE, pp. 1–6. Zhao, Y., Zhai, W., Zhao, J., Zhang, T., Sun, S., Niyato, D., Lam, K.-Y., 2020. A
Suchithra, M., Pai, M.L., 2020. Data mining based geospatial clustering for suitable rec- comprehensive survey of 6g wireless communications. arXiv preprint arXiv:2101.
ommendation system. In: 2020 International Conference on Inventive Computation 03889.
Technologies. ICICT, IEEE, pp. 132–139. Zheng, Y., Xu, Y., Qiu, Z., 2023. Blockchain traceability adoption in agricultural supply
Sung, W.-T., Devi, I.V., Hsiao, S.-J., 2022. Early warning of impending flash flood based chain coordination: An evolutionary game analysis. Agriculture 13 (1), 184.
on aIoT. EURASIP J. Wireless Commun. Networking 2022 (1), 1–18. Zheng, Z., Zhou, Y., Sun, Y., Wang, Z., Liu, B., Li, K., 2021. Federated learning in smart
Tao, W., Zhao, L., Wang, G., Liang, R., 2021. Review of the internet of things cities: a comprehensive survey. arXiv preprint arXiv:2102.01375.
communication technologies in smart agriculture and challenges. Comput. Electron. Zheng, Z., Zhou, Y., Sun, Y., Wang, Z., Liu, B., Li, K., 2022. Applications of federated
Agric. 189, 106352. learning in smart cities: recent advances, taxonomy, and open challenges. Connect.
Tiwari, A., Garg, R., 2022. Adaptive ontology-based IoT resource provisioning in Sci. 34 (1), 1–28.
computing systems. Int. J. Semantic Web Inf. Syst.(IJSWIS) 18 (1), 1–18.
Zhou, J., Zhang, S., Lu, Q., Dai, W., Chen, M., Liu, X., Pirttikangas, S., Shi, Y.,
Tomar, P., Kaur, G., 2021. Artificial Intelligence and IoT-based Technologies for
Zhang, W., Herrera-Viedma, E., 2021. A survey on federated learning and its
Sustainable Farming and Smart Agriculture. IGI Global.
applications for accelerating industrial internet of things. arXiv preprint arXiv:
Tomaszewski, L., Kołakowski, R., 2023. Mobile services for smart agriculture and 2104.10501.
forestry, biodiversity monitoring, and water management: Challenges for 5G/6G
networks. Telecom 4 (1), 67–99.
Torres-Sanchez, R., Martínez-Zafra, M.T., Castillejo, N., Guillamon-Frutos, A., Artes-
Hernandez, F., 2020. Real-time monitoring system for shelf life estimation of fruit Dalhatu Muhammed received his BSc. degree in Computer Science from Usmanu
and vegetables. Sensors 20 (7), 1860. Danfodiyo University Sokoto, Nigeria in 2011 and a Masters in Computer Science from
Tunc, M.A., Gures, E., Shayea, I., 2021. A survey on IoT smart healthcare: Emerging the University of Malaya Malaysia in 2017. He is currently pursuing his PhD at Institut
technologies, applications, challenges, and future trends. arXiv preprint arXiv: Supérieur d’ Electronique de Paris (ISEP), Paris, France under the doctorate school
2109.02042. of Sorbonne University. His research interests include Smart Agriculture, Artificial
Udutalapally, V., Mohanty, S.P., Pallagani, V., Khandelwal, V., 2020. Scrop: A novel Intelligence of Things (AIoT), Machine Learning and Federated Learning. From 2012
device for sustainable automatic disease prediction, crop selection, and irrigation in to 2014 he was a System Analyst/Higher Technical Officer with the Department of
internet-of-agro-things for smart agriculture. IEEE Sens. J. 21 (16), 17525–17538. Computer Science, Kebbi State University of Science and Technology Aliero. He was a
Usmonov, M., Gregoretti, F., 2017. Design and implementation of a LoRa based wireless Graduate Assistant from 2014 to 2017, Assistant lecturer from 2017 to 2019, Lecturer II
control for drip irrigation systems. In: 2017 2nd International Conference on from 2019 to 2022 and Lecturer I from 2022 to date with the Department of Computer
Robotics and Automation Engineering. ICRAE, IEEE, pp. 248–253. Science, Kebbi State University of Science and Technology Aliero.
Verdouw, C., Tekinerdogan, B., Beulens, A., Wolfert, S., 2021. Digital twins in smart
farming. Agric. Syst. 189, 103046.
Vincent, D.R., Deepa, N., Elavarasan, D., Srinivasan, K., Chauhdary, S.H., Iwendi, C., Ehsan Ahvar is a senior staff standardization specialist (Lead Cloud Infrastructure
2019. Sensors driven AI-based agriculture recommendation model for assessing land management) at Nokia. Prior to joining Nokia, he was an associate professor at ESIEA
suitability. Sensors 19 (17), 3667. graduate engineering school, Paris, France. He did a postdoc at Inria and received
Vogeler, I., Cichota, R., Langer, S., Thomas, S., Ekanayake, D., Werner, A., 2023. a Ph.D. degree in computer science and telecommunications from Telecom SudParis
Simulating water and nitrogen runoff with APSIM. Soil Tillage Res. 227, 105593. in co-accreditation with the Sorbonne University (Pierre and Marie Curie University-
Vyas, S., Shabaz, M., Pandit, P., Parvathy, L.R., Ofori, I., 2022. Integration of artificial Paris 6). Ehsan has been involved in several national and international projects and
intelligence and blockchain technology in healthcare and agriculture. J. Food Qual. published more than 50 documents including book chapters, journal articles, conference
2022. papers and patents. His research interests include Cloud computing, Internet of Things
Wang, Y., Ho, I.W.-H., Chen, Y., Wang, Y., Lin, Y., 2021. Real-time water quality (IoT)/Artificial Intelligence of Things (AIoT), 5G/6G with focusing on improving energy
monitoring and estimation in AIoT for freshwater biodiversity conservation. IEEE consumption and carbon emission. He also served as a guest editor for several special
Internet Things J.. issues in different journals.
Wassan, S., Suhail, B., Mubeen, R., Raj, B., Agarwal, U., Khatri, E., Gopinathan, S., Dhi-
man, G., 2022. Gradient boosting for health IoT federated learning. Sustainability
14 (24), 16842. Shohreh Ahvar is a senior research engineer at Nokia and an adjunct associate
Wazid, M., Das, A.K., Park, Y., 2021. Blockchain-envisioned secure authentication professor at Telecom SudParis, France. From 2019 to 2022, Shohreh served as
approach in aIoT: Applications, challenges, and future research. Wirel. Commun. an associate professor at the ISEP graduate engineering school (l’Institut supérieur
Mob. Comput. 2021. d’électronique de Paris), where she was the head of the software engineering and AWS
Wolfert, S., Ge, L., Verdouw, C., Bogaardt, M.-J., 2017. Big data in smart farming–a certification. Prior to joining ISEP, she was a postdoctoral fellow at Inria, Paris. Shohreh
review. Agric. Syst. 153, 69–80. received her PhD degree in computer science and telecommunications from Telecom
Wu, Y.C., Wu, Y.J., Wu, S.M., 2019. An outlook of a future smart city in Taiwan from SudParis, in co-accreditation with Sorbonne university (Paris 6). She was involved in a
post–internet of things to artificial intelligence internet of things. In: Smart Cities: number of research projects, both national and international, including FUSE-IT, IDEAC,
Issues and Challenges. Elsevier, pp. 263–282. USNB, and Next Generation CDNs. She also worked as a collaborating researcher with
Yang, X., Shu, L., Chen, J., Ferrag, M.A., Wu, J., Nurellari, E., Huang, K., 2021. A the Concordia university, Canada, on a project funded by Ericsson. Shohreh published
survey on smart agriculture: Development modes, technologies, and security and more than 40 documents including book chapters, journal articles, conference papers,
privacy challenges. IEEE/CAA J. Autom. Sin. 8 (2), 273–302. and patents in the domains of Cloud computing, 5G/6G, and AIoT/IoT where she
Ye, S.S., 2021. FarmBeats: Microsoft and seeed’s aIoT solution for precision mainly focuses on energy efficiency and sustainability aspects. Shohreh also served as
agriculture and technology democratization for local farmers. Available online: a member of several technical committees at international conferences and workshops.
https://www.seeedstudio.com/blog/2021/11/30/farmbeats-microsoft-and-seeeds-
aiot-solution-for-precision-agriculture-and-technology-democratization-for-local-
farmers/. (Accessed on 22 January 2023). Maria Trocan received her M.Eng. in Electrical Engineering and Computer Science from
Yin, K., Bin, W., Rui, Z., Lin, X., Zhuofu, T., Guofeng, H., Zhiguo, W., Guangqiang, Y., Politehnica University of Bucharest in 2004, the Ph.D. in Signal and Image Processing
2022. DLDP-FL: Dynamic local differential privacy federated learning method based from Telecom ParisTech in 2007 and the Habilitation to Lead Researches from Pierre
on mesh network edge devices. J. Comput. Sci. 63, p.101789. & Marie Curie University (Sorbonne Université) in 2014. In May 2009 she joined ISEP,
Yu, K., Guo, Z., Shen, Y., Wang, W., Lin, J.C.-W., Sato, T., 2021. Secure artificial where between 2011 and 2018 she is responsible of the Signal Processing Program and
intelligence of things for implicit group recommendations. IEEE Internet Things J. between March 2018 and January 2023 she is in charge of the Data Science Program.
9 (4), 2698–2707. She is the Scientific Integrity referent for ISEP. She was Deputy Editor in Chief for IEEE
Yu, C., Shen, S., Zhang, K., Zhao, H., Shi, Y., 2022. Energy-aware device scheduling Open Journal on Circuits and Systems 2018–2021 and is Associate Editor for Springer
for joint federated learning in edge-assisted internet of agriculture things. In: Journal on Signal, Image and Video Processing etc. In 2014 she has been elected IEEE
2022 IEEE Wireless Communications and Networking Conference. WCNC, IEEE, France Vice-President for Student Activities and IEEE Circuits and Systems Board of
pp. 1140–1145. Governors member. She has been re-elected as IEEE France Vice-President for Student
Zhang, X., Ming, H., Jun, X., Tongquan, W., Mingsong, C., Shiyan, H., 2020. Efficient Activities in 2017, Chair of Technical Activities and Membership Development in 2020
federated learning for cloud-based aIoT applications. IEEE Trans. Comput.-Aided and Vice-President of IEEE French Section in 2023. Her research interests are on sparse
Des. Integr. Circuits Syst. 40 (11), 2211–2223. signal models.

26
D. Muhammed et al. Journal of Network and Computer Applications 228 (2024) 103905

Marie-José Montpetit is a well-known Canadian researcher in distributed systems, Dr. Reza Ehsani received his Ph.D. in Biological and Agricultural Engineering from the
networking, digital twinning and applied Artificial Intelligence. From 2019 to 2022 she University of California, Davis, where he worked on precision agriculture applications
was the lead system architect at Ferme d’Hiver a controlled environment vertical farm for high-value crops. He joined the Ohio State University as a faculty member and state
where she developed concepts in AIoT. She now focuses her research on distributed precision agriculture specialist in 2000. His work at the Ohio State University included
intelligent networking. She is the co-chair of the Computing in the Network (COIN) the application of precision agriculture for row crops with a focus on auto-steering
research group at the IRTF. an adjunct professor at École de Technologie Supérieure systems. In 2005, he joined the University of Florida, where he worked on developing
in Montreal, a lecturer at Telecom Paris Sud and an academic consultant at Iowa State tools and techniques for disease and stress detection and management, sensor systems
University on advanced networking for farming and rural areas. for autonomous platforms such as UAVs in precision horticulture, and automation and
machine enhancement for fruit harvesting. Currently, he is a professor of Mechanical
Engineering at the University of California, Merced. His current areas of research
include engineering systems for agriculture, automation and intelligent machines for
agriculture, precision agriculture and horticulture, and mechanical harvesting machines
and robots.

27

You might also like