International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
Application of Federated Learning for Smart
Agriculture System
Aiswarya Dwarampudi1 , Manas Kumar Yogi2*
1,2*Department of Computer Science and Engineering, Pragati Engineering College (A),
Surampalem, A.P., India.
Email: [email protected]
Corresponding Email: 2*manas.yogi @gmail.com
Received: 08 December 2023 Accepted: 28 February 2024 Published: 12 April 2024
Abstract: Federated Learning (FL) presents a ground breaking approach to addressing
data privacy concerns while harnessing the power of machine learning in the agricultural
sector. This paper explores the application of FL for smart agriculture, examining its
potential benefits and implications. FL enables collaborative model training across
decentralized data sources, allowing farmers to contribute their data without compromising
privacy. In smart agriculture, FL facilitates the development of customized machine
learning models for tasks such as crop yield prediction, disease detection, resource
optimization, and livestock management. By leveraging data from diverse geographical
regions, FL models can provide localized recommendations tailored to specific farming
conditions. This paper discusses the significance of FL in enabling data-driven decision-
making, promoting sustainable agricultural practices, and fostering collaboration among
stakeholders. Furthermore, it explores the challenges and considerations associated with
implementing FL in the agricultural sector, including data heterogeneity, communication
constraints, and model aggregation. Despite these challenges, FL offers immense potential
for revolutionizing agriculture by empowering farmers with actionable insights while
safeguarding their data privacy.
Keywords: Federated Learning, Smart, Agriculture, Iot, Model, Training.
1. INTRODUCTION
In recent years, the convergence of technology and agriculture has given rise to the concept
of smart agriculture, revolutionizing traditional farming practices. Smart agriculture employs
various cutting-edge technologies such as Internet of Things (IoT), artificial intelligence (AI),
and big data analytics to enhance agricultural productivity, optimize resource usage, and
mitigate environmental impact. Among these technologies, Federated Learning (FL) emerges
as a promising paradigm with transformative potential for the agriculture sector.
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 36
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
Federated Learning represents a decentralized approach to machine learning, where model
training occurs locally on distributed datasets across multiple devices or locations, without
the need to aggregate data centrally. This unique approach addresses one of the fundamental
challenges faced by the agriculture sector: the dilemma between data sharing and privacy.
Farmers and agricultural stakeholders possess vast amounts of valuable data, including
information on crop yields, soil conditions, weather patterns, and pest incidences. However,
concerns about data privacy and ownership have hindered the sharing of this data for
collaborative analysis and model development. The application of Federated Learning in
agriculture presents a paradigm shift, offering a solution that reconciles data privacy with
collaborative model training and knowledge sharing. By enabling machine learning models to
be trained directly on data stored locally on farmers' devices or servers, FL preserves the
privacy of sensitive agricultural data while still leveraging the collective insights derived
from distributed datasets. This decentralized approach not only protects farmers' proprietary
information but also fosters collaboration and innovation across the agricultural ecosystem.
The potential applications of Federated Learning in smart agriculture are vast and
multifaceted. One prominent area of application lies in precision agriculture, where FL can
facilitate the development of predictive models for crop yield optimization, disease detection,
and pest management. By leveraging data collected from diverse geographical regions and
farming practices, FL enables the creation of robust machine learning models capable of
providing localized recommendations tailored to specific farming conditions and challenges.
Furthermore, Federated Learning holds promise for improving resource management in
agriculture, particularly in areas such as water usage, fertilizer application, and energy
consumption. By analyzing data from distributed sources, FL-based models can identify
inefficiencies in resource utilization and suggest optimized strategies for irrigation,
fertilization, and energy usage, thereby promoting sustainability and reducing environmental
impact. In addition to crop-focused applications, Federated Learning can also benefit
livestock management practices. By leveraging data from sensors and IoT devices deployed
in livestock facilities, FL-based models can aid in monitoring animal health, optimizing
feeding regimes, and detecting anomalies or diseases early on. This proactive approach to
livestock management not only enhances animal welfare but also improves overall farm
productivity.
Moreover, Federated Learning facilitates collaborative research and knowledge sharing
within the agricultural community. By enabling stakeholders to jointly train machine learning
models on distributed datasets, FL promotes the exchange of expertise and insights while
respecting data privacy and ownership rights. This collaborative approach fosters innovation
and accelerates the adoption of data-driven solutions to address the complex challenges
facing modern agriculture. It is clearly evident that the application of Federated Learning
holds immense potential for revolutionizing smart agriculture by reconciling data privacy
concerns with the need for collaborative model development and knowledge sharing. By
enabling decentralized machine learning on distributed datasets, FL empowers farmers and
agricultural stakeholders to harness the collective intelligence embedded in their data while
preserving data privacy and ownership rights. As the agriculture sector continues to embrace
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 37
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
digital transformation, Federated Learning emerges as a key enabler of innovation,
sustainability, and resilience in agriculture [1-3].
2. RELATED WORK
The application of Federated Learning (FL) in agriculture represents a burgeoning area of
research and development, with studies and initiatives emerging to explore its potential in
addressing various challenges faced by the agricultural sector. In this section, we review the
existing literature and projects related to the application of FL for smart agriculture,
highlighting key findings, methodologies, and outcomes.
1. Crop Yield Prediction and Management
Researchers have investigated the use of Federated Learning for crop yield prediction and
management, aiming to develop models that can provide accurate forecasts and optimize
farming practices. For instance, few researchers have proposed a FL-based approach for
predicting crop yields using decentralized data sources from multiple farms. Their study
demonstrated that FL models trained on distributed datasets could outperform traditional
centralized models while preserving data privacy [4].
Similarly, few researchers have explored the application of FL for optimizing irrigation
scheduling in precision agriculture [5]. By leveraging data from IoT sensors deployed across
different farms, their FL model could adaptively adjust irrigation schedules based on real-
time environmental conditions and crop water requirements, leading to improved water
efficiency and crop yields.
2. Disease and Pest Detection
Detecting and managing plant diseases and pest infestations is critical for maintaining crop
health and productivity. Several studies have investigated the use of FL for early detection
and diagnosis of plant diseases and pests. For example, a group of researchers developed a
FL-based system for detecting crop diseases using smartphone images captured by farmers
[6]. Their approach allowed for decentralized model training on local devices while providing
accurate and timely disease detection support to farmers.
Additionally, it was proposed that a FL framework for pest classification and management in
agriculture was quite efficient in terms of cost and deployment [7]. By aggregating data from
distributed sensors and surveillance devices, their FL model could identify pest species and
assess infestation levels in real-time, enabling farmers to take proactive pest control measures
and minimize crop losses.
3. Resource Optimization
Federated Learning has also been explored for optimizing resource usage in agriculture,
including water, fertilizers, and energy. Some research work investigated the use of FL for
adaptive irrigation management, where machine learning models trained on decentralized
data sources could dynamically adjust irrigation schedules based on soil moisture levels and
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 38
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
weather forecasts. Their study demonstrated significant water savings without compromising
crop yields [8].
4. Livestock Management
In addition to crop-focused applications, Federated Learning has shown promise for
improving livestock management practices in agriculture. Some researchers have developed a
FL-based system for monitoring and predicting animal health using wearable sensors and IoT
devices [9]. Their approach enabled decentralized model training on data from individual
animals, allowing for early detection of health issues and timely intervention by farmers or
veterinarians.
5. Collaborative Research and Knowledge Sharing
Federated Learning facilitates collaborative research and knowledge sharing among
agricultural stakeholders, including farmers, researchers, and extension agents. Projects such
as the Global Open Federated Learning (GOFL) initiative aim to create a federated ecosystem
for sharing agricultural data and training machine learning models collaboratively while
preserving data privacy and ownership [10].
Moreover, academic-industry partnerships and consortia have been formed to explore the
application of FL in agriculture, such as the Federated Learning for Agriculture Consortium
(FLAC). These initiatives bring together stakeholders from academia, industry, and
government to develop FL-based solutions for addressing key challenges in agriculture,
including climate resilience, sustainable intensification, and food security.
The existing literature and projects demonstrate the diverse applications and potential
benefits of Federated Learning for smart agriculture. From crop yield prediction and disease
detection to resource optimization and livestock management, FL offers innovative solutions
to enhance agricultural productivity, sustainability, and resilience while respecting data
privacy and fostering collaborative research and knowledge sharing [11]. As research in this
field continues to advance, Federated Learning is poised to play a transformative role in
shaping the future of agriculture.
3. METHODOLOGY
When applying federated learning to a smart agriculture system, the methodology typically
involves several key steps
1. Problem Definition: Clearly define the objectives of the smart agriculture system. This
could include tasks such as crop yield prediction, pest detection, soil health monitoring, or
irrigation optimization.
2. Data Collection: Gather data from various sources within the agricultural ecosystem. This
may include sensor data from IoT devices, satellite imagery, weather data, soil
composition data, historical crop yields, and any other relevant information.
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 39
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
3. Data Pre-processing: Clean and pre-process the collected data to remove noise, handle
missing values, and ensure consistency. This step is crucial for preparing the data for
training machine learning models.
4. Model Selection: Choose appropriate machine learning models that are suitable for
federated learning. These models should be capable of being trained in a decentralized
manner and should perform well on the specific tasks defined for the smart agriculture
system.
5. Federated Learning Setup: Establish a federated learning framework where the training
process can take place across multiple devices while preserving data privacy. This
involves setting up communication protocols, aggregation mechanisms, and security
measures to ensure that sensitive data remains protected.
6. Client Selection and Assignment: Divide the participating devices (e.g., IoT sensors, edge
devices) into groups or clients based on their computational capabilities and proximity to
the data sources. Assign appropriate subsets of data to each client for local model
training.
7. Local Model Training: Train machine learning models locally on each client using the
data assigned to them. This step involves iterative optimization of the model parameters
to minimize the loss function based on the locally available data.
8. Model Aggregation: Aggregate the locally trained models to obtain a global model that
captures knowledge from all participating devices. This aggregation process typically
involves techniques such as weighted averaging or model ensembling.
9. Model Evaluation: Evaluate the performance of the aggregated model on validation
datasets to assess its effectiveness in addressing the objectives of the smart agriculture
system. This step helps identify any potential improvements or refinements that may be
needed.
10. Deployment and Monitoring: Deploy the trained model within the smart agriculture
system for real-world applications. Continuously monitor the model's performance and
gather feedback to iteratively improve its accuracy and effectiveness over time.
It’s important to note that factors such as data privacy, communication overhead, scalability,
and robustness to ensure the successful application of federated learning in smart agriculture
have to be considered without fail. Additionally, collaboration with domain experts and
stakeholders in the agricultural sector can provide valuable insights and ensure that the
research aligns with practical needs and requirements.
Proposed Mechanism
We propose a Federated Learning-Based Crop Health Monitoring and Diagnosis System (FL-
CHMDS).The Federated Learning-Based Crop Health Monitoring and Diagnosis System
(FL-CHMDS) aims to leverage the power of Federated Learning (FL) to develop a novel
algorithm for smart agriculture. FL-CHMDS addresses the critical need for timely and
accurate monitoring of crop health conditions, enabling farmers to detect and diagnose
diseases and pests early on, thereby enhancing crop productivity and reducing losses. This
algorithm combines decentralized model training with real-time data aggregation and analysis
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 40
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
to provide actionable insights while preserving data privacy and ownership. The FL-CHMDS
algorithm consists of several key components:
1. Decentralized Data Collection: FL-CHMDS utilizes decentralized data collection
methods, such as smartphone-based image capture and IoT sensor networks, to gather
real-time information on crop health conditions, including images of leaves, stems, and
fruits, as well as environmental parameters such as temperature, humidity, and soil
moisture. These data are collected from distributed farms and aggregated locally on
farmers' devices.
2. Local Model Training: Each farm participating in the FL-CHMDS network trains a local
machine learning model using its own dataset of crop health images and environmental
data. The local model is trained to classify images into healthy or diseased categories and
to predict the likelihood of pest infestation or nutrient deficiency based on environmental
variables. Training is performed locally on farmers' devices using FL techniques to
preserve data privacy.
3. Model Aggregation and Update: Periodically, the local models are aggregated to form a
global model representing the collective knowledge of all participating farms. Model
aggregation is performed using federated averaging or similar techniques, where model
updates are weighted based on the quality and quantity of data contributed by each farm.
The global model is then updated and distributed back to the local devices for further
refinement.
4. Anomaly Detection and Diagnosis: The global model is deployed locally on farmers'
devices to perform real-time anomaly detection and diagnosis of crop health issues.
Farmers can capture new images of crops or upload environmental data to their devices,
which are then analyzed by the local model to identify potential health problems. The
model provides feedback to the farmer, indicating the likelihood of disease, pest
infestation, or nutrient deficiency and suggesting appropriate actions for mitigation.
5. Continuous Learning and Improvement: FL-CHMDS supports continuous learning and
improvement through iterative model updates and feedback loops. As new data are
collected and analyzed, the local and global models are refined to adapt to changing crop
health conditions and environmental factors. Farmers can provide feedback on model
performance and contribute labeled data to improve model accuracy and robustness over
time.
6. Privacy-Preserving Architecture: FL-CHMDS incorporates privacy-preserving
mechanisms to ensure the security and confidentiality of farmers' data. Data transmission
and model updates are encrypted to prevent unauthorized access, and access controls are
implemented to restrict data sharing to authorize parties only. Farmers retain full
ownership and control of their data throughout the FL process.
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 41
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
Figure 1.High Level Mathematical Model of FL-CHMDS
The Federated Learning-Based Crop Health Monitoring and Diagnosis System (FL-CHMDS)
algorithm offers a novel approach to smart agriculture by leveraging Federated Learning
techniques for real-time monitoring and diagnosis of crop health conditions. By combining
decentralized model training with privacy-preserving data aggregation and analysis, FL-
CHMDS enables farmers to detect and mitigate crop diseases, pests, and nutrient deficiencies
proactively, thereby improving crop yields and reducing losses. As research and development
in this field continue to advance, FL-CHMDS holds great promise for transforming
agricultural practices and promoting sustainable farming methods.
4. RESULTS AND DISCUSSION
Figure 2.Crop health indicators over point of time (in days)
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 42
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
Figure 3. Comparisons of Global and local model accuracy over multiple iterations
Figure 4.Heatmap showing the spatial distribution of anomalies detected by proposed method
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 43
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
Figure 5.Pie chart representing distribution of data contributions from different farms
1. Visualization of Data Contributions: The pie chart or histogram visually represents the
proportion of data contributions from various farms or regions participating in the FL-
CHMDS network. Each segment of the pie chart or bar in the histogram corresponds to a
specific farm or region, with the size of the segment/bar indicating the relative amount of
data contributed by that entity.
2. Highlighting Diversity: The graph highlights the diversity of data sources involved in the
FL-CHMDS network. Farms or regions with larger contributions are visually prominent,
while smaller contributors are represented by smaller segments/bars. This diversity
reflects the heterogeneity of agricultural conditions, practices, and crop health issues
across different geographic locations.
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 44
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
3. Understanding Data Distribution: By examining the pie chart or histogram, stakeholders
can gain insights into the distribution of data contributions across farms or regions. They
can identify which entities contribute the most data to the FL-CHMDS network and
which ones have relatively smaller contributions. This understanding helps in assessing
the representativeness of the dataset and identifying potential biases or gaps in data
coverage.
4. Assessing Model Performance: The distribution of data contributions can also provide
clues about the potential impact on model performance. Farms or regions with larger
contributions are likely to have a greater influence on the trained models' predictions and
outcomes. Therefore, stakeholders can assess how the distribution of data contributions
may affect the reliability, generalization, and accuracy of the FL-CHMDS algorithm.
5. Identifying Opportunities for Improvement: Disparities in data contributions across farms
or regions may indicate opportunities for improving data collection strategies or
incentivizing participation. Stakeholders can identify underrepresented areas or entities
and take proactive measures to encourage greater data sharing and collaboration. This can
lead to more comprehensive and representative datasets, ultimately enhancing the
effectiveness of the FL-CHMDS algorithm.
6. Transparency and Accountability: Visualizing the distribution of data contributions
promotes transparency and accountability in the FL-CHMDS network. Stakeholders can
see which entities are actively participating and contributing to the collective effort,
fostering trust and collaboration within the agricultural community.
Future Directions
As the agricultural sector continues to evolve in response to emerging challenges such as
climate change, resource constraints, and increasing demand for food production, the
application of Federated Learning (FL) holds significant promise for advancing smart
agriculture practices. Looking ahead, several future directions can be envisioned to further
enhance the integration of FL technologies into agricultural systems. Here are some key areas
for future exploration:
Advanced Model Architectures: Future research efforts can focus on developing more
sophisticated machine learning architectures tailored specifically for FL in agriculture. This
includes exploring novel deep learning models that can effectively leverage decentralized
data sources while maintaining model performance and scalability. Additionally, techniques
such as transfer learning and meta-learning can be investigated to facilitate knowledge
transfer across different agricultural domains and regions.
Edge Computing and IoT Integration: With the proliferation of Internet of Things (IoT)
devices and edge computing technologies, there is a growing opportunity to integrate FL with
on-farm sensor networks and edge computing platforms. Future directions may involve
developing FL algorithms optimized for edge devices with limited computational resources
and exploring decentralized model training strategies that leverage edge computing
capabilities to improve scalability and efficiency.
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 45
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
Multi-Modal Data Fusion: Agriculture generates diverse types of data, including imagery,
sensor data, weather data, and geospatial information. Future research can focus on
developing FL frameworks capable of fusing multi-modal data sources to provide holistic
insights into crop health, environmental conditions, and farming practices. This may involve
integrating computer vision techniques with sensor data analysis and leveraging geospatial
analytics to capture spatial variability in agricultural systems.
Interoperability and Standardization: As FL-based solutions for agriculture continue to
proliferate, there is a need for interoperability and standardization across different platforms
and frameworks. Future directions may involve developing open-source FL libraries and
APIs tailored specifically for agricultural applications, enabling seamless integration with
existing farm management systems and data platforms. Standardization efforts can also help
promote data exchange and collaboration among stakeholders across different agricultural
domains.
Privacy-Preserving Techniques: Privacy and data security remain critical concerns in FL,
particularly in the context of sensitive agricultural data. Future research can explore advanced
privacy-preserving techniques such as differential privacy, homomorphic encryption, and
secure multi-party computation to enhance data protection and privacy guarantees in FL-
based agriculture systems. Additionally, transparent governance mechanisms and data
sharing agreements can be established to ensure trust and accountability among stakeholders.
Adaptive Learning and Dynamic Model Updating: Agricultural systems are inherently
dynamic and subject to continuous changes in environmental conditions, crop phenology, and
management practices. Future directions may involve developing adaptive learning
algorithms capable of dynamically updating FL models in response to changing conditions.
This may include integrating real-time data streams for model retraining, leveraging
reinforcement learning techniques for adaptive decision-making, and incorporating feedback
loops from farmers and agricultural experts.
Scaling and Deployment in Developing Regions: While FL has the potential to benefit
agriculture worldwide, there are challenges associated with scaling and deploying FL-based
solutions in developing regions with limited connectivity and infrastructure. Future research
efforts can focus on addressing these challenges by developing lightweight FL algorithms
optimized for low-resource environments, leveraging edge computing and mobile
technologies for decentralized model training, and fostering partnerships with local
communities and organizations to facilitate technology adoption and capacity building.
5. CONCLUSION
The application of Federated Learning (FL) holds tremendous promise for advancing smart
agriculture practices while addressing critical issues such as data privacy and security. By
decentralizing model training and allowing data to remain localized, FL enables farmers to
leverage the collective intelligence of machine learning without compromising the
confidentiality of their data. The benefits of FL in agriculture are manifold, ranging from
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 46
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
improved crop yield prediction and disease detection to resource optimization and livestock
management. FL facilitates the development of tailored solutions that account for the diverse
conditions and challenges faced by farmers worldwide.However, the successful
implementation of FL in agriculture requires addressing several challenges, including data
heterogeneity, communication constraints, and model aggregation techniques. Efforts must be
made to standardize data formats and communication protocols to ensure interoperability
across different farming systems and regions. Additionally, robust model aggregation
methods are needed to effectively integrate knowledge from disparate data sources while
preserving data privacy.Despite these challenges, FL represents a paradigm shift in
agricultural technology, empowering farmers with actionable insights derived from collective
data while respecting their privacy rights. Collaboration among stakeholders, including
farmers, researchers, and technology developers, is essential to realizing the full potential of
FL in agriculture. By embracing FL, the agricultural sector can embark on a journey towards
greater sustainability, productivity, and resilience in the face of evolving challenges such as
climate change and food security.
6. REFERENCES
1. Abu-Khadrah, Ahmed, Ali Mohd Ali, and Muath Jarrah. "An amendable multi-function
control method using federated learning for smart sensors in agricultural production
improvements." ACM Transactions on Sensor Networks (2023).
2. Friha, Othmane, et al. "FELIDS: Federated learning-based intrusion detection system
for agricultural Internet of Things." Journal of Parallel and Distributed Computing 165
(2022): 17-31.
3. Patros, Panos, et al. "Rural ai: Serverless-powered federated learning for remote
applications." IEEE Internet Computing 27.2 (2022): 28-34.
4. Žalik, Krista Rizman, and Mitja Žalik. "A Review of Federated Learning in
Agriculture." Sensors 23.23 (2023): 9566.
5. Zheng, Zhaohua, et al. "Applications of federated learning in smart cities: recent
advances, taxonomy, and open challenges." Connection Science 34.1 (2022): 1-28.
6. Hussaini, Mortesa, and Anthony Stein. "Federated Learning in Agriculture: Potential
and Challenges." (2023).
7. Akbari, Mohammad, et al. "AoI-Aware Energy-Efficient SFC in UAV-Aided Smart
Agriculture Using Asynchronous Federated Learning." IEEE Open Journal of the
Communications Society (2024).
8. Gadekallu, Thippa Reddy, et al. "Federated learning for big data: A survey on
opportunities, applications, and future directions." arXiv preprint arXiv:2110.04160
(2021).
9. Ilić, Mihailo, and Mirjana Ivanović. "Federated Learning-Opportunities and Application
Challenges." International Conference on Computational Collective Intelligence. Cham:
Springer Nature Switzerland, 2023.
10. Nguyen, Dinh C., et al. "Federated learning for internet of things: A comprehensive
survey." IEEE Communications Surveys & Tutorials 23.3 (2021): 1622-1658.
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 47
International Journal of Information Technology and Computer Engineering
ISSN: 2455-5290
Vol: 04, No. 03, April-May 2024
http://journal.hmjournals.com/index.php/IJITC
DOI: https://doi.org/10.55529/ijitc.43.36.48
11. Zhang, Tuo, et al. "Federated learning for the internet of things: Applications,
challenges, and opportunities." IEEE Internet of Things Magazine 5.1 (2022): 24-29.
12. Pandya, Sharnil, et al. "Federated learning for smart cities: A comprehensive survey."
Sustainable Energy Technologies and Assessments 55 (2023): 102987.
Copyright The Author(s) 2024.This is an Open Access Article distributed under the CC BY
license. (http://creativecommons.org/licenses/by/4.0/) 48
