American Journal of Artificial Intelligence

2024, Vol. 8, No. 2, pp. 55-62

[Link]

Research Article

AI-Driven 5G Network Optimization: A Comprehensive

Review of Resource Allocation, Traffic Management, and
Dynamic Network Slicing
Dileesh Chandra Bikkasani1, * , Malleswar Reddy Yerabolu2
1
Department of Technology Management, University of Bridgeport, Bridgeport, USA
2
Independent Researcher, North Carolina, USA

Abstract
The rapid advancement of 5G networks, coupled with the increasing complexity of resource management, traffic handling, and
dynamic service demands, underscores the necessity for more intelligent network optimization techniques. This paper
comprehensively reviews AI-driven methods applied to 5G network optimization, focusing on resource allocation, traffic
management, and network slicing. Traditional models face limitations in adapting to the dynamic nature of modern
telecommunications, while AI techniques—particularly machine learning (ML) and deep reinforcement learning (DRL)—offer
scalable and adaptive solutions. These approaches facilitate real-time optimization by learning from network conditions,
predicting traffic patterns, and managing resources intelligently across virtual network slices. The integration of AI into 5G
networks enhances performance, reduces latency, and ensures efficient bandwidth utilization, which is essential for supporting
emerging applications such as the Internet of Things (IoT), autonomous systems, and augmented reality. Furthermore, this paper
highlights key AI techniques and their applications to 5G challenges, illustrating their potential to drive future innovations in
network management. By laying the groundwork for autonomous network operations in 6G and beyond, this research
emphasizes the transformative impact of AI on telecommunications infrastructure and its role in shaping the future of
connectivity.

Keywords
5G, Telecommunication, Wireless Communication, Artificial Intelligence, Network Performance

1. Introduction
With the evolution of wireless communication came sig- dented data speeds, ultra-low latency, and multi-device con-
nificant advancements in the telecommunications space, with nectivity. The new 5G-NR (New Radio) standard is catego-
data demand increasing 1000-fold from 4G to 5G [1]. Each rized into three distinct service classes: Ultra-Reliable
new generation has addressed the shortcomings of its prede- Low-Latency Communications (URLLC), massive Ma-
cessors, and the advent of the 5th generation of wireless chine-Type Communications (mMTC), and enhanced Mobile
network (5G) technology, in particular, promises unprece- Broadband (eMBB). URLLC aims to provide highly reliable

*Corresponding author:

Received: 24 October 2024; Accepted: 9 November 2024; Published: 28 November 2024

Copyright: © The Author(s), 2024. Published by Science Publishing Group. This is an Open Access article, distributed
under the terms of the Creative Commons Attribution 4.0 License ([Link] which
permits unrestricted use, distribution and reproduction in any medium, provided the original work is properly cited.
American Journal of Artificial Intelligence [Link]

and low-latency connectivity; eMBB focuses on increasing chine-to-Machine (M2M) communication, and smart cities.
bandwidth for high-speed internet access; and mMTC sup- Built on technologies like millimeter-wave (mmWave) spec-
ports many connected devices, enabling IoT on a massive trum, massive multiple-input multiple-output (MIMO) sys-
scale [2]. Optimizing 5G performance is crucial for emerging tems, and network function virtualization (NFV) [3], 5G
applications such as autonomous vehicles, multimedia, aug- promises to revolutionize many industries.
mented and virtual realities (AR/VR), IoT, Ma-

Figure 1. Components of NFV.

Figure 1 illustrates the components of Network Function proaches to network management become inadequate. Effi-
Virtualization (NFV), a key enabler for 5G. NFV decouples cient resource allocation, traffic management, and dynamic
network functions from proprietary hardware, allowing these network slicing [4] are necessary to handle demanding use
functions to run as software on standardized hardware. By cases without compromising speed or reliability. Additionally,
virtualizing network functions—such as firewalls, load bal- with the increase in mobile traffic flow, meeting customer
ancers, and gateways—NFV supports dynamic and scalable demands on time requires addressing the allocation of band-
network management, making it easier to allocate resources width for heterogeneous cloud services [5].
flexibly across different network slices and use cases. This Network resource allocation (RA) in 5G networks plays a
flexibility is critical in managing the growing demands of 5G critical role in optimizing the efficient utilization of spectrum,
applications, where real-time adaptability and resource opti- computing power, and energy to meet the demands of modern
mization are paramount. The significance of NFV lies in its wireless communication. Resource allocation is pivotal for
ability to decouple hardware from software, allowing network data-intensive applications, IoT devices, and emerging tech-
operators to deploy new services and scale existing ones more nologies like AV and AR. It ensures these technologies re-
efficiently. For example, in a 5G network, operators can al- ceive adequate network resources, enhancing overall per-
locate resources dynamically to different virtual network formance and QoS in dynamic and heterogeneous environ-
functions (VNFs), optimizing for the specific needs of ap- ments. Traditional resource allocation relies on channel state
plications such as autonomous vehicles or telemedicine, information (CSI), which necessitates significant overhead
which demand high reliability and low latency. Figure 1 costs, thereby increasing the overall expense of the process [6].
showcases the architectural elements of NFV, including the Section 2 focuses on various AI techniques applied to re-
virtualization layer, hardware resources, and the management source management, highlighting their impact on network
and orchestration functions that control resource allocation slicing, energy efficiency, and overall quality of service (QoS).
and scaling. NFV plays a pivotal role in enabling network By leveraging reinforcement learning (RL), optimization
slicing, a critical feature of 5G, which allows operators to methods, and machine learning (ML) models, these advanced
create virtual networks tailored to specific application re- strategies address the dynamic and complex requirements of
quirements. 5G networks, providing adaptive and intelligent solutions for
However, the complexity and heterogeneity of 5G net- enhanced connectivity and sustainability.
works present several challenges, including quality of service Network slicing allows creating multiple virtual networks
(QoS) provisioning, resource management, and network op- on top of a shared physical infrastructure, each optimized for
timization. As 5G networks scale, traditional rule-based ap- specific service requirements. Network slices can be inde-

56
American Journal of Artificial Intelligence [Link]

pendently configured to support diverse applications with Many Deep Reinforcement Learning (DRL) techniques
varying performance needs, such as low-latency communica- have been applied for network slicing in 5G networks, al-
tion, massive device connectivity, or high-throughput data lowing dynamic resource allocation that enhances throughput
services. Using network slicing, 5G can provide tailored ex- and latency by learning from the network environment.
periences for different user types while maximizing the use of DRL-based approaches can handle the complexity and over-
network resources. This capability is crucial for emerging use head issues of traditional centralized resource allocation
cases like smart factories, telemedicine, and autonomous methods [14]. Balevi and Gitlin (2018) proposed a clustering
systems, where performance requirements can differ signifi- algorithm that maximizes throughput in 5G heterogeneous
cantly across applications. The rest of the paper's sections are networks by utilizing machine learning techniques to improve
structured as follows: The first section covers the general network efficiency and adjust resource allocation based on
understanding of the topic and defines why AI is necessary; real-time network conditions [15]. A Graph Convolution
the second contains the traditional AI-ML methods used in Neural Network (GCN) resource allocation scheme was ex-
current spaces; and the third section consists of the DL-based plored, addressing the problem of managing resource alloca-
techniques. tions effectively. Optimization techniques such as heuristic
methods and Genetic Algorithms (GAs) and are being ex-
plored to solve resource allocation problems efficiently by
2. Data Science and AI Techniques for minimizing interference and maximizing spectral efficiency.
Resource Allocation Genetic algorithms, for instance, utilize evolutionary princi-
ples like selection, crossover, and mutation to evolve solu-
AI has shown promising results in resource allocation tions toward optimal resource allocation configurations.
through continuous learning and adaptation to network Heuristic methods like simulated annealing and particle
changes. Unlike traditional mathematical model-based para- swarm optimization (PSO) are employed to further enhance
digms, Reinforcement Learning (RL) employs a data-driven resource management. The integration of AI-driven algo-
model to efficiently allocate resources by learning from the rithms, such as RL and DRL, into 5G networks enables re-
network environment, thereby improving throughput and al-time, adaptive resource allocation based on changing net-
latency [7]. However, achieving fully distributed resource work conditions and user demands, significantly improving
allocation is not feasible due to virtualization isolation con- network performance and efficiency [16]. By utilizing
straints in each network slice [8]. AI-driven optimization, 5G networks can achieve higher
In 4G LTE and LTE Advanced (LTE/A) networks, efficiency and better manage the interplay between different
IP-based packet switching primarily focuses on managing network elements, ensuring seamless connectivity and high
varying user numbers in a given area. By employing performance.
ML-based MIMO for channel prediction and resource allo- AI and machine learning techniques are revolutionizing
cation, these technologies enhance CSI accuracy through data resource allocation in 5G networks. By shifting from tradi-
compression and reduced processing delays while adapting to tional models to adaptive, data-driven approaches like RL and
channel rates despite imperfect CSI [9]. However, this tech- DRL, these technologies can significantly enhance network
nique proves inefficient due to the complexity and traffic load throughput, reduce latency, and efficiently manage system
of 5G networks. Traditional resource allocation using CSI overhead. As traditional methods struggle with rising com-
struggles with system overhead, which can reach up to 25% of plexity and user demand, AI-driven optimization provides
total system capacity, making it suboptimal for 5G Cloud dynamic solutions that adapt to real-time network conditions,
Radio Access Network (CRAN) applications. The conven- enabling more efficient and effective resource management in
tional method also falters with an increasing number of users the 5G era and beyond.
[10].
Traditional optimization techniques include using an ap-
proximation algorithm to connect end-users with Remote 3. Data Science and AI in Traffic
Radio Heads (RRH). This algorithm estimates the number of Management
end-users linked to each RRH and establishes connections
between end-users, RRHs, and Baseband Units (BBU) [11]. Integrating AI technologies in 5G network traffic man-
Challenges such as millimeter-wave (mmWave) beamform- agement aims to achieve traffic volume prediction, enhance
ing and optimized time-delay pools using hybrid beamform- real-time computational efficiency, and ensure network ro-
ing for single and multiple-user scenarios in 5G CRAN net- bustness and adaptability to fluctuating traffic patterns. Mo-
works have also been explored [12]. In another instance, a bile data traffic is anticipated to grow significantly, with 5G's
random forest algorithm was proposed for resource allocation, share increasing from 25 percent in 2023 to around 75 percent
where a system scheduler validates outputs from a binary by 2029 [17]. This growth and the complexity of 5G networks
classifier; although robust, further research and development necessitate advanced techniques for efficient traffic prediction,
are necessary [13]. crucial for optimizing resource allocation and ensuring net-

57
American Journal of Artificial Intelligence [Link]

work reliability. Machine learning (ML) offers diverse ap- sensor network of five nodes [25]. Advanced techniques such
plications in network traffic management, from predicting as transformer-based architectures leverage self-attention
traffic volumes and identifying security threats to optimizing mechanisms to efficiently process vast amounts of data [26].
traffic engineering (TE). These capabilities enable proactive DRL techniques have been proposed to schedule high-volume
network monitoring, enhanced security measures, and im- flexible traffic (HVFT) in mobile networks. This model uses
proved traffic flow management, leading to efficient and deep deterministic policy gradient (DDPG) reinforcement
resilient network operations [18]. By combining time, loca- learning to learn a control policy for scheduling IoT and other
tion, and frequency information, researchers have identified delay-tolerant traffic, aiming to maximize the amount of
five basic time-domain patterns in mobile traffic, corre- HVFT traffic served while minimizing degradation to con-
sponding to different types of urban areas such as residential, ventional delay-sensitive traffic [27].
business, and transportation hubs [19]. Several studies highlight innovative applications of DRL
Traditional models like ARIMA have been widely used for and AI frameworks in this context. One study introduced a
seasonality because they can model temporal dependencies in DRL approach for decentralized cooperative localization
time series data. The ARIMA model combines autoregression scheduling in vehicular networks [28]. An AI framework
(AR), integration (I), and moving average (MA) components using CNN and RNN enhanced throughput by approximately
to predict future values based on past observations [20]. 36%, though it incurred high training time and memory usage
Variations such as seasonal ARIMA (SARIMA) use spectrum costs [29]. Another DRL model based on LSTM enabled
analysis to describe traffic patterns and predict parameter small base stations to dynamically access unlicensed spectrum
estimation using maximum likelihood methods [21]. The and optimize wireless channel selection [30]. A DRL ap-
seasonal ARIMA (SARIMA) model algorithm describes a proach for SDN routing optimization achieved configurations
procedure for fitting seasonality to traffic data, determining comparable to traditional methods with minimal delays [31].
seasonality periods through spectrum analysis, estimating Work on routing and interference management, often reliant
differencing parameters, and identifying model orders using on costly algorithms like WMMSE, has advanced by ap-
information criteria that can predict the parameter estimation proximating these algorithms with finite-size neural networks,
using maximum likelihood methods. Despite their effective- demonstrating significant potential for improving Massive
ness, ARIMA and SARIMA often struggle with the MIMO systems.
non-linear and complex traffic patterns characteristic of 5G In summary, integrating AI technologies into 5G network
networks. Machine learning models, including Support Vec- traffic management offers significant advancements in mul-
tor Machines (SVM) and Random Forests, have been im- tiple facets, such as traffic prediction, resource allocation, and
plemented to overcome these limitations. SVMs capture network management. Techniques such as ML and DL using
non-linear relationships, particularly in their regression form models like LSTM and advanced frameworks utilizing CNN,
(SVR) [22]. The Random Forest algorithm constructs deci- RNN, and DRL address the complex and dynamic nature of
sion trees by training each one on randomly selected data 5G networks. AI-driven solutions improve network efficiency
points and features, improving prediction accuracy and ro- and reliability by enhancing interference management, spec-
bustness [23]. This approach builds multiple decision trees trum access, and routing capabilities and adapting to varying
and merges them to improve prediction accuracy and ro- traffic patterns and demands. These innovations highlight the
bustness, effectively handling the heterogeneity of 5G net- transformative potential of AI in achieving robust, adaptive,
work traffic. Support Vector Machines focus on maximizing and efficient 5G network operations, paving the way for fu-
the distance from the separating plane to the nearest data ture research and development in this critical field.
points, known as support vectors, using dot products and
kernel functions. This approach enables faster training than
methods like Bagging and Random Forest, which require 4. Network Slicing in 5G: Data Science
using the entire dataset. and AI Approaches
Deep learning (DL) has revolutionized many facets of
network traffic management, including traffic prediction, Network slicing is one of 5G's most transformative features.
estimation, and smart traffic routing. Models like Long It enables the partitioning of a single physical network into
Short-Term Memory (LSTM) networks are particularly ef- multiple virtual networks, each modified and adjusted to meet
fective because they can capture and learn long-term de- specific service requirements. These network slices can be
pendencies in sequential data, making them highly suitable dynamically created, modified, and terminated to optimize
for predicting network traffic. DL also presents promising resources for various applications, ranging from massive IoT
alternatives for interference management, spectrum man- deployments to ultra-reliable low-latency communications
agement, multi-path usage, link adaptation, multi-channel (URLLC). The challenge lies in managing the complexity of
access, and traffic congestion [24]. For instance, an AI creating and maintaining these slices in real time, a task where
scheduler using a neural network with two fully connected Artificial Intelligence (AI) plays a crucial role.
hidden layers can reduce collisions by 50% in a wireless AI technologies are increasingly being adopted in 5G to

58
American Journal of Artificial Intelligence [Link]

automate dynamic network slicing. The traditional manual algorithms like neural networks can predict peak traffic times
approach to network management is insufficient for handling for specific services, enabling proactive resource allocation to
the large-scale, highly heterogeneous environments enabled avoid congestion.
by 5G. AI, particularly machine learning (ML), offers ad- Reinforcement learning, particularly in a multi-agent en-
vanced capabilities in real-time decision-making, predictive vironment, is also becoming popular for resource allocation in
analytics, and adaptive control, which are critical for the ef- network slicing. Multi-agent reinforcement learning (MARL)
ficient deployment and management of network slices. allows different network entities, such as base stations and
AI models predict traffic patterns, analyze network condi- user equipment, to collaborate as independent agents to
tions, and dynamically adjust resource allocation to meet the maximize overall network performance. The result is more
specific needs of each slice. This ensures that slices maintain efficient resource utilization, minimizing waste and ensuring
optimal performance, even under fluctuating traffic and var- that each slice receives the appropriate resources to maintain
ying service demands. Reinforcement learning (RL) and deep its service-level agreements (SLAs).
learning (DL) algorithms are frequently used to handle the Traffic management in network slicing is another area
complex decision-making processes required for slice or- where AI excels. The diversity of services in a 5G network,
chestration. These algorithms can autonomously learn from such as enhanced mobile broadband (eMBB), URLLC, and
network data, optimize resources, and balance loads between massive IoT, demands intelligent traffic prioritization. AI
slices without human intervention. algorithms analyze traffic patterns in real time, enabling the
AI-driven resource allocation plays a critical role in the system to prioritize slices that require lower latency or higher
success of network slicing. Each network slice may have reliability automatically. This dynamic traffic management
distinct bandwidth, latency, and reliability requirements, helps ensure that critical services, like autonomous vehicles or
making it necessary to allocate resources dynamically. AI can remote surgeries, get priority over less critical applications
help predict and pre-allocate resources based on historical like video streaming see Figure 1.
data and real-time network traffic patterns. For instance, ML

Figure 2. Network Slicing in 5G.

AI-powered traffic management can also mitigate conges- self-optimization capability. AI can continuously monitor
tion and improve the overall quality of service (QoS) by re- network performance metrics such as latency, throughput, and
routing traffic through less congested paths or adjusting error rates across different slices. When deviations from ex-
bandwidth allocations. Predictive models, trained on histori- pected performance are detected, AI systems can autono-
cal traffic data, can forecast potential bottlenecks and allow mously adjust configurations, redistribute resources, or even
the network to take preemptive measures, ensuring smooth alter the slice architecture to restore optimal performance.
operations even during peak usage periods. For instance, in cases where a slice serving IoT applications
One key advantage of integrating AI in network slicing is experiences a sudden increase in device connections, AI can

59
American Journal of Artificial Intelligence [Link]

scale the slice’s capacity by reallocating resources from less learning for resource allocation, could offer more adaptable
critical slices. Similarly, slices that require ultra-low latency solutions to 5G’s heterogeneous environments.
can be dynamically reconfigured to prioritize routing through Network slicing in 5G also requires more sophisticated
lower-latency paths. AI-driven orchestration mechanisms. Real-time prediction
AI-driven approaches are fundamental in overcoming the and adaptation of network slices based on AI algorithms will
complexity of network slicing in 5G networks. By leveraging become crucial, particularly in managing different services’
AI technologies like reinforcement learning, neural networks, varying latency, reliability, and bandwidth requirements.
and multi-agent systems, 5G networks can achieve greater Integrating AI models with software-defined networking
efficiency, adaptability, and scalability. AI ensures that net- (SDN) and network function virtualization (NFV) can help
work slices are dynamically created, maintained, and opti- optimize slice management dynamically.
mized, providing tailored services to meet the varying de-
mands of modern digital ecosystems.
6. Conclusion
5. Challenges and Future Directions Integrating AI-driven techniques into 5G networks pre-
sents a transformative approach to overcoming the inherent
Integrating AI in 5G networks for resource allocation, challenges of resource allocation, traffic management, and
traffic management, and network slicing presents significant network slicing. As 5G networks scale in complexity, tradi-
potential and numerous challenges. One major challenge is tional methods struggle to provide the real-time adaptability
the complexity of managing increasingly dense and hetero- required for dynamic, high-performance environments. AI
geneous networks. As 5G supports various applications with models, particularly those based on machine learning (ML)
differing requirements, like eMBB, URLLC, and massive IoT, and deep reinforcement learning (DRL), offer adaptive, da-
the need for real-time optimization of resources becomes ta-driven solutions that can continuously learn from network
critical. Traditional rule-based systems fail to manage dy- conditions to optimize performance, reduce latency, and
namic traffic and user demands efficiently, necessitating manage system overhead.
AI-driven adaptive solutions. However, deploying AI models Resource allocation in 5G is especially critical given the
for real-time decision-making at scale requires significant rise of data-intensive applications like autonomous vehicles,
computational power and efficient learning algorithms to augmented reality, and massive IoT deployments. AI-based
avoid system delays and bottlenecks. A key issue is the methods, such as DRL and genetic algorithms, provide scal-
overhead and latency of AI-based resource allocation, mainly able approaches to efficiently manage spectrum, compute
when using deep reinforcement learning (DRL) models. DRL power, and energy resources. These intelligent methods ad-
systems effectively learn from the network environment and dress the shortcomings of conventional models, such as
make dynamic resource adjustments but often suffer from channel state information (CSI)-based allocation, by offering
high training costs and memory consumption. This can lead to lower overhead and better adaptability to fluctuating condi-
inefficiencies in real-time operations, especially when net- tions. By leveraging AI, 5G networks can dynamically allo-
works are large and involve many interconnected devices, cate resources to meet the needs of different applications,
such as in smart cities or autonomous vehicle networks. from low-latency services to high-throughput data demands.
Moreover, multi-agent reinforcement learning (MARL) Traffic management is another area where AI significantly
methods used in network slicing require extensive coordina- enhances the operation of 5G networks. Through advanced
tion between network entities, which can result in system traffic prediction and real-time analysis, AI models such as
overhead and resource wastage if not correctly managed. LSTM and transformer-based architectures offer sophisti-
Another challenge is the reliance on accurate channel state cated tools to predict traffic patterns and optimize network
information (CSI) for resource allocation. This practice incurs load distribution. These capabilities are crucial in managing
considerable system overhead and is particularly inefficient in the expected exponential increase in mobile data traffic, en-
CRAN and mmWave-based 5G applications. Existing solu- suring efficient bandwidth utilization, and maintaining net-
tions like heuristic algorithms, genetic algorithms, or clus- work robustness even under high demand. Furthermore,
tering techniques provide partial improvements but often fail network slicing, a cornerstone of 5G’s architecture, benefits
to scale effectively as user demand increases. Future direc- immensely from AI’s ability to orchestrate and optimize vir-
tions involve improving the efficiency and scalability of tual network slices in real time. AI techniques such as mul-
AI-based solutions in 5G. Research is needed to optimize ti-agent reinforcement learning (MARL) enable more granu-
learning algorithms to reduce training costs and memory lar control over resource allocation across slices, ensuring
usage, potentially through federated learning or edge compu- each slice meets its specific service-level agreements (SLAs)
ting, where processing is distributed closer to the network while optimizing overall network efficiency.
edge. Additionally, hybrid AI models combining multiple AI’s integration into 5G is not just a complementary tech-
machine learning techniques like convolutional neural net- nology but a necessity to fully realize the potential of
works (CNNs) for traffic prediction and reinforcement next-generation networks. The shift from static, rule-based

60
American Journal of Artificial Intelligence [Link]

systems to intelligent, adaptive algorithms marks a paradigm [6] Imtiaz, S., et al. Random forests resource allocation for 5G
shift that will define future telecommunications, enabling systems: Performance and robustness study. in 2018 IEEE
more resilient, efficient, and scalable network operations that Wireless Communications and Networking Conference
Workshops (WCNCW). 2018. IEEE.
support a wide array of emerging technologies. This conver-
gence of AI and 5G lays the foundation for autonomous [7] Wang, T., S. Wang, and Z.-H. Zhou, Machine learning for 5G
networks and opens new research directions to further en- and beyond: From model-based to data-driven mobile wireless
hance performance, efficiency, and scalability in the era of 6G networks. China Communications, 2019. 16(1): p. 165-175.
and beyond.
[8] Baghani, M., S. Parsaeefard, and T. Le-Ngoc, Multi-objective
resource allocation in density-aware design of C-RAN in 5G.
IEEE Access, 2018. 6: p. 45177-45190.
Abbreviations
[9] Shehzad, M. K., et al., ML-based massive MIMO channel
ML Machine Learning prediction: Does it work on real-world data? IEEE Wireless
DL Deep Learning Communications Letters, 2022. 11(4): p. 811-815.
DRL Deep Reinforcement learning
NR New Radio [10] Chughtai, N. A., et al., Energy efficient resource allocation for
energy harvesting aided H-CRAN. IEEE Access, 2018. 6: p.
URLLC Ultra-Reliable Low-Latency Communications
43990-44001.
MARL Multi-agent Reinforcement Learning
HVFT High-volume Flexible Traffic [11] Zarin, N. and A. Agarwal, Hybrid radio resource management
DDPG Deep Deterministic Policy Gradient for time-varying 5G heterogeneous wireless access network.
mMTC Machine-Type Communications IEEE Transactions on Cognitive Communications and
MIMO Massive Multiple-input Multiple-output Networking, 2021. 7(2): p. 594-608.
Systems [12] Huang, H., et al., Optical true time delay pool based hybrid
CSI Channel state information beamformer enabling centralized beamforming control in
millimeter-wave C-RAN systems. Science China Information
Sciences, 2021. 64(9): p. 192304.
Author Contributions
[13] Lin, X. and S. Wang. Efficient remote radio head switching
Dileesh Chandra Bikkasani is the lead author, and Mal- scheme in cloud radio access network: A load balancing
leswar Reddy Yerabolu is the co-author. The authors read and perspective. in IEEE INFOCOM 2017-IEEE Conference on
approved the final manuscript. Computer Communications. 2017. IEEE.

[14] Gowri, S. and S. Vimalanand, QoS-Aware Resource Allocation

Scheme for Improved Transmission in 5G Networks with IOT.
Conflicts of Interest SN Computer Science, 2024. 5(2): p. 234.
The author declares no conflicts of interest. [15] Bouras, C. J., E. Michos, and I. Prokopiou. Applying Machine
Learning and Dynamic Resource Allocation Techniques in
Fifth Generation Networks. 2022. Cham: Springer
References International Publishing.

[1] An, J., et al., Achieving sustainable ultra-dense heterogeneous [16] Li, R., et al., Intelligent 5G: When cellular networks meet
networks for 5G. IEEE Communications Magazine, 2017. artificial intelligence. IEEE Wireless communications, 2017.
55(12): p. 84-90. 24(5): p. 175-183.

[2] ITU. Setting the Scene for 5G: Opportunities & Challenges. [17] Ericsson. 5G to account for around 75 percent of mobile data
2020 [cited 2024 07/13]; Available from: traffic in 2029. [cited 2024 07/13]; Available from:
[Link] [Link]
ortunities-challenges/ ort/dataforecasts/mobile-traffic-forecast

[3] Sakaguchi, K., et al., Where, when, and how mmWave is used in [18] Amaral, P., et al. Machine learning in software defined
5G and beyond. IEICE Transactions on Electronics, 2017. networks: Data collection and traffic classification. in 2016
100(10): p. 790-808. IEEE 24th International conference on network protocols
(ICNP). 2016. IEEE.
[4] Foukas, X., et al., Network slicing in 5G: Survey and
challenges. IEEE communications magazine, 2017. 55(5): p. [19] Wang, H., et al. Understanding mobile traffic patterns of large
94-100. scale cellular towers in urban environment. in Proceedings of
the 2015 Internet Measurement Conference. 2015.
[5] Abadi, A., T. Rajabioun, and P. A. Ioannou, Traffic flow
prediction for road transportation networks with limited traffic [20] Box, G. E., et al., Time series analysis: forecasting and control.
data. IEEE transactions on intelligent transportation systems, 2015: John Wiley & Sons.
2014. 16(2): p. 653-662.

61
American Journal of Artificial Intelligence [Link]

[21] Shu, Y., et al., Wireless traffic modeling and prediction using [27] Chinchali, S., et al. Cellular network traffic scheduling with
seasonal ARIMA models. IEICE transactions on deep reinforcement learning. in Proceedings of the AAAI
communications, 2005. 88(10): p. 3992-3999. Conference on Artificial Intelligence. 2018.

[22] Kumari, A., J. Chandra, and A. S. Sairam. Predictive flow [28] Peng, B., et al., Decentralized scheduling for cooperative
modeling in software defined network. in TENCON 2019-2019 localization with deep reinforcement learning. IEEE
IEEE Region 10 Conference (TENCON). 2019. IEEE. Transactions on Vehicular Technology, 2019. 68(5): p.
4295-4305.
[23] Moore, J. S., A fast majority vote algorithm. Automated
Reasoning: Essays in Honor of Woody Bledsoe, 1981: p. [29] Cao, G., et al., AIF: An artificial intelligence framework for
105-108. smart wireless network management. IEEE Communications
Letters, 2017. 22(2): p. 400-403.
[24] Arjoune, Y. and S. Faruque. Artificial intelligence for 5g
wireless systems: Opportunities, challenges, and future [30] Challita, U., L. Dong, and W. Saad, Proactive resource
research direction. in 2020 10th annual computing and management for LTE in unlicensed spectrum: A deep learning
communication workshop and conference (CCWC). 2020. perspective. IEEE transactions on wireless communications,
IEEE. 2018. 17(7): p. 4674-4689.

[25] Mennes, R., et al. A neural-network-based MF-TDMA MAC [31] Stampa, G., et al., A deep-reinforcement learning approach for
scheduler for collaborative wireless networks. in 2018 IEEE software-defined networking routing optimization. arXiv
Wireless Communications and Networking Conference preprint arXiv:1709.07080, 2017.
(WCNC). 2018. IEEE.

[26] Vaswani, A., et al., Attention is all you need. Advances in

neural information processing systems, 2017. 30.

62