A Comprehensive Study of Network Slicing in 5G Wireless Networks: Insights, Challenges, and Future Directions

Abstract

Network slicing will be the basis for enabling next- gen wireless networks by allowing multiple virtual networks to share a physical foundation and provide various types of service needs through one system. The goal of this report has been to evaluate and review fifteen (15) research papers (2019 through 2024) with respect to network slicing, their performance (slicing management), and their use of machine learning (ML) and AI- based slicing automation capabilities. A focus was placed on comparing the research publications along multiple dimensions including slice management, resource allocation, quality of ser- vice (QoS), energy efficiency, security, and AI/ML automated application. While the review identified a transition to using ML, RL, and Intent Based Networking for the purpose of automating smart slice orchestration, it also identified gaps in the existing body of research such as a lack of real world validation and scalability issues. Emerging areas were also identified such as 6G native slicing, LLM-assisted selection, and cloud-edge integration; thereby this review represents a structured resource for both researchers and practitioners in the area of wireless communications and virtualization.

Introduction

Network slicing is a transformative technology in 5G and future 6G networks that enables multiple virtual networks (slices) to operate on shared physical infrastructure. Each slice can be customized to meet specific service requirements such as ultra-low latency, high throughput, and massive IoT connectivity, while maintaining guaranteed Quality of Service (QoS) and Service Level Agreements (SLAs). Supported by technologies like Software Defined Networking (SDN), Network Function Virtualization (NFV), and cloud-native orchestration, network slicing has become a core component of modern wireless networks.

Purpose of the Study

The paper reviews and compares 15 research studies published between 2019 and 2024 to examine advances in network slicing, including:

• Resource allocation
• Slice management
• QoS optimization
• Energy efficiency
• Security
• AI and machine learning integration
• Future 6G and LLM-based slicing approaches

Literature Review Highlights

1. Early Research (2019)

• Machine learning was introduced for traffic classification and slice assignment, outperforming traditional heuristic methods.
• Optimization techniques improved slice placement and network reliability.
• Orchestration-based approaches enhanced QoS through reduced latency and increased throughput.
• Most studies relied on simulation-based validation.

2. Energy Efficiency and Security (2020–2021)

• Dynamic resource allocation was proposed to reduce energy consumption in 5G networks.
• Security frameworks were developed to identify threats and controls for network slices.
• Intent-based networking improved automation and faster service provisioning.
• Researchers also began exploring slicing requirements for future 6G networks.

3. AI-Driven Orchestration (2023)

• Machine learning and reinforcement learning became dominant approaches for automated slice management.
• Multi-Agent Deep Deterministic Policy Gradient (MADDPG) significantly reduced deployment costs but required high computational resources.
• Cloud-native orchestration improved scalability and reduced service downtime.
• Practical implementations in Wi-Fi environments demonstrated the feasibility of slice-aware management beyond cellular networks.

4. AI-Native and LLM-Based Approaches (2024)

• AI-powered adaptive slicing systems were proposed for future 6G services.
• Large Language Models (LLMs) were explored for network slice assignment and policy reasoning.
• LLMs showed promise in creating understandable and flexible slice management policies, although large-scale validation remains limited.

Methodologies Used

The reviewed studies employed five main approaches:

1. Machine Learning and Reinforcement Learning
   • Traffic classification, resource allocation, and slice deployment optimization.
   • Improved accuracy, efficiency, and adaptability.
2. Intent-Based Networking and SDN
   • Automated translation of business goals into network configurations.
   • Improved cross-domain orchestration.
3. Optimization-Based Techniques
   • Mathematical models for slice placement and resource management.
4. Survey and Conceptual Studies
   • Focused on architectural frameworks, energy models, and future 6G concepts.
5. Experimental Testbeds
   • Limited real-world validation through 5G-Wi-Fi integrated environments.
6. LLM-Assisted Decision Making
   • Emerging approach using pretrained language models for intelligent slice assignment.

Comparative Analysis

Resource Allocation

• Early studies focused on traffic classification and scheduling.
• Reinforcement learning achieved better optimization but increased computational complexity.
• Machine learning offered a balance between efficiency and practicality.

Quality of Service (QoS)

• Dynamic orchestration significantly reduced latency and improved throughput.
• Intent-based networking accelerated slice provisioning.
• Future 6G systems aim for sub-millisecond latency and real-time IoT support.

AI and Automation

• AI evolved from an optional enhancement to a core requirement for network slicing.
• Recent studies emphasize fully automated and intelligent slice management.
• LLM-based approaches may represent the next generation of network orchestration.

Energy Efficiency

• Dynamic slice allocation can reduce power consumption.
• Energy optimization remains underexplored despite its importance in sustainable networks.

Key Results

The review found that:

• Machine learning consistently outperforms traditional rule-based methods in classification accuracy, resource utilization, and latency reduction.
• Dynamic orchestration improves QoS and service provisioning speed.
• Reinforcement learning reduces deployment costs but introduces computational challenges.
• Intent-based networking simplifies automation and management.
• LLM-assisted slicing shows strong potential for interpretable and flexible decision-making.
• Most reported benefits are based on simulations rather than real-world deployments.

Research Gaps

Several challenges remain:

• Lack of large-scale experimental validation.
• Limited focus on security and energy efficiency.
• High computational requirements for advanced AI models.
• Insufficient research on cloud-edge integration.
•  

Conclusion

In this paper we provide a survey of 15 network slicing contributions, from 2019 to 2024, to understand how solu- tions have evolved from early ML-based classification, to optimization-based placement, to using RL and intent-based orchestration and, finally, toLLM-assisted slice assignment. With each generation of solutions the quality of service (QoS), the resource efficiency and automation increased; however new issues about computational costs and verification processes also emerged. Three observations emerge. First, the use of ML techniques has really been fruitful across multiple independent efforts and has consistently shown good results in classification accuracy, allocation efficiency, and latency. It seems, however, that the investigation of security and energy efficiency is lacking, despite the fact that they are a core requirement in operational practice, and even more critical assliced networkssupport vertical applications of critical importance. Futurework directions: 6G native AI integration with slicing frameworks requires to move from theory to experimental studies. LLM-enabled orchestration presents an interpretabil- ity gain worth investigating, as long as inference latency challenges can be overcome. The cloud-edge integration[11] should be further explored given its increasing importance to service delivery. Security, an ignored aspect of many approaches to date, must be considered from the beginning of slice lifecycle management. In conclusion, network slicing is a technology still in evolution: its theoretical foundations are strong, but practical deployment issues abound. Moving beyond its theoretical po- tential will require coupling sophisticated algorithmic visions with pragmatic testing in real world environments.

References

[1] Z. Wu, Q. Li, J. Li, and X. Chen, “Machine learning based 5G network slicing management and classification,” IEEE Research Paper, 2019. [2] J. Lorincz, T. Garma, and G. Petrovic,“Acomprehensive overview of network slicing for improving the energy efficiency of fifth-generation networks,” MDPI Energies/Sensors, vol. 13, no. 4, pp. 1–35, 2020 [3] T. Wichary, “Network slicing security controls and assurance for verti- cals,” MDPI Future Internet, vol. 13, no. 5, pp. 1–22, 2021. [4] A. Seng, R. Buettner, and J. Hofmann, “Network slice placement in wireless mesh networks,” IEEE Research Paper, 2019. [5] R. Moreira, R. L. Aguiar, and J. Ferreira, “Towards 6G network slicing,”arXiv preprint arXiv:2101.00001, 2021. [6] N. An, T. Kim, J. Park, J. Kwon, and H. Joo, “Slice management for quality of service differentiation in wireless network slicing,” MDPI Electronics, vol. 8, no. 6, pp. 1–18, 2019. [7] K. Abbas, M. Afaq, T. A. Khan, and W. Song, “Slicing the core network and radio access network domains through intent-based networking for 5G networks,” MDPI Sensors, vol. 20, no. 22, pp. 1–24, 2020. [8] S. Venkatapathy, A. Khlass, and R. Guerzoni,“AnE2E network slicing framework for slice creation and deployment using machine learning,”MDPI Future Internet, vol. 15, no. 3, pp. 1–20, 2023 [9] “Redefining 6G network slicing: AI-driven solutions for future use cases,” MDPI Electronics, vol. 13, no. 2, pp. 1–25, 2024. [10] “MADDPG-based deployment algorithm for 5G network slicing,” MDPI Applied Sciences, vol. 13, no. 8, pp. 1–19, 2023. [11] “Evaluation of cloud-based dynamic network scaling and slicing for next-generation wireless networks,” MDPI Electronics, vol. 12, no. 10, pp. 1–22, 2023. [12] A. Napolitano, “Network slicing: An overview,” Research Paper, un- published. [13] “Acomparative study of network slicing techniques for effective utiliza- tion of channel for 5G and beyond 5G networks,” IEEE Access, vol. 11, pp. 1–18, 2023. [14] M. Nerini,“5Gnetwork slicing for Wi-Fi networks,” arXiv preprint arXiv:2301.00001, 2023. [15] S. Sudhakara, “Constrained network slice assignment via large language models,” arXiv preprint arXiv:2401.00001, 2024.

Copyright

Copyright © 2026 Kaivalya Kamble, Nikhil Mishra, Aryan Gurav, Soham Kulkarni, Adesh Borhade, Dr. Asha Durafe. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.