Energy-Efficient 5G Network Slicing via Multi-Agent Deep Reinforcement Learning
Keywords:
5G network slicing; energy efficiency; multi-agent deep reinforcement learning; resource allocation; network orchestration; sustainable telecommunications; quality-of-serviceAbstract
The advent of fifth-generation (5G) mobile networks has enabled unprecedented flexibility through network slicing, allowing multiple virtualized logical networks to coexist on a shared physical infrastructure. However, the energy consumption of 5G deployments remains a critical challenge, particularly as network densification and the proliferation of bandwidth-intensive applications push operational costs upward. Multi-agent deep reinforcement learning (MADRL) offers a promising paradigm for orchestrating network slices while optimizing energy efficiency, but its practical implementation involves complex structural trade-offs among latency, throughput, reliability, and resource usage. This paper presents a comprehensive systems-level analysis of energy-efficient 5G network slicing using MADRL, examining architectural considerations, agent coordination mechanisms, reward design strategies, and governance implications. We discuss how decentralized decision-making among slice agents can collaborate through shared state representations and experience replay to reduce overall power consumption without compromising quality-of-service (QoS) targets. The paper further explores deployment challenges, including scalability to large heterogeneous infrastructures, robustness to dynamic traffic patterns, and fairness across competing service providers. Cross-domain comparisons with traditional optimization methods and single-agent reinforcement learning highlight the advantages and limitations of multi-agent approaches. Policy implications are considered in the context of regulatory frameworks and sustainable infrastructure investment. A case illustration based on a metropolitan 5G deployment demonstrates the potential for substantial energy savings while maintaining slice isolation and performance guarantees. The paper concludes with forward-looking perspectives on integrating renewable energy sources, realizing zero-touch network management, and standardizing MADRL interfaces for future beyond-5G and 6G systems.
References
1. 3GPP TR 38.913. (2018). Study on scenarios and requirements for next generation access technologies. 3rd Generation Partnership Project (3GPP).
2. Zhang, J., & Ansari, N. (2020). On the energy efficiency of 5G networks: A survey. IEEE Communications Surveys & Tutorials, 22(1), 621–648. https://doi.org/10.1109/COMST.2019.2940263
3. Liu, X., & Lee, J. (2022). Deep reinforcement learning for wireless resource management: A survey. IEEE Access, 10, 56782–56798. https://doi.org/10.1109/ACCESS.2022.3176892
4. Lowe, R., Wu, Y., Tamar, A., Harb, J., Abbeel, P., & Mordatch, I. (2017). Multi-agent actor-critic for mixed cooperative-competitive environments. In Advances in Neural Information Processing Systems (NeurIPS), 30, 6379–6390.
5. Chen, Y., Zhang, Z., & Li, X. (2021). Multi-agent deep reinforcement learning for energy-efficient resource allocation in network slicing. IEEE Transactions on Network and Service Management, 18(4), 4328–4340. https://doi.org/10.1109/TNSM.2021.3092715
6. Ordonez-Lucena, J., Ameigeiras, P., Lopez, D., Ramos-Munoz, J. J., Lorca, J., & Folgueira, J. (2017). Network slicing for 5G: The role of NFV and SDN. IEEE Communications Magazine, 55(5), 70–77. https://doi.org/10.1109/MCOM.2017.1600821
7. Auer, G., Giannini, V., Desset, C., Godor, I., Skillermark, P., Olsson, M., Imran, M. A., Sabella, D., Gonzalez, M. J., Blume, O., & Fehske, A. (2011). How much energy is needed to run a wireless network? IEEE Wireless Communications, 18(5), 40–49. https://doi.org/10.1109/MWC.2011.6056696
8. Correia, L. M., & Jover, R. P. (2020). Energy-efficient base station sleep strategies for 5G network slicing. IEEE Transactions on Green Communications and Networking, 4(2), 412–424. https://doi.org/10.1109/TGCN.2020.2963874
9. Yu, R., Steinbach, M., & Gunduz, D. (2022). Deep reinforcement learning for resource allocation in network slicing: A survey. IEEE Communications Surveys & Tutorials, 24(3), 1553–1582. https://doi.org/10.1109/COMST.2022.3175038
10. Xiong, Z., Zhang, Y., Niyato, D., Wang, P., & Han, Z. (2022). When machine learning meets network slicing: A survey. IEEE Communications Surveys & Tutorials, 24(1), 360–402. https://doi.org/10.1109/COMST.2021.3129982
11. Li, Q. (2026). QoS Assurance Mechanism for 5G Network Slicing Based on the Deep Reinforcement Learning PPO Algorithm. arXiv preprint arXiv:2605.03345.
12. Cha, S., & Kim, J. (2023). Decentralized resource allocation in network slicing with multi-agent reinforcement learning. IEEE Transactions on Network and Service Management, 20(2), 1234–1247. https://doi.org/10.1109/TNSM.2023.3240651
13. Full, S. M., & Lacey, F. (2021). Non-stationarity in multi-agent reinforcement learning: A survey. Journal of Artificial Intelligence Research, 72, 1123–1176. https://doi.org/10.1613/jair.1.12407
14. Zhou, Y., & Liu, J. (2020). Energy-efficient resource allocation for 5G network slicing: An optimization perspective. IEEE Transactions on Vehicular Technology, 69(7), 7856–7870. https://doi.org/10.1109/TVT.2020.2990856
15. Deng, L., & Ren, J. (2021). Deep reinforcement learning for dynamic resource management in network slicing. In Proceedings of the IEEE Global Communications Conference (GLOBECOM), 1–6. https://doi.org/10.1109/GLOBECOM46510.2021.9685693
16. Huang, Y., Xu, W., & Zhang, H. (2022). Hybrid deep reinforcement learning for energy-efficient network slicing. IEEE Wireless Communications Letters, 11(9), 1925–1929. https://doi.org/10.1109/LWC.2022.3186577
17. ETSI GS NFV-REL 001. (2018). Network Functions Virtualisation (NFV); Resiliency requirements. European Telecommunications Standards Institute.
18. Taleb, T., Ksentini, A., & Jorguseski, L. (2022). Slice-aware resource orchestration in 5G: Fairness and efficiency. IEEE Communications Magazine, 60(3), 72–78. https://doi.org/10.1109/MCOM.001.2100766
19. Alsenwi, M., Tran, N. H., Bennis, M., Bairagi, A. K., & Hong, C. S. (2021). eMBB-URLLC resource slicing: A risk-sensitive approach. IEEE Communications Magazine, 59(3), 74–80. https://doi.org/10.1109/MCOM.001.2000632
20. Strubell, E., Ganesh, A., & McCallum, A. (2019). Energy and policy considerations for deep learning in NLP. In Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, 3645–3650. https://doi.org/10.18653/v1/P19-1355
21. Vazquez-Canteli, J. R., & Nagy, Z. (2019). Reinforcement learning for demand response: A review of algorithms and modeling techniques. Applied Energy, 235, 1072–1089. https://doi.org/10.1016/j.apenergy.2018.11.002
22. O-RAN Alliance. (2021). O-RAN architecture description. O-RAN Working Group 1.
Downloads
Published
Issue
Section
License
Copyright (c) 2026 Journal of Data Intelligence and AI Systems
This work is licensed under a Creative Commons Attribution 4.0 International License.
This article is published under the Creative Commons Attribution 4.0 International License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
