2026· IEEE Transactions on Network and Service Management· Vol 23, pp. 6507-6522· 0 citations· 58 references
TL;DR
This work proposes EAGLE, a Multi-Agent Reinforcement Learning (MARL) system that autonomously orchestrates the scaling of network resources to meet flow demands and reduce power consumption, and shows that the trained model can “zero-shot generalize” to unseen network topologies that share structural or statistical similarity with the training domain, hence reducing training time and associated energy costs.
Abstract
Network automation and data-driven solutions are critical for modern network efficiency and resilience, especially as traditional manual management struggles to keep pace with increasing demands. The programmability of Software-Defined Networks (SDN) and their integration with Machine Learning (ML) algorithms have significantly advanced this quest, enabling data-driven network configuration. Within this context, autoscaling network resources is essential for efficiently operating softwarized and virtualized networks. However, current autoscaling solutions are limited by their logically centralized nature and their poor model portability across different network domains. To address these challenges, we propose EAGLE, a Multi-Agent Reinforcement Learning (MARL) system that autonomously orchestrates the scaling of network resources to meet flow demands and reduce power consumption. Powered by graph embedding, it generalizes to diverse network settings. We evaluate our solution over a Mininet-based emulator and assess its generalization capabilities on Fabric, a large-scale network testbed. Our results show that our approach can reduce flow completion time (FCT) by up to 40% while achieving up to 15% power savings. In addition, we show that our trained model can “zero-shot generalize” to unseen network topologies that share structural or statistical similarity with the training domain, hence reducing training time and associated energy costs.
This paper presents a comprehensive framework for artificial intelligence (AI)-enabled autonomous network slicing optimization in 6G systems and investigates the application of advanced machine learning paradigms specifically deep reinforcement learning, federated learning, and generative AI to orchestrate dynamic resource provisioning, cross-slice isolation, and proactive SLA (Service Level Agreement) enforcement.
N. P J, Jeeva Jothi· International Journal of Com...· 0 citations
A QoE-aware framework for Multi-Access Edge Computing-enabled Open Radio Access Network (O-RAN) architectures, combining a graph attention network (GAT) encoder, distributed multi-agent DRL, and privacy-preserving FL, while transitioning control from Quality of Service (QoS) to QoE metrics is proposed.
Manoj Prasad Kunasegran, Wai Leong Pang, S. K. Phang· IEEE Access· 0 citations
Future 6G services will require strict performance guarantees, especially in terms of delay, end-to-end (e2e) across multiple network domains including packet and radio segments. While deterministic transport and slice-based capacity allocation can improve segment-level performance, ensuring e2e Network Service (NS) performance remains challenging as it requires making decisions Near–Real-Time (Near-RT) on a per-service basis, which does not fit well within the typical centralized control and orchestration hierarchy. Multi-agent systems (MAS), where a number of distributed agents collaborate, has demonstrated its capabilities for such Near-RT control. Agents equipped with Deep Reinforcement Learning (DRL) engines autonomously made traffic routing decisions based on e2e telemetry measurements. In this paper, we extend such MAS solutions for NS traffic routing focused on covering several issues that appear under frequent NS reconfiguration, e.g., caused by end device mobility. In addition, we define a lifecycle for NS operation that includes the initial MAS deployment, model reconfiguration during operation, and NS reconfiguration. The proposed lifecycle requires the definition of DRL training and validation procedures to produce models ready to be deployed with guaranteed performance under certain network conditions. In addition, model selection algorithms are defined for the lifecycle scenarios. In case of NS reconfiguration, a procedure for probe testing the actual network conditions is proposed to improve model selection. Evaluation across a meaningful set of network and traffic scenarios shows that the MAS is able to maintain e2e delay guarantees under all the lifecycle scenarios.
H. Shakespear-Miles, S. Barzegar, M. Ruiz et al.· IEEE Transactions on Network...· 0 citations
Results affirm that the combination of structural learning, adaptive decision‐making, and automated evaluation on a cloud platform offers a realisable, scalable route to intelligent, autonomous, programmable network management that can be used in next‐generation communication infrastructures.
Muhammad Hasnain, Faisal Naeem, Imran Ghani· Applied AI Letters· 0 citations
The emergence of 5G and 6G advanced ecosystems demands highly adaptive resource management to orchestrate the specialised requirements of eMBB, URLLC, and mMTC network slices. In dense multi-cell environments, capturing complex spatial interdependencies and mitigating dynamic interference is paramount for maintaining Quality of Service (QoS). This paper introduces a robust GNN-DQN framework designed for Rate Splitting Multiple Access (RSMA) based networks. By representing the network topology as a graph, the framework leverages Graph Neural Networks (GNNs) to extract highdimensional spatial features and model inter-cell interference patterns. These insights enable a Deep Q-Network (DQN) agent to perform intelligent resource partitioning and dynamic power splitting of the RSMA common stream. Experimental results demonstrate that the proposed GNN-DQN framework achieves a connectivity success ratio exceeding 90% across all slices, representing an average improvement of over 60% compared to non-graph-based reinforcement learning and supervised baselines. Notably, the framework demonstrates exceptional spectral efficiency, maintaining near-total connectivity while utilising less than 10% of the normalised system bandwidth, a 4× reduction in resource overhead compared to traditional methods. Furthermore, the GNN-driven architecture ensures stable convergence during training, yielding a 1.6× higher system reward score. Our findings validate GNN-DQN as a high-performance, scalable, and resource-efficient paradigm for intelligent orchestration in 5G and 6G networks.
Aya Kh. Ahmed, Nadia Al-Aboody, Hamed S. Al-Raweshidy· International Conference on...· 0 citations
Networks with highly dynamic data transmission demands and network topologies are common in real world. A fundamental problem in such networks is achieving scalable traffic allocation to maximize long-term total throughput under link capacity constraints. However, state-of-the-art (SOTA) works lack scalability. This is primarily due to two reasons in large-scale networks: first, they require solving constrained optimization problems online, which leads to high decision latency; second, they rely on reinforcement learning algorithms for policy optimization, which are inefficient in exploration and challenging to train effectively. To address these issues, we propose the Fast Networked Control (FNC) policy framework, which firstly utilizes parallelizable neural network modules to process the state and generate raw decisions, followed by basic operations such as normalizations and comparisons, which do not require iteration or optimization, to obtain decisions that satisfy the constraints. Hence, FNC policy avoids solving constrained optimization problems and supports parallel execution, significantly reducing decision latency. Furthermore, this policy preserves gradient flow and supports backpropagation, which enable us to design an imitation learning algorithm to efficiently train the policy in an end-to-end manner. Experiments in large-scale networks show that our FNC policy achieves an average 8% improvement in demands satisfaction and 10 times reduction in decision latency versus SOTA works.
Zhaoxing Yang, Guiyun Fan, Anjie Cao et al.· IEEE Transactions on Network...· 0 citations