A Federated Reinforcement Learning Framework via a Committee Mechanism for Resource Management in 5G Networks DOI Creative Commons
Jae-Won Jeong, Joohyung Lee

Sensors, Journal Year: 2024, Volume and Issue: 24(21), P. 7031 - 7031

Published: Oct. 31, 2024

This paper proposes a novel decentralized federated reinforcement learning (DFRL) framework that integrates deep (DRL) with (DFL). The DFRL boosts efficient virtual instance scaling in Mobile Edge Computing (MEC) environments for 5G core network automation. It enables multiple MECs to collaboratively optimize resource allocation without centralized data sharing. In this framework, DRL agents each MEC make local decisions and exchange model parameters other MECs, rather than sharing raw data. To enhance robustness against malicious server attacks, we employ committee mechanism monitors the DFL process ensures reliable aggregation of gradients. Extensive simulations were conducted evaluate proposed demonstrating its ability maintain cost-effective usage while significantly reducing blocking rates across diverse traffic conditions. Furthermore, demonstrated strong resilience adversarial nodes, ensuring operation management. These results validate framework's effectiveness adaptive management, particularly dynamic varied scenarios.

Language: Английский

Presenting the COGNIFOG Framework: Architecture, Building Blocks and Road toward Cognitive Connectivity DOI Creative Commons

Toni Adame,

Emna Amri, G. Antonopoulos

et al.

Sensors, Journal Year: 2024, Volume and Issue: 24(16), P. 5283 - 5283

Published: Aug. 15, 2024

In the era of ubiquitous computing, challenges imposed by increasing demand for real-time data processing, security, and energy efficiency call innovative solutions. The emergence fog computing has provided a promising paradigm to address these bringing computational resources closer sources. Despite its advantages, characteristics pose in heterogeneous environments terms resource allocation management, provisioning, connectivity, among others. This paper introduces COGNIFOG, novel cognitive framework currently under development, which was designed leverage intelligent, decentralized decision-making processes, machine learning algorithms, distributed principles enable autonomous operation, adaptability, scalability across IoT–edge–cloud continuum. By integrating capabilities, COGNIFOG is expected increase reliability next-generation environments, potentially providing seamless bridge between physical digital worlds. Preliminary experimental results with limited set connectivity-related building blocks show improvements network utilization real-world-based IoT scenario. Overall, this work paves way further developments on framework, are aimed at making it more resilient, aligned ever-evolving demands environments.

Language: Английский

Citations

1
EVRM: Elastic Virtual Resource Management framework for cloud virtual instances DOI
Desheng Wang, Yiting Li, Weizhe Zhang

et al.

Future Generation Computer Systems, Journal Year: 2024, Volume and Issue: unknown, P. 107569 - 107569

Published: Nov. 1, 2024

Language: Английский

Citations

1
XAI-driven Model Design for Resource Utilization Forecasting in Cloud-native 6G Networks DOI

Lazaros Liatsas,

Godfrey Kibalya, Angelos Antonopoulos

et al.

Published: July 8, 2024

Language: Английский

Citations

0
Anomaly Detection-Based Multilevel Ensemble Learning for CPU Prediction in Cloud Data Centers DOI
Mustafa Daraghmeh, Anjali Agarwal, Yaser Jararweh

et al.

Published: Aug. 6, 2024

Language: Английский

Citations

0
A Federated Reinforcement Learning Framework via a Committee Mechanism for Resource Management in 5G Networks DOI Creative Commons
Jae-Won Jeong, Joohyung Lee

Sensors, Journal Year: 2024, Volume and Issue: 24(21), P. 7031 - 7031

Published: Oct. 31, 2024

This paper proposes a novel decentralized federated reinforcement learning (DFRL) framework that integrates deep (DRL) with (DFL). The DFRL boosts efficient virtual instance scaling in Mobile Edge Computing (MEC) environments for 5G core network automation. It enables multiple MECs to collaboratively optimize resource allocation without centralized data sharing. In this framework, DRL agents each MEC make local decisions and exchange model parameters other MECs, rather than sharing raw data. To enhance robustness against malicious server attacks, we employ committee mechanism monitors the DFL process ensures reliable aggregation of gradients. Extensive simulations were conducted evaluate proposed demonstrating its ability maintain cost-effective usage while significantly reducing blocking rates across diverse traffic conditions. Furthermore, demonstrated strong resilience adversarial nodes, ensuring operation management. These results validate framework's effectiveness adaptive management, particularly dynamic varied scenarios.

Language: Английский

Citations

0