Improved Deep Reinforcement Learning With Faster Graph Recurrent Convolutional Neural Network-Enabled Adaptive Network Slicing for Tailored Service Delivery in NextGen Networks
IF 1.7 4区 计算机科学Q3 ENGINEERING, ELECTRICAL & ELECTRONIC
{"title":"Improved Deep Reinforcement Learning With Faster Graph Recurrent Convolutional Neural Network-Enabled Adaptive Network Slicing for Tailored Service Delivery in NextGen Networks","authors":"S. Sugapriya, R. Vijayabhasker","doi":"10.1002/dac.6049","DOIUrl":null,"url":null,"abstract":"<div>\n \n <p>Many new use cases and a broad spectrum of vertical businesses are expected to be supported by next-generation wireless networks. Network slicing has been implemented to suit the stringent requirements of different services. By dividing the infrastructure network into several logical networks, this technology enables resource allocation based on services. In dynamic and unpredictable situations, managing resources across several domains and dimensions for end-to-end (E2E) slicing still presents difficulties. Tenant satisfaction requires striking a balance between the trade-off between revenue and the expense of resource allocation. Network slicing has emerged as a fundamental paradigm in next-generation networks to meet the diverse service requirements of various applications and users. However, the dynamic nature of network conditions and service demands poses challenges in efficiently allocating network resources to meet performance objectives. In this paper, a faster graph recurrent convolutional neural network (FGRCNN) with improved deep reinforcement learning (IDRL) is proposed to learn traffic behavior from link and node properties in addition to network structure. To train the FGRCNN model in the IDRL framework without requiring a labeled training dataset, employ the Deep Q-learning technique. This allows the framework to swiftly adjust to changes in traffic dynamics. A system is proposed to analyze real-time network and service data enabling dynamic adaptation of network slices for changing traffic patterns and service requirements. A comprehensive framework is presented that integrates deep learning models with network slicing orchestration mechanisms to achieve tailored service delivery. Through extensive simulations and experiments, the effectiveness approach in optimizing resource utilization is demonstrated, improving service quality and enabling agile network management in next-gen networks. Results highlight the potential of deep learning-enabled adaptive network slicing to support diverse and evolving service demands in future network environments.</p>\n </div>","PeriodicalId":13946,"journal":{"name":"International Journal of Communication Systems","volume":"38 2","pages":""},"PeriodicalIF":1.7000,"publicationDate":"2024-12-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"International Journal of Communication Systems","FirstCategoryId":"94","ListUrlMain":"https://onlinelibrary.wiley.com/doi/10.1002/dac.6049","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0
Abstract
Many new use cases and a broad spectrum of vertical businesses are expected to be supported by next-generation wireless networks. Network slicing has been implemented to suit the stringent requirements of different services. By dividing the infrastructure network into several logical networks, this technology enables resource allocation based on services. In dynamic and unpredictable situations, managing resources across several domains and dimensions for end-to-end (E2E) slicing still presents difficulties. Tenant satisfaction requires striking a balance between the trade-off between revenue and the expense of resource allocation. Network slicing has emerged as a fundamental paradigm in next-generation networks to meet the diverse service requirements of various applications and users. However, the dynamic nature of network conditions and service demands poses challenges in efficiently allocating network resources to meet performance objectives. In this paper, a faster graph recurrent convolutional neural network (FGRCNN) with improved deep reinforcement learning (IDRL) is proposed to learn traffic behavior from link and node properties in addition to network structure. To train the FGRCNN model in the IDRL framework without requiring a labeled training dataset, employ the Deep Q-learning technique. This allows the framework to swiftly adjust to changes in traffic dynamics. A system is proposed to analyze real-time network and service data enabling dynamic adaptation of network slices for changing traffic patterns and service requirements. A comprehensive framework is presented that integrates deep learning models with network slicing orchestration mechanisms to achieve tailored service delivery. Through extensive simulations and experiments, the effectiveness approach in optimizing resource utilization is demonstrated, improving service quality and enabling agile network management in next-gen networks. Results highlight the potential of deep learning-enabled adaptive network slicing to support diverse and evolving service demands in future network environments.
期刊介绍:
The International Journal of Communication Systems provides a forum for R&D, open to researchers from all types of institutions and organisations worldwide, aimed at the increasingly important area of communication technology. The Journal''s emphasis is particularly on the issues impacting behaviour at the system, service and management levels. Published twelve times a year, it provides coverage of advances that have a significant potential to impact the immense technical and commercial opportunities in the communications sector. The International Journal of Communication Systems strives to select a balance of contributions that promotes technical innovation allied to practical relevance across the range of system types and issues.
The Journal addresses both public communication systems (Telecommunication, mobile, Internet, and Cable TV) and private systems (Intranets, enterprise networks, LANs, MANs, WANs). The following key areas and issues are regularly covered:
-Transmission/Switching/Distribution technologies (ATM, SDH, TCP/IP, routers, DSL, cable modems, VoD, VoIP, WDM, etc.)
-System control, network/service management
-Network and Internet protocols and standards
-Client-server, distributed and Web-based communication systems
-Broadband and multimedia systems and applications, with a focus on increased service variety and interactivity
-Trials of advanced systems and services; their implementation and evaluation
-Novel concepts and improvements in technique; their theoretical basis and performance analysis using measurement/testing, modelling and simulation
-Performance evaluation issues and methods.