{"title":"Enhanced-SETL: A multi-variable deep reinforcement learning approach for contention window optimization in dense Wi-Fi networks","authors":"","doi":"10.1016/j.comnet.2024.110690","DOIUrl":null,"url":null,"abstract":"<div><p>In this paper, we introduce the Enhanced Smart Exponential-Threshold-Linear (Enhanced-SETL) algorithm, a new approach that uses the multi-variable Deep Reinforcement Learning (DRL) framework to simultaneously optimize multiple settings of the Contention Window (CW) in IEEE 802.11 wireless networks. Unlike traditional DRL methods that adjust only a single CW parameter, our innovative approach simultaneously optimizes both the CW minimum (<span><math><mrow><mi>C</mi><msub><mi>W</mi><mrow><mi>m</mi><mi>i</mi><mi>n</mi></mrow></msub></mrow></math></span>) and CW Threshold (<span><math><mrow><mi>C</mi><msub><mi>W</mi><mrow><mi>T</mi><mi>h</mi><mi>r</mi><mi>e</mi><mi>s</mi><mi>h</mi><mi>o</mi><mi>l</mi><mi>d</mi></mrow></msub></mrow></math></span>), significantly improving network traffic control. We utilize a Double Deep <em>Q</em>-learning Network (DDQN) for dynamic updates of these CW settings, broadcasted across the dense Wi-Fi networks. This dual adjustment method, coupled with dynamic, data-driven updates, not only enhances throughput, but also reduces collision rates, and ensures fairness access across both static and dynamic wireless environments. Enhanced-SETL achieves a throughput improvement ranging from 3.55% up to 43.73% and from 3.98% up to 30.15% in static and dynamic scenarios over standard protocols and state-of-the-art DRL models, while maintaining a fairness index near 99% across diverse stations, showcasing its effectiveness and adaptability in various network conditions.</p></div>","PeriodicalId":50637,"journal":{"name":"Computer Networks","volume":null,"pages":null},"PeriodicalIF":4.4000,"publicationDate":"2024-08-07","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://www.sciencedirect.com/science/article/pii/S138912862400522X/pdfft?md5=fe5bf107051a7089d2435b2d6188ffa5&pid=1-s2.0-S138912862400522X-main.pdf","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computer Networks","FirstCategoryId":"94","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S138912862400522X","RegionNum":2,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"COMPUTER SCIENCE, HARDWARE & ARCHITECTURE","Score":null,"Total":0}
引用次数: 0
Abstract
In this paper, we introduce the Enhanced Smart Exponential-Threshold-Linear (Enhanced-SETL) algorithm, a new approach that uses the multi-variable Deep Reinforcement Learning (DRL) framework to simultaneously optimize multiple settings of the Contention Window (CW) in IEEE 802.11 wireless networks. Unlike traditional DRL methods that adjust only a single CW parameter, our innovative approach simultaneously optimizes both the CW minimum () and CW Threshold (), significantly improving network traffic control. We utilize a Double Deep Q-learning Network (DDQN) for dynamic updates of these CW settings, broadcasted across the dense Wi-Fi networks. This dual adjustment method, coupled with dynamic, data-driven updates, not only enhances throughput, but also reduces collision rates, and ensures fairness access across both static and dynamic wireless environments. Enhanced-SETL achieves a throughput improvement ranging from 3.55% up to 43.73% and from 3.98% up to 30.15% in static and dynamic scenarios over standard protocols and state-of-the-art DRL models, while maintaining a fairness index near 99% across diverse stations, showcasing its effectiveness and adaptability in various network conditions.
期刊介绍:
Computer Networks is an international, archival journal providing a publication vehicle for complete coverage of all topics of interest to those involved in the computer communications networking area. The audience includes researchers, managers and operators of networks as well as designers and implementors. The Editorial Board will consider any material for publication that is of interest to those groups.