Learning based Spatial Reuse with Adaptive Timestep and Action Space for Dense WLANs

• Item type: Symposium(1)
• 公開日: 2020-06-17
• タイトル: Learning based Spatial Reuse with Adaptive Timestep and Action Space for Dense WLANs
• タイトル / 言語: en
• タイトル: Learning based Spatial Reuse with Adaptive Timestep and Action Space for Dense WLANs
• 言語: eng
• キーワード / 主題Scheme: Other
• キーワード / 主題: 無線・移動体
• 資源タイプ / 資源タイプ識別子: http://purl.org/coar/resource_type/c_5794
• 資源タイプ: conference paper
• 著者所属: Graduate School of Science and Technology, Keio University
• 著者所属: Graduate School of Science and Technology, Keio University,
• 著者所属: Graduate School of Science and Technology, Keio University,
• 著者所属(英): en
• 著者所属(英): Graduate School of Science and Technology, Keio University
• 著者所属(英): en
• 著者所属(英): Graduate School of Science and Technology, Keio University,
• 著者所属(英): en
• 著者所属(英): Graduate School of Science and Technology, Keio University,
• 著者名: Chow, Zhao Wen; Shoto, Sakai; Hiroshi, Shigeno
• 著者名(英): Chow, Zhao Wen; Shoto, Sakai; Hiroshi, Shigeno
• 論文抄録 / 内容記述タイプ: Other
• 論文抄録 / 内容記述: The rapid densification of IEEE 802.11 Wireless Local Area Networks (WLANs) has lead to higher interferences among Basic Service Sets (BSSs) and has negatively impacted their performance. Spatial reuse methods such as Dynamic Sensitivity Control (DSC) or Transmit Power Control (TPC) help mitigate the hidden and exposed terminals issues in these dense deployments. In this work, a Reinforcement Learning (RL) based method with adaptive timestep and action space is proposed to enhance the spatial reuse in dense WLANs. In particular, the problem is modeled through Multi-Armed Bandits (MABs) and the Thompson Sampling strategy is employed. In this scheme, a learner first observes the Received Signal Strengths (RSSs) it can sense and derives a set of Carrier Sense Thresholds (CSTs) from these. It then applies Thompson Sampling with the computed set and updates the model after a specified number of transmissions or a predefined timeout. Simulation results show that the proposed scheme is able to improve the fairness compared to a previous RL scheme while providing a considerable aggregate throughput.
• 論文抄録(英) / 内容記述タイプ: Other
• 論文抄録(英) / 内容記述: The rapid densification of IEEE 802.11 Wireless Local Area Networks (WLANs) has lead to higher interferences among Basic Service Sets (BSSs) and has negatively impacted their performance. Spatial reuse methods such as Dynamic Sensitivity Control (DSC) or Transmit Power Control (TPC) help mitigate the hidden and exposed terminals issues in these dense deployments. In this work, a Reinforcement Learning (RL) based method with adaptive timestep and action space is proposed to enhance the spatial reuse in dense WLANs. In particular, the problem is modeled through Multi-Armed Bandits (MABs) and the Thompson Sampling strategy is employed. In this scheme, a learner first observes the Received Signal Strengths (RSSs) it can sense and derives a set of Carrier Sense Thresholds (CSTs) from these. It then applies Thompson Sampling with the computed set and updates the model after a specified number of transmissions or a predefined timeout. Simulation results show that the proposed scheme is able to improve the fairness compared to a previous RL scheme while providing a considerable aggregate throughput.
• 書誌情報: マルチメディア，分散協調とモバイルシンポジウム2230論文集; 巻 2020, p. 1474-1479, 発行日 2020-06-17
• 出版者 / 言語: ja
• 出版者: 情報処理学会