Scholarlink

<p>The spectrum scarcity problem prompts the need for efficient resource management in Beyond 5&#xa0;G (B5G) and 6&#xa0;G networks. Cognitive radio (CR), leveraging under-utilized licensed spectrum, offers dynamic solutions for spectrum management. Multi-cell configurations, typical in cellular systems, enable frequency reuse within multi-cell CR networks (CRNs), which enhances spectrum utilization. However, designing effective channel-allocation algorithms for these networks is a critical challenge given the dynamic and uncertain nature of demands in each cell and spectrum availability fluctuations due to licensed users’ activities. This paper develops a novel Reinforcement Learning (RL)-based algorithm for channel allocation across multi-cell CRNs that attempts to maximize the number of served users and reduce the blocking probability while ensuring fairness. The proposed algorithm learns spatiotemporal user-behavior patterns through RL by interacting with the CRN environment. Using the learned experience, the proposed algorithm aims to determine the most efficient channel allocation strategy without requiring prior knowledge of the per-cell demand. Specifically, we formulate the channel allocation problem as a profit-maximization discounted-return problem. This problem is then reformulated as a Markov Decision Problem and solved using RL. Compared to reference algorithms, the simulation results demonstrate that the proposed algorithm, which effectively deals with the per-cell demand uncertainties and channel availability, significantly reduces blocking probability, maximizes served users, and ensures fairness.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Adaptive RL-driven spectrum allocation in multi-cell cognitive B5G networks

  • 期刊论文

  • 发表日期：

    2025年3月13日

查看全文
  • Haythem Bany Salameh,
  • Manar Shatara,
  • Rami Halloush,
  • Ahmed Musa,
  • Mohannad Alhafnawi

摘要

The spectrum scarcity problem prompts the need for efficient resource management in Beyond 5 G (B5G) and 6 G networks. Cognitive radio (CR), leveraging under-utilized licensed spectrum, offers dynamic solutions for spectrum management. Multi-cell configurations, typical in cellular systems, enable frequency reuse within multi-cell CR networks (CRNs), which enhances spectrum utilization. However, designing effective channel-allocation algorithms for these networks is a critical challenge given the dynamic and uncertain nature of demands in each cell and spectrum availability fluctuations due to licensed users’ activities. This paper develops a novel Reinforcement Learning (RL)-based algorithm for channel allocation across multi-cell CRNs that attempts to maximize the number of served users and reduce the blocking probability while ensuring fairness. The proposed algorithm learns spatiotemporal user-behavior patterns through RL by interacting with the CRN environment. Using the learned experience, the proposed algorithm aims to determine the most efficient channel allocation strategy without requiring prior knowledge of the per-cell demand. Specifically, we formulate the channel allocation problem as a profit-maximization discounted-return problem. This problem is then reformulated as a Markov Decision Problem and solved using RL. Compared to reference algorithms, the simulation results demonstrate that the proposed algorithm, which effectively deals with the per-cell demand uncertainties and channel availability, significantly reduces blocking probability, maximizes served users, and ensures fairness.