Scholarlink

As the complexity of next-generation mobile communication networks increases, resource allocation becomes one of the fundamental technologies to make a competitive advantage among vendors. Artificial intelligence (AI), particularly reinforcement learning (RL), provides significant opportunities for improvements in this domain. Additionally, the open radio access network (O-RAN) architecture allows seamless integration of AI capabilities into networks through rApp and xApp concepts. This chapter presents two deep reinforcement learning (DRL)-based uplink resource allocation methods, where the first method focuses on maximizing the throughput metrics, while the second one balances the trade-off between throughput and fairness metrics. In the O-RAN architecture, multiple trained DRL models can be orchestrated by rApp policies to dynamically allocate the limited radio resources according to time-varying application requirements. The simulation results demonstrate that the DRL-based uplink resource allocation methods can be effectively utilized to meet different performance objectives specified by high-level policies.

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

Deep Reinforcement Learning Based Uplink Resource Allocation in Open RAN Systems

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Ali Yıldırım,
  • Hasan Anıl Akyıldız,
  • İbrahim Hökelek,
  • Hakan Ali Çırpan

摘要

As the complexity of next-generation mobile communication networks increases, resource allocation becomes one of the fundamental technologies to make a competitive advantage among vendors. Artificial intelligence (AI), particularly reinforcement learning (RL), provides significant opportunities for improvements in this domain. Additionally, the open radio access network (O-RAN) architecture allows seamless integration of AI capabilities into networks through rApp and xApp concepts. This chapter presents two deep reinforcement learning (DRL)-based uplink resource allocation methods, where the first method focuses on maximizing the throughput metrics, while the second one balances the trade-off between throughput and fairness metrics. In the O-RAN architecture, multiple trained DRL models can be orchestrated by rApp policies to dynamically allocate the limited radio resources according to time-varying application requirements. The simulation results demonstrate that the DRL-based uplink resource allocation methods can be effectively utilized to meet different performance objectives specified by high-level policies.