Scholarlink

<p>The efficient utilization of limited radio spectrum remains a critical challenge in cognitive radio (CR) networks, particularly in underlay communication, where secondary users must operate under strict interference constraints. This paper proposes an actor-critic based deep reinforcement learning (DRL) framework for dynamic transmit power control of secondary users, aiming to enhance spectral efficiency while satisfying interference constraints. The proposed framework incorporates a diverse set of DRL algorithms to systematically investigate their effectiveness in continuous power allocation under a dynamically varying wireless environment. The effectiveness of the proposed approach is validated through extensive simulations using key performance metrics, including average achievable capacity, average interference level, training time, and execution time. Furthermore, the DRL-based solutions are compared with a nature-inspired optimization algorithm serving as a near-optimal performance benchmark. Simulation results show that, among the considered DRL methods, TD3 (twin delayed deep deterministic policy gradient) achieves performance close to the benchmark while reducing execution time by approximately 91%, demonstrating the suitability of the proposed framework for real-time underlay communication.</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

Actor-critic based deep reinforcement learning framework for underlay communication

  • 期刊论文

  • 发表日期：

    2026年5月20日

查看全文
  • Vishwas Srivastava,
  • Binod Prasad

摘要

The efficient utilization of limited radio spectrum remains a critical challenge in cognitive radio (CR) networks, particularly in underlay communication, where secondary users must operate under strict interference constraints. This paper proposes an actor-critic based deep reinforcement learning (DRL) framework for dynamic transmit power control of secondary users, aiming to enhance spectral efficiency while satisfying interference constraints. The proposed framework incorporates a diverse set of DRL algorithms to systematically investigate their effectiveness in continuous power allocation under a dynamically varying wireless environment. The effectiveness of the proposed approach is validated through extensive simulations using key performance metrics, including average achievable capacity, average interference level, training time, and execution time. Furthermore, the DRL-based solutions are compared with a nature-inspired optimization algorithm serving as a near-optimal performance benchmark. Simulation results show that, among the considered DRL methods, TD3 (twin delayed deep deterministic policy gradient) achieves performance close to the benchmark while reducing execution time by approximately 91%, demonstrating the suitability of the proposed framework for real-time underlay communication.