Adaptivity, reconfigurability and intelligence are key features of the next-generation wireless networks to meet the increasingly diverse quality of service (QoS) requirements of the future applications. Conventional protocol designs, however, struggle to provide flexibility and agility to changing radio environments, traffic types and different user service requirements. In this paper, we explore the potential of deep reinforcement learning (DRL), in particular Proximal Policy Optimization (PPO), to design and configure intelligent and application-specific medium access control (MAC) protocols. We propose a framework that enables the addition, removal, or modification of protocol features to meet individual application needs. The DRL channel access policy design empowers the protocol to adapt and optimize in accordance with the network and radio environment. Through extensive simulations, we demonstrate the superior performance of the learned protocols over legacy IEEE 802.11ac in terms of throughput and latency.

翻译：适应性、可重构性与智能性是下一代无线网络的关键特征，以满足未来应用日益多样化的服务质量需求。然而，传统协议设计难以应对变化的无线环境、业务类型及不同用户服务需求带来的灵活性与敏捷性挑战。本文探索了深度强化学习（特别是近端策略优化）在设计和配置智能应用特定介质访问控制协议方面的潜力。我们提出一个框架，支持添加、移除或修改协议特征以满足个体应用需求。基于深度强化学习的信道接入策略设计使协议能够根据网络与无线环境进行自适应优化。通过大量仿真实验，我们证明学习到的协议在吞吐量和时延方面优于传统IEEE 802.11ac协议。