Scholarlink

This chapter first reviews the fundamentals of DRL and representative DRL algorithms, including DQN and its variants for discrete action spaces, as well as DDPG and PPO for continuous control in the actor-critic framework. We then briefly analyze their design principles, applicability, and computational characteristics, highlighting their potential applications and generic frameworks for various control problems in wireless networks, including decentralized access control, resource allocation in MEC systems, and UAV-assisted wireless networks. Furthermore, we focus on practical limitations of DRL and discuss potential interplays between model-based optimization and model-free DRL for accelerating learning performance. The following chapters will demonstrate these interplays by providing specific case studies with integrated algorithm designs for various control problems in wireless networks.

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

Introduction

  • 章节

  • 发表日期：

    2026年

查看全文
  • Shimin Gong,
  • Dusit Niyato,
  • Bo Gu,
  • Kaibin Huang

摘要

This chapter first reviews the fundamentals of DRL and representative DRL algorithms, including DQN and its variants for discrete action spaces, as well as DDPG and PPO for continuous control in the actor-critic framework. We then briefly analyze their design principles, applicability, and computational characteristics, highlighting their potential applications and generic frameworks for various control problems in wireless networks, including decentralized access control, resource allocation in MEC systems, and UAV-assisted wireless networks. Furthermore, we focus on practical limitations of DRL and discuss potential interplays between model-based optimization and model-free DRL for accelerating learning performance. The following chapters will demonstrate these interplays by providing specific case studies with integrated algorithm designs for various control problems in wireless networks.