摘要
针对飞行自组网(Flying Ad Hoc Network,FANET)在通信空白场景下存在的高时延问题,提出了一种深度强化学习(Deep Reinforcement Learning,DRL)辅助的双跳信息增强路由协议(Double-Hop Information Enhanced Routing Protocol,DHRP)。为了实现有效的路由决策,采用马尔可夫决策过程(Markov Decision Process,MDP)对路由行为进行建模,在状态空间设计中结合了节点位置信息与链路信道容量,并综合考虑了双跳范围内的网络信息,以深度值网络为核心,在融合实时网络状态动态调整机制的奖励函数引导下,做出最优下一跳路由决策。实验结果表明,在通信空白场景下,DHRP相较于现有的路由方案,显著降低了FANET的平均端到端时延。此外,在不同节点规模和网络拥塞条件下,DHRP均表现出优越的适应性和鲁棒性,通过对动态网络环境的实时感知与智能决策机制,有效保障了整体网络性能。
To address the challenge of high end-to-end delay in Flying Ad Hoc Network(FANET)under communication blackout scenarios,this paper proposes a Deep Reinforcement Learning(DRL)-assisted Double-Hop Information Enhanced Routing Protocol(DHRP).The proposed protocol models the routing process as a Markov Decision Process(MDP)to enable effective decision-making.In constructing the state space,it incorporates both node location information and link channel capacity,while considering network information within a two-hop neighborhood.Centered on a deep value network,the protocol employs a reward function that reflects real-time network dynamics to guide the agent in selecting the optimal next-hop node.Simulation results show that,compared to existing approaches,DHRP significantly reduces the average end-to-end delay in FANET under communication blackout conditions.Furthermore,DHRP demonstrates strong adaptability and robustness across various node densities and levels of network congestion by leveraging real-time environmental awareness and an intelligent decision-making mechanism to maintain overall network performance.
作者
郭歆莹
李明
朱春华
GUO Xinying;LI Ming;ZHU Chunhua(Key Laboratory of Grain Information Processing and Control of the Ministry of Education,Henan University of Technology,Zhengzhou 450001,China;Key Experiment on Intelligent Perception and Decision-making of Grain Storage Information in Henan Province,Henan University of Technology,Zhengzhou 450001,China;Henan Province Engineering Research Center for Intelligent Monitoring and Application of Grain Conditions,Henan University of Technology,Zhengzhou 450001,China)
出处
《无线电通信技术》
北大核心
2025年第5期929-939,共11页
Radio Communications Technology
基金
国家自然科学基金(61901159)
河南工业大学青年骨干教师培养计划(21420104)。
作者简介
郭歆莹,女,(1989-),博士,副教授,硕士生导师。主要研究方向:5G/6G智能通信、无人机通信;李明,男,(2001-),硕士研究生。主要研究方向:无人机自组网、强化学习;朱春华,女,(1976-),博士,教授,博士生导师。主要研究方向:宽带无线通信、通信信号处理、智能信息处理。
