摘要
飞行自组织网络FANET的高度动态性和不稳定性,导致通讯节点易失效、易受到攻击、欺骗等,为飞行自组织网络的可靠通信带来了巨大挑战.针对飞行自组织网络的高动态性以及链路不可靠问题,提出了一种基于强化学习动态特征驱动的可靠路由发现算法,并设计了一种基于强化学习的可靠路由协议RR-AODV(Reliable Reinforcement AODV).首先,通过距离和相对运动速度矢量双约束条件进行可靠节点预筛选,预判并规避短期链路中断风险,有效避免因节点高速移动导致的频繁路径重建.其次,通过时间、质量、能量、空间四维联合链路评估方法,建立可靠邻居链路,改善了动态网络中节点易失效、链路易中断等问题.最后,通过Q-learning算法构建拓扑感知的智能决策引擎,将洪泛式路由发现改进为动态链路特征驱动的精准单播探索,解决了AODV协议洪泛广播的资源浪费、链路不可靠的问题.经实验表明,相较于AODV、AODV-ETX和P-AODV协议,该协议在投递率和路由开销性能上均有改善,能够更好地适应飞行自组织网络.
The high dynamicity and instability of Flying Ad Hoc Networks(FANET)leading to frequent node failures,vulnerability to attacks,and susceptibility to spoofing,posing significant challenges to reliable communication.To address network dynamics and link unreliability in FANET,this paper proposes a reinforcement learning-based reliable routing discovery algorithm and designs the RRAODV(Reinforcement-based Reliable AODV)protocol.First,a dual-constraint mechanism(distance and relative velocity vectors)pre-filters reliable nodes,preemptively mitigating short-term link interruptions and reducing frequent path reconstructions caused by high-speed mobility.Second,a four-dimensional link assessment framework(time,quality,energy,and space)establishes robust neighbor links,enhancing resilience against node failures and link disruptions in dynamic environments.Finally,a topology-aware intelligent decision engine powered by Q-learning replaces flooding-based route discovery with dynamic link feature-driven unicast probing,resolving AODV's inherent issues of resource waste and unreliable links.Experimental results demonstrate that compared to AODV,AODV-ETX,and P-AODV,the proposed RR-AODV protocol achieves higher packet delivery ratios and lower routing overhead,proving its superior adaptability to the demanding environment of FANET.
作者
李姝
谢睿
冯永新
LI Shu;XIE Rui;FENG Yongxin(School of Equipment Engineering,Shenyang Ligong University,Shenyang 110159,China)
出处
《小型微型计算机系统》
北大核心
2025年第8期1861-1868,共8页
Journal of Chinese Computer Systems
基金
辽宁省属本科高校基本科研业务费专项资金面上项目(JYTMS20230186)资助。
