基于启发式Q学习的FANET可信路由算法

无人机自组织网络（FANET）是实现无人机自主集群的关键技术，其通过各无人机节点来完成协同通信。但节点的高机动性、网络结构的开放性造成FANET拓扑变化频繁，容易遭受恶意攻击。为此，提出一种基于启发式Q学习的可信路由算法HQTR。将FANET中的路由选择问题映射为有限马尔科夫决策过程，针对路由层面的黑洞攻击与泛洪攻击，引入数据包转发率与路由请求发送速率，通过模糊推理计算节点的信任值，同时考虑节点的邻居关系，提出一种模糊动态信任奖励机制。结合单跳链路状况设计启发式函数，采用改进的ε-贪婪策略来平衡利用-探索过程，引导当前节点选择最优可信下一跳节点。仿真结果表明，相对AOMDV、TEAOMDV与ESRQ算法，HQTR算法能够有效应对黑洞攻击与RREQ泛洪攻击，降低节点高速运动与网络规模变化所造成的影响，提高数据包投递率与吞吐量，减少路由开销与平均端到端时延。

Trusted Routing Algorithm Based on Heuristic Q-Learning for FANET

The Flying Ad hoc Network(FANET) is the key technology for realizing UAV autonomous clusters.It completes cooperative communication through UAV nodes.However, owing to the high mobility of nodes and the openness of the network structure, FANET topology changes frequently and is vulnerable to malicious attacks.Therefore, a trusted routing algorithm HQTR based on heuristic Q-learning is proposed.The routing problem in the FANET is mapped to a finite Markov decision process.To mitigate black hole and flooding attacks at the routing level, a packet forwarding rate and a routing request sending rate are introduced.The trust value of the node is calculated via fuzzy reasoning, and considering the neighbor relationship of the node, a fuzzy dynamic trust reward mechanism is proposed.Combined with the single hop link condition, a heuristic function is designed, and an improved greedy strategy is used to balance the utilization-exploration process to facilitate the current node in selecting the best trusted next hop node.Simulation results show that compared with AOMDV, TEAOMDV, and ESRQ algorithms, the HQTR algorithm can effectively address black hole and RREQ flooding attacks, reduce the effects of high-speed node movements and network scale changes, improve the packet delivery rate and throughput, and reduce the routing overhead and average end-to-end delay.