Sensing Information Assisted Routing Scheme for UAV Networks

Flying Ad-Hoc Network (FANET) is widely used in network communication services in military, emergency relief, and environmental monitoring. A good routing protocol can provide a guarantee for its reliable transmission in harsh communication conditions. Due to the unique high mobility and frequent topology changes of UAV nodes, traditional routing protocols cannot meet the needs of establishing efficient and reliable paths in the UAV networking process. Therefore, this paper proposes a sensing information-assisted routing algorithm for UAV networks. Utilizing the perception information obtained from the interaction between the UAV and the environment, the movement prediction method is introduced into the routing strategy. Through the prediction of the node movement state, the relative position between nodes can be judged, and the survival time of the link can be calculated. Moreover, in route maintenance, it is possible to determine whether a link is disconnected based on the relative position and movement status of nodes, and re-build routes before disconnection, reducing the packet loss rate during transmission and improving network performance. At the same time, the reinforcement learning method Q-Learning is used to assist routing decisions. Most existing Q-Learning-based protocols use fixed parameters. In the scheme proposed in this paper, Q-Learning parameters can be adaptively adjusted according to network conditions. Comprehensively measure multiple indicators such as transmission delay, channel conditions, and link quality, increase routing metrics, and perform multi-objective optimization to find the optimal path between the source and the target.