Communication-Aware Hierarchical Routing for FANET: A Reinforcement Learning Approach

As an essential flexible and real-time wireless communication infrastructure, flying ad hoc network is regarded as a key enabler for low-altitude intelligent networks (LAIN). However, the deployment of drone networks still faces challenges due to high node mobility, dynamic topology, and limited energy resources. To address these issues, this paper proposes CA-HRF, a communication-aware hierarchical routing framework that combines fuzzy clustering with reinforcement learning in a two-layer architecture. Specifically, a communication-aware fuzzy C-means algorithm enables adaptive cluster formation, while a lightweight proximal policy optimization approach trains cluster head agents for inter-cluster forwarding based on a multi-objective reward function. Meanwhile, complexity analysis and simulation results show that CA-HRF outperforms benchmark protocols, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">e.g.</i>, ad hoc on-demand distance vector (AODV), AODV with k-means clustering, and flat reinforcement learning-based routing, in terms of packet delivery ratio, end-to-end delay, and throughput.