Cross-Layer Queue-Stability-Oriented Optimization for Multi-UAV-Assisted Vehicular Communications

This paper considers a multi-UAV-assisted vehicular communication system in which multiple unmanned aerial vehicles (UAVs) support bidirectional communications for multiple moving ground vehicles while maintaining wireless links to ground base stations. Unlike conventional UAV communication designs that mainly emphasize throughput, coverage, or trajectory efficiency, our focus is on a cross-layer queue-aware problem. In particular, we explicitly account for dynamic traffic arrivals, physical queue evolution, backhaul stress, and delay constraints. To deal with the strong coupling among vehicle–UAV association, UAV–BS backhaul selection, UAV mobility, and uplink/downlink resource allocation, we develop a hierarchical cross-layer framework. At the slow timescale, the framework determines communication association and backhaul selection. At the fast timescale, it adjusts UAV motion and communication resource allocation according to queue states, delay urgency, and link quality. The distinguishing feature is that the same physical- and virtual-queue pressures link these two timescales, affecting both the discrete UL/DL relay and BS-selection decisions and the continuous mobility and access–backhaul resource splits. We also introduce virtual queues to track backhaul pressure and delay violations, so that the proposed method can regulate long-term service stability and quality-of-service performance instead of relying only on instantaneous communication metrics. Simulation results under multiple representative scenarios and traffic load levels show that the proposed approach achieves better overall performance than representative benchmark schemes in terms of queue backlog, backhaul stress, and delay-related metrics, while maintaining competitive throughput. These results show the importance of explicitly handling access–backhaul coupling and queue-aware delay regulation in dynamic multi-UAV vehicular communication systems.