Predictive routing for dynamic UAV networks

The emerging Flying Ad-Hoc Networks (FANETs) provide efficient infrastructure solutions for a wide range of military and commercial applications. These networks are typically composed of high-speed Unmanned Aerial Vehicles (UAVs) and form dynamic network topologies. As such, conventional routing algorithms do not efficiently accommodate these continued topology changes and exhibit poor delay performances. In this paper, we introduce an optimal routing algorithm for UAV networks with queued communication systems based on Dijkstra's shortest path algorithm as a primary step towards developing a fully predictive communication platform. The core idea is to incorporate the anticipated locations of intermediate nodes during a transmission session into the path selection criterion. We further evaluate the impact of measurement uncertainties on choosing the optimal path and on the resulting end-to-end delay. The simulation results confirm the superior delay performance of the proposed algorithm compared to the conventional routing algorithms. The enhancement is more significant, when the network size is larger, the relative node velocities are higher, and the average waiting times in transmit buffers are longer.