End-to-end Multipath Optimization for Reliable Aerial Connectivity

Next-generation Aerial Vehicles (AVs) demand multipath Air-to-Ground (A2G) connectivity to maintain reliable communications for safety-critical services. While utilizing redundant links improves the achievable Quality of Service (QoS), the link usage must be optimized to minimize the overall cost of connectivity. In this paper, we provide an optimization framework that minimizes the communication cost while ensuring to meet the QoS demands of the aerial application. Given the available access links and their QoS capabilities, the model selects a set of links that can jointly meet the required flow capacity, reliability and latency constraints with redundant transmission. Studying the Remote Piloting (RP) use case, we realistically implement the scenario in a flight and channel simulator, and feed the optimizer with the measured link quality from the simulation. As the optimization model is non-convex, we also linearize it, and benchmark them against conventional heuristic and brute-force algorithms. The results show that while the linearized model is favorable for online optimization scenarios due to fast computation times, the nonlinear model with Taylor approximations achieves more accurate latency behavior, and can be suitable for latency-critical applications in offline scenarios. Although dual cellular connectivity can suffice the QoS demands most of the time, excessive number of link switches imply a demand to orchestrate the access networks in a unified way to minimize connectivity disruptions.