Unmanned aerial vehicle traffic network design with risk mitigation

As unmanned aerial vehicle (UAV) technology continues to advance, more retailers are considering UAVs as a viable alternative to ground-based transportation for package delivery. While the literature has investigated UAV-based delivery, it often assumes UAV movement within a point-to-point Euclidean distance from origin to destination, due to the size and mobility of UAVs. However, this straight path movement raises significant air traffic management concerns, making it impractical as drone deployment scales up. In this research, we investigate the optimal UAV service network when UAV travel is restricted to flying over street networks in urban areas, as proposed by our industry partner AirMatrix. The goal is to determine the optimal subnetwork — consisting of street segments over which UAV travel is permitted — that balances demand fulfillment and risk mitigation. We propose a bi-objective mixed integer programming model for UAV network design which minimizes risk while maximizing satisfied demand subject to budgetary constraints. To model drone travel risk, we propose six risk measures motivated by practical regulatory considerations and traffic risk research, aimed at reducing total risk, worst-case risk, or risk variance within the network. We further propose a two-stage stochastic programming variant to design a UAV network that is robust to uncertain demand. We show that for one of the proposed risk measures, the stochastic problem and the mean-value deterministic problem result in equivalent optimal UAV network decisions. To the best of our knowledge, the proposed models are the first in the literature to explicitly incorporate risk considerations into UAV traffic network design. Extensive computational experiments are conducted in a case study of suburban Miami, using real street network data, to evaluate the performance of the various risk objectives with respect to the overall network design, path-based and arc-based performance metrics. The results indicate that the total risk and the total risk deviation objectives achieve the best overall performance and effectively minimize risk across various metrics at the least cost.