Who Pays for the Drones? A Two-Part Market for Multi-Provider UAV Edge Computin

Abstract Uncrewed aerial vehicles (UAVs) can extend edge computing to crowds and areas that ground infrastructure cannot serve. However, a UAV incurs a cost for every minute it is airborne, whether or not it computes. This paper examines how that airtime can be funded when several independent providers sell computation through per-task auctions. It is proved that, when deployed drones compete on price for each task, prices that serve every worthwhile task and never exceed users’ valuations cannot cover airtime costs. Simulations confirm this dilemma: pricing tasks at marginal cost causes losses of up to $243 per hour with 20 drones, whereas including airtime in task prices turns away valuable tasks and still causes losses with 15–20 drones. A two-part mechanism is proposed: drone availability is paid in a periodic capacity market, and each task is priced in a second-price auction. Capacity offers are compared by cost per unit of service and paid threshold prices, which makes truthful bidding optimal in each period and keeps payments within the value at stake. In 3,950 simulations, the mechanism keeps providers profitable at every fleet size evaluated and, with 20 drones, yields $113–215 per hour more welfare than full-cost pricing. With an optimally sized fleet, full-cost pricing and an adaptation of a published mechanism are 2–4% more efficient, so the mechanism’s value lies in sustainable provider profits and robustness to a fleet size not known in advance. Because users or a sponsor must fund the capacity payments, provider profitability does not imply that users gain.