An auction approach to aircraft bandwidth scheduling in non-terrestrial networks

Internet Service has witnessed only limited deployment on commercial flights, where network infrastructure is lacking and subscription fees are high. It is natural to design an efficient scheduling strategy for Internet service access on flights, while maximizing social welfare. There are challenges to use either antennas on the ground or geostationary (GEO) satellites in space as Internet Service Provider (ISP) equipment. First, preemptive-purchase and resource-allocation in post-paid plans may not satisfy the dynamic demand for large amounts of flight-ISP machine links. Second, finding an optimal, dynamic, online solution based on flight demands is NP-hard. Third, existing solutions often fail to maximize either seller profit or social welfare. To address the aforementioned challenges, we formulate an online bandwidth scheduling integer linear program (ILP) among flights, antennas, and GEO satellites in a Non-Terrestrial Network (NTN). We transform the ILP into an exponential optimization problem, which is then relaxed, with its dual formulated. We iteratively solve a series of dual subproblems in polynomial time, towards social welfare maximization with a good competitive ratio in typical network settings. In empirical studies, our algorithm achieved a much better competitive ratio than the theoretical worst-case guarantee, and clearly outperforms alternative algorithms.