Two-Stage Game-Theoretic Resource Allocation for Satellite-Assisted Low-Altitude Networks

Abstract

Low-altitude aerial wireless networks (LAWNs) enabled by uncrewed aerial vehicles (UAVs) can provide proximity access and edge computing for delay-sensitive Internet of Things (IoT) services, but they face fast channel variation, mobility-induced service degradation, and limited terrestrial support. This article studies a satellite-assisted LAWN with a dense UAV access layer serving massive IoT devices, a low Earth orbit (LEO) layer offering quota-limited backhaul and collaborative computing, and a medium Earth orbit (MEO) orchestration layer that coordinates inter-UAV sharing with low signaling overhead. We develop an end-to-end latency model for two-tier task execution under mobility-dependent effective satellite resources and formulate a per-period mixed-integer benchmark that maximizes aggregate UAV profit using a contract-style quality-of-service (QoS) score that penalizes both mean delay and tail timeout risk via conditional value at risk (CVaR). To obtain an implementable solution under incomplete information, we design a two-stage closed-loop scheme: 1) a quota-bidding game with virtual pricing allocates shared LEO backhaul service rates and computing across UAVs; and 2) a distributed association mechanism updates UE-UAV matching using lightweight UAV broadcasts and local link measurements. Monte Carlo simulations show consistent gains over strong baselines across capacity, mobility, and reliability-demand regimes, improving low-latency success by 5%–9% in representative settings and reducing tail delay by tens of milliseconds while lowering tail-risk metrics and improving objective-aligned profit.