Intelligent latency-oriented optimization for multi-UAV-assisted mobile edge computing in space-air-ground integrated networks

Unmanned aerial vehicles (UAV)-assisted mobile edge computing (MEC) in space-air-ground integrated networks (SAGINs) provide a promising solution for enhancing communication, computing, and storage services for the increased number of Internet of Things (IoT) devices. However, jamming attacks and co-channel interference severely degrade the network performance due to wide field of line of sight. In this work, we propose a resource scheduling method to jointly optimize the channel selection, UAV deployment and task offloading to minimize both the communication and computing latency, under the malicious jamming attacks and resource constraints. Considering the highly dynamic and complex nature of the SAGIN environment and the multi-UAV collaboration framework, we then design an advanced anti-jamming-driven multi-agent deep reinforcement learning (MADRL) method based on the multi-agent twin-delayed deep deterministic policy gradient (MATD3) algorithm. This method adaptively adjusts the resource scheduling strategy to enhance the network’s ability to withstand jamming attacks, reduce total network latency, and maintain real-time services even under unfavorable conditions. Simulation results show that our proposed method significantly outperforms existing benchmark methods in terms of latency reduction, signal-to-interference-plus-noise ratio (SINR) improvement, and overall network robustness under jamming attacks. For example, the proposed MATD3-SAGAJ method reduces latency by about 10% and improves the SINR satisfaction ratio from around 91% to nearly 95% compared to the current optimal benchmark method.