30 / 2024-08-09 11:27:24

An Resource Allocation Strategy in Space-Air-Ground Integrated Network

unmanned aerial vehicles,task offloading,Low Earth orbit satellites,deep reinforcement learning,Space-Ground integrated network framework

Conventional unmanned aerial vehicles (UAVs) suffer from constrained computational resources, hindering their ability to undertake extensive data processing and promptly fulfill real-time tasks, particularly in areas like face identification. To address this challenge, we have devised a space-air-ground integrated network for data offloading, emphasizing the utilization of Low Earth orbit (LEO) satellites to assist in UAV data offloading. This framework introduces a collaborative computation offloading strategy tailored for LEO satellites and UAVs, wherein computationally intensive data segments are transmitted to the satellite for processing. This methodology effectively curtails transmission delays and offers greater flexibility in task allocation. Recognizing the dynamic interplay between satellite and UAV networks, we have framed the collaborative computation offloading challenge as a Distributed Markov Decision Process and developed a computation offloading algorithm leveraging Multi-Agent Proximal Policy Optimization (MAPPO) to derive efficient offloading strategies. Our simulation outcomes underscore the substantial improvement in system computing performance and latency reduction achieved by our proposed approach.