DRL-Based Optimisation for Task Offloading in Space-Air-Ground Integrated Networks: A Reliability-Driven Approach

• Published in: IEEE International Conference on Communications (2003) pp. 6958 - 6963
• Main Authors: Van Huynh, Dang, Khosravirad, Saeed R., Cotton, Simon L., Dobre, Octavia A., Duong, Trung Q.
• Format: Conference Proceeding
• Language: English
• Published: IEEE 08.06.2025
• Subjects: Channel allocation; Edge computing; Optimization; Quality of service; Reliability engineering; Servers; Simulation; Space-air-ground integrated networks; Training; Vehicle dynamics
• ISSN: 1938-1883
• DOI: 10.1109/ICC52391.2025.11161527

This paper addresses the problem of reliable task offloading in space-air-ground integrated network (SAGIN) based edge computing systems. Specifically, we aim to maximise the successful task offloading ratio for ground users communicating with a satellite's edge server. In our network topology, end-to-end communications are facilitated by relay unmanned aerial vehicles (UAVs). The formulated problem jointly optimises task offloading portions and bandwidth allocations for both ground-to-air and air-to-space links, subject to quality-of-service (QoS) requirements, transmission rates, system bandwidth, and the computing capacity of the satellite's edge server. To solve the formulated complex non-linear, non-convex, and mixed-integer problem, we propose an efficient solution underpinned by a deep reinforcement learning (DRL). Simulation results demonstrate the effectiveness of the proposed method, which achieves stable training performance and an optimised reliable offloading ratio compared to benchmark schemes.