DRL Framework for Joint Task Offloading and Resource Allocation in Terahertz Aerial-Space MEC Networks

Mobile edge computing (MEC) architectures based on Terahertz (THz) aerial-space networks are attractive for the efficient computation and low-latency transmission of space exploration tasks delivered by LEO satellites. Specifically, UAVs are deployed to provide MEC services, which can reduce the computational burden of ground stations (GSs). In this THz MEC network, a DRL framework is proposed to solve the NP-hard long-term resource efficiency (RE) maximization via joint task offloading and resource allocation. The proposed DRL framework maximizes RE by learning from the interaction between the THz MEC environment and the DRL agent. Additionally, the multi-task mechanism is adopted in the DRL agent to train offloading and allocation sub-tasks cooperatively.