Less is More: Channel-Aware Semantic Satellite Access Network Slicing

Satellites equipped with computing capabilities play a crucial role as access platforms for 5G and NextG (beyond 5G) non-terrestrial networks (NTNs). They facilitate the continuous execution of resource-intensive edge-assisted deep learning (DL) tasks offloaded from user equipment (UEs) in remote areas. To manage this effectively, satellite access network (SAN) resources must be carefully 'sliced', taking into account both the limited energy availability and the inherent scarcity of SAN resources. Existing SAN slicing approaches tend to use semantic communications for UE data transmission but overlook the impact of varying channel qualities between UEs and satellites. A cyclic dependency between the inter-slice and intra-slice resource schedulers makes it challenging to incorporate channel awareness at both levels. Armed with this insight, in this paper, we propose channel-aware semantic SAN (CASemSAN), a semantic SAN slicing algorithm that considers the channel conditions for NextG AI-native NTNs. It not only compresses tasks' data according to their semantics but also exploits the channel conditions of a SAN to support more tasks while still minimizing overall energy consumption. After analyzing the characteristics of this optimization problem, we propose an online greedy CASemSAN slicing algorithm to approximate its optimal solution. Extensive experiments verify the effectiveness of CASemSAN in energy saving and its ability to support a substantial number of tasks, compared with other baselines.