LAUSR

Space-air-ground integrated networks (SAGINs) provide global information sharing, large-scale coverage, and ubiquitous collaboration architecture, and alleviate the pressure of rapid traffic growth on terrestrial wireless networks. On the other hand, the emergence of artificial intelligence (AI) technology in all walks of life has attracted significant attention on intelligent services. However, frequent link errors and dynamic connections in SAGINs aggravate the issues of data failure and computation slowdown, and constrain the improvement of AI service efficiency. In this article, we propose a novel coded stor-age-and-computation (CSC) architecture, which can offer reliable storage and flexible computation offloading to accelerate distributed machine learning. Through several case studies, we demonstrate that the designed CSC-AI system can realize reliable massive data retrieval and fast computation offloading in SAGINs. Finally, we identify future research issues for CSC-AI system optimization in SAGINs.