Artificial Intelligence Service by Satellite Networks based on Ensemble Learning with Cloud-Edge-End Integration

Abstract

Satellite network is the necessary supplement to ground network for wireless coverage for remote area. At the same time, Deep Neural Networks (DNN) based artificial intelligence (AI) applications has been widely applied in many aspects of people's life, but most DNN models are deployed in cloud due to their high complexity. Since cloud is far from end users in remote area, cloud based AI service cannot meet the latency requirement for users at remote area. Satellite based edge computing is a promising solution to this problem, but the limited resources on satellite platform cannot support complex models. In this paper, we propose to integrate the cloud, satellite edges and the end users to form an AI service framework, and apply ensemble learning to provide efficient and reliable AI service. Two case studies on image classification task are performed to show the effectiveness of the proposed scheme.