How Young Developers Are Putting AI Code on Satellites with Cloud‑Native Platforms

Hundreds to thousands of kilometers above Earth, low‑orbit satellites are becoming the playground for young innovators eager to see their code fly in space.

In June, China Youth Daily, Beijing University of Posts and Telecommunications, and Huawei Cloud launched the “Code to Space” call for projects, inviting university students to develop innovative applications for satellites.

Guided by Professor Liu Fangming of Huazhong University of Science and Technology, doctoral student Chen Shutong and the OpenEdgeNeXt team won the championship with their system “CoStar: A Neural‑Network‑Based Dynamic Splitting Star‑Ground Collaborative Remote‑Sensing Image Analysis System.”

The “TianSuan Constellation” program, jointly initiated by BUPT Shenzhen Institute and Tianyi Research Institute, deployed its first pilot satellite in December 2021 and aims to complete a network by 2023. Huawei Cloud, as a co‑builder, introduced cloud‑native and edge‑cloud collaborative concepts to space computing, providing platforms such as KubeEdge, the AI framework Sedna, elastic cloud servers, and the MindSpore AI engine.

On November 9, 2022, the competition results were announced. Eleven teams from across China received awards; the OpenEdgeNeXt team claimed the champion title, while teams from Huazhong University, Wuhan University, BUPT, China Telecom Research Institute, and Beijing Border Space Technology also earned excellence awards.

“Our code is no longer confined to the lab; it now runs on real satellites, tackling practical problems and achieving better real‑world impact,” said Professor Liu.

CoStar splits a remote‑sensing neural‑network model between onboard satellite hardware and ground‑station data centers, enabling fine‑grained, adaptive, and diversified star‑ground collaboration for applications such as disaster monitoring and resource exploration. The dynamic splitting mechanism improves inference efficiency by 2.97× and dramatically lowers average bandwidth demand.

Commercially, the solution is projected to save over 100 million CNY annually.

Other participating teams demonstrated cloud‑native, edge‑AI solutions: a fire‑detection system using a “cloud‑edge‑device” collaborative architecture, a ship‑detection system for maritime security, and the “Screen Peng” star‑ground proxy that creates an IP‑transparent link between satellites and ground servers, simplifying deployment of any IP‑based application in space.

These projects illustrate a new “industry‑academia integration” model, where universities contribute cutting‑edge research in cloud computing, edge computing, and artificial intelligence to accelerate satellite intelligence and serve critical societal needs.