From space sensing to space computing: China's AI satellites take off

China sent two AI-powered hyperspectral satellites into orbit on Wednesday morning, marking a step forward in bringing AI computing capabilities into space.

The Smart Dragon-3 carrier rocket lifted off at 10:38 am on Wednesday from the offshore waters near Haiyang, east China's Shandong Province, carrying the Oriental Smart Eye (OSE) hyperspectral satellites 01 and 02 into their planned orbits. The offshore launch mission was conducted by the Taiyuan Satellite Launch Center.

From space sensing to space computing

Unlike conventional remote-sensing satellites that send raw data back to Earth for processing, the two satellites can analyze data directly in orbit.

Each satellite has onboard AI computing capability of up to 400 TOPS, or 400 trillion AI operations per second.

"A high-performance laptop has about 20 to 30 TOPS of computing power. That means a single hyperspectral AI satellite has computing power equivalent to 20 to 30 high-performance laptops," said Song Bingchen, Star.ai's co-founder.

"With an onboard 'brain,' we can process the data in space and transmit results instead of images," he added.

"Each satellite can generate tens to hundreds of gigabytes of imaging data per day," said Peng Feiyu, a Satellite AI Algorithm Engineer. "It can process 15 seconds of imagery in orbit within just 10 to 15 minutes, and transmit only the analyzed results to the ground. The downlinked data is reduced to just tens of kilobytes or tens of megabytes."

What makes the AI satellites different?

The two OSE satellites feature a spatial resolution of five meters, an imaging swath of 300 kilometers and 22 calibrated spectral channels.

They can not only capture the shape and color of objects, but also can identify the unique "spectral fingerprints" of individual pixels, providing more detailed information about what they observe.

"The two satellites are benchmarked against leading hyperspectral satellites currently in orbit," said Wang Mi, a professor at Wuhan University. "They can conduct hyperspectral imaging over land during the day and capture light scenes associated with human activity across cities during the night."

How AI-powered satellites put to work?

The two satellites will work with the OSE Gaofen-1 satellite already in orbit. Their data can support applications in agriculture, forestry, environmental protection, marine monitoring and mineral resources.

Instead of simply sending images to users, AI-powered satellites can analyze the data and provide recommendations.

"For disaster monitoring, conventional satellites may take hours to provide the latest information, while our satellites could potentially provide warnings and decision-making guidance within minutes," said Song.

"In agriculture, intelligent satellites can provide specific recommendations, such as how to sow crops and carry out agricultural operations, instead of simply providing images," Song added.

Satellite 01, part of a space cooperation project between China and Indonesia, is expected to help safeguard agricultural production in Indonesia by precisely tracking crop growth and assessing disaster risks, according to the developer.

Satellite 02, part of a space cooperation project between China and Uzbekistan, is designed to support the landlocked country's full-cycle monitoring of cotton cultivation from sowing to harvesting.

High-quality data combined with onboard computing capabilities is a key step in bringing space computing from concept to reality. More importantly, the launch marks the beginning of a much larger plan.

The OSE constellation is planned to eventually comprise 258 satellites, with deployment scheduled for completion by 2030.

An integrated "satellite-rocket-ship" model

The launch also highlighted the development of China's commercial space industry.

The rocket, satellites and launch vessel involved in the mission were all produced or developed within the Dongfang Aerospace Port industrial park in Yantai, creating an integrated "satellite-rocket-ship" model.

"The integrated model is expected to improve offshore launch capabilities, accelerate satellite constellation deployment and support the large-scale development of commercial spaceflight," said Wang Chao, an official with the Shandong Provincial Department of Industry and Information Technology.

(With input from Xinhua)