Amazon has been testing machine learning software in space for the past 10 months so that it can independently analyse Earth observation photographs and only transmit the best ones to Earth.
Over the previous ten years, the use of Earth observation satellites has greatly increased. Hundreds of satellites, both public and private, orbit the globe and keep an eye on its surface, watching for indications of climate change as well as enemy state activity. The large amount of data that these satellites collect makes it difficult to deliver it all to Earth due to the dearth of ground stations and the bandwidth that is available. But how might the best and most pertinent photographs be picked out by a satellite to relay home?
To solve this issue, Amazon collaborated with its partners, Italian space start-up D-Orbit and computing technology company Unibap, to develop artificially intelligent software that runs directly on an orbiting satellite and can choose which images to send to Earth.
“Using Amazon Web Services (AWS) software to perform real-time data analysis onboard an orbiting satellite, and delivering that analysis directly to decision makers via the cloud, is a definite shift in existing approaches to space data management”, “It also helps push the boundaries of what we believe is possible for satellite operations. Providing powerful and secure cloud capability in space gives satellite operators the ability to communicate more efficiently with their spacecraft and deliver updated commands using AWS tools they’re familiar with.”
Max Peterson, Vice president, AWS
The test was conducted using the ION satellite, which D-Orbit launched in January 2022. According to the announcement from AWS, during the testing, the Unibap-built machine learning payload processed “huge volumes of space data directly onboard” the satellite. (Machine learning is the term used to describe software algorithms that can recognise patterns in historical data and make judgments without being explicitly told what to do.) The system makes use of AWS machine learning models, which instantly analyse satellite imagery that has been obtained, and the AWS IoT Greengrass cloud management and analytics system, which can function even when there is a lack of connectivity.
“We want to help customers quickly turn raw satellite data into actionable information that can be used to disseminate alerts in seconds, enable onboard federated learning for autonomous information acquisition, and increase the value of data that is downlinked”, “Providing users real-time access to AWS edge services and capabilities on orbit will allow them to gain more timely insights and optimize how they use their satellite and ground resources.”
Fredrik Bruhn, Unibap, Chief evangelist in digital transformation and co-founder
The experiment’s machine learning software successfully recognised objects like atmospheric clouds, billows of smoke from wildfires, as well as ground- and sea-based buildings. According to AWS, the software also managed to speed up and improve the delivery process by up to 42% by reducing the size of the images transmitted to Earth.