For the past 10 months, Amazon has been testing machine learning software in space that can independently analyze Earth observation images and send only the best ones back to Earth.
Over the past decade, Earth observation satellites have experienced a massive boom. Hundreds of satellites, both public and private, circle the planet and monitor its surface for signs of climate change as well as the actions of enemy states. The volume of data that these satellites collect is so large that it is difficult to send them to Earth due to the limited number of ground stations and available bandwidth. But how do you get the satellite to choose the best and most relevant images to send home?
Amazon and its partners, Italian space startup D-Orbit and computing technology developer Unibap, have teamed up to demonstrate a solution to this problem – AI software running directly on an orbiting satellite that can make its own decisions about which photos to transmit to Earth. .
Related: Microsoft teams up with SpaceX to launch Azure Space to take cloud computing to the last frontier
“Using Amazon Web Services (AWS) software to analyze data in real time on board an orbiting satellite and delivering that analysis directly to decision makers via the cloud is a definite shift in the way space data is managed.” – Max Peterson, AWS. vice president, the message says (will open in a new tab). “It also helps to push the boundaries of what we think is possible for satellite operations. Providing powerful and secure cloud experiences in space gives satellite operators the ability to more effectively communicate with their spacecraft and deliver updated commands with the AWS tools they are familiar with. “
The experiment was carried out on the D-Orbit ION satellite, which was launched in January 2022. During the test, a machine learning payload built by Unibap processed “large amounts of space-based data directly on board” the satellite, AWS said in a statement. (Machine learning refers to software algorithms that can learn from patterns in the source data to make decisions without following explicit instructions.) The system uses AWS machine learning models that analyze received satellite imagery in real time, as well as AWS cloud management IoT Greengrass. and an analytics system that can work even during periods with limited connectivity.
“We want to help customers quickly turn raw satellite data into actionable information that can be used to distribute warnings in seconds, enable embedded federated learning for autonomous intelligence gathering, and add value to downlink data.” — Fredrik Bruhn, Chief Evangelist in digital transformation and co-founder of Unibap, the statement said. “Giving users real-time access to AWS edge services and in-orbit capabilities will allow them to get more timely information and optimize the use of their satellite and ground resources.”
During the experiment, the machine learning program successfully identified objects such as atmospheric clouds and puffs of smoke from forest fires, as well as buildings on land and ships at sea. The software has also been able to reduce the size of images transmitted to Earth by up to 42%, AWS said in a statement, making the delivery process faster and more efficient.
The satellite is still in orbit, continuing its experiments.
Follow Tereza Pultarova on Twitter @TerezaPultarova (will open in a new tab). Follow us on Twitter @Spacedotcom (will open in a new tab) and on facebook (will open in a new tab).