For the past ten months, Amazon has been testing machine learning software in space. This software can independently analyze earth observation images and send only the best to earth.
earth observation satellite We’ve experienced a huge boom in the last decade. Hundreds of satellites, both government-owned and private, orbit the Earth, monitoring its surface and watching for signs. climate change not only, enemy activityThe amount of data these satellites acquire is enormous and all of it earth Difficult due to the limited number of ground stations and available bandwidth. But how can satellites choose the best, most relevant images to send home?
Amazon and its partners, Italian space startup D-Orbit and computing technology developer Unibap, have partnered to demonstrate a solution to this problem. This is artificial intelligence software that runs directly on the orbiting satellite and can independently decide which photos to send. Earth.
Related: Microsoft partners with SpaceX to launch Azure Space, bringing cloud computing to the final frontier
“Using Amazon Web Services (AWS) software to perform real-time data analysis aboard orbiting satellites and delivering that analysis directly to decision makers via the cloud is an existing approach to space data management. ,” said Max Peterson, vice president of AWS statement (opens in new tab)“It also helps us push the boundaries of what we believe is possible with satellite operations.By providing powerful and secure cloud capabilities in space, satellite operators will be able to You will be able to communicate more efficiently and deliver updated commands using familiar AWS tools.”
The experiment was performed on D-Orbit’s ION satellite, which launched in January 2022. During testing, machine learning payloads built with Unibap processed “massive amounts of space data” aboard the satellite, AWS said in a statement. (Machine learning refers to software algorithms that can learn from patterns in previous data to make decisions without following explicit instructions.) The system uses AWS machine learning to analyze acquired satellite imagery in real time. Leverage learning models and AWS IoT Greengrass cloud management. An analytics system that can run even during periods of limited connectivity.
“We enable our customers to transform raw satellite data into immediate actionable information, deliver alerts in seconds, enable on-board federated learning for autonomous information acquisition, and enable downlinked data “By giving users real-time access to AWS edge services and capabilities in orbit,” said the co-founder of Digital Transformation and Unibap in a statement. , users will be able to gain more timely insights and optimize the use of satellite and ground resources.”
During experiments, the machine learning software successfully identified objects such as atmospheric clouds and wildfire smoke billows, buildings on the ground, and ships at sea. The software has also successfully reduced the size of images sent to Earth by up to 42%, increasing the speed and efficiency of the delivery process, AWS said in a statement.
The satellite is still in orbit and continues experiments.
Follow Tereza Pultalova on Twitter @TerezaPultarova (opens in new tab). follow us on Twitter @space.com (opens in new tab) and Facebook (opens in new tab).
