en.Wedoany.com Reported - Loft Orbital announced on June 23 that it has reached an agreement with NASA's Jet Propulsion Laboratory (NASA JPL) to test artificial intelligence software on spacecraft to improve the automation of Earth science monitoring. The test, part of NASA's funded project "Federated Autonomous Measurement" (FAME), began this month on one of Loft Orbital's spacecraft, with its onboard computer running an AI model developed by JPL. Additional tests are planned for 2027 and 2028 using future Loft Orbital spacecraft.
The core goal of the project is to automate the "tip-and-cue" process. This process typically uses image recognition features from one spacecraft, followed by subsequent observations from another spacecraft at higher detail or using other observation techniques. Currently, this process often requires downlinking images to the ground for analysis. Paul Lasserre, General Manager of Loft Orbital's AI business, stated that by processing at the edge, spacecraft can capture, perceive, understand, and send insights about real-time conditions on Earth without downlinking large amounts of data. The AI model, trained on a "very large corpus," can identify various objects without needing to specify targets in advance.
The project faces two technical challenges: integrating infrastructure on satellites, including sensors and processors capable of receiving and processing images in real time, and determining whether open-source AI models can operate within satellite hardware constraints, requiring the model to be sufficiently small. Lasserre noted that small, high-performance, multimodal reasoning models have only recently emerged—something that did not exist six months ago—while space infrastructure has also gained the necessary capabilities. The convergence of these elements has made the project possible. He also mentioned that institutions like NASA, which have held this vision for some time and recognize that now is the moment for these different elements to come together, have provided support for the project. Lasserre, who has a background in AI for ground applications, said he had to reassess AI's capabilities in space given the computational limitations on spacecraft, but he believes these limitations are not true constraints, and with adequate preparation, the most advanced models can be run.
Lasserre described an operational vision: a spacecraft equipped with always-on sensors continuously collects data and uses AI models to search for features of interest, marking them and transmitting detailed information to other spacecraft via inter-satellite links for rapid response. This "patrol mode" was previously difficult to achieve due to processing and connectivity bottlenecks. Application examples include quickly detecting and tracking wildfires or marine pollution events, as well as rapid information acquisition in security, military, and intelligence domains. Loft Orbital plans to expand its capabilities through a series of 10 satellites called Altair, equipped with multiple sensors, edge computing devices for AI model operations, and inter-satellite links. Lasserre believes that if one has to wait hours for the next satellite, the entire value chain of real-time insights and autonomy becomes meaningless, and this turning point suddenly makes the value significantly higher both commercially and for governments.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
