AI Sweden launches models into space
AI Sweden and our partners have taken a groundbreaking step in the realm of AI for space: For the first time, models developed in the Swedish Space Data Lab are now in orbit.
“AI Sweden's networking capabilities enabled us to establish a valuable connection with Aleksis Pirinen, researcher at RISE. This collaboration allowed us to adapt Aleksis’ model for cloud optical thickness-based cloud masks, developed under the Swedish Data Lab initiative, into an application for our SpaceCloud platform,” says Unibap’s Carl Bondesson.
The SpaceEdge project, a collaboration between Unibap and AI Sweden, is the world’s first open testbed for space app development. The platform allows developers to upload their apps to Unibap’s SpaceCloud in orbit.
“The work we conducted in AI Sweden’s Edge Lab made it possible to adapt previously developed machine learning models and perform proof-of-concept work before launching them into orbit,” says Pirinen.
Photo credits: SpaceX
Edge Lab, SpaceEdge, and Swedish Space Data Lab
- Edge Lab is a physical infrastructure at AI Sweden.
- SpaceEdge is a project where AI Sweden hosts 4 ix-10 computers from Unibap in the Edge Lab, to run preliminary experiments for apps before sending them in orbits.
- Swedish Space Data Lab is a hub for earth observation data users and a community-building initiative.
Unibap is spearheading efforts to provide more private and public organizations access to space data through the development of an innovative "app platform" for satellites. This platform will streamline the deployment and utilization of AI applications in space, fostering a new era of accessible and efficient space data management.
Deploying a model from our lab into space for the first time is a significant milestone. Space presents a harsh and resource-constrained environment with limited electricity, computing power, and bandwidth. Successfully running models in such conditions opens the doors to numerous other space-based applications that more partners can benefit from.
Chiara Ceccobello
Project manager, AI Sweden
Earth observation data from satellites is transforming decision-making across multiple fields. The applications are vast and vital: improving the accuracy of weather forecasting and advancing climate change research, enhancing disaster response for forest fires and floods, and optimizing crop management in agriculture, to name a few. This data is becoming indispensable for tackling some of the world's most pressing challenges.
The rise of mini and cube-satellites is opening new frontiers in space exploration. These compact satellites serve as ideal platforms for AI-based applications in decentralized and distributed computing environments. By processing data onboard the satellites, AI models can significantly reduce the need for data transmission to Earth, enabling rapid, real-time responses to dynamic situations and marking a transformative shift in space data utilization.
“AI Sweden’s expertise in decentralized learning and AI will focus on laying the foundation for real-time decision-making, building robust solutions for many different fields,” says Chiara Ceccobello.
AI Sweden’s Work on AI in Space
There are two main projects that are space-based as of now: SpaceEdge and Swedish Space Data Lab.
SpaceEdge is an SNSA-funded project where AI Sweden is collaborating with Unibap to explore the capability of their machines to deploy, train, and do inference with AI models in orbit. The goal is to achieve smooth operations on Unibaps’ hardware for third-party applications. We are interested in testing AI models and how a decentralized learning setup can help reduce the communication bottleneck and enable real-time actions/alerts.
Swedish Space Data Lab is a project that has been going on since 2019, with recurrent funding from Vinnova and SNSA and a consortium that has been evolving along with funding cycles and has now reached a much larger size (10+ organizations). AI Sweden has been in it since the beginning and up until last year were the project managers of it, and today RISE has the project management role.
The aim is to connect the ecosystem of earth observation (EO) data users and lower the bar for the utilization of these data. Every cycle there has been a number of pilot studies to showcase how to use AI to analyze EO and, in parallel, we contribute to the development of the platform Digital Earth Sweden.
For more information, contact
Select reading:
The latest advancements in decentralized learning
Introducing AI Labs special lectures
Follow us
AI Sweden Newsletter
Sign up for the latest in Swedish AI
AI Sweden is formally hosted by Lindholmen Science Park AB.