Satellite density in Earth orbit has increased exponentially lately. The lower cost of small satellites allows governments, researchers and personal corporations to launch and operate around 2,877 satellites in 2023 alone. This includes the increasing activity of satellites in geostationary Earth orbit (GEO), bringing with it technologies with global impact, from broadband web to climate monitoring. However, with the multiple advantages of those satellite-based technologies come increased safety risks and environmental concerns. To prevent collisions and other disasters, more accurate and efficient methods of monitoring and modeling satellite behavior are urgently needed.
To address this challenge, the MIT Astrodynamics, Space Robotic, and Controls Laboratory (ARCLab) launched the MIT ARCLab Prize for AI Innovation in Space: a first-of-its-kind competition that requires participants to make use of AI to characterize satellite patterns of life (PoLs) – the long-term behavioral history of a satellite in orbit – using purely passively collected information. After the decision for entries last fall, 126 teams used machine learning to construct algorithms that might label and timestamp the behavioral modes of GEO satellites over a six-month period, competing for accuracy and efficiency.
With support from the U.S. Department of the Air Force and the MIT AI Accelerator, the competition has a complete prize fund of $25,000. A team of judges from ARCLab and MIT Lincoln Laboratory evaluated the submissions based on clarity, novelty, technical depth, and reproducibility, awarding each entry a rating out of 100 points. The judges have now announced the winners and runners-up:
First prize: David Baldsiefen — Team Hawaii2024
With a winning rating of 96, Baldsiefen will receive $10,000 and shall be invited to offer a poster session with the ARCLab team on the Advanced Maui Optical and Space Surveillance Technologies (AMOS) Conference in Hawaii in the autumn. One reviewer noted: “Clear and concise report with excellent ideas equivalent to the label encoding of the localizer. Decisions in regards to the architectures and have development are well-founded. The provided code can be well documented and structured, allowing for simple reproducibility of the experiments.”
Second prize: Binh Tran, Christopher Yeung, Kurtis Johnson, Nathan Metzger — Team Millennial-IUP
With a rating of 94.2, Y, Millennial-IUP will receive $5,000 and will even join the ARCLab team on the AMOS conference. One reviewer said, “The models they chose made sense and were justified, they made impressive efforts to enhance efficiency… They used physical principles to support their models and this appeared to be reproducible. Overall, it was an easy-to-understand, concise report without a number of technical jargon.”
Third prize: Isaac Haik and Francois Porcher — Team QR_Is
With a rating of 94, Haik and Porcher share the third prize of $3,000 and are also invited to the AMOS conference together with the ARCLab team. One reviewer noted, “This informative and interesting report describes the mix of ML and signal processing techniques in a compelling way, supported by informative charts, tables, and sequence diagrams. The writer identifies and describes a modular approach to class detection and his evaluation of feature usefulness, which, as he appropriately notes, just isn’t equally useful for all classes… Any lack of operational experience is compensated by a transparent and detailed discussion of the advantages and pitfalls of the methods they used and a discussion of their findings.”
Teams in fourth through seventh place will each receive $1,000 and a certificate of excellence.
“The goal of this competition was to encourage an interdisciplinary approach to problem-solving within the space domain by inviting AI development experts to use their skills on this recent context of orbital capability. And all of our winning teams really delivered – bringing technical prowess, novel approaches and expertise to a really impressive round of submissions,” says Professor Richard Linares, Head of ARCLab.
Active modeling with passive data
Throughout the time a GEO satellite spends in orbit, operators issue commands to place it into different behavioral modes – position keeping, longitudinal translation, end-of-life behavior, etc. Satellite Patterns of Life (PoLs) describe in-orbit behavior, which consists of sequences of natural and non-natural behavioral modes.
ARCLab has developed a groundbreaking benchmarking tool for characterizing life patterns of geosynchronous satellites and Satellite-based dataset for identifying life patterns (SPLID), consisting of real and artificial space object data. Participants within the challenge used this tool to create algorithms that use AI to model the behavior of a satellite in orbit.
The goal of the MIT ARCLab Prize for AI Innovation in Space is to encourage engineers and enthusiasts to bring innovation and recent skills to established aerospace challenges. The team intends to carry the competition in 2025 and 2026 to explore other topics and invite AI experts to use their skills to recent challenges.
