Data Fusion for Space Applications Problem Statement

How might we take advantage of the latest in information analysis and data fusion to provide insight into space objects’ past, current, and predicted trajectories, capabilities, and intentions?


The world is entering into a second space age, this time driven by economic opportunity as much as national or military interest. Concepts that seemed far-fetched a decade ago, like global internet from space, space tourism, and a permanent presence at the moon, now look like near-term realities. Each of these exciting opportunities, though, presents a special kind of challenge to those tasked with tracking objects in space. To achieve internet from space, constellations of thousands of satellites are needed – each constellation will be roughly the number of active satellites we are currently tracking – which will drive a need for increased capacity and automation of our space surveillance network. An increased human presence in space will drive an elevated sense of urgency around the need to track space objects well. Finally, the interest by many nations and commercial entities to go to the moon and beyond means that we need to be able to track more objects, further away, on more complicated trajectories than anything we have to consider today. Government, commercial, and academic sensor networks have emerged in a variety of phenomenologies to track space objects — radar, optical, infrared, and passive RF networks on the ground are complemented by a small but growing number of on-orbit sensors. There is enormous potential to harness this nascent big data to detect, track, identify, and characterize satellites.


The Catalyst Accelerator is seeking small businesses and startups with commercially viable data analytics to address the Air Force’s and Department of Defense’s needs in space situational awareness. We are seeking approaches to analyze or fuse existing data containing space object positions and signal levels (in the optical, radar, RF, IR, or other) as a function of time, along with any other freely available data or metadata (e.g., weather, astronomical data, published satellite data). The data analytics would provide relevant space object characteristics such as, but not limited to: assured object identification, object taxonomy, pattern of life behavior analysis, identification and prediction of collaborative behaviors, and detection of changes in activity, behavior, health, etc. Techniques should be prepared to address data sets that are sparse, irregular, and of disparate phenomenologies, representing objects in any earth-centric or cis-lunar orbit. Because of the disparate nature of potential SSA data sources, we are also seeking validation and trust techniques capable of identifying bad data, faulty sensors, mistagged objects, other outliers, and techniques to assess datasets and make recommendations for further collection.