Quickly assimilate tracking data of new orbiting space objects while simultaneously shrinking the error ellipsoids in order to create more room to operate in space. To do this we need to know where all satellites and debris are and where they will be for the next seven days.

Current capabilities are inadequate to monitor the space objects population to avoid collisions. E.g., current practice did not prevent the Iridium-Cosmos collision that increased space debris by 10%. (False alarms are also bad and avoidance maneuvers are both costly and risky.) The biggest obstacles to achieving effective collision avoidance are inaccurate prediction of Sun-Earth system dynamics and atmospheric drag, and current manually intensive techniques for assimilating tracking data.


  • Probabilistic SSA/cataloging using Bayesian learning, for monitoring of a much bigger population and faster handling of rapid changes in population, i.e., better inference and statistics
  • Thermospheric density forecast modeling improvements to eliminate major source of density uncertainty from the lower atmosphere, using NRLMSISE-00, WACCM-X (0-500km), NRL-HDAS (high-alt data assimilation; 0-100km), MERRA (0-50km): new physics, better modeling, HPC performance
  • Key solar extreme ultraviolet (EUV) irradiance forecasts to eliminate major source of density uncertainty from above, using NRLSSI, NRLEUV, GONG and HYPERION, for new data and—for the first time—incorporation of physics beyond empirical modeling


  • Validation against observations and comparison with current capabilities
  • Proof-of-Principle capability ready for prototyping
  • Enhancements to the NRLMSIS-E empirical thermosphere model which has significant commercial, military and academic demand.
  • Anticipated payoff: 85% reduction in 7-day position uncertainty volumes and capability to handle 10x increase in space catalog