STELLAR SCIENCE LTD. CO. — Department of Defense SBIR Phase I: AF151-083

Description

ABSTRACT:Satellite operators heavily rely on the Joint Space Operations Center (JSpOC) to maintain an updated catalog of all tracked space objects and to regularly perform conjunction analyses to predict potential on-orbit collisions. In the near future, the number of tracked objects is projected to grow by an order of magnitude. The corresponding increase in collision warnings produced by the existing conjunction analysis process will easily overwhelm the current system. The three keys to a collision prediction process that can successfully handle the projected growth in data are: accurate data representation, scalable screening techniques, and automated decision processes. Stellar Science proposes a new conjunction analysis system based on the Adaptive Entropy-based Gaussian-mixture Information Synthesis (AEGIS) method, a mature technology that has been demonstrated to accurately propagate and update orbit estimates and uncertainties and compute collision probabilities. We will combine the AEGIS method with innovative techniques based on spatiotemporal data indexing to screen for potential conjunctions, and build an automated decision framework based on machine learning algorithms to provide JSpOC with advanced capabilities to handle its future SSA missions.BENEFIT:Timely and accurate conjunction filtering is a critical need of the Joint Space Operations Center (JSpOC). The collision warnings produced by the JSpOC are of critical importance in protecting U.S. and allied spacecraft against destructive collisions, and in protecting the lives of astronauts during space flight. By improving the accuracy of how orbit uncertainty is represented, propagated, and updated, the proposed collision prediction system will produce fewer false collision warnings, without missing high-risk conjunction events. Combined with conjunction filter techniques based on spatiotemporal data structures, the proposed system will accurately and efficiently process a catalog of 100,000 objects and screen for high probability conjunction events. Finally, automated screening using machine learning techniques will further eliminate the low-risk events, and reduce the number of conjunctions requiring manual review by operators and consideration of risk-mitigating actions.