Linked Contract Record

EAGER: AI APPLIED TO MITIGATION OF SATELLITE INTERFERENCE WITH OPTICAL ASTRONOMY -THE 10-YEAR VERA C. RUBIN OBSERVATORY LEGACY SURVEY OF SPACE AND TIME (LSST) SURVEY BEGAN IN MID-2026. HOWEVER, A MASSIVE SURGE IN LOW EARTH ORBIT (LEO) SATELLITE LAUNCHES, DRIVEN BY OVER A MILLION PENDING LICENSE APPLICATIONS FOR NEW ULTRA-BRIGHT CONSTELLATIONS, SEVERELY THREATENS THE QUALITY OF RUBIN IMAGERY. SOME OF THE PROPOSED SATELLITES ARE SO BRIGHT, AND THEIR NUMBER SO VAST, THAT THEY WILL PRODUCE A DIFFUSE SKY GLOW WHEREIN MUCH OF THE INTENDED LSST SCIENCE MAY BECOME IMPOSSIBLE. FURTHERMORE, THESE SATELLITES GENERATE FLARES AND GLINTS THAT CAN EASILY BE CONFUSED WITH REAL ASTROPHYSICAL EVENTS, ULTIMATELY RESULTING IN BOGUS TRANSIENT ALERTS AND ARTIFICIAL FAINT GALAXIES. CONSEQUENTLY, THE NEED FOR MITIGATION IS CRITICAL, AND THE TIMESCALE FOR DEVELOPMENT IS SHORT. THIS PROJECT WILL USE ARTIFICIAL INTELLIGENCE (AI) TO DEVELOP TECHNIQUES OF ADDRESSING THESE CHALLENGES. THE PROJECT WILL LEVERAGE STATE-OF-THE-ART AI TO MITIGATE THE SYSTEMATIC ERRORS THAT DEGRADE IMAGE QUALITY AND THREATEN THE SCIENTIFIC IMPACT OF THE LSST. WORKING IN CLOSE COLLABORATION WITH ASTRONOMERS, SPACEX, AND OTHER CONSTELLATION OPERATORS, THE INITIATIVE UTILIZES GOOGLE DEEPMIND GEMINI 3.5 PRO AS ITS CORE AI ENGINE. TRAINED ON LAB IMAGES OF SATELLITE STREAKS ON A REALISTIC SKY, A NOVEL AI STREAK DETECTOR ALREADY PRODUCES HIGH ACCURACY AND COMPLETENESS. HOWEVER, THIS PROJECT WILL IMPROVE THE PERFORMANCE. FIRST, USING RUBIN LSSTCAM IMAGING DATA THE AI WILL GENERATE A COMPREHENSIVE DATABASE CATALOGING THE NUMBER, ORIENTATION, FLUX PROFILE, AND LOCATION OF STREAKS ON EACH IMAGE. SECOND, THE ENGINE WILL BE FURTHER TRAINED ON A SPECIALIZED HARDWARE SIMULATOR OF LSST SATELLITE STREAKS TO DRASTICALLY IMPROVE THE ACCURACY AND COMPLETENESS OF AUTOMATED DETECTION. FINALLY, THE PROJECT WILL DEVELOP ADVANCED CHARACTERIZATIONS OF THESE STREAKS TO ASSIST RESEARCHERS WITH SYSTEMATIC ERROR ANALYSIS AND DATA MASKING, AND SUPPORT SATELLITE OPERATORS IN THEIR REFLECTIVITY MITIGATION EXPERIMENTS. ULTIMATELY, THIS WORK WILL DELIVER CALIBRATED MONITORING FOR A SATELLITE POPULATION APPROACHING ONE MILLION, PROVIDING THE GLOBAL ASTRONOMY COMMUNITY WITH AN INCREASED UNDERSTANDING OF SYSTEMATIC ERRORS AND VASTLY IMPROVED DATA RELIABILITY. THIS AWARD REFLECTS NSF'S STATUTORY MISSION AND HAS BEEN DEEMED WORTHY OF SUPPORT THROUGH EVALUATION USING THE FOUNDATION'S INTELLECTUAL MERIT AND BROADER IMPACTS REVIEW CRITERIA.- SUBAWARDS ARE NOT PLANNED FOR THIS AWARD.

EAGER: AI APPLIED TO MITIGATION OF SATELLITE INTERFERENCE WITH OPTICAL ASTRONOMY -THE 10-YEAR VERA C. RUBIN OBSERVATORY LEGACY SURVEY OF SPACE AND TIME (LSST) SURVEY BEGAN IN MID-2026. HOWEVER, A MASSIVE SURGE IN LOW EARTH ORBIT (LEO) SATELLITE LAUNCHES, DRIVEN BY OVER A MILLION PENDING LICENSE APPLICATIONS FOR NEW ULTRA-BRIGHT CONSTELLATIONS, SEVERELY THREATENS THE QUALITY OF RUBIN IMAGERY. SOME OF THE PROPOSED SATELLITES ARE SO BRIGHT, AND THEIR NUMBER SO VAST, THAT THEY WILL PRODUCE A DIFFUSE SKY GLOW WHEREIN MUCH OF THE INTENDED LSST SCIENCE MAY BECOME IMPOSSIBLE. FURTHERMORE, THESE SATELLITES GENERATE FLARES AND GLINTS THAT CAN EASILY BE CONFUSED WITH REAL ASTROPHYSICAL EVENTS, ULTIMATELY RESULTING IN BOGUS TRANSIENT ALERTS AND ARTIFICIAL FAINT GALAXIES. CONSEQUENTLY, THE NEED FOR MITIGATION IS CRITICAL, AND THE TIMESCALE FOR DEVELOPMENT IS SHORT. THIS PROJECT WILL USE ARTIFICIAL INTELLIGENCE (AI) TO DEVELOP TECHNIQUES OF ADDRESSING THESE CHALLENGES. THE PROJECT WILL LEVERAGE STATE-OF-THE-ART AI TO MITIGATE THE SYSTEMATIC ERRORS THAT DEGRADE IMAGE QUALITY AND THREATEN THE SCIENTIFIC IMPACT OF THE LSST. WORKING IN CLOSE COLLABORATION WITH ASTRONOMERS, SPACEX, AND OTHER CONSTELLATION OPERATORS, THE INITIATIVE UTILIZES GOOGLE DEEPMIND GEMINI 3.5 PRO AS ITS CORE AI ENGINE. TRAINED ON LAB IMAGES OF SATELLITE STREAKS ON A REALISTIC SKY, A NOVEL AI STREAK DETECTOR ALREADY PRODUCES HIGH ACCURACY AND COMPLETENESS. HOWEVER, THIS PROJECT WILL IMPROVE THE PERFORMANCE. FIRST, USING RUBIN LSSTCAM IMAGING DATA THE AI WILL GENERATE A COMPREHENSIVE DATABASE CATALOGING THE NUMBER, ORIENTATION, FLUX PROFILE, AND LOCATION OF STREAKS ON EACH IMAGE. SECOND, THE ENGINE WILL BE FURTHER TRAINED ON A SPECIALIZED HARDWARE SIMULATOR OF LSST SATELLITE STREAKS TO DRASTICALLY IMPROVE THE ACCURACY AND COMPLETENESS OF AUTOMATED DETECTION. FINALLY, THE PROJECT WILL DEVELOP ADVANCED CHARACTERIZATIONS OF THESE STREAKS TO ASSIST RESEARCHERS WITH SYSTEMATIC ERROR ANALYSIS AND DATA MASKING, AND SUPPORT SATELLITE OPERATORS IN THEIR REFLECTIVITY MITIGATION EXPERIMENTS. ULTIMATELY, THIS WORK WILL DELIVER CALIBRATED MONITORING FOR A SATELLITE POPULATION APPROACHING ONE MILLION, PROVIDING THE GLOBAL ASTRONOMY COMMUNITY WITH AN INCREASED UNDERSTANDING OF SYSTEMATIC ERRORS AND VASTLY IMPROVED DATA RELIABILITY. THIS AWARD REFLECTS NSF'S STATUTORY MISSION AND HAS BEEN DEEMED WORTHY OF SUPPORT THROUGH EVALUATION USING THE FOUNDATION'S INTELLECTUAL MERIT AND BROADER IMPACTS REVIEW CRITERIA.- SUBAWARDS ARE NOT PLANNED FOR THIS AWARD.