Dynamic Tasking of Networked Sensors Using Covariance Information

Abstract : A comprehensive high-fidelity simulation environment of networked optical space surveillance sensors has been created under an effort called TASMAN (Tasking Autonomous Sensors in a Multiple Application Network). One of the first studies utilizing this environment was focused on a novel resource management approach, namely covariance-based tasking. Under this scheme the state error covariance of resident space objects (RSO), sensor characteristics, and sensor-target geometry were used to determine the effectiveness of future observations in reducing the uncertainty in orbit estimates. The different observation effectiveness metrics evaluated in this study predicted the amount of error reduction in the position, velocity, or semi-major axis estimate for the RSO. These observation effectiveness metrics were used to schedule the most effective times for the sensors to observe the RSOs in different sensor tasking scenarios. The tasking scenarios included fully distributed sensor schedulers, a fully centralized network scheduler, and a baseline case that did not use observation effectiveness in order to roughly mirror the current Space Surveillance Network tasking. The different tasking and scheduling techniques were compared by evaluating their impact on the accuracy of the orbit estimates in the RSO catalog. The observation effectiveness metrics measuring reduction in position or velocity error all substantially reduced the errors in the catalog estimates when compared to the baseline tasking scenario and would enable a more efficient use of the sensor network for catalog maintenance.