Space-Based Sensor Tasking Using Deep Reinforcement Learning

Abstract To maintain a robust catalog of resident space objects (RSOs), situational awareness (SSA) mission operators depend on ground- and space-based sensors to repeatedly detect, characterize, track in orbit. Although some are capable monitoring large swaths the sky with wide fields view (FOVs), others—such as maneuverable optical telescopes, narrow-band imaging radars, or satellite laser-ranging systems—are restricted relatively narrow FOVs must slew at finite rate from object during observation. Since there many that FOV sensor could choose observe within its field regard (FOR), it schedule pointing direction duration using algorithm. This combinatorial optimization problem is known sensor-tasking problem. In this paper, we developed deep reinforcement learning agent task narrow-FOV low Earth orbit (LEO) proximal policy The sensor’s performance—both singular acting alone, but also complement network taskable, ground-based sensors—is compared greedy scheduler across several figures merit, including cumulative number RSOs observed mean trace covariance matrix all observable scenario. results simulations presented discussed. Additionally, an LEO SSA different orbits evaluated discussed, well various combinations sensors.