The Space Force urgently needs to accelerate space-asset tasking in the wake of adversarial threats. The 2021 Russian ASAT test and Chinas SJ-21 repositioning of Compass G-2 demonstrate malicious co-orbital ASAT capabilities, currently unmatched by USA capabilities. Algorithms capable of reinforcement learning to enable rapid sensor tasking and re-tasking is essential to ensuring the USSF can continue to operate in a proven contested environment. Todays sensor tasking operations rely primarily on human-in-the-loop systems to assign tasks to spacecraft. A reinforcement learning algorithm integrated with a model-based simulation of on-orbit assets can accelerate planning for remote rendezvous, proximity, and sensing operations. Stevens Institute of Technology (first-place winner of the 2021 Hyperspace Challenge) has successfully demonstrated a reinforcement learning algorithm to efficiently allocate ground-based sensor tasks for commercial and DoD use cases. Turion Space Corporation (TSC) proposes a Phase I STTR in collaboration with Stevens Institute of Technology to adapt and extend this reinforcement learning approach with a predictive model that optimizes orbital sensor task allocation to reduce resource usage and extend asset lifetime.
