A. Khan
BosonQ Psi Corp (BQP), New York, United States
Keywords: SDA, QAPINN, PINN, Propagation, AI, Edge, Autonomy, Orekit, Maneuver, Surrogate, Conjunction, Uncertainty
Real-time autonomous satellite operations, including conjunction assessment, collision avoidance, and rendezvous and proximity operations, are increasingly constrained by the limited computational resources available on flight hardware. Existing orbital propagation solutions typically trade accuracy for speed, as analytical propagators such as SGP4 offer rapid execution but reduced fidelity, while numerical tools such as Orekit provide high accuracy at significantly higher computational cost. As satellite autonomy expands from low Earth orbit (LEO) to cislunar regimes, resource-efficient propagation and maneuver-planning capabilities become essential. This work investigates Physics-Informed Neural Networks (PINNs) and Quantum-Assisted PINNs (QA-PINNs) as surrogate models for high-fidelity orbital propagation on embedded systems. The proposed PINN incorporates orbital dynamics directly into training through physics-informed losses and hard constraints, ensuring physically consistent predictions even with sparse or noisy data. The QA-PINN extends this framework by replacing the classical hidden layer with a variational quantum circuit, reducing parameter counts while maintaining predictive performance. Performance is benchmarked against SGP4 and Orekit on a Jetson Nano platform representative of CubeSat-class hardware. The results provide a systems-level assessment of scalable, low-SWaP autonomy technologies that enable resilient, real-time onboard orbit prediction and maneuver decision support for future national security and commercial space missions.