A Rolling Horizon Quantum Optimization for Attitude-Aware Satellite Task Scheduling

Abstract

Earth Observation Satellite (EOS) scheduling is a highly constrained, NP-hard optimization challenge requiring the maximization of task priority while strictly adhering to kinematic (slew time) and resource (battery) limitations. Hybrid quantum-classical approaches have shown promising results to solve these problems, but standard global Quadratic Unconstrained Binary Optimization (QUBO) formulations scale very poorly with increasing problem size as allocating variables across continuous time slots generates an exponential search space which quickly surpasses the exact verification horizon (~ 50 variables). Hence, it remains a critical challenge to establish quantum advantage or even basic computational feasibility on Noisy Intermediate Scale Quantum (NISQ) hardware.

To resolve this scalability bottleneck, this thesis proposes a novel hybrid quantum-classical framework integrating Quantum Approximate Optimization Algorithm (QAOA) with Rolling Horizon Optimization (RHO). This approach decouples the entire mission timeline into discrete orbital passes, partitioning the global scheduling problem into tractable, quantum sub-problems. To bridge temporal constraints across the orbital passes, a classical physics architecture is employed which tracks and dynamically updates the continuous variables subsequently across quantum evaluations, such as the available battery level (Sk) and solar recharge rates. Moreover, to reduce the QUBO matrix size, in turn the whole system size, a targeted strategy is employed which assigns qubits strictly to kinematically valid temporal task slots.

The simulation results are benchmarked using real-world orbital data from the ISRO’s Cartosat-3 satellite, demonstrating that the RHO-QAOA framework successfully bounds the peak hardware complexity, while a monolithic global formulation exhibits an exponential memory growth which quickly crashes the standard statevector simulators. The RHO architecture maintains a flat memory footprint determined by the densest single orbit, peaking just at 19 qubits for a densely packed 48-tasks mission. Moreover, benchmarking against a Priority-Driven Greedy heuristic reveals that the quantum rolling horizon approach yields a 120.4% increase in total mission priority, successfully avoiding the local-minima traps of short-sighted classical schedulers to achieve denser schedule packing. Ultimately, this research provides a highly scalable, hardware-efficient blueprint for deploying quantum optimization on continuous, time-dependent aerospace operations.