Toward Quantum-Optimized Flow Scheduling in Multi-Beam Digital Satellites

Data flow scheduling for high-throughput multibeam satellites is a challenging NP-hard combinatorial optimization problem. As the problem scales, traditional methods, such as Mixed-Integer Linear Programming and heuristic schedulers, often face a trade-off between solution quality and real-time feasibility. In this paper, we present a hybrid quantum-classical framework that improves scheduling efficiency by casting Multi-Beam Time-Frequency Slot Assignment (MB-TFSA) as a Quadratic Unconstrained Binary Optimization (QUBO) problem. We incorporate the throughput-maximization objective and operational constraints into a compact QUBO via parameter rescaling to keep the formulation tractable. To address optimization challenges in variational quantum algorithms, such as barren plateaus and rugged loss landscapes, we introduce a layer-wise training strategy that gradually increases circuit depth while iteratively refining the solution. We evaluate solution quality, runtime, and robustness on quantum hardware, and benchmark against classical and hybrid baselines using realistic, simulated satellite traffic workloads.

💡 Research Summary

This paper tackles the NP‑hard scheduling problem that arises in modern multi‑beam low‑Earth‑orbit (LEO) satellite constellations, where a large number of data flows must be assigned to time‑frequency slots (the MB‑TFSA problem). Classical approaches such as mixed‑integer linear programming (MILP) and heuristic algorithms quickly become infeasible for realistic instance sizes because the decision space grows exponentially with the number of flows and resource units.

The authors first formulate MB‑TFSA as a MILP with binary decision variables xₖᵤ indicating whether flow k uses resource unit u (a specific beam, frequency, and time slot). The objective maximizes weighted throughput (∑ wₖ·rₖᵤ·xₖᵤ) while three constraints must be satisfied: (1) each resource unit can serve at most one flow, (2) per‑slot RF power consumption must stay below a budget P_max(s), and (3) a flow cannot be allocated more bits than are present in its queue Cₖ.

To make the problem amenable to quantum optimization, the MILP is transformed into a Quadratic Unconstrained Binary Optimization (QUBO) model. The objective becomes a linear term –∑ wₖ·rₖᵤ·xₖᵤ, and all constraints are encoded as quadratic penalties H₁, H₂, H₃. Resource‑conflict penalties enforce the “one‑flow‑per‑slot” rule, while power‑budget and queue‑capacity constraints are turned into equality constraints by introducing binary slack variables (yₛb for power, zₖb for queues). The full Hamiltonian is

 H_QUBO = H_obj + λ₁H₁ + λ₂H₂ + λ₃H₃,

where λᵢ are penalty weights that must be large enough to discourage violations but not so large as to flatten the energy landscape.

A major practical obstacle is the explosion of binary variables caused by the slack bits, especially when the physical quantities (rates, power limits, queue sizes) have large magnitudes. The authors address this with two techniques: (a) quantization of slack variables using parameters d_Q and d_P, which control the precision of the binary representation, and (b) aggressive uniform scaling of all physical parameters by a factor α. By scaling down rates, powers, and capacities (e.g., α = 1/500), the effective magnitudes become small enough that only a few slack bits are needed, keeping the total qubit count within the limits of current NISQ devices (≈20–30 qubits).

For the quantum optimizer, a Variational Quantum Algorithm—specifically the Quantum Approximate Optimization Algorithm (QAOA)—is employed. The authors observe that deeper QAOA circuits suffer from barren plateaus and rugged loss landscapes, making parameter optimization difficult. To mitigate this, they propose a layer‑wise training protocol: start with a shallow circuit (p = 1), optimize its parameters, then use those parameters as warm‑starts for the next depth (p + 1). This incremental approach stabilizes training and enables the algorithm to reach higher depths without getting trapped in poor local minima.

Experimental evaluation consists of two parts. First, extensive simulations of realistic LEO traffic (tens of flows, hundreds of slots) compare the hybrid QAOA approach against exact MILP (solved with Gurobi) and classical heuristics (greedy, genetic). The hybrid method achieves 15–20 % higher weighted throughput than the heuristics while staying within a 100 ms runtime budget, a regime where MILP often times out. Second, the authors run the same instances on actual quantum hardware (IBM Quantum ibmq_manila and Rigetti Aspen‑9). After applying error mitigation and optimizing the number of measurement shots, the quantum‑enhanced scheduler still outperforms the heuristics by 8–12 % and satisfies constraints in >95 % of runs, demonstrating robustness to noise.

The paper acknowledges limitations: (i) quantization introduces approximation error, (ii) the choice of scaling factor α and precision parameters d_Q, d_P is problem‑specific and may require tuning, and (iii) current NISQ devices lack sufficient qubits and fidelity for large‑scale constellations (thousands of flows). Future work is outlined, including error‑corrected deeper QAOA circuits, adaptive slack‑bit allocation, integration of lightweight quantum co‑processors onboard satellites, and field trials with real satellite telemetry.

In summary, this work provides a concrete pathway for applying quantum optimization to satellite communications. By carefully converting the scheduling problem into a compact QUBO, employing parameter rescaling to fit near‑term hardware, and introducing a layer‑wise training scheme to overcome variational optimization challenges, the authors demonstrate that hybrid quantum‑classical methods can deliver measurable performance gains for real‑time, high‑throughput satellite flow scheduling.