WorksPhysics and Astronomy
EntangledEdge: A Distributed System for Orbital Quantum Sensing and Learning
Abstract
An orbital quantum sensor is more precise than classical and holds its information in a quantum state's superposition and entanglement, which classical measurement collapses. Learning on quantum-sensed data therefore requires a distributed system in which the data stay quantum from sensor to inference. We present EntangledEdge, the first system architecture to unify orbital quantum sensing and learning end-to-end, with a quantum data plane and a classical control plane. It routes each event to a channel-matched model in a capacity-diverse bank, adapts its detection threshold without labels, and aggregates across stations by hardware fidelity. Evaluated with a hybrid prototype (real quantum processors, simulated orbit-to-ground link) over satellite telemetry, EntangledEdge achieves high-quality inference at bounded per-event cost, improves quantum processor utilization, and scales with fleet size.
CCS Concepts: • Computer systems organization → Quantum computing.
Introduction
Quantum sensing in orbit. A quantum sensor uses a qubit probe to interact with a physical field and records the result in its quantum state. For a growing set of space-sensing tasks, the information of interest is held in the quantum structure of that state, in its superposition and entanglement, which a classical measurement collapses before it can be used. For example, entanglement-enhanced gradiometry and networked clock comparison are two such tasks [92, 94]. A learner with coherent access to the states needs exponentially fewer samples than any classical learner working from measured values [34, 35], so these types of tasks can benefit from an end-to-end quantum pipeline, from the orbital probe through to ground processing.
Quantum sensors also exceed classical sensitivity. A qubit probe reaches sensitivities orders of magnitude beyond fluxgate, mechanical, and oscillator-based instruments. Atomic magnetometers, for instance, measure at the femtotesla scale, while fluxgate instruments operate near the picotesla scale.
An atomic-transition reference also removes the periodic recalibration classical space instruments require [20, 37]. This is why quantum sensors can be used for space tasks like geomagnetic field mapping, space-weather characterization, and gravity-anomaly tracking for ice-mass and hydrology [20].
Recent hardware makes an orbital quantum system feasible. Orbital quantum sensors have flown as first demonstrations, with classical readout [42, 90]. Neutral atom platforms are commercially available and are good candidates for deployed sensors. Because the sensed quantum state has to reach the ground intact, the orbit-to-ground link has to be quantum, which teleportation over a pre-shared entangled pair provides. Quantum networking has reached multi-node operation over fiber, and this transfer works between non-neighboring nodes [32, 65] and at Hertz rates across a city [76]. In orbit, the link has been demonstrated as a proof of concept. The Micius mission distributed entanglement to two ground sites 1200 km apart and demonstrated ground-to-satellite teleportation of a single-photon qubit [45, 69, 95], and a 23-kg microsatellite payload reproduced satellite-to-ground transmission with portable ground stations [43]. On the ground, deployed quantum processors with tens to hundreds of qubits can run the learning and inference [93]. EntangledEdge is a system architecture for this chain, whose stages have been demonstrated separately.
Learning on orbital quantum-sensed data as a distributed system. These parts form an end-to-end pipeline that learns and infers on the quantum-sensed states. Orbital edge computing established satellite constellations as distributed computers with a classical data plane [16, 21, 22, 44], and a quantum data plane raises challenges that classical remedies do not solve (Table 4 in Sec. 6 positions EntangledEdge against existing classes). Sensed states cannot be buffered or copied, since no-cloning forbids it and coherence times are far shorter than the gaps between ground-station passes [11], so a state that reaches a busy receiver is lost. The atmospheric channel distorts each downlinked state by a different amount per event, and the ground processors are physically distinct with noise that drifts between recalibrations. This variation shifts the score distribution over time, and labels for learning cannot be obtained without destroying the state, so the system has to track it without them.
| Class | Crosses the link | Stored or re-sent? | EntangledEdge adds |
| Classical FL [52] | Parameters | Yes | Non-copyable quantum data |
| Satellite edge, satellite FL [22, 50] | Sensed data; parameters | Yes, buffered on board | Unbufferable quantum data; heterogeneous QPUs |
| Quantum networks [32, 45]; DQC [56, 74] | Prepared states; entangled photons | Re-prepared, re-sent | Sensed input that cannot be re-sent; a learner; model bank |
| Distributed quantum sensing [49] | Measurement outcomes | Yes | State kept unmeasured across the link; heterogeneous QPUs |
| Hybrid QML [13, 31] | Nothing; one QPU | Re-prepared | Input over a noisy quantum link; model bank; many QPUs |
| EntangledEdge | Sensed quantum state; parameters | No | — |
Table 4. EntangledEdge against existing distributed systems.
This paper is accepted at the 2026 ACM SIGOPS Annual Technical Conference (SIGOPS ATC 2026), November 15–18, 2026, Hong Kong, China.
The EntangledEdge system design. EntangledEdge addresses these challenges in one end-to-end design with a quantum data plane and a classical control plane. An orbital sensor teleports each state to a ground station, which recovers and processes it, and the stations share only classical model updates (Figure 1). Because the channel changes per event, the system keeps several models of different capacities and routes each state to the one whose channel suits. It sets its own detection cutoff and updates it as the scores shift, so it needs no labels. Across ground stations, each weights the others by how reliably they recover, so drift and mismatch between processors do not skew the combined system.
Figure 1. An orbital quantum sensor teleports its data qubit to a ground station’s QPUs through a pre-shared Bell pair and a classical radio channel, where an ansatz performs learning and inference on the recovered state.
Overall, we make the following contributions:
Problem and system architecture. We formulate learning on orbital quantum-sensed data as a distributed-systems problem in which the data cannot be stored, copied, or re-sent, and present EntangledEdge, an end-to-end system architecture for orbital quantum sensing, learning, and inference.
Design to address quantum-specific challenges. To address the per-event channel and the unlabeled, drifting stream, EntangledEdge introduces a channel-conditioned multi-model bank, a label-free dynamic threshold, and a fidelity-weighted aggregation, which together add robustness to drifting processor/channel noise and station outages.
Hybrid-prototype evaluation. We build and evaluate EntangledEdge on superconducting processors (IBM Heron and Nighthawk), neutral atom sensors (QuEra Aquila), and a simulated orbit-to-ground link, over satellite-telemetry.
Quantum model for detection. We develop autoencoder methods to operate on sensed data and show they improve inference quality over a state-of-the-art quantum autoencoder (Quorum [48]) and federated alternatives in our setting.
Systems performance and scaling. EntangledEdge delivers its detection gains without hurting latency or throughput, improves processor utilization compared to alternatives, is robust across different operating conditions, and scales across sensors, ground stations, and model sizes.
EntangledEdge’s implementation is open-sourced at https://doi.org/10.5281/zenodo.22909579. We hope it helps others to reproduce our results and build on this work.
Background and Motivation
A satellite-based quantum sensing pipeline has two components. An orbital sensor node is a satellite payload carrying qubits that interact with a physical field. A ground QPU is an optical ground station that decodes the sensed state and processes it on a quantum processing unit (QPU). The two are connected by a quantum link and a classical radio link. Similar partitioning of work between satellite and ground stations appears in orbital edge computing for classical compute-based satellite payloads [21, 22, 50].
Qubits, gates, and quantum machine learning
A qubit is a two-level quantum system that can be in a superposition of its basis states and . Two or more qubits can occupy entangled states. The simplest example is , in which measuring either qubit fixes the outcome of the other. Operations on qubits are unitary gates, single-qubit rotations , and the two-qubit entangling controlled-NOT (CNOT), applied as a circuit that ends in measurement. A Quantum Machine Learning (QML) model is a parameterized circuit , an ansatz, whose gate angles form a trainable vector . Training proceeds in a hybrid quantum-classical loop. Input data, delivered directly to the processor, is loaded into a register. The ansatz acts on the register, and observables are measured to produce a scalar cost . A classical optimizer updates the parameters along the negative gradient , with each partial derivative estimated from two runs of the same circuit at shifted angles, the parameter-shift rule [75].
Quantum sensing
A quantum sensor uses a qubit as a probe, prepared in the superposition and exposed to the field for a free-evolution time . The field shifts the relative phase between the and components at a rate proportional to its strength, and the qubit ends in , where the accumulated phase records the integrated field [20]. Entangling probes before exposure improves the error scaling from the classical to the Heisenberg limit [26, 37]. A long coherence time lets the probe integrate signal, so sensitivity scales roughly as . Recent work has improved both for sensors of different modalities [20, 37, 46].
Orbital quantum sensors. Orbital quantum sensors open a class of Earth-observation and space sensing tasks that classical sensors cannot reach. Quantum probes operate near the shot-noise floor and reach magnetic, gravitational, and frequency sensitivities orders of magnitude beyond fluxgate, mechanical, or oscillator-based instruments. In the setup of Figure 1, each sensed qubit is teleported from orbit to a ground QPU. When sensor and ground QPU modalities differ, a transducer maps the sensed state onto a photonic flying qubit. Entanglement between optical and microwave photonic modes has been demonstrated on integrated hardware [53, 54].
Networking via quantum teleportation
Teleportation moves a quantum state across the link without sending the qubit itself. The two parties pre-share an entangled qubit pair (a Bell pair). To transfer a data qubit, the sender performs a Bell-basis measurement on it together with its half of the Bell pair; the entanglement projects the receiver’s qubit into a Pauli-rotated copy of the original. The sender transmits the two-bit outcome over an ordinary radio link, and the receiver applies the matching Pauli correction to recover the data qubit. The data qubit itself never crosses the optical channel, making teleportation the elementary primitive for quantum networking [86].
Teleportation has been demonstrated on quantum hardware over fiber and between ground and orbit, so the satellite-ground topology this paper relies on already exists as a proof of concept [32, 43, 45, 69, 95]. Atmospheric turbulence distorts the optical phase and spatial mode of each downlinked photon, and the receiving telescope sees a polarization frame [30]. At the qubit level, these effects show up as random single-qubit errors on each teleported qubit, and the errors grow with the zenith angle and the slant-path length.
Design choices. A design choice is where to place the QPU. Hosting the QPU on the satellite would be ideal but is not yet practical at small-satellite scale. A quantum processor with the gate depths and fidelities a learning workload needs depends on infrastructure that does not fit on board today. Sensor payloads sit at the more permissive end of this spectrum. We therefore place the QPU at the ground station for the rest of this paper. Placing the QPU in orbit instead would remove the satellite-to-ground link and its atmospheric noise, although a quantum link would still remain wherever a sensor and its processor are apart.
Given that the link is needed, the next question is whether it needs to carry quantum information at all. The alternative is to measure the sensed state at the satellite and ship classical outcomes down, which suffices for point estimation but collapses the superposition and entanglement a learner uses to tell states apart. Quantum learners with coherent access reach exponentially lower sample complexity than any classical learner [34, 35, 47], so the quantum link carries strictly more information. The workload for the rest of this section is anomaly detection on satellite telemetry, where streams are large, and labeled events are rare, which is exactly the unsupervised regime in which this advantage shows up.
Empirical evidence. Figure 2 confirms this on OPS-SAT [71] telemetry data as the anomaly fraction sweeps from 6% to 20%. All three pipelines share the same bottleneck and training budget. The fully quantum pipeline (sense → teleport → quantum autoencoder detection) holds the highest F1 at every operating point. The classical-transmission pipeline measures quantum state at the satellite into a 25-dimensional vector of single-qubit X/Y/Z and two-qubit ZZ observables, transmits it, and runs a classical autoencoder on the ground. The hybrid pipeline takes the same satellite-side measurement and transmission, then re-encodes the feature vector into a quantum state on the ground, and runs the same quantum autoencoder. Both alternatives lose because the satellite-side measurement collapses , destroying the superposition and entanglement structure that separates anomalous from normal states. An all-classical baseline (classical sensor on the satellite, classical processing on the ground) trails all three pipelines by more than 15% F1 at every operating point. That gap reflects the sensitivity headroom quantum sensors carry at these low signal levels (§ 2.2).
Figure 2. The fully quantum pipeline leads. The others measure the state at the satellite and lose the quantum structure. Shaded bands are across-seed standard deviations.
Coordination across ground stations
Coverage and ground tracks. A real deployment runs over many satellites and ground stations, and the orbital geometry constrains when each satellite-ground link is available. A station communicates with a satellite only while the satellite is above its local horizon. As the satellite orbits, the curve of points directly beneath it (its ground track) sweeps the surface, and a station sees the satellite only when the ground track passes through its coverage disk. Figure 3 (top) shows the shared ground track of two satellites in the same near-polar orbital plane, with four representative ground stations drawn at their 21-degree-elevation coverage disks. The track is global; each station’s coverage disk is a small patch.

Figure 3. Ground tracks of two satellites (NOAA-20, NOAA-21), ground-station coverage disks, and contact windows. Passes are short, sparse, and not simultaneous across stations.
Temporal fragmentation across stations. The bottom row of Figure 3 unrolls these contacts in time. Each row corresponds to one ground station over a 580-minute observation window, and the bars within a row mark the intervals during which each of the two satellites is above that station’s horizon. The pattern drives the design: contact windows at any station are short and sparsely scattered. A pass over New York rarely coincides with a pass over Tokyo, and at any one instant the satellite delivers data to at most one station.
A natural workaround would be to buffer the sensed states on the satellite and downlink them later, but this is not an option for quantum data. Qubit coherence times are on the order of milliseconds to seconds in the best laboratory quantum memories, orders of magnitude shorter than the gaps between contact windows. Sensing and transmission therefore happen in the same brief window, and a learning workload that depends on the full distribution of sensed states has to combine what multiple stations see.
Comparison of coordination architectures. Three natural coordination architectures fit this setting. Isolated trains a separate model at each ground station on that station’s own data, with no exchange between stations. Centralized routes the sensed data from every station to a single server and trains one model on the pooled union. Federated keeps each station’s data local, trains a model at each station on its own shard, and exchanges only classical model parameters between stations for aggregation. We use federated here as a label for the distributed architecture. The design choice is driven by communication cost and the noise budget of each architecture’s data path, not by data privacy or security.
Table 1 compares the three architectures on a smaller OPS-SAT subset anomaly-detection [71] with a single quantum autoencoder; § 5 evaluates the full dataset, averaged over 100 calibration days. Federated reaches the highest F1. Isolated trails because each station’s local shard is small and biased toward the orbital patches it sees, so the model overfits to a slice of the trajectory. Centralized trails for the opposite reason. Pooling routes each station’s quantum state onward over a second ground-to-central link, and the extra hop adds atmospheric noise to the training data on top of the satellite-to-ground noise the others pay. Federated avoids both failure modes: like isolated, it pays only the satellite-to-ground link, and like centralized each station’s model sees data from across all stations. The cross-station sharing happens by classical parameter updates. Isolated’s recall lands
| Architecture | F1 | Precision | Recall | Accuracy |
| Isolated | 0.51 | 0.39 | 0.76 | 0.70 |
| Centralized | 0.44 | 0.30 | 0.82 | 0.58 |
| Federated | 0.54 | 0.45 | 0.70 | 0.77 |
Table 1. Anomaly-detection performance of three coordination architectures on a smaller OPS-SAT subset; the complete dataset under full per-event noise is in Sec. 5.
a few points above federated’s because its overfitted shard flags more candidates as anomalies, but the same overfitting collapses its precision and leaves F1 (the right metric for sparse anomalies) and accuracy well below federated.
Summary of observations
Orbital quantum sensors reach precision and calibration stability classical instruments cannot. The sensed state has to remain quantum end to end, since measuring at the satellite collapses the state. Ground-station contacts are short, fragmented across stations, and unbuffered (decoherence time is orders of magnitude shorter than contact-window), so a federated architecture with a single quantum link per station and cross-station sharing of classical model parameters offers the most benefit.
Design of EntangledEdge
EntangledEdge operationalizes the observations of § 2 into a distributed system. We define an event as one teleported quantum state received at a ground station, together with the station’s telemetry at the time of reception. In our design, the data plane is quantum end-to-end, the control plane is classical, and channel variability, hardware drift, and event loss are first-order design constraints.
At one level of abstraction, EntangledEdge works as follows. Each ground station runs a small bank of capacity-diverse learning models, and picks one per event from the channel condition that delivered it. The selected model’s score becomes a flag through a streaming statistical fit at the station, so the threshold tracks the drift in the score distribution. Across stations, only classical parameters move, weighted by how reliable each station’s hardware is, so noisy backends do not drag the global model. Events arriving on a busy QPU are dropped, and offline stations do not contribute.
Design Challenges and Requirements
EntangledEdge’s goal is to learn on quantum-sensed data without any classical intermediate, from the orbital probe through to the ground inference. Training and deployment share the same channel, QPU mix, and drop policy, so the system operates within a single regime that subsumes both. This sharing is forced – a sensed state can be neither stored nor labeled, so no offline training set can exist, and learning has to happen online, on the stream the system serves. Five challenges shape this regime, none of which are solved by classical remedies, shaping EntangledEdge’s requirements.
C1. Teleported states cannot wait. Qubit coherence times are orders of magnitude below the gaps between contact windows (Sec. 2.4), and the no-cloning theorem [91] of quantum states forbids buffering. A sensed state that arrives while the receiving QPU is occupied is lost since there is no queue. Stations can also go offline due to weather or maintenance.
C2. The sensor itself is noisy. Platform-specific imperfections (Sec. 2.2) corrupt the state before it enters the link, and the noise structure differs across sensor modalities. The input the learner receives is already mixed at the source.
C3. The atmospheric channel varies per event. The Pauli error rates each downlinked qubit incurs depend on satellite position, slant geometry, and the station’s local turbulence profile. A single station encounters a different effective channel from one event to the next, and different stations observe different distributions over their respective passes. No single model capacity remains adequate across this range.
C4. Ground hardware is heterogeneous and noisy. Each station hosts a physically distinct QPU with its own gate errors, coherence times, and coupling map, and these characteristics drift between recalibration windows. A uniform treatment of every station’s gradient contribution can let the noisiest backends bias the global model.
C5. Labels are not reliably available. Labeling a sensed state at the satellite requires destructive measurement, and no-cloning forbids retaining a copy for downstream use.
The challenges interact. First, C1, C2, and C3 jointly rule out approaches that increase per-event compute to compensate for channel noise. Under the no-buffering constraint, additional compute per event increases the drop rate over the contact window. The capacity dimension compounds this. A model sized for mild channels overfits the broadening from heavy dephasing and depolarization, and a model robust to heavy noise underfits clean channels. Second, C2–C4 introduce continuous drift. Sensor, channel, and backend noise vary across events, which a fixed rule fails to track.
C1 and C5 have no classical analogs. For C2–C4, classical systems calibrate or denoise sensors, retransmit after channel errors, and reweight stored contributions from heterogeneous hardware. All of these need data that can be stored, copied, or re-sent, which a sensed state cannot, so noise should be handled on the state itself, within the per-event budget (Sec. 3.2). These challenges and their interactions impose six requirements on EntangledEdge. C1 forbids buffering, so the pipeline has to tolerate dropped events (R1). C5 forbids per-event labels, so the system has to operate without external supervision (R2). The cross-station heterogeneity and drift of C4 make uniform aggregation unsafe, so contributions need a per-station trust signal refreshed each round
(R3). C1, C2, and C3 jointly leave no compute headroom per event for noise compensation, so the system has to maintain robustness to input and channel noise within a fixed per-event budget (R4). Continuous drift across C2–C4 leaves any fixed reference stale, so the system has to track its operating distribution online (R5). Stations can also go offline through weather, maintenance, or recalibration (C1, C4), so the system has to withstand station downtime and backend recalibration without compromising the global model (R6).
The EntangledEdge Distributed System
The primary contribution of EntangledEdge is a distributed system that learns from sensed quantum data that cannot be stored, copied, or re-sent, and that arrives over an intermittent noisy link. It targets orbital quantum sensing workloads of the kind introduced in Sec. 2. These workloads share three properties. Streams are mostly nominal with a minority of anomalous events. Labels at the source require destructive measurement and are not available. The atmospheric channel and the ground hardware vary continuously across events and across ground stations.
EntangledEdge addresses these with two components. Each ground station runs a per-event pipeline. A federated coordination layer maintains a single global model across stations by sharing only classical model parameters between rounds. Two design decisions distinguish EntangledEdge from a basic federated quantum learning setup. At the station, a capacity-diverse model bank with channel-conditioned per-event routing handles a channel range that no single model capacity covers. Across stations, a quantum fidelity-weighted circular aggregation handles heterogeneous and drifting QPUs that a uniform federated average cannot.
The Per-Event Pipeline at a Ground Station
At each station, every event undergoes the same processing (Fig. 4). Station telemetry produces a channel-condition estimate. The estimate selects one of several models held at the station. The selected model produces a reconstruction score from the teleported quantum state. A streaming threshold converts the score into a binary flag.
Label-free reconstruction via a Quantum Autoencoder.
The workloads with quantum orbital sensing use cases fall in the category of unsupervised anomaly detection problems for two reasons. First, no labels are produced at the source. Labeling at the satellite requires destructive measurement (C5), which is incompatible with transmitting the quantum state for downstream processing. Simulated anomaly labels at training time diverge from what the orbital sensor produces in deployment, and a model trained on them is miscalibrated against the real stream. Second, the channel, the sensor, and the ground hardware drift continuously, so the system has to keep learning over its lifetime and keep segregating normals from anomalies as the score distribution shifts under that drift. EntangledEdge operates on the unlabeled stream the system observes (R2), and the training signal has to come from a quantity the model itself produces on each event.
Among unsupervised approaches, three standard scalar signals exist for anomaly detection. Reconstruction error from an autoencoder asks whether each input can be compressed through a bottleneck and recovered. Likelihood under a generative model asks whether each input is typical under a learned density. Distance from a learned centroid asks whether each input sits near a learned reference state. EntangledEdge uses reconstruction error. Of the three, it is the only one that imposes a structural compression constraint on the model. Each input has to be recoverable from a compact latent representation, which gives the model an inductive bias toward the structure of the normal distribution. Likelihood estimation requires an explicit density model that is expensive to learn on quantum states. A learned centroid requires a metric on states and a single anchor, which fits a normal distribution only at low spread and collapses under the broader normal distribution this workload produces. The per-event scalar from reconstruction at training is also the scalar the threshold consumes at deployment.
For quantum-sensed data, the autoencoder has to be a Quantum Autoencoder (QAE) [70]. A classical autoencoder would receive a classical feature vector from the satellite, which requires measuring the quantum state at the source and collapses the entanglement structure that separates normal from anomalous states (Figure 2). The QAE applies a parameterized encoder to the quantum state (Figure 4). The encoder partitions the result into latent qubits (compressed representation) and trash qubits (residual). When compression succeeds, the trash qubits carry no information about the input and approach a fixed reference state. Reconstruction quality is defined as the closeness of the trash qubits relative to that reference state.
Figure 4. Classical telemetry alone selects one capacity-tier QAE on the QPU, with SPSA gradient estimation on a classical node. The inset details the label-free SWAP-test reconstruction, which yields a scalar score without measuring the input state.
Reading out the trash qubits directly would destroy the quantum state and prevent downstream use of the latent representation. The QAE instead uses a SWAP test, a quantum primitive that returns the closeness between two quantum registers as a single scalar without disturbing the registers it compares. The test uses an auxiliary qubit and a controlled-SWAP between the trash register and the reference register. A single measurement on the auxiliary qubit produces an outcome whose probability grows as the trash register deviates from the reference. This rate is the per-event reconstruction score, low when the encoder has compressed the input into the latent register and high otherwise. No projective measurement of the input itself occurs, so the entanglement structure of the teleported state is preserved end to end.
Capacity-diverse multi-model bank with channel-conditioned routing.
R4 requires robustness across the per-event channel range within a fixed compute budget, and no single QAE capacity provides it. Figure 5 shows this for two QAE capacities at opposite ends of the model-size axis. score is plotted against the per-link channel noise parameter . The large capacity has a higher at low , where the model has headroom to learn the reconstruction structure, and degrades at high , where it overfits the broadening that heavy dephasing and depolarization induce. The small capacity is flat across the range and outperforms the large capacity at high . The pattern generalizes to any pair across the model-size axis, and our evaluation confirms this.

Figure 5. score against the per-link channel noise parameter for two QAE capacities at opposite ends of the model-size axis. The large capacity has higher at low . The small capacity has a higher at high . No single capacity covers the range. The pattern generalizes to other pairs on the axis.
EntangledEdge therefore maintains a capacity-diverse bank of QAEs at each station. The router selects the bank member from a channel-condition estimate,
The equation combines two factors. The first, , is the per-link noise scaling (the sweep in Figure 5). The second, , is the atmospheric scintillation factor. here is the Rytov variance, a standard measure of optical turbulence along the line of sight. It is computed from the station-satellite zenith angle and the station’s atmospheric turbulence strength through the Hufnagel–Valley refractive-index profile (discussed in § 4). Clean links give . All three inputs are obtained without invoking the quantum state, so the router commits to a model before any quantum operation runs. The range is partitioned into zones bijectively mapped to the bank members. Low routes to a higher-capacity member and high to a lower-capacity member. This zone-to-member mapping follows the capacity-noise pattern of Figure 5. The bank’s members are trained on the same federated distribution under the same atmospheric, sensor, and QPU noise. Capacity is the only variable that differs across them, so each is a generalist at its scale and not a specialist for a single channel band. The selected model’s score is the flag for the event, but the other bank members still run. Each scores the event, takes a local optimizer step against the reconstruction loss, and updates its own streaming threshold. Without this, an idle model’s weights and threshold would drift away from the channel statistics, and when conditions later brought it back into routing range, neither would match the live distribution. The compute cost of keeping all members active is reserved in the per-event budget. The router commits to one model and does not blend predictions across the bank (Figure 4). Different channel conditions are best served by different capacities, and a blended ensemble does not recover the capacity advantage at either end. Hard routing also avoids the extra hyperparameters a soft-mixture or overlapping-zone scheme would introduce.
Label-free dynamic thresholding. A trained bank member produces low reconstruction scores for samples it can reconstruct (normals) and high scores for samples it cannot (anomalies), so the score distribution has two clusters. The bank is trained on a normal-dominant mix without filtering anomalous signal, so both clusters have non-zero spread and overlap near the boundary. A fixed cutoff fails because channel and QPU drift shift both cluster means and spreads. Three timescales of drift apply. The atmospheric channel varies per event with the satellite’s geometry, the local turbulence, and the weather. The orbital sensor’s own noise drifts more slowly. The ground QPU drifts between recalibration windows. The two-cluster structure motivates a two-component mix model for score distribution, one component per cluster. EntangledEdge uses a Gaussian for each component, since each cluster is unimodal and Gaussian parameters fit from observed scores in closed form. The resulting 2-component Gaussian mixture (GMM-2) places the threshold at the score where the two component densities cross
where is the Gaussian density at with mean and variance , and are the mixing weights of the low-score (normal) and high-score (anomalous) clusters, are their cluster means, and are their cluster variances. The crossing is the score at which a sample is equally likely under either component, and it has a closed-form solution, a quadratic in with the relevant root in . Both components are re-estimated from observed scores. The crossing moves with them and the threshold tracks drift (Figure 6). Each station maintains a rolling reservoir of recent reconstruction scores, indexed per station because each station’s geometry produces a distinct distribution per model, as the bank members score differently. The GMM-2 is re-fit periodically from this reservoir, not on every event. Per-event refit would track short bursts of noisy events and oscillate. New events evict oldest scores, refreshing the fit.
Figure 6. Per-station QAE supplies both the score and, via the GMM-2 refit on the rolling reservoir, the threshold the score is judged against, so classification adapts to drift without labels.
Federated Coordination Across Stations
The per-event pipeline runs independently at every station, but a single model that generalizes across the orbit has to incorporate data from all stations. Different stations receive events from different satellite passes (Figure 3), and relaying sensed states to a central station is not viable. A second quantum hop compounds the atmospheric noise, and the satellite cannot measure the state for classical transmission without collapsing the entanglement structure QAE requires. EntangledEdge therefore trains in parallel at each station and exchanges only classical model parameters between rounds.
Each training round begins with the server broadcasting the current global weights to all stations. The station scores its incoming events through its model bank, takes local optimizer steps to reduce the SWAP-test costs, and at round close returns its updated weights together with its ESP signal (defined below). The server combines the contributions via ESP-weighted FedAvg, and the cycle repeats.
ESP-weighted circular aggregation. Noisier QPUs produce contaminated gradient updates, and uniform FedAvg that treats all stations equally would let those updates bias the global model. The aggregator therefore needs a per-station trust signal. The signal has to be computable without labels, refresh per round to track recalibration (R6), and capture gradient quality instead of sample count.
We use Expected Success Probability (ESP), the calibration-derived probability that a circuit executes correctly on a given backend. ESP is the product of three terms covering gate errors, amplitude relaxation, and dephasing, defined as – ESP , where is the calibrated error rate of gate , is circuit depth, and the average gate duration. ESP is computed from each backend’s live calibration and is cheap to recompute per round. Alternatives like total variation distance (TVD) need the ideal output distribution per circuit, far more expensive to characterize. At round , the aggregator combines each station ’s locally trained weights with mixing weight proportional to that station’s ESP. The trainable parameters are gate rotation angles, so aggregation is on the unit circle,
An arithmetic mean breaks down for angles near the wraparound. The circular form is topologically correct (Figure 7). ESPs are recomputed per round against each station’s live calibration, so the mixing weights track QPU drifts.
Figure 7. Stations return gate-angle parameters weighted by their expected success probability (ESP). The global update is their ESP-weighted circular mean (the vector sum on the circle), so a low-ESP station pulls it less.
Robustness to station outages and QPU recalibration. Stations can go offline through weather, maintenance, or hardware faults. EntangledEdge handles this in two ways. Events the offline station would have received are rerouted to the nearest operational station whose coverage window includes the pass; otherwise they are dropped (R1). The federated round proceeds with the reduced station set, and ESP-weighted aggregation renormalizes over the remaining contributors. When the station returns, it pulls the current global parameters in the next round and reinitializes its streaming threshold from incoming scores. QPUs can recalibrate between events. An IBM backend, for example, may expose a new coupling map and updated gate errors. EntangledEdge tolerates recalibration because the trained QAE parameters are the rotation angles. On a recalibration, the station retranspiles the circuit against the live backend and continues with the same parameters. The ESP signal incorporates the new gate errors and propagates them into the next round’s aggregation weights without retraining.
Scope and Generality
Teleporting sensed states to multiple ground stations is tied to the satellite setting, as no single station sees the whole orbit (Figure 3). The model bank applies to any noisy quantum link, terrestrial or orbital. One model capacity cannot serve all events (Figure 5). Fidelity-aware aggregation applies to any heterogeneous QPUs that share one model. These three mechanisms are not an arbitrary composition. Each one is forced by a constraint. None of these mechanisms reads a task’s labels or metric. Hence, the pipeline supports the sensing tasks of Sec. 1 beyond the anomaly detection and classification we evaluate (Sec. 4.3). If the QPU were placed in orbit, only the satellite-to-ground link would go away. Federation would run across satellites, the model bank would be conditioned on sensor noise, and fidelity-weighted aggregation would stay because QPUs drift.
Putting It All Together
End-to-end, the orbital sensor prepares its qubits in the sensing superposition, exposes them to the field, and teleports each phase-encoded state to the ground station whose contact window covers it (Sec. 2.3), where the station applies the radio-link Pauli correction to recover it on its QPU. The QAE is a parameterized ansatz (Sec. 2.1) whose rotation-gate angles are its trainable parameters. Before any quantum operation, the channel-conditioned router selects the bank member for the event (R4). A new event is dropped if the QPU is busy (R1), since coherence times forbid buffering. Otherwise the selected QAE produces the SWAP-test reconstruction score (R2) and the streaming GMM-2 threshold converts it to a binary flag (R5). Everything else runs on a classical compute node co-located with each station, which updates from the per-event reconstruction loss via the SPSA gradient estimation (§4.2.2), maintains the rolling score reservoir, performs the periodic GMM-2 refit, and exchanges parameter vectors with the federation server. At round close, every station returns its updated and its ESP from live calibration. ESP-weighted circular aggregation merges the contributions on the unit circle (R3), and the merged global model broadcasts at the next round’s start.
EntangledEdge is an online learning system that runs training and inference jointly on every event. After the selected QAE scores an event, its local optimizer takes a training step against the reconstruction loss, and the other bank members also score it and take their own training steps so they stay calibrated to the live channel. Each score enters the rolling reservoir that feeds the periodic GMM-2 refit, so the threshold at event reflects scores through event and the model has taken one more step than at event . EntangledEdge thus tracks the three drift timescales the design assumes — the per-event channel, the slower orbital sensor, and the inter-recalibration QPU. Under station outages and QPU recalibrations (R6), failover reroutes affected events with ESP renormalization, and recalibration retranspiles the ansatz against the live backend without retraining.
Implementation and Methodology
Hardware setup
EntangledEdge combines real hardware with simulation. The orbits come from published orbital data, and the orbit-to-ground link is simulated. Table 2 summarizes our prototype.
Table 2. Real and simulated parts of our hybrid prototype.
Satellite and ground-station QPUs
The configuration spans up to eight low-Earth-orbit satellites carrying quantum sensors and up to eight ground-station QPUs. Orbits are propagated from published two-line element sets with the standard SGP4 propagator [83].
The main experiments use four stations, Fairbanks, Wallops, Madrid, and Singapore, which run on real superconducting IBM hardware (Heron and Nighthawk).
Satellite-to-ground quantum channel
The transmitted state crosses the atmosphere to the receiving QPU and arrives with turbulence-induced channel noise [81]. We model it as a per-qubit composition of two completely positive trace-preserving (CPTP) maps, with strengths fixed per event by slant-path turbulence, which follows the Hufnagel–Valley (HV) profile [3, 84], set by a per-station boundary-layer strength () and standard in satellite quantum-channel work [6, 10, 25, 28]. From it we compute the Rytov variance through the slant-path integral [3], with taken from SGP4 [83], and convert it via the exponential decay of coherence with accumulated phase variance [3], to per-qubit dephasing () and depolarization () channels applied to every qubit of the -qubit density matrix [57]. This link is simulated in our prototype for evaluation.
Input sensor
For the main experiments, we assume the density matrix as the sensing input, since no orbital quantum sensor is yet deployed.
Neutral atom sensor. We use QuEra Aquila [93], a real neutral atom device whose Rydberg states respond sharply to weak fields and stay coherent long enough to read the shift out (§ 2.2). A controlled Hamiltonian prepares each atom in a ground–Rydberg superposition. The sensed field shifts the atom’s transition frequency, so over the exposure the relative phase winds at the shifted rate while the populations stay fixed. The accumulated phase, read out by interference, is the field value, and longer exposure within the coherence time gives a finer reading (Figure 8). From Aquila’s calibration we fit a three-component noise spectral density [46] to Ramsey and CPMG sequences on the Aquila device through AWS Braket, and derive a single-atom Kraus channel, amplitude damping at its spontaneous-emission and dephasing matched to the coherence loss over the exposure window. We apply it to each of ’s atoms before the atmospheric channel (§ 4.3), so an event accumulates sensor, channel, and QPU noise in turn.
Figure 8. Field sensing on QuEra Aquila. Atoms start in a ground–Rydberg superposition , and the field imparts a phase of onto that encodes the reading.
Model and training
Model bank
The capacity-diverse bank (§ 3.2.1) holds three quantum autoencoders, large (), medium (), and small (). All use an EfficientSU2 [1] encoder with full entanglement and alternating rotations, followed by the SWAP test of § 3.2.1. uses four qubits at depth four, three qubits at depth three, and three qubits at depth one. carries a two-qubit trash space compared against , and and carry a single trash qubit against .
Training
Each client trains its local copy with Adam at learning rate 0.03 on the IBM hardware (§ 4.2), at 1024 shots. Since the parameter-shift rule of § 2.1 is prohibitively expensive. We instead estimate the gradient with Simultaneous Perturbation Stochastic Approximation (SPSA) [77], a standard estimator under hardware noise [24, 38, 88]. We track the area under the precision-recall curve (PR-AUC) across rounds, a threshold-independent metric for anomaly detection [19, 72, 73].
Datasets and encoding
We evaluate on OPS-SAT [71] of § 2 and SMAP [36], widely used satellite telemetry benchmarks. Each segment is summarized by a handful of statistical features and fit to the -length amplitude vector on -qubit memory expects.
OPS-SAT ships as pre-extracted segments, while SMAP is raw per-channel telemetry. For SMAP, we take a subset of channels and cut each into non-overlapping windows, summarized with the same features as OPS-SAT [71]. Since SMAP anomalies are defined per channel, we standardize features per channel against that channel’s anomaly-free reference and draw a split at OPS-SAT’s anomaly rate (). The same channels, windows, features, and split serve EntangledEdge and every baseline. The one SMAP-specific change is the encoding: it keeps each window’s magnitude.
Supervised image tasks. We also run classification tasks, MNIST (3 versus 4) and FashionMNIST (bag versus sandal), for a generality check. The federated topology, noise, and router banks are unchanged; only the SWAP-test reconstruction objective becomes a supervised readout. Each image is flattened as above, into a balanced split. The other sensing tasks named in § 1 are supported but not evaluated (§ 3.5.1).
Competing Solutions
Isolated. Each station runs its own QAE on its own shard under its own channel and QPU noise and shares nothing.
Centralized. Every station forwards its sensed states to one server, which trains a single QAE on the pooled data under one QPU’s calibration. Each state crosses the satellite-to-ground link and a second, lossier ground-to-central hop, so the central model trains on doubly attenuated states.
FedAverage. The federated protocol with plain FedAvg, replacing the ESP-weighted aggregation of Sec. 3.2.2.
Quorum [48]. Quorum is a standalone quantum autoencoder anomaly detector with no notion of an orbital channel, sensing, or federation, but it is the closest prior quantum method, so we port it into EntangledEdge’s setting. It scores through SWAP-test reconstruction but trains no encoder: it fixes an ensemble of encoders and scores each sample by its reconstruction-cost deviation from a bucketed ensemble mean. We run it on the same channel and per-station QPU noise as EntangledEdge, with a larger shot budget.
Two of its mechanisms cannot run on a single teleported state. Its score needs a second copy of the input to SWAP-test against, which no-cloning forbids [91], so we use EntangledEdge’s single-state trash-qubit SWAP test (§ 3.2.1). It also re-selects a random feature subset per member, but the ground station holds only the sensed state, so we drop feature selection. We otherwise follow Quorum as published and bucket its events by the same the router uses (§ 3.2.1).
Evaluation Protocol
Event stream. We evaluate on a real-telemetry event stream over one orbital window across the constellation of satellites and ground stations (§ 4.2). We encode the data (§ 4.6) in the neutral atom quantum sensors as the sensors’ output state and treat it as the sensing input (§ 4.4). We then simulate the orbital sensor-to-ground-station pass, propagating the orbit and applying the atmospheric channel (§ 4.3) to each downlinked state. All events are drawn from the held-out test partition, and every method is scored on the same stream.
Serving model. We model each station as an M/D/1/1 server, so any event that arrives while it is busy is dropped, the drop policy of C1 (§ 3.1). We use IBM superconducting QPUs to process the events. An event’s service time is one inference plus online training over the models the method runs on.
Metrics. Our system metrics are throughput, latency, and processor utilization, with utilization normalized to the in-view active window, and we also track drop rate. For detection quality, we report F1 score, precision, recall, and accuracy, with F1 as the primary measure. F1 is the metric of the anomaly-detection task (not of EntangledEdge’s mechanism (§ 3.5.1)). Isolated is the baseline, so we report both the percent improvement over it and the raw metric values.
Evaluation
We first establish the end-to-end effectiveness of EntangledEdge, then examine its robustness, scaling, and the ablations that explain why it works in practice.
Effectiveness of EntangledEdge
We compare EntangledEdge against the isolated, centralized, and FedAverage baselines on the held-out stream, scored in streaming mode (§ 4.8). All four converge in PR-AUC on both OPS-SAT and SMAP (Figure 9, top), and EntangledEdge reaches the top rank on each, with FedAverage close on SMAP. Isolated plateaus lowest, since each station overfits the single orbital patch it sees. Centralized pools across stations, but its second ground-to-central quantum hop adds noise to the training data and slows convergence. EntangledEdge shares only classical parameters, so it gets the pooled coverage without the hop, and its ESP-weighted aggregation down-weights noisy backends, which FedAverage lacks.
Figure 9. EntangledEdge converges to the top PR-AUC rank on both datasets (top), and serves events at latency and throughput near the single-model baselines, while centralized is slowed by its teleportation hop (bottom).
EntangledEdge also reaches the highest F1 in terms of raw value, 0.63 on OPS-SAT and 0.48 on SMAP (Table 3) against 0.52 and 0.42 for FedAverage, and leads on precision and accuracy. Each entry is a mean over 100 days of calibration drift with a small standard deviation, so the ranking is stable. As improvement over isolated (Figure 10), the gain sits almost entirely in precision while recall drops a few points. The reason is calibration. An overfit isolated model flags more candidates from a miscalibrated distribution, so it trades precision for recall. EntangledEdge instead routes each event to the bank member matched to its channel and judges the score against a threshold that re-fits to the drifting distribution, which separates normal from anomalous cleanly and suppresses false positives. On a sparse-anomaly stream, this high-precision point is the useful one, and F1 settles in our favor. The same table reports MNIST and FashionMNIST, where F1 stays above 0.80 under the identical setup, so the pipeline holds beyond the anomaly workload.
| Dataset | Statistic | F1 | Precision | Recall | Accuracy |
| OPS-SAT | Mean | 0.63 | 0.59 | 0.66 | 0.83 |
| OPS-SAT | Standard Deviation (%) | 4.3 | 6.8 | 2.4 | 2.9 |
| SMAP | Mean | 0.48 | 0.37 | 0.67 | 0.71 |
| SMAP | Standard Deviation (%) | 2.0 | 2.2 | 1.5 | 1.9 |
| MNIST | Mean | 0.81 | 0.82 | 0.80 | 0.82 |
| MNIST | Standard Deviation (%) | 3.0 | 3.5 | 4.5 | 2.9 |
| FashionMNIST | Mean | 0.83 | 0.83 | 0.84 | 0.84 |
| FashionMNIST | Standard Deviation (%) | 4.5 | 2.2 | 6.8 | 3.4 |
Table 3. Detection quality of EntangledEdge under QPU calibration drift, as mean and standard deviation across runs.

Figure 10. EntangledEdge gains most in precision and dips in recall, but the precision gain outweighs it, so F1 (major metric) still leads (expressed as % improvement over isolated).
Every method runs online, so each event costs one inference and an SPSA update per model it runs (Figure 9, bottom, under the M/D/1/1 model at ms). The single-model baselines run only the large model at 0.30 s per event. EntangledEdge runs all three bank members on every event, so each stays calibrated and ready when the channel drifts. Because the bank is capacity-diverse and the small and medium models are cheaper, this readiness costs 0.57 s, 1.9× the single-model figure and well below the 3× of three equal models. Centralized runs the same single-model compute yet is slowest at 0.78 s, because every off-station event pays a distance-scaled teleportation delay to the one central QPU, so latency is set by the topology more than by the bank size.
The same choice to train the whole bank locally raises active-window utilization (Figure 11), where EntangledEdge
Figure 11. EntangledEdge has the highest utilization due to training every bank member on each event, and centralized is lowest because one QPU does all the learning.
reaches 51.6%, about twice that of FedAverage and isolated. EntangledEdge keeps every QPU working through its short contact window, the efficient use of hardware that is otherwise idle most of the orbit. Centralized instead routes all learning to one QPU, and its other three only receive and forward states, so most of the fleet sits idle.
EntangledEdge also leads Quorum on all four metrics on OPS-SAT under matched channel and QPU noise (Figure 12, Sec. 4.4), with F1 at 0.63 against 0.44. Quorum reaches 0.62 F1 in an ideal noise-free run, so its capacity is not what holds it back. It freezes random encoders and never trains, so it cannot track the channel and QPU drift the deployment stream carries. EntangledEdge trains online and refits its threshold to that drift, so its F1 holds while Quorum falls. A static detector cannot stay calibrated in this orbital setting.

Figure 12. EntangledEdge against Quorum on OPS-SAT data. EntangledEdge leads on every metric by adapting online to noise the zero-training competing technique cannot track.
Robustness of EntangledEdge
We stress EntangledEdge along three axes – the QPU backend at each station, the QPU and atmospheric noise levels, and the sensor noise itself. Every panel reports F1 improvement over the isolated per-station baseline on OPS-SAT.
EntangledEdge results in the largest F1 improvement in every configuration (Figure 13). The left figure runs all four stations on a single shared backend, one backend at a time, and the right figure runs different combinations of backends across the stations. FedAverage trails well behind in both, and Centralized falls below the isolated baseline on the weakest backend. EntangledEdge stays ahead because its ESP-weighted aggregation reads each backend’s live calibration and weights its update accordingly, so a weak or mismatched backend contributes less and does not set the global model. The system is therefore robust across backends.
Figure 13. EntangledEdge improves F1 score over isolated baseline across backends on OPS-SAT, with every station on one backend (left) and a mix of backends across stations (right).
We scale the QPU and atmospheric noise around the calibrated level (Figure 14). Under QPU noise scaling, EntangledEdge’s F1 improvement is roughly constant and largest at the low-noise end, because its ESP weighting and federated averaging keep the noisier backends from degrading the shared model. That low-noise end matters most, since it approximates the fault-tolerant (FTQC) hardware to come. EntangledEdge still improves over isolated there, so its value does not depend on today’s high noise. Under atmospheric scaling, EntangledEdge shows the largest improvement at each severity and remains positive across the range, because its router re-matches the model to the channel as conditions change. Centralized is most affected on the channel, since its second ground-to-central hop compounds atmospheric noise, falling below the isolated baseline at the harshest settings.
Figure 14. EntangledEdge’s F1 improvement over the isolated baseline under QPU noise scaling (left) and atmospheric noise scaling (right) on OPS-SAT.
The other experiments leave sensor noise out, since it perturbs every method’s input alike and does not change their ranking (§ 4.4). Here we inject the real neutral-atom sensor noise into the state , on top of the channel and QPU noise the pipeline already applies. The top panel of Figure 15 reports the total variation distance (TVD) between each event’s clean and sensor-perturbed input state, which isolates the sensor’s own contribution to the noise, and the sensor distorts different inputs by widely varying amounts. We sort events by that distortion from light to heavy. EntangledEdge still improves F1 over isolated by about 21% overall and stays positive in nearly every chunk across that range (bottom). The gain holds because EntangledEdge’s advantage comes from matching capacity to the channel and sharing across stations, which the input-side perturbation leaves intact.
Figure 15. EntangledEdge keeps improving F1 over the isolated baseline as sensor distortion grows (OPS-SAT). Chunks are ordered by mean TVD from the clean input (top), and the improvement stays positive across nearly all of them (bottom).
Why does EntangledEdge Work ?
We examine where EntangledEdge’s gains come from (Figure 16). The router spreads events across all three model sizes (left). The large model takes most traffic, but the medium and small models carry a meaningful share, because the router matches every event to its channel and the channel varies (§ 3.2.1).
Figure 16. Every design component matters (result for OPS-SAT data). The router uses all three capacities, and ablating the bank, the dynamic threshold, or ESP weighting lowers F1.
To capture the benefit of each of EntangledEdge’s three main design components, we remove one at a time and report the F1 drop against the full system (Figure 16, right). No single model or two-model pair matches the full bank, and the large
model alone is the worst single choice despite taking the most traffic, because it suits the events routed to it but degrades on the noisier ones the smaller models handle. Replacing the dynamic GMM-2 threshold (§ 3.2.1) with a fixed cutoff gives the largest drop, because the score distribution drifts with the channel and only a re-fitting threshold tracks it. Replacing ESP-weighted aggregation (§ 3.5) with uniform FedAverage lets the noisier QPUs bias the global model. Each component contributes to EntangledEdge’s effectiveness.
Scaling of EntangledEdge
We scale EntangledEdge along the sensing rate and the fleet size (Figure 17). Running every configuration at the full event scale is too expensive, so the sensing-rate sweep uses a shortened simulation window, and the fleet sweep grows its event volume with the configuration.
Figure 17. EntangledEdge scales on both axes (OPS-SAT). A higher sensing rate raises utilization and throughput (top), and a larger fleet adds throughput in parallel (bottom), while per-event latency and F1 improvement stay roughly constant.
As the sensing rate increases from to (Figure 17, top), more events arrive per visibility window, so QPU utilization increases from 51% to 81% and throughput roughly doubles. Per-event latency remains at 0.57 s, since the M/D/1/1 service time does not depend on the arrival rate, and F1 improvement remains near 22 to 27%. EntangledEdge handles the higher load by using otherwise-idle QPU time, without increasing latency or reducing detection quality.
As the fleet size increases from to satellites and ground stations (Figure 17, bottom), throughput increases with it, from a few hundred events to about 3,700 at , because each station serves its own events in parallel. Per-event latency remains at 0.57 s and utilization remains between 45 and 52% across all nine combinations, since the federated design has no central server to serialize the work. F1 improvement remains positive throughout, from 13 to 28%. Adding satellites and stations increases throughput without raising per-event latency, so EntangledEdge scales out.
Finally, we vary the large model’s qubit count from 4 to 10 (medium and small fixed) on ibm_pittsburgh hardware, and compare against the large model trained under the isolated baseline on OPS-SAT. Federation improves F1 at every size, and the gain grows with capacity over 6 to 10 qubits, from 8.4% at 6 to 12.9% at 8 and 18.3% at 10, because a larger model better uses federation’s cross-station data, with the largest effect at low channel noise. This trend generalizes beyond, showing EntangledEdge can scale to larger models.
Limitations
Orbital quantum sensors that emit their state into a quantum link are not yet deployed. We therefore apply Aquila-measured sensor noise to telemetry-encoded states and simulate the orbit-to-ground link (Table 2). EntangledEdge is thus evaluated as a system architecture on a hybrid prototype, ahead of a deployable orbital stack.
Related Works
Orbital Edge Computing. Orbital edge computing treats a satellite constellation as a distributed system that computes onboard [22]. Later systems run ML inference in orbit [21, 44, 96], detection-driven capture [16], and federated learning [41, 50, 67, 68] and flown on an onboard accelerator [27]. Other work reduces orbit-to-ground latency through distributed ground stations, task splitting, and scheduling [17, 40, 79, 85]. Separate lines place compute in orbit as microdatacenters [8, 9] and LEO cloud and serverless platforms [7, 62, 63]. All stay classical end-to-end, and none of these works operate on quantum-sensed data that requires state preservation.
Quantum Sensing, Networking, and Learning. Quantum field sensing surpasses the classical limit in magnetometry, gravimetry, and time-frequency metrology [20, 61], with advances on atomic and optical platforms [4, 5, 12, 18, 20, 37, 92, 94]. Quantum networks distribute entanglement and transfer states [86], with teleportation [76, 82] and entanglement distribution [78, 87] over fiber and multi-node testbeds [66]. Satellites reach ground stations [45, 95] under studied schedules [23, 33, 51, 58, 89], and been proposed to entangle ground quantum sensors [39], and fiber testbeds target distributed quantum computing [56, 74]. For learning, variational classifiers run on quantum hardware [31, 34], through a hybrid quantum-classical loop [13] whose parameters extend to quantum federated learning [2, 15, 29, 60, 64], including over satellite–ground links [59]. Quantum learning has also been applied to LEO communication signals [80] and Earth-observation imagery [55]. None of these works learns from an orbital sensed state that stays unmeasured. What they move across a link is classical data, model parameters, prepared states, or entanglement.
Why existing remedies do not transfer? Satellite edge computing, teleportation, and federated learning recover from a missed contact or a downed station by storing data, re-sending it, or holding it until it can be labeled [14, 22, 52] (Table 4). A sensed quantum state allows none of these. A missed pass loses the state, a failed transfer cannot be repeated, and labeling destroys it. Because these mechanisms do not carry over, link noise reaches the model directly, which is why we need the model bank and ESP-weighted aggregation.
Conclusion
EntangledEdge is the first system architecture to unify orbital quantum sensing and learning. Its data plane stays quantum from the orbital sensor to ground inference, and it improves detection quality, maintains high system utilization, and scales out across fleets without raising per-event latency.
Acknowledgments. We thank our shepherd and the reviewers for their feedback. This research used resources of the National Energy Research Scientific Computing Center (NERSC), a DOE Office of Science User Facility supported under Contract No. DE-AC02-05CH11231, using NERSC award DDR-ERCAP0037923. This work was supported by the Kahlert School of Computing and the SCI Institute at the University of Utah. Rice University and DOE Quantum Testbed Finder Award DE-SC0024301 supported the Rice authors.
Artifact Appendix
Abstract
Our artifact consists of the trained models of the paper and the code that evaluates and trains them. We claim that EntangledEdge improves detection quality, latency, throughput, and QPU utilization over the isolated, centralized, and FedAverage baselines and outperforms Quorum (Sec. 5), holds its gain on other QPU backends and under the sensor noise (Sec. 5.2), and scales with the sensing rate and the fleet size (Sec. 5.4). The artifact contains the trained autoencoders of every method, the inputs they are evaluated on, and one script per figure that regenerates it and prints every value next to the paper’s. The latest version of the archive also provides the real-hardware path, which needs an IBM Quantum account and AWS Braket access. By default, the circuits run in Qiskit Aer as noisy simulations under the backends’ calibration data, and the sensor noise comes from the shipped spectroscopy, so no quantum hardware account is needed in that case.
Artifact check-list (meta-information)
Program: The EntangledEdge package: the autoencoder bank, router, thresholds, channel and QPU noise models, sensor channel, serving model, federated training, and one script per figure.
Model: The trained autoencoders of every method on OPS-SAT and SMAP with their training logs; the scores of Quorum’s 50 members.
Data set: OPS-SAT and NASA SMAP as published, the two event streams, the IBM calibration snapshots, and the Aquila sensor noise.
Hardware: In simulation, an NVIDIA GPU (our run times are for four A100s). On real hardware, an IBM Quantum account and AWS Braket access to QuEra Aquila.
How much disk space required (approximately)?: 32 GB
How much time is needed to prepare workflow (approximately)?: minutes
How much time is needed to complete experiments (approximately)?: 1 h 15 min for the main figures on four A100s, about 3 hours more for Quorum and Figs. 15 and 17; 15 to 20 hours per dataset to retrain the models. On real hardware, the time depends on the backends’ queues.
Publicly available?: Yes: doi.org/10.5281/zenodo.22909579 (latest version; simulation and real hardware)
Code licenses (if publicly available)?: MIT
Description
How to access
Download EntangledEdge-Artifact.zip from doi.org/10.5281/zenodo.22909579 and unpack it. Every module runs from the root of the archive as python -m entanglededge.<module> and lists its options with –help.
Hardware dependencies
In simulation, an NVIDIA GPU with a CUDA 12 driver; a scoring job needs 1 GB of GPU memory, 3 GB of host memory, and 4 CPU threads, and the sensor spectroscopy on the Bloqade emulator needs 16 CPU cores. On real hardware, an IBM Quantum account with access to at least one backend (by default the scripts use the backends’ calibration snapshots; –ibm-backends names others) and AWS Braket access to QuEra Aquila.
Software dependencies
Linux x86-64, an NVIDIA driver with CUDA 12 support, and Miniconda; setup.sh installs everything else (requirements.txt). On real hardware, an IBM Quantum token and AWS credentials; the sensor spectroscopy on the Bloqade emulator also needs Julia with Bloqade.
Data sets
All inputs ship in data/. It holds the OPS-SAT and SMAP datasets of Sec. 4.3, with the raw SMAP telemetry in smap_raw/. events/ holds the event streams of both datasets, backends/ the calibration data of the four IBM backends as Aer noise models, and sensor/ the Aquila noise spectroscopy and the sensor channel fitted to it.
Models
The artifact ships every trained model so that EntangledEdge can be tested without the long training jobs (models/). The scores of the 50 members of the Quorum ensemble (Sec. 4.4) are shipped at models/opsat/Quorum-50members/.
Installation
From the root of the unpacked archive:
./setup.sh
conda activate entanglededge
python reproduce/kick_the_tires.py
The smoke test takes about 7 minutes on one GPU and ends with All stages ran. For real hardware, add the IBM Quantum token and the AWS credentials as in README Sections 9 and 10.
Experiment workflow
Training. To train the models of a dataset (
–dataset smapfor SMAP), one trainer per model with the arguments of the paper’s runs, in Aer under the stations’ calibration data:
python reproduce/train_flagship.py \
--dataset opsat --gpus 0,1,2,3
or on the IBM backends:
python reproduce/train_flagship.py \
--dataset opsat --qpu ibm --ibm-shots 512
Either writes the models with their training logs to results/<dataset>/; configs/<dataset>.yaml selects them for the deployment below in place of the shipped ones in models/.
2. Deployment. The trained models serve the held-out event streams of § 4.8: every event is scored, routed, thresholded, and run through the serving model. This one run yields Figure 9 to Figure 12 and Figure 16. In Aer, reproduce_all.sh runs the figure scripts, and the backend sweep of point 3, in one go:
./reproduce_all.sh
On the IBM backends, only the scoring script changes, with --qpu ibm; the other four run no circuits:
python reproduce/figure10.py --qpu ibm
python reproduce/figure9.py
python reproduce/figure11.py
python reproduce/figure12.py
python reproduce/figure16.py
where --ibm-max-events limits the scoring to the first events of each stream and a rerun continues from the last finished job. Quorum’s 50 members score the same OPS-SAT stream; the artifact ships the scores of that run, and
python reproduce/quorum_ensemble.py --gpus 0,1,2,3
python reproduce/figure12.py \
--members-dir outputs/figure12/members
produces them again in Aer. Every figure lands as PDF in figures/, and comparison_with_paper.txt lists every value of the run next to the paper’s.
3. Robustness. The deployment is run again under other QPU backends (Figure 13) and under the sensor noise (Figure 15). The backend sweep, in Aer under each configuration’s calibration data, or with each station on the backend of its configuration:
python reproduce/figure13.py --gpus 0,1,2,3
python reproduce/figure13.py --qpu ibm
The sensor noise is the channel fitted to the noise spectroscopy of one Aquila atom (§ 4.4), on the noisy Bloqade emulator of Aquila:
python -m entanglededge.sensor.spectroscopy \
--backend bloqade --shots 10000 \
--out outputs/sensor/noise_spectroscopy.npz
or measured on the device through AWS Braket (--backend aquila --shots 1000); the spectral density, the channel, and the deployment under it follow:
python -m entanglededge.sensor.nsd \
--spectroscopy \
outputs/sensor/noise_spectroscopy.npz \
--out outputs/sensor/nsd_fit.json
python -m entanglededge.sensor.channel \
--nsd outputs/sensor/nsd_fit.json \
--out outputs/sensor/sensor_channel.npz
python reproduce/figure15.py --gpus 0,1,2,3 \
--sensor-channel \
outputs/sensor/sensor_channel.npz
Without --sensor-channel, figure15.py uses the shipped channel in data/sensor/; with --qpu ibm, its circuits run on the IBM backends.
4. Scaling. The deployment is run on the thirteen event streams of § 5.4, four sensing rates and nine fleet sizes, which the script generates:
python reproduce/figure17.py --gpus 0,1,2,3
--qpu ibm runs their circuits on the IBM backends, as in point 2.
Evaluation and expected results
1. Training. Retrained models are statistically, not bitwise, equivalent to the shipped ones, since the shot sampling is not seeded; run through the deployment, they are expected to reproduce the ranking of the methods within the tolerance of point 2.
2. Deployment. The figures are expected to show the paper’s trends, with EntangledEdge ahead of every baseline and of Quorum on F1, and comparison_with_paper.txt lists every value next to the paper’s. In Aer, the metrics are expected within a few percent of the paper’s values. On the IBM backends, the outputs are the same files, with the backend, the shots, and the job IDs recorded in each run’s manifest; this path was not part of the artifact evaluation, and we state no tolerance for it.
3. Robustness. EntangledEdge is expected to keep the largest F1 improvement over Isolated in every backend configuration (Figure 13; the artifact uses ibm_fez in place of Torino, whose calibration data are not available) and a positive improvement in nearly every chunk as the sensor distortion grows (Figure 15). A spectroscopy measured on Aquila enters through the same fit and channel as the shipped one.
4. Scaling. Utilization and throughput are expected to rise with the sensing rate, throughput with the fleet size, while per-event latency and the F1 improvement over Isolated stay roughly constant and positive; the improvement comes out higher than the paper’s, which used an earlier version of the event pipeline.
References
- [1][n. d.]. EfficientSU2 (latest version) | IBM Quantum Documentation — quantum.cloud.ibm.com. https://quantum.cloud.ibm.com/docs/en/api/qiskit/qiskit.circuit.library.EfficientSU2.
- [2]Kwabena Addo, Musasa Kebaya Evans Eshiemogie Ojo. 2025. Federated quantum machine learning for distributed cybersecurity in multi-agent energy systems. Energies 18, 20 (2025), 5418.
- [3]Larry C Andrews and Ronald L Phillips. 2005. Laser beam propagation through random media. Laser Beam Propagation Through Random Media: Second Edition (2005).
- [4]Nabeel Aslam, Hengyun Zhou, Elana K Urbach, Matthew J Turner, Ronald L Walsworth, Mikhail D Lukin, and Hongkun Park. 2023. Quantum terms for biomedical applications. Nature Reviews Physics 5, 3 (2023), 157–169.
- [5]Steven D Bass and Michael Doser. 2024. Quantum sensing for particle physics. Nature Reviews Physics 6, 5 (2024), 329–339.arxiv.org/abs/2305.11518
- [6]Satya Ranjan Behera and Urbasi Sinha. 2024. Estimating the link budget of satellite-based Quantum Key Distribution (QKD) for uplink transmission through the atmosphere. EPJ Quantum Technology 11, 1 (2024), 66.
- [7]Vaibhav Bhosale, Ada Gavrilovska, and Ketan Bhardwaj. 2024. Krios: Scheduling abstractions and mechanisms for enabling a LEO compute cloud. In Proceedings of the 2024 ACM Symposium on Cloud Computing. 322–340.DOI
- [8]Nathan Bleier, Rick Eason, Michael Lembeck, and Rakesh Kumar. 2025. Architecting space microdatacenters: A system-level approach. In 2025 IEEE International Symposium on High Performance Computer Architecture (HPCA). IEEE, 1304–1319.DOI
- [9]Nathaniel Bleier, Muhammad Husnain Mubarik, Gary R Swenson, and Rakesh Kumar. 2023. Space microdatacenters. In Proceedings of the 56th Annual IEEE/ACM International Symposium on Microarchitecture. 900–915.DOI
- [10]Jean-Philippe Bourgoin, Evan Meyer-Scott, Brendon L Higgins, B Helou, Chris Erven, Hannes Huebel, B Kumar, D Hudson, Ian D’Souza, Ralph Girard, et al. 2013. A comprehensive design and performance analysis of low Earth orbit satellite quantum communication. New Journal of Physics 15, 2 (2013), 023006.
- [11]Vladimir Bužek and Mark Hillery. 1996. Quantum copying: Beyond the no-cloning theorem. Physical Review A 54, 3 (1996), 1844.DOI
- [12]Christophe Cassens, Bernd Meyer-Hoppe, Ernst Rasel, and Carsten Klempt. 2025. Entanglement-enhanced atomic gravimeter. Physical Review X 15, 1 (2025), 011029.DOI
- [13]Marco Cerezo, Andrew Arrasmith, Ryan Babbush, Simon C Benjamin, Suguru Endo, Keisuke Fujii, Jarrod R McClean, Kosuke Mitarai, Xiao Yuan, Lukasz Cincio, et al. 2021. Variational quantum algorithms. Nature Reviews Physics 3, 9 (2021), 625–644.
- [14]Mahdi Chemi and Walid Saad. 2022. Quantum federated learning with quantum data. In ICASSP 2022-2022 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP). IEEE, 8617–8621.
- [15]Samuel Yen-Chi Chen and Shinjae Yoo. 2021. Federated quantum machine learning. Entropy 23, 4 (2021), 460.
- [16]Zhuo Cheng, Bradley Denby, Kyle McCleary, and Brandon Lucia. 2024. EagleEye: Nanosatellite constellation design for high-coverage, high-resolution sensing. In Proceedings of the 29th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 1. 117–132.DOI
- [17]Zhuo Cheng and Brandon Lucia. 2025. Geoduck: Nanosatellite Constellation Scheduling for Low Latency Event Detection. In Proceedings of the 23rd ACM Conference on Embedded Networked Sensor Systems. 547–559.DOI
- [18]Scott Crawford, Roman A Shugayev, Hari P Paudel, Ping Lu, Madhava SyamIal, Paul R Ohodnicki, Benjamin Chorpening, Randall Gentry, and Yuhua Duan. 2021. Quantum sensing for energy applications: Review and perspective. Advanced Quantum Technologies 4, 8 (2021), 2100049.DOI
- [19]Jesse Davis and Mark Goadrich. 2006. The relationship between Precision-Recall and ROC curves. In Proceedings of the 23rd international conference on Machine learning. 233–240.DOI
- [20]C. L. Degen, F. Reinhard, and P. Cappellaro. 2017. Quantum sensing. Rev. Mod. Phys. 89 (Jul 2017), 035002. Issue 3. doi:10.1103/RevModPhys.89.035002DOI
- [21]Bradley Denby, Krishna Chintalapudi, Ranveer Chandra, Brandon Lucia, and Shadi NoghabI. 2023. Kodan: Addressing the computational bottleneck in space. In Proceedings of the 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 3. 392–403.DOI
- [22]Bradley Denby and Brandon Lucia. 2020. Orbital edge computing: Nanosatellite constellations as a new class of computer system. In Proceedings of the Twenty-Fifth International Conference on Architectural Support for Programming Languages and Operating Systems. 939–954.
- [23]Ashutosh Jayant Dikshit, Naga Lakshmi Anipeddi, Prajit Dhara, Saikat Guha, Deirdre Kilbane, Leandros TassIulas, Don Towsley, and Nitish K Panigrahy. 2025. Scheduling in Quantum Satellite Networks: Fairness and Performance Optimization. arXiv preprint arXiv:2512.07108 (2025).arxiv.org/abs/2512.07108
- [24]Maria-Andrea Filip. 2024. Fighting noise with noise: a stochastic projective quantum eigensolver. Journal of Chemical Theory and Computation 20, 14 (2024), 5964–5981.DOI
- [25]Masoud Ghalai and Stefano Pirandola. 2022. Quantum communications in a moderate-to-strong turbulent space. Communications Physics 5, 1 (2022), 38.
- [26]Vittorio Giovannetti, Seth Lloyd, and Lorenzo Maccone. 2006. Quantum metrology. Physical review letters 96, 1 (2006), 010401.pubmed.ncbi.nlm.nih.gov/16486424
- [27]Gianluca Guffrida, Luca Fanucci, Gabriele Meoni, Matej Batic, Léonie Buckley, Aubrey Dunne, Chris Van Dijk, Marco Esposito, John Hefele, Nathan Vercruyssen, et al. 2021. The Φ-Sat-1 mission: The first on-board deep neural network demonstrator for satellite earth observation. IEEE Transactions on Geoscience and Remote Sensing 60 (2021), 1–14.
- [28]Tasio Gonzalez-Raya, Stefano Pirandola, and Mikel Sanz. 2024. Satellite-based entanglement distribution and quantum teleportation with continuous variables. Communications Physics 7, 1 (2024), 126.
- [29]Jason Han, Nicholas S DBrita, Daniel Leeds, Jianqiang Li, Jason Ludmir, and Tirthak Patel. 2025. Layerwise federated learning for heterogeneous quantum clients using quorus. arXiv preprint arXiv:2510.06228 (2025).arxiv.org/abs/2510.06228
- [30]Xuan Han, Hai-Lin Yong, Ping Xu, Kui-Xing Yang, Shuang-Lin Li, Wei-Yang Wang, Hua-Jian Xue, Feng-Zhi Li, Ji-Gang Ren, Cheng-Zhi Peng, et al. 2020. Polarization design for ground-to-satellite quantum entanglement distribution. Optics Express 28, 1 (2020), 369–378.DOI
- [31]Vojtěch Havlíček, Antonio D Dórcoles, Kristan Temme, Aram W Harrow, Abhinav Kandala, Jerry M Chow, and Jay M Gambetta. 2019. Supervised learning with quantum-enhanced feature spaces. Nature 567, 7747 (2019), 209–212.DOI
- [32]SLN Hermans, Matteo Pompili, HKC Beukers, Simon Baier, Johannes Borregaard, and Ronald Hanson. 2022. Qubit teleportation between non-neighbouring nodes in a quantum network. Nature 605, 7911 (2022), 663–668.DOI
- [33]Md Zakir Hossain, Nitish K Panigrahy, Walter O Krawec, Don Towsley, and Bing Wang. 2025. Opportunistic Scheduling for Single-downlink Satellite-based Quantum Key Distribution. arXiv preprint arXiv:2512.12514 (2025).arxiv.org/abs/2512.12514
- [34]Hsin-Yuan Huang, Michael Broughton, Jordan Cotler, Sitan Chen, Jerry Li, Masoud Mohseni, Hartmut Neven, Ryan Babbush, Richard Chung, John Preskill, et al. 2022. Quantum advantage in learning from experiments. Science 376, 6598 (2022), 1182–1186.DOI
- [35]Hsin-Yuan Huang, Michael Broughton, Masoud Mohseni, Ryan Babbush, Sergio Boixo, Hartmut Neven, and Jarrod R McClean. 2021. Power of data in quantum machine learning. Nature communications 12, 1 (2021), 2631.
- [36]Kyle Hundman, Valentino Constantinou, Christopher Laporte, Ian Colwell, and Tom Soderstrom. 2018. Detecting Spacecraft Anomalies Using LSTMs and Nonparametric Dynamic Thresholding. In Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining (London, United Kingdom) (KDD ’18). Association for Computing Machinery, New York, NY, USA, 387–395. doi:10.1145/3219819.3219845DOI
- [37]Anastashia Jeabrielli, Chenxu Liu, Keyi Yin, Samuel Stein, Erik Lentz, Yufei Ding, and Ang Li. 2026. Status: A unified system architecture for spatial temporal quantum sensing. ACM Transactions on Quantum Computing 7, 2 (2026), 1–34. ... microwave photonic qubits. Physical Review X 14, 3 (2024), 031055.
- [38]Abhinav Kandala, Antonio Mezzacapo, Kristian Temme, Maika Takita, Markus Brink, Jerry M Chow, and Jay M Gambetta. 2017. Hardware-efficient variational quantum eigensolver for small molecules and quantum magnets. nature 549, 7671 (2017), 242–246.DOI
- [39]Uman Khalid, Muhammad Shohibul Ulum, Moe Z Win, and Hyundong Shin. 2024. Integrated satellite-ground variational quantum sensing networks. IEEE Communications Magazine 62, 10 (2024), 20–27.DOI
- [40]Zeqi Lai, Qian Wu, Hewu Li, Mingyang Lv, and Jianping Wu. 2021. Orbitcast: Exploiting mega-constellations for low-latency earth observation. In 2021 IEEE 29th International Conference on Network Protocols (ICNP). IEEE, 1–12.DOI
- [41]Chengjia Lei, Shaohua Wu, Yi Yang, Jiayin Xue, Dawei Chen, Pengfei Duan, and Qinyu Zhang. 2025. Joint Partitioning, Allocation, and Transmission Optimization for Federated Learning in Satellite Constellations via Multi-Task MARL. IEEE Transactions on Mobile Computing (2025).DOI
- [42]Jinting Li, Xi Chen, Danfang Zhang, Wenzhang Wang, Yang Zhou, Meng He, Jie Fang, Lin Zhou, Chuan He, Junjie Jiang, et al. 2025. Realization of a cold atom gyroscope in space. National Science Review 12, 4 (2025), nwa f012.
- [43]Yang Li, Wen-Qi Cai, Ji-Gang Ren, Chao-Ze Wang, Meng Yang, Liang Zhang, Hui-Ying Wu, Liang Chang, Jin-Cai Wu, Biao Jin, et al. 2025. Microsatellite-based real-time quantum key distribution. Nature 640, 8057 (2025), 47–54.DOI
- [44]Jiacheng Liu, Xiaozhi Zhu, Tongqiao Xu, Xiaofeng Hou, and Chao Li. 2025. {SpaceExit}: Enabling Efficient Adaptive Computing in Space with Early Exits. In 2025 USENIX Annual Technical Conference (USENIX ATC 25). 1343–1358.
- [45]Chao-Yang Lu, Yuan Cao, Cheng-Zhi Peng, and Jian-Wei Pan. 2022. Micius quantum experiments in space. Reviews of Modern Physics 94, 3 (2022), 035001.
- [46]Jason Ludmir, Nicholas S DiBrita, Jason Han, and Tirthak Patel. 2026. SpinTune: Improving the Reliability of Quantum Sensor Networks for Practical Quantum-Classical Utility. arXiv preprint arXiv:2605.04416 (2026).arxiv.org/abs/2605.04416
- [47]Jason Ludmir, Ian Martin, Nicholas S. DiBrita, and Tirthak Patel. 2026. QuFoundry: Generating Data with Quantum Properties for Quantum Machine Learning Utility. arXiv:2509.21622 [cs.ET] https://arxiv.org/abs/2509.21622
- [48]Jason Zev Ludmir, Sophia Rebello, Jacob Ruiz, and Tirthak Patel. 2025. Quorum: Zero-Training Unsupervised Anomaly Detection using Quantum Autoencoders. In 2025 62nd ACM/IEEE Design Automation Conference (DAC). IEEE, 1–7.arxiv.org/abs/2504.13113
- [49]Benjamin K Malia, Yufan Wu, Julián Martínez-Rincón, and Mark A Kasevich. 2022. Distributed quantum sensing with mode-entangled spin-squeezed atomic states. Nature 612, 7941 (2022), 661–665.DOI
- [50]Bho Matthiesen, Nasrin Razmi, Israel Leyva-Mayorga, Armin Dekorsy, and Petar Popovski. 2024. Federated Learning in Satellite Constellations. IEEE Network 38, 2 (2024), 232–239. doi:10.1109/MNET.132.2200504DOI
- [51]Ronald Maule, Nitish K Panigrahy, Naga Lakshmi Anipeddi, Prajit Dhara, Deirdre Kilbane, Md Zakir Hossain, Walter O Krawiec, Don Towsley, and Bing Wang. 2024. Fair and efficient scheduling strategies for satellite assisted quantum key distribution systems. In 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), Vol. 1. IEEE, 1788–1798.
- [52]Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. 2017. Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics. PMLR, 1273–1282.arxiv.org/abs/1602.05629
- [53]Srujan Meesala, David Lake, Steven Wood, Piero Chiappina, Changchun Zhong, Andrew D Beyer, Matthew D Shaw, Liang Jiang, and Oskar Painter. 2024. Quantum entanglement between optical and microwave photonic qubits. Physical Review X 14, 3 (2024), 031055.arxiv.org/abs/2312.13559
- [54]Srujan Meesala, Steven Wood, David Lake, Piero Chiappina, Changchun Zhong, Andrew D Beyer, Matthew D Shaw, Liang Jiang, and Oskar Painter. 2024. Non-classical microwave–optical photon pair generation with a chip-scale transducer. Nature Physics 20, 5 (2024), 871–877.arxiv.org/abs/2303.17684
- [55]Artur Miroszewski, Jakub Nalepa, Agata M Wijata, Jakub Mielczarek, Bertrand Le Saux, and Alessandro Sebastiani. 2026. Quo Vadis, Quantum Machine Learning?: Quantum kernel methods meet Earth observation. IEEE Geoscience and Remote Sensing Magazine (2026).DOI
- [56]Inder Monga, Erhan Saglamyurek, Ezra Kissel, Hartmut Haffner, and Wenji Wu. 2023. QUANT-NET: A testbed for quantum networking research over deployed fiber. In Proceedings of the 1st Workshop on Quantum Networks and Distributed Quantum Computing. 31–37.
- [57]Michael A Nielsen and Isaac L Chuang. 2010. Quantum computation and quantum information. Cambridge university press.
- [58]Nitish K Panigrahy, Prajit Dhara, Don Towsley, Saikat Guha, and Leandros Tassiulas. 2022. Optimal entanglement distribution using satellite based quantum networks. In IEEE INFOCOM 2022–IEEE Conference on Computer Communications Workshops (INFOCOM WKSHPS). IEEE, 1–6.arxiv.org/abs/2205.12354
- [59]Soohyun Park, Soyi Jung, and Joongheon Kim. 2024. Dynamic quantum federated learning for satellite-ground integrated systems using slimmable quantum neural networks. IEEE Access 12 (2024), 58239–58247.DOI
- [60]Soohyun Park, Hyunsoo Lee, Seok Bin Son, Soyi Jung, and Joongheon Kim. 2025. Quantum federated learning with pole-angle quantum local training and trainable measurement. Neural Networks 187 (2025), 107301.DOI
- [61]Luca Pezze, Augusto Smerzi, Markus K Oberthaler, Roman Schmied, and Philipp Treutlein. 2018. Quantum metrology with nonclassical states of atomic ensembles. Reviews of Modern Physics 90, 3 (2018), 035005.DOI
- [62]Tobias Pfandzelter, Nikita Bauer, Alexander Leis, Corentin Perdrizet, Felix Trautwein, Trever Schirmer, Osama Abboud, and David Bermbach. 2025. Trabant: A Serverless Architecture for Multi-Tenant Orbital Edge Computing. arXiv preprint arXiv:2504.08337 (2025).arxiv.org/abs/2504.08337
- [63]Tobias Pfandzelter and David Bermbach. 2024. Komet: A serverless platform for low-earth orbit edge services. In Proceedings of the 2024 ACM Symposium on Cloud Computing. 866–882.DOI
- [64]Atit Pokharel, Ratun Rahman, Shaba Shaon, Thomas Morris, and Dinh C Nguyen. 2025. Differentially private federated quantum learning via quantum noise. In 2025 IEEE International Conference on Quantum Computing and Engineering (QCE), Vol. 1. IEEE, 1559–1565.arxiv.org/abs/2508.20310
- [65]Matteo Pompili. 2021. Multi-Node Quantum Networks with Diamond Qubits. Ph. D. Dissertation. Delft University of Technology, Netherlands.DOI
- [66]Matteo Pompili, Sophie L N Hermans, Simon Baier, Hans K C Beukers, Peter C Humphreys, Raymond N Schouten, Raymond FL Vermeulen, Marijn J Tiggelman, Laura dos Santos Martins, Bas Dirkse, et al. 2021. Realization of a multinode quantum network of remote solid state qubits. Science 372, 6539 (2021), 259–264.DOI
- [67]Nasrin Razmi, B h o Matthiesen, Armin Dekorsy, and Petar Popovski. 2022. Ground-assisted federated learning in LEO satellite constellations. IEEE Wireless Communications Letters 11, 4 (2022), 717–721.DOI
- [68]Nasrin Razmi, Bho Matthiesen, Armin Dekorsy, and Petar Popovski. 2024. On-Board Federated Learning for Satellite Clusters With Inter-Satellite Links. IEEE Transactions on Communications 72, 6 (2024), 3408–3424. doi:10.1109/TCOMM.2024.3356429DOI
- [69]Ji-Gang Ren, Ping Xu, Hai-Lin Yong, Liang Zhang, Sheng-Kai Liao, Juan Yin, Wei-Yue Liu, Wen-Qi Cai, Meng Yang, Li Li, et al. 2017. Ground-to-satellite quantum teleportation. Nature 549, 7670 (2017), 70–73.DOI
- [70]Jonathan Romero, Jonathan P Olson, and Alan Aspuru-Guzik. 2017. Quantum autoencoders for efficient compression of quantum data. Quantum Science and Technology 2, 4 (2017), 045001.DOI
- [71]Bogdan Ruczacak, Krzysztof Kotowski, David Evans, and Jakub Nalepa. 2025. The OPS-SAT benchmark for detecting anomalies in satellite telemetry. Scientific Data 12, 1 (2025), 710.
- [72]Takaya Saito and Marc Rehmsmeier. 2015. The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets. PloS one 10, 3 (2015), e0118432.DOI
- [73]Sebastian Schmidl, Phillip Wenig, and Thorsten Papenbrock. 2022. Anomaly detection in time series: a comprehensive evaluation. Proc. VLDB Endow. 15, 9 (May 2022), 1779–1797. doi:10.14778/3538598. 3538602
- [74]Damian Schon, Prathwiraj Umesh, You-Wei Cheah, Se-Young Yu, Ezra Kissel, Raju Valivarthi, Erhan Saglamyurek, Lavanya Ramakrishnan, Wenji Wu, Alp Sipahigil, et al. 2024. The QUANT-NET testbed development and preliminary results. In 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), Vol. 1. IEEE, 1799–1808.DOI
- [75]Maria Schuld, Ville Bergholm, Christian Gogolin, Josh Izaac, and Nathan Killoran. 2019. Evaluating analytic gradients on quantum hardware. Physical Review A 99, 3 (2019), 032331.DOI
- [76]Si Shen, Chenzhi Yuan, Zichang Zhang, Hao Yu, Ruiming Zhang, Chuanrong Yang, Hao Li, Zhen Wang, You Wang, Guangwei Deng, et al. 2023. Hertz-rate metropolitan quantum teleportation. Light: Science & Applications 12, 1 (2023), 115.
- [77]James C Spall. 2002. Multivariate stochastic approximation using a simultaneous perturbation gradient approximation. IEEE transactions on automatic control 37, 3 (2002), 332–341.
- [78]Arian J Stolk, Kian L van der Enden, Marie-Christine Slater, Ingmar te Raa-Derckx, Pieter Botma, Joris Van Rantwijk, JJ Benjamin Biemond, Ronald AJ Hagen, Rodolf W Herfst, Wouter D Koek, et al. 2024. Metropolitan-scale heralded entanglement of solid-state qubits. Science advances 10, 44 (2024), eadp6442.
- [79]Bill Tao, Om Chabra, Ishani Janjeva, Jindranil Gupta, and Deepak Vasish t. 2024. Known knowns and unknowns: Near-realtime earth observation via query bifurcation in serval. In 21st USENIX Symposium on Networked Systems Design and Implementation (NSDI 24), 809–824.
- [80]Ki-Chang Tong, Yeong-Gyun Jung, Hyeon-Su Yu, Young-Hwan Youn, Myung-Sun Baek, Gyeong-Ho Lee, and Hyoung-Kyu Song. 2026. Hybrid quantum-classical neural networks for low earth orbit satellite communications in 6G. Scientific Reports (2026).
- [81]Phuc V Trinh, Alberto Carrasco-Casado, Hideki Takenaka, Mikio Fujiwara, Mitsuo Kitamura, Masahide Sasaki, and Morio Toyoshima. 2022. Statistical verifications and deep-learning predictions for satellite-to-ground quantum atmospheric channels. Communications Physics 5, 1 (2022), 225.DOI
- [82]Raju Valivarthi, Marcel li Grimau Puigibert, Qiang Zhou, Gabriel H Aguilar, Varun B Verma, Francesco Marsili, Matthew D Shaw, Sae Woo Nam, Daniel Oblak, and Wolfgang Tittel. 2016. Quantum teleportation across a metropolitan fibre network. Nature Photonics 10, 10 (2016), 676–680.DOI
- [83]David Vallado, Paul Crawford, Ricahrd Hujsak, and Thomas Sean Kelso. 2006. Revisiting spacetrack report# 3. In AIAA/AAS astrodynamics specialist conference and exhibit. 6753.DOI
- [84]George C Valley. 1980. Isoplanatic degradation of tilt correction and short-term imaging systems. Applied Optics 19, 4 (1980), 574–577.DOI
- [85]Deepak Vashist, Jayanth Shenoy, and Ranveer Chandra. 2021. L2D2: Low latency distributed downlink for LEO satellites. In Proceedings of the 2021 ACM SIGCOMM 2021 Conference. 151–164.
- [86]Stephanie Wehner, David Elkouss, and Ronald Hanson. 2018. Quantum internet: A vision for the road ahead. Science 362, 6412 (2018), eaam9288.
- [87]Sören Wengerowsky, Siddarth Koduru Joshi, Fabian Steinlechner, Julien R Zichi, Sergiy M Dobrovolskiy, Rene Van der Molen, Johannes W Los, Val Zwiller, Marijn AM Versteegh, Alberto Mura, et al. 2019. Entanglement distribution over a 96-km-long submarine optical fiber. Proceedings of the National Academy of Sciences 116, 14 (2019), 6684–6688.DOI
- [88]Marco Wiedmann, Marc Hölle, Maniraman Periyasamy, Nico Meyer, Christian Ufrecht, Daniel D Scherer, Axel Plinge, and Christopher Mutschler. 2023. An empirical comparison of optimizers for quantum machine learning with spsa-based gradients. In 2023 IEEE International Conference on Quantum Computing and Engineering (QCE), Vol. 1. IEEE, 450–456.
- [89]Albert Williams, Nitish K Panigrahy, Andrew McGregor, and Don Towsley. 2024. Scalable scheduling policies for quantum satellite networks. In 2024 IEEE International Conference on Quantum Computing and Engineering (QCE), Vol. 1. IEEE, 1760–1769.arxiv.org/abs/2405.09464
- [90]Jason R Zacket, Holger Ahlers, David C Aveline, Patrick Boegel, Sofia Botsi, Eric Charron, Ethan R Elliott, Naceur Gaaloul, Enno Giese, et al. 2024. Pathfinder experiments with atom interferometry in the Cold Atom Lab onboard the International Space Station. Nature Communications 15, 1 (2024), 6414.
- [91]William K Wootters and Wojciech H Zurek. 2009. The no-cloning theorem. Physics Today 62, 2 (2009), 76–77.
- [92]Shuhe Wu, Guzhi Bao, Jinxian Guo, Jun Chen, Wei Du, Minwei Shi, Peiyu Yang, Liqing Chen, and Weiping Zhang. 2023. Quantum magnetic gradiometer with entangled twin light beams. Science Advances 9, 15 (2023), eadg1760.DOI
- [93]Jonathan Wurtz, Alexei Bylinskii, Boris Braverman, Jesse Amato-Grill, Sergio H Cantu, Florian Huber, Alexander Lukin, Fangli Liu, Phillip Weinberg, John Long, et al. 2023. Aquila: Quera’s 256-qubit neutral-atom quantum computer. arXiv preprint arXiv:2306.11727 (2023).arxiv.org/abs/2306.11727
- [94]YA Yang, Maya Miklos, Yee Ming Tso, Stella Kraus, Joonseok Hur, and Jun Ye. 2025. Clock precision beyond the Standard Quantum Limit at 10-18 level. Physical Review Letters 135, 19 (2025), 193202.
- [95]Juan Yin, Yuan Cao, Yu-Huai Li, Sheng-Kai Liao, Liang Zhang, Ji-Gang Ren, Wen-Qi Cai, Wei-Yue Liu, Bo Li, Hui Dai, et al. 2017. Satellite-based entanglement distribution over 1200 kilometers. Science 356, 6343 (2017), 1140–1144.arxiv.org/abs/1707.01339
- [96]Qiyang Zhang, Xin Yuan, Ruolin Xing, Yiran Zhang, Zimu Zheng, Xiao Ma, Mengwei Xu, Schahram Dustdar, and Shangguang Wang. 2024. Resource-efficient in-orbit detection of earth objects. In IEEE INFOCOM 2024-IEEE Conference on Computer Communications. IEEE, 551–560.arxiv.org/abs/2402.01675












