A federated quantum feature-sketching pipeline for traffic-stability gating using Pauli expectations and NGSIM trajectories, SUMO predictive gating with sensing noise, and IBM Quantum hardware is presented.
Abstract
Roadside units (RSUs) supporting connected and autonomous vehicle corridors need compact models to decide when cooperative maneuvers should be rewarded, deferred, or disabled. Raw sensor streams and neural network weight checkpoints are poorly suited to bandwidth-limited, privacy-sensitive roadside learning. This paper presents $\texttt{FedQML-Edge}$, a federated quantum feature-sketching pipeline for traffic-stability gating. Each RSU constructs a traffic-state summary and sends circuit inputs to a quantum computer; Pauli expectations form a nonlinear sketch processed by a logistic classifier. Only classifier updates are shared with an aggregator, whose head supports reward gating. Raw observations, vehicle records, event traces, and quantum sketches remain private. We evaluate the method using NGSIM trajectories, SUMO predictive gating with sensing noise, and IBM Quantum hardware. On NGSIM, the Pauli sketch reduces test log loss by $14.4\%$ relative to the strongest matched classical sketch. On SUMO, it approaches larger MLPs in stable-window recall while using $7-28$ times less communication per round.
The rapid growth of sensor-equipped edge devices calls for machine learning paradigms that are both lightweight and privacy-preserving. Federated learning (FL) enables distributed AI on edge data but still demands heavy forward and backward propagation on resource-constrained sensors. Split learning (SL) offloads computation to a server, yet its intermediate activations can leak private information. Homomorphic encryption closes this gap at the cost of prohibitive computation and bloated communication. We propose HFL-Lite, a hierarchical federated learning framework that achieves practical privacy without cryptographic primitives. Each sensor encodes its raw data into a compact sparse feature code via a lightweight sensor encoder. Only this code, which is insufficient to reconstruct the original data, leaves the device. An edge server synthesizes surrogate training samples from the uploaded codes, trains a teacher model, and distills knowledge back to the sensors. Because the sensors perform only one-way uploads, backward propagation, gradient transmission, and plaintext activation leakage are all eliminated. Experiments on the UCI HAR dataset show that HFL-Lite reduces per-round communication to 0.5 KB and sensor-side latency to 9 ms while delivering competitive accuracy.
Cangming Liang, Kuan Ching Li, Zulong Diao et al.· Tsinghua Science and Technol...· 0 citations
Federated learning enables multiple organizations to collaboratively analyze data while retaining local control over their datasets, making it attractive for applications such as healthcare. Distributed quantum computing provides a natural framework for such workflows, allowing geographically separated quantum processors to execute joint computations using shared entanglement while preserving data privacy. In this work, we investigate the feasibility of satellite-enabled distributed quantum computing for federated quantum learning. As a representative application, we consider a distributed distance-based quantum classifier in which multiple parties contribute local data through quantum operations. To support this application, we develop a hybrid space-ground quantum network architecture in which satellites distribute entanglement between distant ground stations. The communication layer is combined with a noise-aware neutral-atom processor model, enabling a system-level analysis that captures both network and hardware constraints. Simulation results across varying network sizes, feature dimensions, and coherence regimes show that classifier performance is jointly determined by communication resources, processor noise, and data representation. In low-coherence regimes, decoherence destroys the classifier’s discriminative signal, whereas high-coherence regimes reveal limitations arising from feature-space conditioning and feature redundancy. These results demonstrate that satellite quantum networks could support distributed quantum learning over long distances, while highlighting the importance of coherence time, entanglement-distribution performance, and learning-aware data encoding for future large-scale deployments.
J. Boschero, Rares Adrian Oancea, L. Mazzarella et al.· Entropy· 0 citations
Applications in industrial and smart-infrastructure Wireless sensor networks (WSNs) in the field are increasingly expected to employ predictive intelligence, which, even when operating under dynamic configurations, heterogeneity of hardware, and strong privacy requirements. The proposed paper is DynaQuAI, quantum-inspired, edge-intelligent framework that can predict real-time failures on constrained sensor nodes regarding their resources. The proposed system employs a probabilistic state encoding and oscillatory exploration schedule—mathematically inspired by quantum superposition analogies—to improve reinforcement learning exploration in sparse and noisy environments. These techniques are entirely classical algorithms implemented with standard trigonometric and probabilistic operations, requiring no quantum hardware. Such quantum-inspired algorithms enhance coverage of state-space and convergence faster than when using the usual deep RL models, with no computational demand. To support distributed nodes collaborative learning, DynaQuAI uses a lightweight federated approach on training paired with parameter aggregation through secure masks. Privacy is ensured through secure aggregation via pairwise parameter masking, which prevents the aggregation server from observing individual node updates while recovering the correct aggregate. The framework is additionally compatible with local differential privacy (ε_total ≤ 9.8 over 150 rounds at δ = 4 × 10⁻⁶) for deployments requiring formal statistical privacy guarantees. We evaluated DynaQuAI through a simulation-based testbed emulating 500 heterogeneous sensor nodes with ARM Cortex-A7/M4 computational profiles over a simulated IEEE 802.15.4 mesh network. The evaluation uses the Kaggle Predictive Maintenance Dataset augmented with synthetic fault injections to create realistic distributed learning scenarios. Through experiments, this is, which has a sensitive aspect, demonstrated at 33% more fault-prediction accuracy over federated deep-learning baselines, 25% reduced energy consumption over adaptive quantum-inspired exploration, and 40% reduced convergence during policy learning. Its system can maintain real-time inference latency of 85 ms with minimally more than 5% of the available computational resources, thus suitable to be deployed in the long term. Individual contributions of quantum-inspired encoding, exploration scheduling and privacy-enhanced system of aggregation mechanisms are verified through ablation studies. In general, DynaQuAI has made easily extendable, privacy conscious and robust architectural learning offerings to next generation IoT ecosystems where high reliability and sustainability are essential.
We present Q-OPSEC, an adaptive middleware that uses supervised, unsupervised and reinforcement learning to select cryptographic strategies from classical, post-quantum and quantum-assisted (QKD) options. The selection is modeled as a multi-objective MDP that balances security, latency, computational and energy cost, and compliance. A negotiator and registry enforce hard constraints, handle endpoint compatibility and fallbacks, and store empirical cost profiles. Experiments in simulated smart environments and hardware-in-the-loop tests show high success rates ($>95 \%$) and context-aware adaptation. Limitations include reliance on simulated QKD channels, limited device profiling, empirically tuned hyperparameters, and evaluation in high-performance environments, which may not reflect IoT constraints; future work targets real QKD integration, broader benchmarking, robust RL methods, federated learning and explainability.
Darlan Noetzold, J. L. V. Barbosa, Juan F. de Paz et al.· International Conference on...· 0 citations
Federated fine-tuning is bottlenecked by communication: FedAvg and pseudo-gradient schemes transmit a payload that scales with the model, and gradient compression shrinks it by only a constant factor. We take a different lever. Mapping networks generate a network's weights from a small trainable latent through a frozen affine projection; because the map is shared and affine, averaging latents is exactly averaging the generated weights. We turn this into a practical low-bandwidth federated channel with two changes: a low-rank, seed-regenerable factorisation of the projection (cutting generator memory from ~80 GB to ~10 MB), and a delta formulation $\theta = \theta^{\mathrm{pre}} + U V^{\top} z$ that learns an additive correction around a shared centrally-pretrained base -- federated fine-tuning, which is what makes the method work at scale. A frozen orthogonal classifier head further removes the head from the payload while improving accuracy. On CIFAR-100 with ResNet-18+GroupNorm, our method (FLITE, Federated Low-rank Iterative Training Engine) communicates 1,280 floats (~5 KB) per client per round -- an 8718x reduction -- and reaches 74.67%, within ~0.5 pp of full-weight FedAvg. The averaging identity holds to floating-point precision ($6 \times 10^{-8}$); the method sits one to two orders of magnitude below PowerSGD and top-k on the bandwidth-accuracy Pareto; it matches or exceeds full-weight FedAvg under strong non-IID skew. int4 latents reach 648 bytes per round at unchanged accuracy, whereas int4 full-weight FedAvg collapses to chance.
A green quantized FSL (GQ-FSL) framework that incorporates stochastic quantization for both local collaborative training and wireless transmissions and enables large-scale DNN deployment on resource-constrained devices, achieving superior energy efficiency compared to quantized federated learning and full-precision FSL.