Abstract In various applications of optical communications, such as quantum key distribution (QKD) systems, the task of polarization control in optical fiber arises. A polarization controller (PC) on the receiver side is used to solve this problem. In our work, we investigate two machine learning approaches to polarization control: supervised learning (SL) and reinforcement learning (RL). We generalised analytical solution of the polarization control problem using SL-approach. Supervised learning algorithm was trained on simulations and then was validated in experimental setup. We compared supervised approach with analytical solution. The advantage of RL approach based on the capability of approximating nonlinear effects related to the imperfections of detection process. For the effective operation of polarization control algorithms, knowledge of the parameters of the PC is required. Reinforcement learning agent also eliminates control hardware calibration procedures, since the agent recognize necessary system parameters based on the experience of interacting with the environment. RL-agent was firstly pretrained on simulations and then was fine-tuned on the practical QKD system. We demonstrated that RL-agent’s tuning provides the 10% quantum bit error rate (QBER) on our setup, and converges in 10 steps.
Abstract In various applications of optical communications, such as quantum key distribution (QKD) systems, the task of polarization control in optical fiber arises. A polarization controller (PC) on the receiver side is used to solve this problem. In our work, we investigate two machine learning approaches to polarization control: supervised learning (SL) and reinforcement learning (RL). We generalised analytical solution of the polarization control problem using SL-approach. Supervised learning algorithm was trained on simulations and then was validated in experimental setup. We compared supervised approach with analytical solution. The advantage of RL approach based on the capability of approximating nonlinear effects related to the imperfections of detection process. For the effective operation of polarization control algorithms, knowledge of the parameters of the PC is required. Reinforcement learning agent also eliminates control hardware calibration procedures, since the agent recognize necessary system parameters based on the experience of interacting with the environment. RL-agent was firstly pretrained on simulations and then was fine-tuned on the practical QKD system. We demonstrated that RL-agent’s tuning provides the 10% quantum bit error rate (QBER) on our setup, and converges in 10 steps.
In various applications of optical communications, such as quantum key distribution (QKD) systems, the task of polarization control in optical fiber arises. A polarization controller (PC) on the receiver side is used to solve this problem. We investigate two machine learning approaches to polarization control: supervised learning (SL) and reinforcement learning (RL). We generalized analytical solution of the polarization control problem using SL approach. Supervised learning algorithm was trained on simulations and then was validated in experimental setup. We compared supervised approach with analytical solution. The advantage of RL approach is based on the capability of approximating nonlinear effects related to the imperfections of detection process. For the effective operation of polarization control algorithms, knowledge of the parameters of the PC is required. Reinforcement learning agent also eliminates control hardware calibration procedures, since the agent recognizes necessary system parameters based on the experience of interacting with the environment. RL agent was firstly pretrained on simulations and then was fine-tuned on the practical QKD system. We demonstrated that RL agent’s tuning provides the 10% quantum bit error rate on our setup and converges in 10 steps.
Abstract Polarization drift in fiber and free-space optical links is a major factor in the dynamic increase of bit error rate in polarization-coded quantum key distribution (QKD) systems. A dynamic polarization compensation method applicable to both links is a challenge. Here we propose a universally applicable real-time polarization compensation method, that the Muller parameters of the optical links are first detected using a polarization detector, and then the optimal parameters of the controller are obtained by gradient descent algorithm. Simulation results indicate advantages over current methods, with fewer waveplates, faster speed, and wider applicability for various optical links. In equivalent experiments of both satellite and fiber optical links, the average polarization extinction ratio of 27.9 dB and 32.2 dB are respectively achieved. The successful implementation of our method will contribute to the real-time polarization design of fiber and free-space QKD systems, while also contributing to the design of laser-based polarization systems.
Quantum Key Distribution (QKD) offers theoretically unbreakable security by leveraging quantum mechanics. However, practical implementation is challenged by environmental vulnerabilities, noise, and hardware imperfections. Recently, Machine Learning (ML) has emerged as a powerful tool to address these limitations and enhance the real-world viability of QKD systems. In this survey, we review ML techniques applied to improve QKD security and performance across five applications. First, parameter optimization, covering signal calibration, polarization alignment, phase stabilization, modulation state tuning, and post-processing enhancements to maximize secure key generation and minimize error rates. Second, attack detection, where ML models identify and classify quantum threats such as photon-number-splitting and Trojan-horse attacks. Third, protocol selection, leveraging ML to dynamically choose QKD protocols based on operational conditions. Fourth, key performance prediction of core metrics such as Secret Key Rate (SKR) and Quantum Bit Error Rate (QBER). Finally, quantum network management, optimizing large-scale QKD deployments through intelligent routing, node management, and resource allocation. Performance improvements are evaluated using accuracy, reduced QBER, and increased SKR. While ML shows significant potential for finance, government, and defense applications, challenges remain in scalability, computational demands, and real-world testing. Ongoing work should focus on lightweight, generalizable models and standardized benchmarks for practical ML-enhanced QKD deployment.
Abstract The crucial task for polarization-encoding fiber QKD is to compensate polarization drift occurring in a quantum channel. To solve this problem, the receiver usually uses a polarization controller. For proper operation, this device must be efficiently managed in real-time. In this work, a gradient-descent-based algorithm is proposed to solve this problem. The algorithm was implemented and tested on a QRate commercial QKD fiber system, that utilizes BB84-protocol. Low and stable QBER has been obtained during a day of continuous operation.
Abstract The practical deployment of Continuous Variables Quantum Key Distribution (CV-QKD) systems benefits from existing optical fiber telecommunication infrastructures. However, optical fibers introduce random variations in the state of polarization, which degrades the system's performance.We consider a CV-QKD system featuring a polarization diversity heterodyne receiver and the constant modulus algorithm (CMA) to compensate for the polarization drifts in the quantum channel. Our setup can effectively realign Alice's quantum signal with Bob's local oscillator for polarization drift variances below 10−10 .This value is compatible with most experimental implementations, allowing for accurate estimation of the channel transmission and excess noise parameters. This validates the approximation of real-world implementations under polarization drift to an ideal scenario without polarization drift, ensuring that the security proof remains applicable for polarization drift variances below $ 10^{-10} $.Our results establish operational limits for passive polarization drift compensation using a polarization diversity receiver combined with digital CMA.This enables long-term stability in \linebreak CV-QKD systems, eliminating the need for active polarization controllers and manual adjustments.
Polarization drift due to propagation medium birefringence is a primary cause of increased quantum bit error ratio (QBER) in polarization-encoded quantum key distribution (QKD). The random nature of polarization drift and environmental instabilities cause it to vary over time, necessitating algorithms for effective polarization drift compensation in polarization-encoded optical fiber QKD links. I propose an algorithm based on processing time-multiplexed control pulses by unscented Kalman filter that minimizes the number of additional measurements that interrupt key exchange. I evaluated the robustness of the algorithm under various conditions, including random polarization drift velocities, and additional noise due to depolarization for weak-coherent BB84-like protocols, to determine the applicability limits of the proposed algorithm and its overhead.