To address the issue of goods position estimation in medium and large intelligent warehouse systems, this paper proposes an indoor visible light positioning model based on contrastive learning and received signal strength (RSS), namely the contrastive learning transformer (CLTf) model. First, hundreds of optical power values are filtered to select the highest-intensity light-emitting diode (LED) signals for constructing the optical power vector. Then, the Transformer model is employed to capture long-sequence dependencies, while contrastive learning techniques are integrated to mine anchor point prior knowledge for feature representation optimization. The simulation results show that in a medium to large warehouse space of 50 m×20 m×3 m, the average positioning errors of the CLTf model on the 1st to 5th shelves are 0.292, 0.344, 0.375, 1.133, and 2.471 cm, respectively, with a positioning accuracy of centimeter level, significantly better than traditional methods.
To improve the global optimization capability of submarine cable routing planning algorithm, reduce cumulative costs and risks, and improve algorithm efficiency, a machine learning assisted (MLA) multi-objective optimization algorithm for submarine cable routing planning is proposed. Utilizing the advantages of reinforcement learning, MLA autonomously iterates learning, synchronously optimizes costs and risks, considers parameters such as seabed topography and water depth, and adopts Pareto frontier as the convergence evaluation criterion. It is compared and verified with traditional ant colony optimization (ACO) algorithm. The experimental results show that under the same risk level, the algorithm can reduce the laying cost by 27.45%, and its optimal solution cumulative risk is only 25% of the ACO algorithm, and the convergence speed is improved by more than 330 times. In addition, most of its Pareto solutions are located at the forefront, which is significantly better than the discrete distribution of the ACO algorithm solution set.
To address the nonlinear channel issues caused by atmospheric turbulence and scattering effects in free space optical (FSO) communication systems and enhance the transmission reliability of optical communication systems in the sixth generation mobile communication (6G) era, this paper reviews the research progress of machine learning (ML) in improving FSO channel estimation. It compares and analyzes the applications of deep learning and non-deep learning methods in FSO channel estimation, demonstrating the advantages of ML in enhancing estimation accuracy and system performance. Finally, the challenges and future development trends of ML in FSO communication are discussed, highlighting the significant potential of ML algorithms in FSO systems and exploring future research directions.
A multiple input multiple output (MIMO) neural network equalization algorithm based on channel attention mechanism (MIMO-NNE-CAM) is proposed to address the problem of mode crosstalk in mode division multiplexing optical transmission systems. This algorithm introduces a channel attention mechanism to focus the neural network on more important channel features, achieving effective signal equalization. In order to verify the performance of the algorithm, a the third mock examination mode division multiplexing system is built on the VPI Transmission simulation platform for testing. The experimental results show that, under the condition of a bit error rate (BER) of 1×10-3, the MIMO-NNE-CAM algorithm achieves performance gains of 1.3 dB and 3.1 dB compared to the original MIMO-NNE and least mean square (LMS) algorithms, respectively. Moreover, it maintains stable bit error rate performance even under strong coupling conditions, demonstrating faster convergence speed and enhanced anti-coupling capability.
To address the issues of artificial intelligence (AI) model bias and insufficient feature learning caused by imbalanced optical network datasets, this paper proposes a fault identification algorithm for optical line terminal (OLT) equipment based on deep cross network (DCN) and multi-task learning (MTL). First, potential faults are assessed using standardized mean Manhattan distance, with high-similarity samples labeled as poor-quality data. Subsequently, a DCN-MTL model is constructed, incorporating high-order feature interactions to enhance learning capability, while utilizing poor-quality detection as an auxiliary task to optimize the training of the primary fault detection task. Experimental results demonstrate that, compared to traditional deep neural networks, the proposed algorithm achieves improvements of 1.15% in accuracy, 11.83% in recall, 6.39% in F1-score, and 5.91% in area under the curve (AUC) under the same data volume, with all metrics exceeding 0.95. This validates the algorithm's strong detection capability in scenarios with scarce fault data.
To improve the design efficiency and performance of fiber amplifiers, this paper systematically investigates the application of artificial intelligence algorithms in fiber amplifier design, focusing on the roles of metaheuristic algorithms and neural networks in addressing three core challenges: inverse design, forward modeling, and dynamic control. The research demonstrates that metaheuristic algorithms (including genetic algorithms, particle swarm optimization, and simulated annealing) effectively optimize multidimensional parameters such as fiber length and pump configurations through natural evolution or swarm intelligence simulations. Neural networks, with their superior nonlinear modeling capabilities, enable efficient solutions for gain spectrum prediction, quality of transmission (QoT) estimation, and pulse evolution simulation, showing significant computational speed advantages over conventional numerical methods. Furthermore, the integration of metaheuristic algorithms with neural network technologies achieves adaptive real-time control in optical networks, successfully addressing the dynamic bandwidth requirements of emerging applications like video streaming and cloud computing. Finally, prospects for the application of artificial intelligence algorithms in fiber amplifiers are discussed.
The imaging quality of optical systems is affected by aberrations caused by deviations between the actual optical path and the ideal optical path. Zernike polynomials, as an effective tool for describing aberrations, can describe and analyze the characteristics of optical aberrations. This paper reviews the research progress at home and abroad, and focuses on analyzing the mathematical expression forms of standard Zernike polynomials, Zernike circular polynomials, and Zernike annular polynomials, clarifying their corresponding relationships with typical aberrations such as spherical aberration, coma, and astigmatism. Research shows that Zernike circular polynomials can efficiently represent the distribution of axisymmetric aberrations through the characteristics of orthogonal basis functions in polar coordinates, while annular polynomials are suitable for describing off-axis aberrations.
To further enhance the sensitivity of photonic crystal fiber (PCF), a novel PCF structure incorporating both circular and rectangular air holes is designed. The proposed configuration comprises a four-layer cladding with circular air holes and a core region embedded with five rectangular air holes. Utilizing the full-vector finite element method with perfectly matched layer (PML) boundary conditions, the optical properties of the PCF are systematically simulated and analyzed when infiltrated with three analytes: water, ethanol, and benzene. The results demonstrate that at the wavelength of 1.55 m, the designed PCF achieves sensitivities of 71.8%, 74.5%, and 75.6% for the three analytes, representing improvements of 1.26~7.97 times, 1.24~6.2 times, and 1.2~5.6 times, respectively, compared to existing PCF. Furthermore, the structure exhibits high birefringence on the order of 10-3 and ultralow confinement loss below 10-5 dB/cm.
To address the low classification and recognition rate of radio frequency fingerprints in internet of things (IoT) terminal devices, this paper proposes an improved convolutional neural network-gated recurrent unit (CNN-GRU) method based on comb filtering for radio frequency signal "gene" classification and recognition. First, the time-frequency characteristics of RF signals are enhanced using a comb filter to construct a unique "gene map" for each device. Second, the traditional one-dimensional CNN is expanded into a three-layer two-dimensional structure, combined with a dual-layer GRU to achieve joint time-frequency feature extraction and sequence modeling. Finally, hybrid pooling and exponential linear unit (ELU) activation functions are introduced to optimize feature representation. Experimental results show that the proposed method achieves a identification accuracy of 100% in simulated data and 95.52% in real-world data, outperforming traditional algorithms by 5%-22%, significantly enhancing the security and manage-ment efficiency of IoT devices.
To address the orbital angular momentum (OAM) recognition problem with imbalanced label distribution, a weighted extreme learning machine (WELM) recognition model based on the particle swarm optimization (PSO) algorithm is proposed. This model jointly optimizes the input weights and biases of WELM using PSO algorithm, enhancing the stability and robustness of WELM. A comparative analysis was conducted on the performance of the PSO-WELM model against support vector machine (SVM), deep learning (DL), and backpropagation artificial neural network (BP-ANN) models. Experimental results show that the PSO-WELM model can correctly identify minority-class and majority-class OAM beams under weak turbulence intensity. Under moderate turbulence intensity, all evaluation metrics of the PSO-WELM model outperform those of the comparative methods, demonstrating the feasibility and effectiveness of the model in recognizing imbalanced OAM beams.
To address the survivability requirements of service transmission in elastic optical network (EON) and the spectrum fragmentation caused by dynamic allocation and release of spectrum resources, a time-frequency fragmentation-aware survivable multi-path resource allocation (TFFA-SMRA) algorithm is proposed. This algorithm employs multi-path transmission technology to ensure service survivability and calculates link weights by integrating link length and spectrum resource status to identify candidate routes, thereby alleviating service blocking issues caused by link resource bottlenecks. Additionally, the algorithm introduces a fragmentation-aware spectrum allocation mechanism based on a time-frequency domain matching metric. Simulation results show that under a specific network topology with a traffic load of 400 Erlang, compared to the load balancing and time-frequency (LB-TF) algorithm, the TFFA-SMRA algorithm reduces the service blocking rate by 15.7% while improving spectrum utilization by 8.52%.
To reduce the dimensionality constraints of neural network decoders for polar codes during the training phase, a partitioned successive cancellation (SC) decoder based on fully connected neural networks (FCNN) is designed. By dividing the polar code decoding tree into two regions and processing each with differently parameterized FCNNs, the need for large-scale training data is reduced. The simulation results show that in an additive white Gaussian noise (AWGN) channel, when the signal-to-noise ratio (SNR) is between 1 to 5 dB, the performance of the FCNN-SC decoder approaches that of the SC decoding algorithm. When the SNR is between 1.5 to 3 dB, the FCNN-SC decoder achieves approximately 0.5 dB coding gain compared to the FCNN decoder, and requires a smaller dataset during the training phase, being roughly half the size needed for the FCNN decoder.
To address the issues of excessive path lengths and spectrum resource wastage in virtual optical network embedding (VONE), a fragmentation-aware virtual optical network cooperative embedding (FA-VONE) algorithm is proposed. This algorithm enhances the success rate of virtual node embedding through a node ranking strategy and adopts a cooperative node and link embedding approach to reduce optical path hops and spectrum consumption. During the link embedding phase, an optical path resource evaluation strategy is employed to improve the success rate of virtual link embedding. In the spectrum allocation phase, a spectrum fragmentation metric is designed to optimize spectrum utilization. Simulation results on NSFNET and Indian Network topologies demonstrate that FA-VONE excels in virtual request acceptance rate, spectrum utilization, and substrate network revenue.
To address the trade-off between high throughput and low latency in scheduling algorithms for fiber channel switches, this paper proposes a time-triggered queue proportional sampling (lp-QPS) algorithm. The algorithm introduces a long-queue prioritization mechanism to preferentially match virtual output queue (VOQ) exceeding the threshold, while employing queue proportional sampling for cyclic scheduling of the remaining queues. Comparative experiments with other iterative scheduling algorithms demonstrate that the lp-QPS algorithm achieves 100% throughput and superior latency performance under four traffic models, including both Bernoulli and burst arrival processes. These findings provide valuable insights for optimizing the performance of fiber channel switches.
To address the issues of traditional forward design for structural colors, such as limited optimization parameters, time-consuming computations, and non-tunable static structural colors, an inverse design method for structural colors of metasurfaces based on the Bayesian optimization algorithm is proposed. By introducing tunable phase-change materials to design nanoantennas and combining the Bayesian optimization algorithm with the finite-difference time-domain method, the structural color parameters of the metasurface are simulated and optimized. The designed structure utilizes Mie resonances in reflection mode to generate structural colors, while reversible color tuning is achieved through phase transitions of the phase-change material. Simulation results demonstrate that the proposed structural color device exhibits dynamic tunability of metasurface colors, with color differences of 63.30, 69.30, and 54.21 achieved at wavelengths of 450 nm, 545 nm, and 660 nm, respectively, along with angle-sensitive characteristics.
To address the dual requirements of gain and stability in high-power fiber laser system, a high-gain and high-stability fiber amplifier is designed. The optical path employs single-fiber double-pass dual-stage amplification technology combined with an all-polarization-maintaining fiber structure to improve gain medium utilization. Additionally, a circuit system based on automatic power control and closed-loop feedback is developed, achieving dynamic regulation through a peak sampling circuit and precision pump driving circuit. Experimental results show that the system achieves a maximum gain of 28.25 dB, with closed-loop control reducing output energy fluctuations from 2.21% to 0.42%, improving stability by 81%. Moreover, it reduces the required number of pump lasers while delivering a pulse energy output of 70.2 nJ at a pump current of 260 mA.
To improve the accuracy and robustness of modulation format identification (MFI) in elastic optical network (EON), this paper proposes an MFI method based on Stokes space and a Stacking model. The method extracts one-dimensional probability distribution features of the three axes in Stokes space using kernel density estimation to construct a 240-dimensional feature vector. A genetic algorithm is employed to optimize the combination of base models and meta-models in the Stacking model, while Bayesian optimization is used for hyperparameter tuning, enhancing classification performance under low signal-to-noise ratios. Simulation results show that, within an optical signal-to-noise ratio (OSNR) range of 5~30 dB, the model achieves a macro-average area under the receiver operating characteristic curve (AUC) of 1. The identification accuracy exceeds 98.5% for modulation formats such as polarization-division multiplexing binary phase-shift keying (PDM-BPSK) and polarization-division multiplexing quadrature phase-shift keying (PDM-QPSK), with an average accuracy improvement of 2.05%~5.63% compared to benchmark models like XGBoost and TabNet. Additionally, 100% identification precision is achieved at an OSNR of 18 dB.
To explore the critical role and technological trends of optical communication networks in smart city development, this study systematically analyzes the application scenarios and performance of three key technologies: free-space optical communication (FSO), integrated sensing and communication (ISAC) optical networks, and visible light communication (VLC). Furthermore, based on the latest advancements in novel optical fibers, optoelectronic device manufacturing, and quantum communication technologies, future development directions are proposed. The analysis reveals that FSO offers high-speed transmission advantages for cross-domain connectivity and emergency communications but suffers from weather-dependent instability. ISAC-enabled optical networks facilitate real-time monitoring in scenarios such as traffic and seismic activity through fiber-optic sensing, yet exhibit high false alarm rates in complex environments. VLC achieves centimeter-level precision and interference-free transmission in indoor positioning and medical applications but faces challenges in hardware cost. Looking ahead, hybrid hollow-core and multi-core fiber architectures, co-packaged optics (CPO), and quantum key distribution (QKD) are expected to drive smart city advanc ements in network infrastructure, energy-efficient device interconnection, and security enhancement, respectively.