Method and system for self-optimization of optical communication device parameters based on deep learning

Abstract

The present application provides a method and system for self-optimization of optical communication device parameters based on deep learning, which relates to the field of optical communication equipment management. The method includes: based on the node modulation model in the optical communication device, adaptively determining the corresponding device modulation parameters according to the environmental perception data; based on the central server, receiving the model parameters of the node modulation model and the corresponding environmental modulation data group and device communication indicators from each optical communication device; based on the central server, using the received data to perform federated training optimization on the global modulation model to update the model parameters of the global modulation model; based on the updated model parameters of the global modulation model, updating each node modulation model. As a result, not only the risk of modulation error or inadaptability to environmental changes of a single device is reduced, but also the various optical communication devices in the entire optical communication network can work together to achieve parameter optimization of the entire network.

Description

Technical Field

[0001] The present application relates to the technical field of optical communication equipment management, and in particular to a method and system for self-optimization of optical communication device parameters based on deep learning. Background Art

[0002] As the core of modern communication networks, optical communication technology has been widely used in data transmission around the world, relying on its advantages of high bandwidth, low latency and long-distance transmission. However, with the rapid growth of data volume and the increase in the complexity of communication networks, optical communication equipment faces more and more challenges in maintaining high-performance and stable operation. In particular, in order to ensure the quality of data transmission, the setting or optimization of various parameters in optical communication equipment needs to be precisely configured.

[0003] At present, the parameter setting method for optical communication devices is generally optimized based on the pre-set parameter thresholds in a relatively static network environment. However, the static setting method cannot flexibly match real-time environmental changes. For example, during optical fiber transmission, dynamic changes caused by factors such as temperature changes, signal attenuation, and optical fiber bending lack flexibility and adaptability. When facing different scenarios, it is unable to adaptively adjust the optimal parameter settings, resulting in a significant decrease in device performance.

[0004] In response to the above problems, the industry has not yet proposed a better technical solution. Summary of the invention

[0005] The present application provides a method, system, storage medium, computer program product and electronic device for self-optimization of optical communication device parameters based on deep learning, which is used to at least solve the problem that the static parameter setting method in the current related technology cannot support adaptive parameter adjustment in different scenarios and lacks flexibility.

[0006] In a first aspect, an embodiment of the present application provides a method for self-optimization of optical communication device parameters based on deep learning, comprising: based on a node modulation model in the optical communication device, adaptively determining corresponding device modulation parameters according to environmental perception data; the parameter types of the environmental perception data include signal-to-noise ratio, bit error rate, polarization mode dispersion, ambient temperature and ambient interference noise; the device modulation parameters include any one of the following: switching modulation mode, adjusting coding rate or adjusting modulation depth; based on a central server, receiving model parameters of a node modulation model and corresponding environmental modulation data groups and device communication indicators from each optical communication device; the environmental modulation data group includes multiple historical environmental perception data and corresponding historical device modulation parameters; the device communication indicators include long-term bit error rate, average signal-to-noise ratio, long-term throughput, average delay and accumulated jitter; based on the central server, using the received model parameters of the node modulation model of each optical communication device and the corresponding environmental modulation data groups and device communication indicators, the global modulation model is federated trained and optimized to update the model parameters of the global modulation model; based on the updated model parameters of the global modulation model, the corresponding node modulation models of each of the optical communication devices are updated.

[0007] In a second aspect, an embodiment of the present application provides a deep learning-based optical communication device parameter self-optimization system, including: an adaptive modulation unit, for adaptively determining corresponding device modulation parameters according to environmental perception data based on a node modulation model in the optical communication device; the parameter types of the environmental perception data include signal-to-noise ratio, bit error rate, polarization mode dispersion, ambient temperature and ambient interference noise; the device modulation parameters include any one of the following: switching modulation mode, adjusting coding rate or adjusting modulation depth; a data collection unit, for receiving model parameters of the node modulation model and corresponding environmental modulation data groups and device communication from each optical communication device based on a central server. indicators; the environmental modulation data group includes multiple historical environmental perception data and corresponding historical equipment modulation parameters; the equipment communication indicators include long-term bit error rate, average signal-to-noise ratio, long-term throughput, average delay and accumulated jitter; a global model updating unit is used to perform federal training optimization on the global modulation model based on the central server, using the model parameters of the node modulation models of each optical communication device received and the corresponding environmental modulation data group and equipment communication indicators to update the model parameters of the global modulation model; a node model updating unit is used to update the corresponding node modulation model of each optical communication device based on the updated model parameters of the global modulation model.

[0008] According to a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the method for self-optimization of optical communication device parameters based on deep learning of any embodiment of the present application.

[0009] In a fourth aspect, an embodiment of the present application provides a storage medium on which a computer program is stored, characterized in that when the program is executed by a processor, the steps of the method for self-optimization of parameters of an optical communication device based on deep learning of any embodiment of the present application are implemented.

[0010] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps of the method for self-optimization of optical communication device parameters based on deep learning of any embodiment of the present application.

[0011] The method and system for self-optimization of optical communication device parameters based on deep learning provided by the present application can produce at least the following technical effects:

[0012] (1) By introducing deep learning technology, environmental perception data is analyzed and predicted based on the node modulation model, and adaptive adjustment of device modulation parameters is achieved. Compared with the traditional static threshold parameter optimization method, it can dynamically respond to real-time environmental changes (such as temperature changes, signal attenuation, fiber bending, etc.), thereby dynamically and flexibly optimizing modulation parameters in different scenarios, ensuring that optical communication equipment maintains high-performance and stable operation in a changing environment.

[0013] (2) By collecting and utilizing equipment communication indicators such as long-term bit error rate, average signal-to-noise ratio, long-term throughput, average delay and accumulated jitter, the central server can continuously monitor and evaluate the communication quality of each optical communication device, and integrate the node modulation model parameters and environmental data of each optical communication device. Through federated training, the global modulation model is optimized and updated, so that the global modulation model is continuously updated and adapted to the actual application environment, thereby improving the communication quality and stability of optical communication equipment.

[0014] (3) In this technical solution, the node modulation model of each optical communication device is federated with the central server for training. By sharing model parameters and environmental modulation data sets, each device can learn and optimize the modulation strategy from a global perspective. As a result, the global modulation model can be better generalized to different devices and different environments, avoiding overfitting to the specific scenario of a single device and improving the applicability and robustness of the model in multiple environments.

[0015] Through this technical solution, the modulation model of each optical communication device is coordinated by a central server, so that the modulation parameters of each device can be optimized based on global information. This not only reduces the risk of modulation errors or inadaptability to environmental changes in a single device, but also enables the various optical communication devices in the entire optical communication network to work together to achieve full network parameter optimization, further improving the efficiency and reliability of the entire communication network. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1 A flowchart showing an example of a method for self-optimization of optical communication device parameters based on deep learning according to an embodiment of the present application is shown;

[0018] Figure 2 Shown according to Figure 1 An exemplary operation flow chart of step S120 in FIG.

[0019] Figure 3 An operation flow chart of an example of determining a session key according to an embodiment of the present application is shown;

[0020] Figure 4 An operation flow chart of an example of updating modulation models of various nodes based on model parameters of a global modulation model according to an embodiment of the present application is shown;

[0021] Figure 5 A structural block diagram of an example of a deep learning-based optical communication device parameter self-optimization system according to an embodiment of the present application is shown;

[0022] Figure 6 It is a schematic structural diagram of an embodiment of an electronic device of the present application. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0024] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved shall comply with the provisions of relevant laws and regulations and shall not violate public order and good morals.

[0025] Figure 1 A flowchart of an example of a method for self-optimization of optical communication device parameters based on deep learning according to an embodiment of the present application is shown.

[0026] Regarding the executor of the method of the embodiment of the present application, it can be any controller or processor with computing or processing capabilities. By introducing deep learning and federated training methods, not only the adaptive ability of the modulation parameters of the optical communication equipment is improved, but also the modulation model is continuously optimized from a global perspective through the central server, which significantly improves the communication quality and stability of the equipment, and ultimately realizes the coordinated optimization of optical communication equipment in the entire network, thereby improving the operating efficiency and reliability of the overall network.

[0027] In some examples, it may be integrated and configured in an electronic device, terminal or server through software, hardware or a combination of software and hardware, and the types of the terminal, electronic device or server may be diverse, such as a mobile phone, tablet computer or desktop computer, etc.

[0028] like Figure 1 As shown, in step S110, based on the node modulation model in the optical communication device, the corresponding device modulation parameters are adaptively determined according to the environmental perception data.

[0029] In some embodiments, a node modulation model is deployed in each optical communication device (e.g., optical transmitter) in the optical communication network. The node modulation model is an adaptive model based on deep learning, which can receive and process a variety of environmental perception data in real time. Specifically, the node modulation model will first obtain real-time input of environmental perception data, which includes multiple key parameters such as signal-to-noise ratio, bit error rate, polarization mode dispersion, ambient temperature, and environmental interference noise, identify the potential impact of current environmental changes on optical communication equipment, and predict and output corresponding device modulation parameters, that is, adaptively select the optimal device modulation parameters, which can be switching modulation mode (e.g., switching from QPSK (Quadrature Phase Shift Keying) to QAM (Quadrature Amplitude Modulation, orthogonal amplitude modulation)); or, adjusting the modulation depth, such as changing the size or shape of the modulation constellation; or, adjusting the coding rate to optimize the balance between transmission rate and bit error rate.

[0030] Therefore, by introducing a deep learning adaptive node modulation model, optical communication equipment can dynamically respond to environmental changes and adjust equipment parameters in real time to ensure optimal communication quality under changing external conditions.

[0031] In step S120, based on the central server, model parameters of the node modulation model and corresponding environmental modulation data groups and device communication indicators are received from each optical communication device.

[0032] Here, the environmental modulation data group includes multiple historical environmental perception data and corresponding historical device modulation parameters; the device communication indicators include long-term bit error rate, average signal-to-noise ratio, long-term throughput, average delay and accumulated jitter.

[0033] In some embodiments, the central server serves as the core node of global control and regularly receives the current model parameters of the node modulation model, the environmental modulation data set, and the equipment communication indicators from each optical communication device. This can reflect the modulation strategy and the modulation performance effect achieved by the equipment under different environmental conditions in different time periods, and can be used as an important reference for optimizing the global model.

[0034] In step S130, based on the central server, the global modulation model is federatedly trained and optimized using the model parameters of the node modulation models of each optical communication device received and the corresponding environmental modulation data group and device communication indicators to update the model parameters of the global modulation model.

[0035] Here, the central server uses the data collected from each optical communication device to perform federated training. Through distributed deep learning algorithms, the global model is optimized without collecting all the original data. Under the premise of protecting the privacy data of the device users of each optical communication device, the model parameters of each device can be integrated to update the parameters of the global modulation model. Through centralized data collection and processing, the central server can fully grasp the operating status and historical performance of each optical communication device, and then optimize the global modulation model, so that the optimized global modulation model can more comprehensively reflect the modulation strategy under different environmental conditions, thereby improving the adaptive adjustment capability of the entire network.

[0036] In step S140, the modulation model of the corresponding node of each optical communication device is updated based on the model parameters of the updated global modulation model.

[0037] Here, the federated learning framework is used to synchronously update the node modulation model parameters of the central server and each optical communication device without sharing data access rights, which can improve the performance of the global model while ensuring data privacy.

[0038] In some embodiments, the central server and each optical communication device maintain a deep learning model with the same structural type, and after the central server completes the optimization of the global modulation model, it sends the updated model parameters to each optical communication device, so that each optical communication device immediately updates the local node modulation model after receiving these parameters. Exemplarily, the update process may include replacing the old model parameters, loading new weights, etc. The updated node modulation model can be put into use immediately and continue to adaptively adjust the modulation parameters of the device, thereby realizing a rapid upgrade of the global optimal modulation strategy and improving its own ability to cope with complex communication environments.

[0039] Through the embodiments of the present application, through the federated learning of the central server and the distribution of global optimization parameters, each optical communication device can optimize the adaptive adjustment strategy through the updated global model, thereby improving the overall coordination and communication efficiency of the optical communication network, so that each device can use the latest modulation strategy while reducing network interference and performance bottlenecks caused by inconsistent parameters.

[0040] For the node modulation model or the global modulation model, it can adopt various types of deep learning models, such as MLP (Multilayer Perceptron) model, CNN (Convolutional Neural Network) or LSTM (Long Short-Term Memory), etc., to quickly respond to environmental changes and generate modulation parameters in real time, thereby ensuring the stable operation of optical communication equipment in a dynamic environment.

[0041] It should be noted that in optical communication systems, environmental perception data usually has significant time series characteristics, such as signal-to-noise ratio, bit error rate, polarization mode dispersion and other parameters that are dynamically adjusted over time. In view of this, in some examples of the embodiments of the present application, the node modulation model adopts a long short-term memory network based on a self-attention mechanism, which can capture the long-term dependencies in these data, and at the same time enhance the focus on key time steps through the self-attention mechanism, thereby generating more accurate device modulation parameters.

[0042] Specifically, the node modulation model includes a cascaded input layer, LSTM layer, self-attention layer and output layer. The input layer is used to receive time series input. The LSTM layer can effectively capture the long-term dependencies in the time series through its unique gating mechanism, and can effectively avoid the gradient vanishing problem in the traditional RNN (Recurrent Neural Network) when learning long time series. The output results of the LSTM layer are processed by the self-attention layer, and the self-attention mechanism is introduced. The importance of each time step is determined by calculating the attention weight of each time step, and the output of the LSTM is weighted and summed according to these weights, which enables the model to automatically identify and focus on the time step that is most critical to the generation of device modulation parameters. The self-attention mechanism can dynamically adjust the weight of the time step under different communication environments, so that the model can respond to the changes in the real-time environment more accurately, while reducing the performance fluctuations caused by environmental changes. The output layer can use a fully connected layer, and the feature representation weighted by the self-attention mechanism can be input into the fully connected layer to generate the corresponding device modulation parameters. Through the embodiments of the present application, the node modulation model adopts LSTM based on the self-attention mechanism. The combination of LSTM and the self-attention mechanism enables the model to adaptively adjust the modulation parameters when facing a complex and changeable communication environment, thereby improving the communication quality and efficiency of the system.

[0043] In some examples of the embodiments of the present application, the node modulation model adopts a comprehensive loss function that integrates prediction error loss, attention smoothing loss, and output change penalty loss to optimize from multi-objective prediction.

[0044] More specifically, the loss function of the node modulation model is expressed as follows:

[0045] , formula (1);

[0046] , formula (2);

[0047] , formula (3);

[0048] , formula (4);

[0049] , formula (5);

[0050] In the formula, represents the loss function of the node modulation model; represents the prediction error loss term, represents the attention smoothing loss term, represents the penalty loss term for modulation parameter changes, , and are the adjustment coefficients of the corresponding loss items respectively; is the total number of samples in the data sample set, For the samples at time step The actual modulation parameters, For the samples at time step The predicted modulation parameters, represents the Huber loss function, represents the Huber loss threshold; Indicates The total number of time steps for each sample, For the samples at time step The attention weight of Indicates samples at time step The predicted modulation parameters.

[0051] Here, the node modulation model can be trained and optimized using a data sample set, where each sample Corresponding to a time series data, these data are in time step There is a corresponding real modulation parameter and predict modulation parameters The Huber loss will be applied at each time step The error between the predicted value and the true value is then aggregated over all samples and time steps. The attention smoothing loss should also consider each time step in the time series. In addition, the modulation parameter change penalty loss calculates the modulation parameter change between consecutive time steps for each sample.

[0052] Through the embodiment of the present application, in the prediction error loss term The use of Huber loss function can effectively balance the accuracy and robustness of the model. Specifically, when dealing with small errors, Huber loss is expressed as mean square error (MSE), which ensures the prediction accuracy of the model; when dealing with large errors, it is expressed as absolute value error (MAE), which enhances the robustness of the model in the face of noise and outliers. As a result, the model can cope with various interferences and data anomalies in the optical communication environment and maintain a high prediction accuracy.

[0053] Attention-based smoothing loss By smoothing the attention weights, the model is prevented from over-relying on the data of a single time step, thereby enhancing the robustness of the model when processing long time series data, ensuring that the model can comprehensively consider the important information in the entire time series, rather than just relying on the data of certain prominent time steps, thereby improving the decision quality and robustness of the model.

[0054] It should be noted that in optical communication devices, frequent and drastic changes in modulation parameters may cause fluctuations in communication quality or even cause communication interruption. Penalty loss based on modulation parameter changes , by suppressing the drastic changes of modulation parameters between consecutive time steps, the temporal continuity and smoothness of the modulation parameters output by the model are guaranteed, so that the model can output smoother modulation parameter optimization information, thereby maintaining the efficiency and reliability of the communication system.

[0055] In addition, the adjustment coefficients of each loss term in the comprehensive loss function ( , and ) can be reasonably set according to business needs, so that the contribution of each loss term can be balanced during model training, thereby accelerating convergence and avoiding excessive influence of a certain loss term during training.

[0056] Through the embodiment of the present application, the three loss terms of prediction error, attention smoothing and modulation parameter change are combined together to form a comprehensive loss function It provides a comprehensive optimization strategy, which effectively ensures the outstanding performance of the model in multiple dimensions such as accuracy, stability, and adaptability, thereby improving the overall parameter self-optimization capability of the optical communication device.

[0057] against Figure 1 Implementation details of step S120 in some implementations, by combining symmetric encryption with asymmetric encryption, the model parameters of the node modulation model and the corresponding environmental modulation data group and device communication indicators are encrypted and transmitted in segments.

[0058] Figure 2 Shown according to Figure 1 An example operation flow chart of step S120 in FIG.

[0059] like Figure 2 As shown, in step S210, based on the optical communication device, the data to be transmitted is divided into multiple data segments, and the data to be transmitted includes model parameters of the node modulation model and the corresponding environmental modulation data group and device communication indicators.

[0060] In some embodiments, the data to be transmitted (i.e., model parameters of the node modulation model and the corresponding environmental modulation data group and device communication indicators) can be segmented based on a predetermined data block size. In addition, the data can be segmented according to different types or priorities of the data content. For example, key model parameters can be placed in a separate data segment to ensure their priority transmission.

[0061] In step S220, based on the optical communication device, each data segment is encrypted using a unique corresponding session key to obtain a corresponding encrypted data segment.

[0062] Here, each session key is determined by asymmetric encryption of a unique AES key generated by the AES (Advanced Encryption Standard) algorithm and the public key of the central server. The AES algorithm is a symmetric encryption algorithm that is widely used to protect the security of sensitive information. It has high encryption capabilities, efficient computing speed and good stability.

[0063] In some implementations, a unique AES key is first generated for each data segment, and then each AES key is encrypted using an asymmetric encryption algorithm (such as an RSA algorithm) using the public key of the central server. Specifically, the public key and private key of the central server are generated using the RSA algorithm, which is an asymmetric encryption algorithm. As a result, even if the session key is intercepted during transmission, it cannot be decrypted or abused due to the lack of the private key of the central server, thereby further improving the security of data transmission and preventing criminals from stealing the model parameters of the adaptive modulation model by decoding and deciphering the transmitted data, which helps optical communication operators to continue to maintain their leading technological advantages.

[0064] In step S230, based on the optical communication device, each encrypted data segment is transmitted in parallel to the central server through multiple communication channels.

[0065] Here, the communication channel can be dynamically adjusted based on the channel status. In order to improve transmission efficiency, the optical communication device will use multiple channels for data transmission at the same time. These channels can be different physical paths or spectrum resources. In addition, during the transmission process, the optical communication device will monitor the channel status in real time and dynamically adjust the transmission path of the data segment according to the channel quality to ensure that the data segment can be transmitted to the central server at the fastest speed, thereby improving the transmission efficiency of the optical communication network.

[0066] In addition, by dividing the data into multiple data segments and using multiple communication channels for parallel transmission, the data transmission paths can be effectively dispersed, reducing the risks that may be caused by single-path transmission, and increasing the robustness and reliability of data transmission.

[0067] In step S240, based on the central server, each encrypted data segment is decrypted and combined using a private key to receive the model parameters of the node modulation model and the corresponding environmental modulation data group and device communication indicators.

[0068] In some embodiments, after the central server receives each encrypted data segment from the optical communication device, it first decrypts the encrypted session key using the private key of the RSA algorithm to obtain the AES key corresponding to each data segment. Subsequently, each encrypted data segment is decrypted using these AES keys to restore the original unencrypted data segment. After the data decryption is completed, the central server performs a hash data integrity check and reassembles all data segments into a complete original data set to ensure that the model parameters of the node modulation model, the environmental modulation data group, and the device communication indicators can be restored without loss.

[0069] In the embodiments of the present application, the efficiency and security of data transmission in the optical communication network are significantly improved through data segmentation, encryption, parallel transmission, decryption and reorganization. By combining symmetric encryption with asymmetric encryption, the key data of the optical communication device is encrypted and transmitted in segments. The session key generated by the AES algorithm is used to encrypt the data segment, and the session key itself is encrypted using the public key of the central server through the RSA algorithm. Based on the double encryption mechanism, the core confidential data is effectively prevented from being intercepted and cracked during transmission, ensuring the confidentiality and integrity of the data. In addition, even if the encrypted data segment is intercepted by a third party, the meaningful original data cannot be restored without decryption with the private key, which effectively guarantees the security of the system.

[0070] Figure 3 An operational flow chart of an example of determining a session key according to an embodiment of the present application is shown.

[0071] like Figure 3 As shown, in step S310, an initial key seed corresponding to the data segment is randomly generated based on a random number generator.

[0072] In an optical communication device, the encryption security of a data segment depends on the generated session key, and the generation process of the session key begins with a randomly generated initial key seed. First, the optical communication device uses a high-quality random number generator (RNG) to randomly generate an initial key seed. The random number generator should meet strong randomness and unpredictability to ensure that the generated key seed has sufficient entropy, thereby providing security for subsequent encryption. Exemplarily, the random number generator can use a hardware random number generator (HRNG) or a pseudo random number generator (PRNG).

[0073] In step S320, the initial key seed is adjusted based on the environmental perception data detected in real time to generate a target key seed.

[0074] Here, before generating the session key, the current environmental perception data (such as signal-to-noise ratio, bit error rate, ambient temperature, etc.) is introduced into the key generation process to dynamically adjust the initial key seed. For example, the environmental perception data is used as input, and the initial key seed is corrected through certain mathematical operations (such as weighted average, linear transformation or nonlinear mapping) to generate the target key seed. In this way, the key seed generation process can be adaptively adjusted according to the current environmental changes, making the generated key seeds more diverse, thereby further improving the key security.

[0075] In step S330, the target key seed is processed based on the AES algorithm to determine a corresponding initial session key.

[0076] In some implementations, the target key seed is used as input to the AES algorithm to generate an initial session key through a series of encryption operations including round functions, byte substitution, row shift, column obfuscation, and round key addition.

[0077] In step S340, the initial session key is asymmetrically encrypted using the public key of the central server to determine the corresponding session key.

[0078] In some implementations, an asymmetric encryption algorithm RSA algorithm is used for encryption processing, which has high security and computational efficiency. Through the asymmetric encryption process, it is ensured that only the central server holding the corresponding private key can decrypt the session key, which greatly reduces the risk of key or data leakage and ensures the security of link data transmission.

[0079] Through the embodiments of the present application, environmental perception data is introduced to dynamically adjust the initial key seed, so that the generated target key seed can reflect the current communication environment. This dynamic adaptability makes each generated session key closely related to the current environmental conditions, increases the unpredictability and uniqueness of the key, and significantly improves the security of the key. In addition, even if the attacker has mastered some historical data, it is difficult to predict or reproduce the new session key, thereby effectively resisting replay attacks and other common cryptographic attacks. In addition, the optical communication device is first generated by the AES algorithm, and then asymmetric encryption is performed using the public key of the central server to generate a session key, ensuring the security of the session key during transmission. Even if the key is intercepted during transmission, it cannot be decrypted and used, which greatly improves the data transmission security between the optical communication device and the central server, and especially ensures the confidentiality of the transmitted model parameters.

[0080] Regarding the implementation details of updating the model parameters of the global modulation model through federated training optimization, in some examples of the embodiments of the present application, the model parameters of the received modulation models of each node are weighted averaged by using a federated averaging algorithm to update the model parameters of the global modulation model.

[0081] More specifically, the basic communication indicators of each optical communication device are standardized so that they can be effectively compared in the weight calculation, for example, by using the minimum-maximum standardization method. Then, the comprehensive performance score of each optical communication device is calculated, which is calculated by the weighted average method. Then, in order to make the newer data have a greater impact on the weight, the time decay factor is introduced. The time decay factor is a function of time, which usually adopts an exponential decay model. Combined with the time decay factor, the final weight of each optical communication device is calculated. Finally, the model parameters of each optical communication device are weighted summed to obtain the model parameters of the corresponding global modulation model.

[0082] In this way, by introducing the time decay factor, the global model can adapt to the dynamically changing communication environment, and newer data has a greater impact on the model, ensuring that the global model can respond quickly to environmental changes and improve real-time performance.

[0083] Specifically, the model parameters of the global modulation model may be updated by the following formula:

[0084] , formula (6);

[0085] , formula (7);

[0086] , formula (8);

[0087] , formula (9);

[0088] In the formula, represents the model parameters of the global modulation model, For the The weight of the optical communication device, For the The model parameters uploaded by the optical communication device, is the total number of optical communication devices in the optical communication network, Optical communication device Overall performance rating of represents the weight coefficient of the bit error rate, Optical communication device Standardized bit error rate of represents the weight coefficient of the signal-to-noise ratio, Optical communication device The normalized signal-to-noise ratio; represents the weight coefficient of throughput, Optical communication device Normalized throughput of represents the weight coefficient of delay, Optical communication device The normalized delay of represents the weight coefficient of jitter, Optical communication device The normalized jitter of represents the time attenuation factor of the optical communication device, Optical communication device The data timestamp, represents the decay rate coefficient; is the normalization factor, Optical communication device The overall performance score of Optical communication device The data timestamp.

[0089] Here, by introducing the time decay factor, the newer data has a greater impact on the model update. This mechanism ensures that the global modulation model can quickly respond to changes in the environment and device status, thereby improving the real-time performance of the model. In a rapidly changing communication environment, the model can adjust the modulation parameters in a timely manner, optimize the communication quality, reduce the bit error rate and delay, and improve the real-time performance of the overall system.

[0090] The calculation of the comprehensive performance score combines multiple device communication indicators (such as bit error rate, signal-to-noise ratio, throughput, latency, jitter) and is standardized to fairly reflect the overall performance of each device. Through these weighted indicators, the system is able to identify and prioritize devices with excellent performance, thereby reducing the adverse effects of noisy data or abnormal devices on the global model. As a result, the global model exhibits higher robustness and reliability in the face of various communication environments.

[0091] Based on the above The weight calculation mechanism of each optical communication device is used to reasonably evaluate and utilize the uploaded data. Devices with higher performance and newer data are given higher weights, thus playing a greater role in the global model update. This type of data utilization not only improves the efficiency of model updates, but also avoids the negative impact of low-quality data on model updates, ensuring that the update process of the global model is always based on high-quality and highly relevant data.

[0092] Through the embodiment of the present application, the model parameters uploaded by each optical communication device are weighted averaged to generate the parameters of the global modulation model. , using the weighted average method to score the comprehensive performance of each optical communication device and the time decay factor Calculating weights , ensuring that the model update process can better reflect the actual performance and data freshness of each device. By assigning more weight to devices with high performance and the latest data (for example, device modulation parameters generated in nearby time), the global model can adapt to the current communication environment more accurately and stably.

[0093] It should be noted that for the attenuation rate coefficient in the above formula (8) , which is a key parameter that affects how the time decay factor decreases over time. In one example of the embodiment of the present application, it can be manually set according to the system business requirements. For example, if the data needs to retain a high influence in the system for a long time, The value of should be set smaller and decay slower. On the contrary, if the data is only important in the short term, It should be set larger to decay faster.

[0094] In some examples of the embodiments of the present application, The value of can also be adaptively adjusted according to actual conditions through intelligent algorithms to ensure that the system can adaptively optimize the data attenuation rate in different operating environments.

[0095] In some embodiments, the system can monitor and analyze the rate of change of input data in real time, for example, the signal-to-noise ratio, bit error rate, or other key performance indicators of the monitoring data. If the system detects that the input data has changed significantly in a short period of time, it means that the current environment is more dynamic and the system should respond to the latest data more quickly, so Should be enlarged.

[0096] Specifically, the decay rate coefficient It is adaptively adjusted by the following formula:

[0097] , formula (10);

[0098] , formula (11);

[0099] , formula (12);

[0100] In the formula, Indicates that at time step Aggregated environmental perception data, Indicates An optical communication device at time step Environmental perception data; and They represent the aggregated environment perception data at time step and The value of Indicates the size of the sliding window; Sense the rate of change of data for the aggregation environment; is the base decay rate coefficient, is the attenuation adjustment coefficient.

[0101] Here, the calculation of the change rate can be achieved through the sliding window method. The sliding window is a time window of fixed length. The system performs statistical analysis on the changes in the data within the window. In addition, according to the data change rate The size of the system is dynamically adjusted The greater the rate of change, The bigger it is, the smaller it is.

[0102] In the embodiment of the present application, the environmental sensing data of each communication device in the optical communication network is aggregated and the change rate is dynamically calculated. , the system can automatically adjust the attenuation rate coefficient according to changes in the overall environment Therefore, through the adaptive adjustment mechanism, the global modulation model can respond to the fluctuations of the external environment in a timely manner, quickly adjust the parameters when the environment changes drastically, and ensure the stability and reliability of communication quality.

[0103] In addition, in a complex and changing communication environment, dynamic real-time calculation and adjustment The value of enables the system to respond more quickly to current communication conditions. For example, when the environment changes suddenly (such as an increase in noise levels, a decrease in signal quality, etc.), the system can quickly reduce its reliance on old data and focus on new input data, thereby adjusting communication parameters more quickly, ensuring the stability and reliability of the system during long-term operation, reducing the risk of communication interruption or performance degradation, and ensuring that the system is always in the best operating state.

[0104] Figure 4 An operational flow chart of an example of updating modulation models of various nodes based on model parameters of a global modulation model according to an embodiment of the present application is shown.

[0105] like Figure 4 As shown, it involves data interaction between the central server 100 and the optical communication device 200. Here, the central server 100 sends the model parameters of the updated global modulation model to each optical communication device 200, so that these devices update their local node modulation models using the new model parameters. In addition, in order to optimize the transmission efficiency and ensure the synchronization of the model, the incremental update and differential synchronization mechanism is adopted.

[0106] Specifically, the central server 100 performs the following operations:

[0107] In step S410, a model version number is generated based on the model parameters of the updated global modulation model.

[0108] In some embodiments, when the central server completes the update of the global modulation model, a unique model version number is generated for the update. The version number is usually generated by a timestamp or an incremental serial number to ensure that each model update can be uniquely identified. After the version number is generated, the central server records the version number together with the updated model parameters to perform version control in the subsequent parameter distribution process, such as version rollback or rapid configuration of new nodes in the optical communication network. In addition, the generation of the model version number can also be used to distinguish different versions of modulation models, so that the optical communication device can identify whether the model version currently used has expired or whether a version update is required.

[0109] In step S420, model parameter difference information between model parameters corresponding to each optical communication device is calculated respectively according to the updated model parameters of the global modulation model.

[0110] Specifically, after the global model is updated, the central server needs to calculate the difference between the old and new global model parameters and transmit only the difference parts to each optical communication device. The difference part represents the part that has changed in the global model, and only the updated part needs to be transmitted, thereby reducing the amount of data transmitted. In some embodiments, the model parameter difference information generally includes relevant model parameters such as model weights or biases that need to be adjusted. Through differentiated calculation, the central server only needs to send down the parameter part that has changed, rather than the entire model parameter set, which can greatly reduce the amount of data transmission and improve transmission efficiency.

[0111] In step S430, AES encryption is performed on each model parameter difference information to generate corresponding encrypted difference parameters.

[0112] In some embodiments, in order to ensure the security of the transmission process, the central server needs to symmetrically encrypt the difference parameters using the AES algorithm, and then transmit the encrypted difference parameters to the corresponding optical communication device. By encrypting the difference parameter information using the AES algorithm, the confidentiality and integrity of the data during the transmission process are ensured.

[0113] In step S440, each encrypted difference parameter and model version number is sent to a corresponding optical communication device.

[0114] Here, after the difference parameter information is encrypted, the central server sends each encrypted data packet together with the newly generated model version number to the corresponding optical communication device. In addition, the data packet is transmitted through a secure communication channel to ensure that the data will not be tampered with or lost during transmission.

[0115] The optical communication device 200 performs the following operations:

[0116] In step S450, when a difference is detected between the local model version number of the optical communication device and the received model version number, the encrypted difference parameter is decrypted using the AES algorithm to restore the corresponding model parameter difference information.

[0117] In some embodiments, after receiving the encrypted difference parameters and model version number sent by the central server, each optical communication device will first check whether the locally stored model version number is consistent with the received version number. If it is detected that the local version number is consistent with the received version number, it indicates that the node modulation model is already the latest version; if it is inconsistent, it indicates that the node modulation model of the device needs to be updated. At this time, the received encrypted difference parameters are decrypted by the AES algorithm to restore the original difference parameter information.

[0118] In step S460, the model parameters of the corresponding node modulation model are updated according to the restored model parameter difference information, and the local model version number is updated according to the received model version number.

[0119] In some embodiments, after successfully decrypting and recovering the difference parameter information, the optical communication device will use this information to update the model parameters of the local node modulation model, such as modifying the part of the model parameters corresponding to the difference parameter information, to ensure the accuracy and efficiency of the update. In addition, after the update is completed, the optical communication device updates its locally stored model version number to the latest version number received. Therefore, in order to ensure that the local model of each optical communication device can be synchronized to the latest version, the central server and the optical communication device need to maintain the version number of the model parameters. Each time the model is updated, the version number is superimposed accordingly, and the parameter update information is synchronized to each device with an out-of-sync version through a differential update mechanism.

[0120] Through the embodiments of the present application, only the difference parts of the global model parameters are transmitted through the differential synchronization mechanism and the incremental update mechanism, which significantly reduces the burden of data transmission. Especially in large-scale optical communication networks, it can improve the transmission efficiency of the overall system and ensure that the local model of each optical communication device is synchronized to the latest version as soon as possible under the premise of minimizing network resource consumption, thereby reducing the performance loss caused by synchronization lag. In addition, through incremental updates and version control, optical communication devices can gradually update to the latest model version even when they are out of sync or in poor network conditions, reducing the impact of single point failures on the overall performance of the system and enhancing the robustness of the system.

[0121] It should be noted that in the incremental update mechanism, the data transmitted is the difference of the global model parameters, not the complete model parameters. Therefore, even if the attacker successfully intercepts and cracks the model parameter difference information symmetrically encrypted by the AES algorithm, what he can get is only the parameter difference information relative to the old model, not the complete model parameters, which ensures the overall security of the model and can effectively maintain the market advantage of the deep learning-based adaptive parameter modulation technology of optical communication service operators.

[0122] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of actions combined, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the present application. In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0123] Figure 5 A structural block diagram of an example of a deep learning-based self-optimization system for optical communication device parameters according to an embodiment of the present application is shown.

[0124] like Figure 5 As shown, the optical communication device parameter self-optimization system 500 based on deep learning includes an adaptive modulation unit 510, a data collection unit 520, a global model updating unit 530 and a node model updating unit 540.

[0125] The adaptive modulation unit 510 is used to adaptively determine the corresponding device modulation parameters according to the environmental perception data based on the node modulation model in the optical communication device; the parameter types of the environmental perception data include signal-to-noise ratio, bit error rate, polarization mode dispersion, ambient temperature and ambient interference noise; the device modulation parameters include any one of the following: switching modulation mode, adjusting coding rate or adjusting modulation depth.

[0126] The data collection unit 520 is used to receive the model parameters of the node modulation model and the corresponding environmental modulation data group and equipment communication indicators from each optical communication device based on the central server; the environmental modulation data group includes multiple historical environmental perception data and corresponding historical equipment modulation parameters; the equipment communication indicators include long-term bit error rate, average signal-to-noise ratio, long-term throughput, average delay and accumulated jitter.

[0127] The global model updating unit 530 is used to perform federal training optimization on the global modulation model based on the central server, using the model parameters of the node modulation models of each optical communication device received and the corresponding environmental modulation data group and device communication indicators, so as to update the model parameters of the global modulation model.

[0128] The node model updating unit 540 is used to update the corresponding node modulation model of each of the optical communication devices based on the model parameters of the updated global modulation model.

[0129] In some embodiments, an embodiment of the present application provides a non-volatile computer-readable storage medium, in which one or more programs including execution instructions are stored, and the execution instructions can be read and executed by an electronic device (including but not limited to a computer, a server, or a network device, etc.) to execute the steps of any of the above-mentioned methods for self-optimization of optical communication device parameters based on deep learning in the present application.

[0130] In some embodiments, the embodiments of the present application also provide a computer program product, which includes a computer program stored on a non-volatile computer-readable storage medium, and the computer program includes program instructions. When the program instructions are executed by a computer, the computer performs any step of the above-mentioned method for self-optimization of optical communication device parameters based on deep learning.

[0131] In some embodiments, the embodiments of the present application also provide an electronic device, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the steps of the method for self-optimization of optical communication device parameters based on deep learning.

[0132] Figure 6 is a schematic diagram of the hardware structure of an electronic device for executing a method for self-optimizing parameters of an optical communication device based on deep learning provided by another embodiment of the present application, such as Figure 6 As shown, the device includes:

[0133] One or more processors 610 and memory 620, Figure 6 A processor 610 is taken as an example.

[0134] The apparatus for executing the method for self-optimization of optical communication device parameters based on deep learning may further include: an input device 630 and an output device 640 .

[0135] The processor 610, the memory 620, the input device 630 and the output device 640 may be connected via a bus or other means. Figure 6 The example of connecting through bus is taken in the following.

[0136] The memory 620, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer executable programs and modules, such as program instructions / modules corresponding to the optical communication device parameter self-optimization method based on deep learning in the embodiment of the present application. The processor 610 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 620, that is, the optical communication device parameter self-optimization method based on deep learning in the above method embodiment is implemented.

[0137] The memory 620 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 620 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the memory 620 may optionally include a memory remotely arranged relative to the processor 610, and these remote memories may be connected to the electronic device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0138] The input device 630 can receive input digital or character information and generate signals related to user settings and function control of the electronic device. The output device 640 can include a display device such as a display screen.

[0139] The one or more modules are stored in the memory 620, and when executed by the one or more processors 610, perform the optical communication device parameter self-optimization method based on deep learning in any of the above method embodiments.

[0140] The above-mentioned product can execute the method provided in the embodiment of the present application, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not fully described in this embodiment, please refer to the method provided in the embodiment of the present application.

[0141] The electronic devices of the embodiments of the present application exist in various forms, including but not limited to:

[0142] (1) Mobile communication equipment: This type of equipment is characterized by having mobile communication functions and its main purpose is to provide voice and data communications. This type of terminal includes: smart phones, multimedia phones, functional phones, and low-end phones.

[0143] (2) Ultra-mobile personal computer devices: These devices belong to the category of personal computers, have computing and processing functions, and generally also have mobile Internet access features. These terminals include: PDA, MID and UMPC devices, etc.

[0144] (3) Portable entertainment devices: These devices can display and play multimedia content. They include audio and video players, handheld game consoles, e-books, smart toys, and portable car navigation devices.

[0145] (4) Other onboard electronic devices with data interaction functions, such as on-board devices installed in vehicles.

[0146] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0147] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the relevant technology, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0148] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for self-optimization of optical communication device parameters based on deep learning, comprising: Based on the node modulation model in the optical communication device, the corresponding device modulation parameters are adaptively determined according to the environmental perception data; The parameter types of the environmental perception data include signal-to-noise ratio, bit error rate, polarization mode dispersion, ambient temperature and ambient interference noise; the device modulation parameters include any one of the following: switching modulation mode, adjusting coding rate or adjusting modulation depth; Based on the central server, model parameters of the node modulation model and corresponding environmental modulation data groups and device communication indicators are received from each optical communication device; the environmental modulation data group includes a plurality of historical environmental perception data and corresponding historical device modulation parameters; the device communication indicators include long-term bit error rate, average signal-to-noise ratio, long-term throughput, average delay and accumulated jitter; Based on the central server, using the received model parameters of the node modulation models of each optical communication device and the corresponding environmental modulation data group and device communication indicators, the global modulation model is federatedly trained and optimized to update the model parameters of the global modulation model; Based on the updated model parameters of the global modulation model, updating the modulation model of the corresponding node of each of the optical communication devices; The method of using the received model parameters of the node modulation models of each optical communication device and the corresponding environmental modulation data group and device communication index to perform federated training optimization on the global modulation model to update the model parameters of the global modulation model includes: The model parameters of the received modulation models of each node are weighted averaged using a federated averaging algorithm to update the model parameters of the global modulation model: In the formula, represents the model parameters of the global modulation model, For the The weight of the optical communication device, For the The model parameters uploaded by the optical communication device, is the total number of optical communication devices in the optical communication network, Optical communication device Overall performance rating of represents the weight coefficient of the bit error rate, Optical communication device Standardized bit error rate of represents the weight coefficient of the signal-to-noise ratio, Optical communication device The normalized signal-to-noise ratio; represents the weight coefficient of throughput, Optical communication device Normalized throughput of represents the weight coefficient of delay, Optical communication device The normalized delay of represents the weight coefficient of jitter, Optical communication device The normalized jitter of Optical communication device The time decay factor, Optical communication device The data timestamp, represents the decay rate coefficient; is the normalization factor, Optical communication device The overall performance score of Optical communication device The data timestamp.

2. The method according to claim 1, wherein: The method of receiving the model parameters of the node modulation model and the corresponding environmental modulation data group and the device communication index from each optical communication device based on the central server includes: The model parameters of the node modulation model and the corresponding environmental modulation data group and device communication indicators are encrypted and transmitted in segments by combining symmetric encryption and asymmetric encryption, including: Based on the optical communication device, the data to be transmitted is divided into a plurality of data segments; the data to be transmitted includes model parameters of the node modulation model and the corresponding environmental modulation data group and device communication indicators; Based on the optical communication device, each of the data segments is encrypted using a unique corresponding session key to obtain a corresponding encrypted data segment; each of the session keys is determined by asymmetric encryption of a unique AES key generated by an AES algorithm and a public key of a central server; Based on an optical communication device, each of the encrypted data segments is transmitted in parallel to the central server through multiple communication channels; Based on the central server, each encrypted data segment is decrypted and combined using a private key to receive the model parameters of the node modulation model and the corresponding environmental modulation data group and device communication indicators; the public key and private key of the central server are generated by the RSA algorithm.

3. The method according to claim 2, wherein: The session key is determined by: Based on a random number generator, randomly generate an initial key seed corresponding to the data segment; Adjusting the initial key seed based on the real-time detected environmental perception data to generate a target key seed; Processing the target key seed based on an AES algorithm to determine a corresponding initial session key; The initial session key is asymmetrically encrypted using the public key of the central server to determine the corresponding session key.

4. The method according to claim 1, wherein: The decay rate coefficient is adaptively adjusted by the following formula: In the formula, Indicates that at time step Aggregated environmental perception data, Indicates An optical communication device at time step Environmental perception data; and They represent the aggregated environment perception data at time step and The value of Indicates the size of the sliding window; Sense the rate of change of data for the aggregation environment; is the base decay rate coefficient, is the attenuation adjustment coefficient.

5. The method according to any one of claims 1 to 4, wherein: The updating of the modulation model of the corresponding node of each optical communication device based on the updated model parameters of the global modulation model comprises: Based on the central server, perform the following operations: generating a model version number based on the model parameters of the updated global modulation model; According to the updated model parameters of the global modulation model, respectively calculating model parameter difference information between the model parameters corresponding to each optical communication device; Performing AES encryption on each of the model parameter difference information to generate corresponding encrypted difference parameters; sending each of the encrypted difference parameters and the model version number to a corresponding optical communication device; Based on the optical communication device, the following operations are performed: In the case of detecting that there is a difference between the local model version number of the optical communication device and the received model version number, decrypting the encrypted difference parameter using an AES algorithm to restore corresponding model parameter difference information; The model parameters of the corresponding node modulation model are updated according to the restored model parameter difference information, and the local model version number is updated according to the received model version number.

6. The method according to claim 5, wherein: The node modulation model adopts a long short-term memory network based on a self-attention mechanism; The loss function of the node modulation model is expressed by the following formula: In the formula, represents the loss function of the node modulation model; represents the prediction error loss term, represents the attention smoothing loss term, represents the penalty loss term for modulation parameter changes, , and are the adjustment coefficients of the corresponding loss items respectively; is the total number of samples in the data sample set, For the The true modulation parameters of samples at time step, For the samples at time step The predicted modulation parameters, represents the Huber loss function, represents the Huber loss threshold; Indicates The total number of time steps for each sample, For the The attention weight of samples at time step; Indicates samples at time step The predicted modulation parameters.

7. A deep learning-based optical communication device parameter self-optimization system, comprising: An adaptive modulation unit, used to adaptively determine corresponding device modulation parameters according to environmental perception data based on a node modulation model in the optical communication device; The parameter types of the environmental perception data include signal-to-noise ratio, bit error rate, polarization mode dispersion, ambient temperature and ambient interference noise; the device modulation parameters include any one of the following: switching modulation mode, adjusting coding rate or adjusting modulation depth; A data collection unit, for receiving, based on a central server, model parameters of a node modulation model and a corresponding environmental modulation data group and device communication indicators from each optical communication device; the environmental modulation data group includes a plurality of historical environmental perception data and corresponding historical device modulation parameters; the device communication indicators include a long-term bit error rate, an average signal-to-noise ratio, a long-term throughput, an average delay and an accumulated jitter; A global model updating unit, configured to perform federated training optimization on a global modulation model based on the central server and using the received model parameters of the node modulation models of each optical communication device and the corresponding environmental modulation data group and device communication index, so as to update the model parameters of the global modulation model; A node model updating unit, configured to update the corresponding node modulation model of each of the optical communication devices based on the model parameters of the updated global modulation model; The method of using the received model parameters of the node modulation models of each optical communication device and the corresponding environmental modulation data group and device communication index to perform federated training optimization on the global modulation model to update the model parameters of the global modulation model includes: The model parameters of the received modulation models of each node are weighted averaged using a federated averaging algorithm to update the model parameters of the global modulation model: In the formula, represents the model parameters of the global modulation model, For the The weight of the optical communication device, For the The model parameters uploaded by the optical communication device, is the total number of optical communication devices in the optical communication network, Optical communication device Overall performance rating of represents the weight coefficient of the bit error rate, Optical communication device Standardized bit error rate of represents the weight coefficient of the signal-to-noise ratio, Optical communication device The normalized signal-to-noise ratio; represents the weight coefficient of throughput, Optical communication device Normalized throughput of represents the weight coefficient of delay, Optical communication device The normalized delay of represents the weight coefficient of jitter, Optical communication device The normalized jitter of Optical communication device The time decay factor, Optical communication device The data timestamp, represents the decay rate coefficient; is the normalization factor, Optical communication device The overall performance score of Optical communication device The data timestamp.

Images