Satellite communication data transmission method and device

Abstract

The invention discloses a satellite communication data transmission optimization method, comprising: collecting data generated in a satellite communication process to form a satellite communication data set, providing basic data support for model training and evaluation; preprocessing the satellite communication data set to form a satellite communication training data set, providing high-quality data for subsequent model training; improving a Transformer model to form an improved Transformer model; using the satellite communication training data set to train the improved Transformer model, adopting an improved Adam optimizer and an adaptive learning rate adjustment strategy to train the improved Transformer model; evaluating the improved Transformer model through indicators, and continuously optimizing the model structure and parameters according to the evaluation results; integrating the trained improved Transformer model into a satellite communication system, verifying its performance through a real-time data transmission experiment, and further optimizing it according to actual application requirements.

Description

Technical Field

[0001] The present invention relates to the field of satellite communication technology, and in particular to a satellite communication data transmission method and device. Background Art

[0002] Satellite communication systems are widely used in global communications, navigation, remote sensing, and military fields. With the increase in data transmission demand, how to efficiently and reliably transmit large amounts of data has become a major challenge facing satellite communication systems. Traditional data transmission methods are difficult to meet the needs of modern communications due to bandwidth limitations, delays, signal interference, and other issues. Therefore, studying efficient data transmission optimization methods has important practical significance.

[0003] In recent years, deep learning technology has made significant progress in various fields. In particular, the Transformer model has been widely used in natural language processing, computer vision and other fields due to its powerful parallel processing capabilities and self-attention mechanism. Based on this, the present invention proposes a satellite communication data transmission optimization method based on an improved Transformer, which improves the efficiency and reliability of data transmission by improving the Transformer model. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the purpose of the present invention is to propose a satellite communication data transmission optimization method based on an improved Transformer, which improves the efficiency and reliability of satellite communication data transmission and reduces the delay and packet loss rate of data transmission by improving the structure and training method of the Transformer model.

[0005] A satellite communication data transmission optimization method proposed by the present invention comprises the following steps:

[0006] Collect data generated during satellite communications to form satellite communication data sets, providing basic data support for model training and evaluation;

[0007] Preprocessing the satellite communication data set to form a satellite communication training data set to provide high-quality data for subsequent model training;

[0008] Improve the Transformer model to form an improved Transformer model;

[0009] Using the satellite communication training data set to train the improved Transformer model, adopting an improved Adam optimizer and an adaptive learning rate adjustment strategy to train the improved Transformer model;

[0010] Evaluate the improved Transformer model through indicators, and continuously optimize the model structure and parameters according to the evaluation results;

[0011] The trained improved Transformer model is integrated into the satellite communication system, its performance is verified through real-time data transmission experiments, and further optimized according to actual application requirements.

[0012] Furthermore, the satellite communication data set is preprocessed, specifically: data standardization, normalization, noise removal, and data set division.

[0013] Furthermore, improvements are made on the basis of the Transformer model to form an improved Transformer model, specifically: on the basis of the Transformer model, the self-attention mechanism is improved, the position encoding is optimized, and the integrated learning framework is formed.

[0014] Furthermore, the improved Transformer model is trained using the satellite communication training data set, specifically: by introducing regularization technology and data enhancement technology, the performance of the model on the training set and the validation set needs to be monitored during the training process to prevent the model from overfitting and improve the generalization ability of the model.

[0015] Furthermore, the indicators include: any one or a combination of transmission efficiency, delay time, packet loss rate, and bit error rate.

[0016] According to the satellite communication data transmission optimization method of the embodiment of the present invention, the satellite communication data transmission is optimized by improving the Transformer model, which has the following beneficial effects:

[0017] Improve data transmission efficiency: Through the improved self-attention mechanism and position encoding method, the model's ability to process long sequence data is improved, thereby improving the efficiency of data transmission. Reduce data transmission delay: The use of an integrated learning framework and a dynamic weight adjustment method enables the model to transmit data more quickly, reducing data transmission delays. Enhance transmission reliability: By introducing regularization technology and data enhancement methods, the generalization ability and robustness of the model are improved, and the packet loss rate and bit error rate during data transmission are reduced. Adapt to a variety of application scenarios: The method of the present invention has strong versatility and scalability, can adapt to different types of satellite communication data transmission needs, and has broad application prospects.

[0018] In view of the above problems, the present invention further proposes a satellite communication data transmission optimization device, which comprises:

[0019] The data set collection module is used to collect data generated during satellite communication to form a satellite communication data set, providing basic data support for model training and evaluation;

[0020] A data preprocessing module is used to preprocess the satellite communication data set to form a satellite communication training data set to provide high-quality data for subsequent model training;

[0021] The model improvement module is used to improve the Transformer model to form an improved Transformer model;

[0022] A model training module, used to train the improved Transformer model using the satellite communication training data set, and adopt an improved Adam optimizer and an adaptive learning rate adjustment strategy to train the improved Transformer model;

[0023] A model evaluation and optimization module, used to evaluate the improved Transformer model through indicators, and continuously optimize the model structure and parameters according to the evaluation results;

[0024] The system integration and application module integrates the trained improved Transformer model into the satellite communication system, verifies its performance through real-time data transmission experiments, and further optimizes it according to actual application requirements.

[0025] The satellite communication data transmission optimization device according to the embodiment of the present invention optimizes the transmission of satellite communication data by improving the Transformer model, which has the following beneficial effects:

[0026] Improve data transmission efficiency: Through the improved self-attention mechanism and position encoding method, the model's ability to process long sequence data is improved, thereby improving the efficiency of data transmission. Reduce data transmission delay: The use of an integrated learning framework and a dynamic weight adjustment method enables the model to transmit data more quickly, reducing data transmission delays. Enhance transmission reliability: By introducing regularization technology and data enhancement methods, the generalization ability and robustness of the model are improved, and the packet loss rate and bit error rate during data transmission are reduced. Adapt to a variety of application scenarios: The method of the present invention has strong versatility and scalability, can adapt to different types of satellite communication data transmission needs, and has broad application prospects.

[0027] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] The above and / or additional aspects and advantages of the present invention will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0029] Figure 1 is a schematic diagram of a transmission optimization method according to an embodiment of the present invention;

[0030] Figure 2 is a flow chart of a transmission optimization method according to an embodiment of the present invention. DETAILED DESCRIPTION

[0031] Embodiments of the present invention are described in detail below. The embodiments described with reference to the accompanying drawings are exemplary. Embodiments of the present invention are described in detail below. Specific implementation method one:

[0033] The technical solution of the present invention is: the satellite communication data transmission optimization method based on improved Transformer, the specific process is as follows Figure 1 and Figure 2 As shown, it includes the following six steps:

[0034] S1:

[0035] Collect data generated during satellite communications to form satellite communication data sets, providing basic data support for model training and evaluation; specifically: rich data sets are the basis for model training, and various types of data generated during satellite communications are collected to provide basic data support for model training and evaluation.

[0036] S2:

[0037] The satellite communication data set is preprocessed to form a satellite communication training data set to provide high-quality data for subsequent model training; specifically, the collected satellite communication data is preprocessed, including data standardization, normalization, noise removal and other operations to provide high-quality data for subsequent model training.

[0038] S3:

[0039] Improvements are made on the basis of the Transformer model to form an improved Transformer model; specifically: improvements are made on the basis of the traditional Transformer model, mainly including improving the self-attention mechanism, optimizing position encoding, and integrating the learning framework.

[0040] S4:

[0041] The improved Transformer model is trained using the satellite communication training data set, and the improved Adam optimizer and the adaptive learning rate adjustment strategy are used to train the improved Transformer model; specifically: the improved Adam optimizer and the adaptive learning rate adjustment strategy are used to train the improved Transformer model. During the training process, by introducing regularization technology and data enhancement methods, it is necessary to monitor the performance of the model on the training set and the validation set to prevent the model from overfitting and improve the generalization ability of the model.

[0042] S5:

[0043] The improved Transformer model is evaluated through indicators, and the model structure and parameters are continuously optimized according to the evaluation results; specifically: the model is evaluated through a series of indicators (such as transmission efficiency, delay time, packet loss rate, etc.), and the model structure and parameters are continuously optimized according to the evaluation results.

[0044] S6:

[0045] The trained improved Transformer model is integrated into the satellite communication system, its performance is verified through real-time data transmission experiments, and further optimized according to actual application requirements. Specifically: the trained improved Transformer model is integrated into the satellite communication system, its performance is verified through real-time data transmission experiments, and further optimized according to actual application requirements.

[0046] A further technical solution of the present invention is that the purpose of collecting the S1 data set is to obtain comprehensive satellite communication data in order to provide sufficient samples for model training and evaluation. The collected data mainly includes the following categories:

[0047] Communication log data: records various log information generated during satellite communication, including communication connection establishment, data packet sending and receiving, signal strength, etc.

[0048] Network performance data: records the performance indicators of the communication network, such as bandwidth utilization, latency, packet loss rate, bit error rate, etc.

[0049] Environmental data: records the impact of environmental factors on signal transmission during the communication process, such as weather conditions, satellite positions, ground receiving station locations, etc.

[0050] User behavior data: records the user's behavior data during the communication process, such as data request frequency, data packet size, communication duration, etc.

[0051] In order to obtain high-quality data, multiple data collection methods need to be adopted to ensure the comprehensiveness and accuracy of the data.

[0052] Logging system: A logging system is deployed in the satellite communication system to automatically record various log information generated during the communication process. The system should have high accuracy and reliability, and be able to capture and store communication data in real time.

[0053] Network monitoring tools: Use network monitoring tools to monitor the performance indicators of the communication network in real time. Through sensors installed on communication equipment, data such as bandwidth utilization, delay time, packet loss rate, bit error rate, etc. are collected in real time and transmitted to the data storage system through the network.

[0054] Environmental monitoring system: Environmental monitoring equipment is installed on satellites and ground receiving stations to collect real-time environmental data such as weather conditions, satellite positions, and ground receiving station positions. The environmental monitoring system should have anti-interference capabilities to ensure data accuracy.

[0055] User behavior tracking tool: embed a behavior tracking tool in the user's terminal device to record the user's behavior data during the communication process. The tool should protect the user's privacy and ensure that the data collection process is legal and compliant.

[0056] A further technical solution of the present invention is: the S2 pre-processing data mainly includes data standardization, data normalization, noise removal, data set division, etc.

[0057] Data standardization: Standardize satellite communication data and map the data to the same dimensional range to eliminate dimensional differences between different features and facilitate subsequent model training.

[0058] Data normalization: Normalize the standardized data and scale the data to the range of 0-1 to speed up the convergence of the model.

[0059]

[0060] Where X′ is the normalized data, X is the data before normalization, and X max , X min They are the maximum and minimum values ​​of a feature of the sample.

[0061] Noise removal: Use filters and other methods to remove noise from data to improve data quality and reliability.

[0062] Dataset partitioning: The dataset is divided into a training set and a test set. The commonly used partitioning ratio is 80% training set and 20% test set. The training set is used for model training, and the test set is used for model evaluation.

[0063] A further technical solution of the present invention is: the improvement of the S3 model mainly includes the following parts:

[0064] Improved self-attention mechanism

[0065] In the traditional Transformer model, the self-attention mechanism captures the dependencies within the sequence by calculating the correlation between each position in the input sequence. In order to improve the flexibility and effectiveness of the self-attention mechanism, this paper proposes an improved multi-head self-attention mechanism and introduces a dynamic weight adjustment method.

[0066] Multi-head self-attention mechanism: By introducing multiple independent attention heads, the model can capture information in the input sequence from different perspectives. Each attention head calculates self-attention independently, and then the outputs of multiple attention heads are concatenated to enhance the expressiveness of the model.

[0067]

[0068] head i =Attention(QW i Q ,KW i K ,VW i V )

[0069] MultiHead(Q,K,V)=Concat(head1,…,head h )W O

[0070] in, represents the dimension of the Key matrix, Q, K, and V are the Quary matrix, Key matrix, and Value matrix after the input data is linearly transformed. and

[0071] is the weight matrix.

[0072] Dynamic weight adjustment: When calculating the attention weight, a dynamic weight adjustment method is introduced to adaptively adjust the attention weight according to the characteristics of the input sequence to enhance the model's ability to capture dependencies at different time steps.

[0073]

[0074] Among them, α and β are the dynamic weight scaling parameter and dynamic weight translation parameter in the adaptive attention mechanism, which can be continuously learned through model training.

[0075] Positional encoding optimization

[0076] The traditional Transformer model uses a fixed position encoding method to represent the position information of each position in the input sequence. However, the fixed position encoding method has certain limitations when processing long sequence data. To this end, the present invention designs a new position encoding method.

[0077] Relative position encoding: The relative position encoding method is used to calculate the position encoding according to the relative distance between each position in the input sequence, so that the model can better capture the dependencies in long sequence data.

[0078] R ij =RelativePosEmb(ij)

[0079] RelativePosEmb(ij)=E[i-j+M]

[0080] Among them, R ij represents the relative position encoding matrix, M represents the maximum relative position difference, and E represents the embedding matrix.

[0081] Learnable positional encoding: Introduce learnable positional encoding parameters and continuously optimize the positional encoding through the model training process to improve the model's adaptability and generalization performance.

[0082] P=[p1,p2,…,p n ]

[0083] Among them, P represents the learnable position encoding matrix with dimension (n, d), n is the sequence length, and d is the feature dimension.

[0084] Ensemble learning framework

[0085] In order to improve the generalization ability and robustness of the model, the present invention introduces the idea of ​​ensemble learning, constructs multiple sub-models, and integrates them through weighted voting.

[0086] Sub-model construction: Based on the improved Transformer model, multiple sub-models are constructed. Each sub-model adopts different initialization parameters and training strategies to capture different features in the input data.

[0087] Weighted voting: Weighted voting is performed on the outputs of multiple sub-models, and the weights are adaptively adjusted according to the performance indicators of the sub-models to improve the prediction accuracy and robustness of the final model.

[0088] A further technical solution of the present invention is: the S4 model training includes using an improved Adam optimizer, regularization technology, and data enhancement technology.

[0089] The improved Adam optimizer is used to optimize the model parameters. Based on the traditional Adam optimizer, the improved Adam optimizer introduces an adaptive learning rate adjustment strategy to enable the model to converge faster.

[0090] m t =β1m t-1 +(1-β1)g t

[0091]

[0092] Among them, θ t represents the model parameters at step t, g t represents the gradient of the tth step, α represents the learning rate, β1 and β2 represent the exponential decay rate of the momentum term, ò represents a constant to prevent division by zero, and λ represents the weight decay coefficient.

[0093] Adaptive learning rate adjustment: During the model training process, the learning rate is dynamically adjusted according to the gradient changes, so that the model uses a larger learning rate in the early stage of training to speed up convergence; and a smaller learning rate in the later stage of training to improve the stability and accuracy of the model.

[0094]

[0095] Among them, η t represents the learning rate at step t, η init represents the initial learning rate, γ represents the decay rate, and t decay Indicates the number of decay steps.

[0096] In order to prevent the model from overfitting, regularization technology is introduced during the model training process.

[0097] Dropout: Add a Dropout layer between each layer of the model to randomly discard the output of some neurons with a certain probability, thereby improving the generalization ability of the model.

[0098] L2 regularization: Adding L2 regularization term to the loss function prevents the model from overfitting the training data by penalizing the size of model parameters.

[0099] Data Augmentation

[0100] In order to improve the robustness of the model, data augmentation methods are used during model training to randomly transform the input data to generate more training samples.

[0101] Data augmentation: Random sampling, translation, rotation and other operations are performed on the original data to generate more training samples and improve the robustness of the model.

[0102] Noise injection: Add a certain proportion of noise to the original data to enhance the model's ability to handle noisy data and improve the model's stability and reliability.

[0103] A further technical solution of the present invention is: the S5 model evaluation and optimization evaluates the model through a series of indicators, including transmission efficiency, delay time, packet loss rate, bit error rate, etc., and continuously optimizes the model structure and parameters according to the evaluation results to improve the performance of the model.

[0104] Transmission efficiency: measures the model's ability to transmit data per unit time, and is evaluated by calculating the amount of data transmitted per second.

[0105]

[0106] Among them, Et represents the transmission efficiency (unit: bits per second, bps). D represents the total amount of data transmitted (unit: bits, bits). T represents the total time used for transmission (unit: seconds, seconds).

[0107] Latency: A measure of the delay in data transmission, evaluated by calculating the time from when data is sent to when it is received.

[0108] Packet loss rate: measures the proportion of data packets lost during data transmission, and is evaluated by calculating the ratio of lost data packets to total data packets.

[0109]

[0110] Where, Pl represents the packet loss rate, Nl represents the number of lost packets, and Nt represents the total number of transmitted packets.

[0111] Bit error rate: measures the proportion of erroneous data generated during data transmission, and is evaluated by calculating the ratio of erroneous data to total data.

[0112]

[0113] Where Be represents the bit error rate, Ne represents the number of error bits generated during transmission, and Nb represents the total number of bits transmitted.

[0114] Structural optimization: Adjust the number of model layers, the number of neurons in each layer, etc. to improve the model's expressiveness and computational efficiency.

[0115] Parameter optimization: Tune the model’s hyperparameters through grid search, random search, and other methods to find the best parameter combination.

[0116] A further technical solution of the present invention is: the S6 system integration and application integrates the improved Transformer model into the satellite communication system, connects with the existing system through an interface, realizes real-time transmission and processing of data, verifies the performance of the improved Transformer model in satellite communication data transmission through a series of practical applications, and further optimizes it according to actual application requirements.

[0117] Interface design: Design standardized interfaces to ensure that the improved Transformer model can be seamlessly connected with the existing system to achieve efficient data transmission and processing.

[0118] System deployment: Deploy the improved Transformer model into the satellite communication system and conduct actual data transmission experiments to verify its performance.

[0119] Application scenario selection: Select typical satellite communication application scenarios, such as satellite remote sensing data transmission, satellite navigation data transmission, etc., for actual verification.

[0120] Performance testing: Through actual data transmission experiments, the performance of the improved Transformer model in different application scenarios is tested, and its transmission efficiency, delay time, packet loss rate and other indicators are evaluated.

[0121] Effect of the invention: Compared with the prior art, the present invention has the following beneficial effects:

[0122] Improve data transmission efficiency: Through the improved self-attention mechanism and position encoding method, the model's ability to process long sequence data is improved, thereby improving the efficiency of data transmission.

[0123] Reduce data transmission delay: The use of an integrated learning framework and dynamic weight adjustment method enables the model to transmit data more quickly, reducing data transmission delays.

[0124] Enhanced transmission reliability: By introducing regularization technology and data enhancement methods, the generalization ability and robustness of the model are improved, and the packet loss rate and bit error rate during data transmission are reduced.

[0125] Adaptable to various application scenarios: The method of the present invention has strong versatility and expansibility, can adapt to different types of satellite communication data transmission requirements, and has broad application prospects. Specific implementation method 2:

[0127] The present invention also provides a satellite communication data transmission optimization device, the device comprising:

[0128] The data set collection module is used to collect data generated during satellite communication to form a satellite communication data set, providing basic data support for model training and evaluation;

[0129] A data preprocessing module is used to preprocess the satellite communication data set to form a satellite communication training data set to provide high-quality data for subsequent model training;

[0130] The model improvement module is used to improve the Transformer model to form an improved Transformer model;

[0131] A model training module, used to train the improved Transformer model using the satellite communication training data set, and adopt an improved Adam optimizer and an adaptive learning rate adjustment strategy to train the improved Transformer model;

[0132] A model evaluation and optimization module, used to evaluate the improved Transformer model through indicators, and continuously optimize the model structure and parameters according to the evaluation results;

[0133] The system integration and application module integrates the trained improved Transformer model into the satellite communication system, verifies its performance through real-time data transmission experiments, and further optimizes it according to actual application requirements.

[0134] The satellite communication data transmission optimization device according to the embodiment of the present invention optimizes the transmission of satellite communication data by improving the Transformer model, which has the following beneficial effects:

[0135] Improve data transmission efficiency: Through the improved self-attention mechanism and position encoding method, the model's ability to process long sequence data is improved, thereby improving the efficiency of data transmission. Reduce data transmission delay: The use of an integrated learning framework and a dynamic weight adjustment method enables the model to transmit data more quickly, reducing data transmission delays. Enhance transmission reliability: By introducing regularization technology and data enhancement methods, the generalization ability and robustness of the model are improved, and the packet loss rate and bit error rate during data transmission are reduced. Adapt to a variety of application scenarios: The method of the present invention has strong versatility and scalability, can adapt to different types of satellite communication data transmission needs, and has broad application prospects.

[0136] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example.

[0137] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

Claims

1. A satellite communication data transmission optimization method, characterized in that: include: Collect data generated during satellite communications to form satellite communication data sets, providing basic data support for model training and evaluation; Preprocessing the satellite communication data set to form a satellite communication training data set to provide high-quality data for subsequent model training; Improve the Transformer model to form an improved Transformer model; Using the satellite communication training data set to train the improved Transformer model, adopting an improved Adam optimizer and an adaptive learning rate adjustment strategy to train the improved Transformer model; Evaluate the improved Transformer model through indicators, and continuously optimize the model structure and parameters according to the evaluation results; The trained improved Transformer model is integrated into the satellite communication system, its performance is verified through real-time data transmission experiments, and further optimized according to actual application requirements.

2. The satellite communication data transmission optimization method according to claim 1, characterized in that: The satellite communication data set is preprocessed, specifically: data standardization, normalization, noise removal, and data set division.

3. The satellite communication data transmission optimization method according to claim 1, characterized in that: Improvements are made on the basis of the Transformer model to form an improved Transformer model. Specifically, on the basis of the Transformer model, improvements are made to the self-attention mechanism, position encoding optimization, and integrated learning framework.

4. The satellite communication data transmission optimization method according to claim 1, characterized in that: The improved Transformer model is trained using the satellite communication training data set, specifically: by introducing regularization technology and data enhancement technology, the performance of the model on the training set and the validation set needs to be monitored during the training process to prevent the model from overfitting and improve the generalization ability of the model.

5. The satellite communication data transmission optimization method according to claim 1, characterized in that: The indicators include: any one or a combination of transmission efficiency, delay time, packet loss rate, and bit error rate.

6. A satellite communication data transmission optimization device, characterized in that: include: The data set collection module is used to collect data generated during satellite communication to form a satellite communication data set, providing basic data support for model training and evaluation; A data preprocessing module is used to preprocess the satellite communication data set to form a satellite communication training data set to provide high-quality data for subsequent model training; The model improvement module is used to improve the Transformer model to form an improved Transformer model; A model training module, used to train the improved Transformer model using the satellite communication training data set, and adopt an improved Adam optimizer and an adaptive learning rate adjustment strategy to train the improved Transformer model; A model evaluation and optimization module, used to evaluate the improved Transformer model through indicators, and continuously optimize the model structure and parameters according to the evaluation results; The system integration and application module integrates the trained improved Transformer model into the satellite communication system, verifies its performance through real-time data transmission experiments, and further optimizes it according to actual application requirements.

Images