Dynamic compression method and device for network transmission data

Abstract

The invention relates to the technical field of automobile control, and discloses a dynamic compression method and device for network transmission data, and the method comprises the steps: determining target monitoring data based on the transmission data of a first time period of a target communication network; when a target item in the target monitoring data meets a compression parameter adjustment condition, generating feedback data according to the target monitoring data; using the dynamic compression model, taking the feedback data as input, generating a target compression parameter of a compression algorithm used for transmitting the data, updating parameter setting of the compression algorithm by adopting the target compression parameter, and using the compression algorithm with the updated parameter setting, and processing transmission data of the target communication network in a second time period after the first time period. According to the invention, the compression parameter of the subsequent transmission data of the communication network can be adjusted rapidly, efficiently and dynamically according to the real-time transmission data of the communication network, and the compression efficiency and data transmission performance of the subsequent transmission data are improved.

Description

Technical Field

[0001] The present disclosure relates to the field of automobile control technology, and in particular to a method and device for dynamic compression of network transmission data. Background Art

[0002] In modern automotive control systems, the CAN network, as the core technology for communication between control units (ECUs), is becoming increasingly important. With the increasing complexity of vehicle functions, such as the vehicle's autonomous driving function and smart entertainment function, the amount of data transmitted by the CAN network has increased significantly, and the requirements for network bandwidth and real-time performance have become increasingly higher.

[0003] In the related technology, the data transmitted in the CAN network usually adopts the traditional data compression technology, which can alleviate the pressure of network bandwidth to a certain extent. However, since this technology uses a fixed compression strategy to compress data, when the actual characteristics of the data or the network status change dynamically, it may lead to low compression efficiency, and even increase transmission delay in some cases, affecting the response speed and stability of the overall system. Therefore, how to ensure the response speed and stability of the overall system when the actual characteristics of the data or the network status change dynamically has become an urgent problem to be solved. Summary of the invention

[0004] In view of this, the present disclosure provides a method for dynamic compression of network transmission data to solve the problem of ensuring the response speed and stability of the overall system when the actual characteristics of the data or the network status change dynamically.

[0005] In a first aspect, the present disclosure provides a method for dynamic compression of network transmission data, the method comprising: determining target monitoring data based on transmission data of a target communication network in a first time period; the target monitoring data comprising at least one of the following: transmission delay, packet loss rate and compression efficiency; generating feedback data according to the target monitoring data when the target item in the target monitoring data meets the compression parameter adjustment condition; using a dynamic compression model, taking the feedback data as input, generating target compression parameters of a compression algorithm used for the transmission data, the target compression parameters comprising at least one of the following: compression rate, block size, number of iterations; using the target compression parameters to update parameter settings of the compression algorithm, and using the compression algorithm with the updated parameter settings to process transmission data of the target communication network in a second time period after the first time period.

[0006] In a second aspect, the present disclosure provides a dynamic compression device for network transmission data, the device comprising: a data determination module, used to determine target monitoring data based on the transmission data of a first time period of a target communication network; the target monitoring data comprises at least one of the following: transmission delay, packet loss rate and compression efficiency; a data generation module, used to generate feedback data according to the target monitoring data when the target item in the target monitoring data meets the compression parameter adjustment condition; a parameter generation module, used to use a dynamic compression model, take the feedback data as input, and generate target compression parameters of the compression algorithm used for the transmission data, the target compression parameters comprising at least one of the following: compression rate, block size, number of iterations; a parameter updating module, used to update the parameter settings of the compression algorithm using the target compression parameters, and a data processing module, used to use the compression algorithm with the updated parameter settings to process the transmission data of the target communication network in a second time period after the first time period.

[0007] In a third aspect, the present disclosure provides a vehicle, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the method for dynamic compression of network transmission data of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0008] In a fourth aspect, the present disclosure provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for dynamic compression of network transmission data of the first aspect or any corresponding embodiment thereof.

[0009] In a fifth aspect, the present disclosure provides a computer program product, including computer instructions, which are used to enable a computer to execute the method for dynamic compression of network transmission data according to the first aspect or any corresponding embodiment thereof.

[0010] The dynamic compression method and device for network transmission data provided by the above-mentioned embodiments of the present disclosure determine the target monitoring data based on the transmission data of the first time period of the communication network collected in real time, and clarify the monitoring object of concern; then, when the target item in the target monitoring data of focus meets the compression parameter adjustment condition, it means that the network status does not meet expectations, and at this time, it is triggered to generate feedback data suitable for analysis and modeling based on the target monitoring data; then, the feedback data is processed using a dynamic compression model to obtain target compression parameters corresponding to the current network status and meeting expectations, and the target compression parameters are used to adjust one or more of the compression rate, block size, and number of iterations of the compression algorithm for the transmission data; then, the target compression parameters are used to update the parameter settings of the compression algorithm, and the updated compression algorithm is further used to process the transmission data of the communication network in the second time period after the first time period, so that the compression parameters of the subsequent transmission data of the communication network can be adjusted quickly, efficiently and dynamically according to the real-time transmission data of the communication network, thereby improving the compression efficiency and data transmission performance of the subsequent transmission data.

[0011] In addition, due to the use of intelligent monitoring of target monitoring data and intelligent triggering of compression parameter adjustment when the target items in the target monitoring data meet the compression parameter adjustment conditions, adaptive adjustment of compression algorithm parameters is achieved, human intervention is reduced, and the intelligence level of the dynamic compression method is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the specific embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the specific embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0013] Figure 1 is a flowchart of a method for dynamic compression of network transmission data according to an embodiment of the present disclosure;

[0014] Figure 2 is a flow chart of another method for dynamic compression of network transmission data according to an embodiment of the present disclosure;

[0015] Figure 3 is a flowchart diagram of a flowchart of an exemplary application scenario of a method for dynamic compression of network transmission data according to an embodiment of the present disclosure;

[0016] Figure 4 It is an exemplary architecture of a dynamic compression model of a dynamic compression method for network transmission data according to an embodiment of the present disclosure;

[0017] Figure 5is a flowchart of a training and deployment process of a dynamic compression model of a dynamic compression method for network transmission data according to an embodiment of the present disclosure;

[0018] Figure 6 is a structural block diagram of a dynamic compression device for network transmission data according to an embodiment of the present disclosure;

[0019] Figure 7 It is a schematic diagram of the hardware structure of the vehicle according to the embodiment of the present disclosure. DETAILED DESCRIPTION

[0020] In order to make the purpose, technical solution and advantages of the embodiments of the present disclosure clearer, the technical solution in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are part of the embodiments of the present disclosure, not all of the embodiments. Based on the embodiments in the present disclosure, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present disclosure.

[0021] According to an embodiment of the present disclosure, an embodiment of a method for dynamic compression of network transmission data is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in an electronic integrated circuit such as a set of computer executable instructions, and the electronic integrated circuit is connected as a data processing node to a serial communication network capable of distributed real-time control. In addition, although a logical sequence is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than that here.

[0022] In this embodiment, a dynamic compression method for network transmission data is provided, which can be used for the above-mentioned electronic integrated circuit, such as a node unit connected to a serial communication network capable of realizing distributed real-time control, etc. Figure 1 is a flow chart of a method for dynamic compression of network transmission data according to an embodiment of the present disclosure. Figure 1 As shown, the process includes the following steps:

[0023] Step S101, determining target monitoring data based on transmission data of a target communication network in a first time period.

[0024] In this embodiment, the target communication network may be a serial communication network capable of implementing distributed real-time control. The target communication network may connect multiple node units (i.e., nodes that transmit data), such as a controller area network (CAN), a FlexRay network, a media-oriented system transport (MOST) network, an in-vehicle Ethernet, etc. The target monitoring data include: transmission delay, packet loss rate, and compression efficiency.

[0025] The transmission data of the first time period of the communication network refers to the transmission data in the target communication network collected in real time by the data collection module before the current moment. The transmission data may include: the timestamp, ID, data length and data content of each data frame.

[0026] The target monitoring data is monitoring data generated for the transmission data of the above-mentioned communication network and has a clear focus, and may include at least one of the following: transmission delay, packet loss rate, and compression efficiency. After obtaining the transmission data of the target communication network in the first time period, the above-mentioned execution subject may determine the target monitoring data based on the transmission data of the communication network in the first time period.

[0027] Among them, transmission delay (Latency) refers to the time interval from the sending to the receiving of a data packet.

[0028] Packet Loss Rate is the ratio of data packets lost during the transmission process.

[0029] Compression Ratio is the ratio of the size of the compressed data to the size of the original data in a data packet.

[0030] The execution subject may calculate and generate each item of data in the target monitoring data according to the transmission data in the first time period and the definition of each item in the target monitoring data.

[0031] Step S102: when the target item in the target monitoring data meets the compression parameter adjustment condition, feedback data is generated according to the target monitoring data.

[0032] In this embodiment, the target item refers to at least one item in the target monitoring data, that is, the target item may be one, two or three items of the transmission delay, packet loss rate and compression efficiency included in the target monitoring data.

[0033] The compression parameter adjustment condition is a pre-set condition, which can be reflected by using existing technology or future technology to measure whether the measurement value of the above target item meets the threshold, whether it meets the preset ratio or whether it meets the preset distribution.

[0034] For example, the compression parameter adjustment condition may be that the target item meets a threshold value. When the target item is one item, the compression adjustment condition may be whether the metric value of the target item meets the threshold value. When the target item is multiple items, the compression adjustment condition may be whether the metric value of each item in the multiple items meets the threshold value corresponding to each item, or the compression adjustment condition may be whether the metric value of any item in the multiple items meets the threshold value corresponding to the item.

[0035] When the target item in the target monitoring data meets the compression parameter adjustment condition, the compression parameter update of the subsequent compression algorithm is triggered.

[0036] When judging whether the target item meets the compression parameter adjustment conditions, the metric parameters can be generated for the data sequences of the transmission delay and compression efficiency, respectively, so as to use the metric parameters to judge whether the compression parameter adjustment conditions are met in the subsequent time. The metric parameter can be a numerical value or numerical distribution that reflects the characteristics of the data sequence.

[0037] For example, the target item in the target monitoring data meets the compression parameter adjustment conditions, which may include: the central tendency measurement value of the transmission delay is greater than the transmission delay threshold; the central tendency measurement value of the compression efficiency is less than the compression efficiency threshold; the packet loss rate is greater than the packet loss rate threshold, etc. Among them, the central tendency measurement value refers to an indicator used to measure the tendency of data in the data set to aggregate, which may include the mean, median, and mode, etc.

[0038] Feedback data is data that can identify network status characteristics and can be generated by preprocessing one or more of the target monitoring data. That is, the original data can be reviewed, screened, and sorted for one or more of the target monitoring data to obtain feedback data so that it is suitable for analysis and modeling.

[0039] The above-mentioned execution subject can trigger the generation of feedback data according to the pre-set compression parameter adjustment conditions when monitoring that the target item meets the compression parameter adjustment conditions.

[0040] Step S103: using the dynamic compression model and taking the feedback data as input, generating target compression parameters of the compression algorithm used for transmitting the data.

[0041] In this embodiment, the dynamic compression model is a pre-trained model used to dynamically adjust the output target compression parameters according to the target monitoring data in the data transmission network. The model is good at processing data with strong time correlation and can predict the corresponding target compression parameters according to the input feedback data.

[0042] In some specific examples, the dynamic compression model can be implemented using models such as the long short-term memory network LSTM model and its variants, the liquid neural network LNNS, the closed continuous time neural network CFC, the recurrent neural network RNN ​​and other variants.

[0043] The target compression parameter refers to the expected compression parameter of the compression algorithm generated based on the currently input feedback data, and may include but is not limited to: one, two or three of the compression rate, the compression block size and the number of compression iterations.

[0044] It is understandable that those skilled in the art can set the data items included in the target compression parameters based on the above target compression parameters according to actual application scenarios and / or their own experience, so as to efficiently achieve the desired adjusted compression parameters.

[0045] For example, for different compression algorithm types, the unique parameters of the compression algorithm can be adjusted. For example, for a dictionary compression algorithm, a hash table algorithm, and a sliding window compression algorithm, the target compression parameter can also include at least one of the following items of the compression algorithm: dictionary size, hash table size, and window size.

[0046] Step S104: using the target compression parameters to update the parameter settings of the compression algorithm.

[0047] In this embodiment, after determining the target compression parameter in step S103, the execution subject may adopt the target compression parameter to update the parameter setting of the local compression algorithm, thereby updating the compression parameter of the compression algorithm and obtaining the compression algorithm with the updated parameter setting.

[0048] Step S105 , using the compression algorithm with updated parameter settings to process transmission data of the target communication network in a second time period after the first time period.

[0049] In this embodiment, the compression parameters of the compression algorithm updated according to the network status characteristics can be used to process the data packets that need to be transmitted in the time period after the current moment, so that these compression parameters can at least include: one, two or three of the compression rate, compression block size, and compression iteration number. In some examples, the compression parameters can also include: unique parameters of different types of compression algorithms, such as one or more of the dictionary size, hash table size, and window size.

[0050] The dynamic compression algorithm for network transmission data in this embodiment determines the target monitoring data based on the transmission data of the first time period of the communication network collected in real time, and clarifies the monitoring object of concern; then, when the target item in the target monitoring data of focus meets the compression parameter adjustment condition, it means that the network status does not meet expectations, and at this time, it is triggered to generate feedback data suitable for analysis and modeling based on the target monitoring data; then, the feedback data is processed using a dynamic compression model to obtain target compression parameters corresponding to the current network status and meeting expectations, and the target compression parameters are used to adjust one or more of the compression rate, block size, and number of iterations of the compression algorithm for transmission data; then, the target compression parameters are used to update the parameter settings of the compression algorithm, and the updated compression algorithm is further used to process the transmission data of the communication network in the second time period after the first time period, so that the compression parameters of the subsequent transmission data of the communication network can be adjusted quickly, efficiently and dynamically according to the real-time transmission data of the communication network, thereby improving the compression efficiency and data transmission performance of the subsequent transmission data.

[0051] In addition, due to the use of intelligent monitoring of target monitoring data and intelligent triggering of compression parameter adjustment when the target items in the target monitoring data meet the compression parameter adjustment conditions, adaptive adjustment of compression algorithm parameters is achieved, human intervention is reduced, and the intelligence level of the dynamic compression method is improved.

[0052] Please refer to Figure 2 , Figure 2 A flow chart of another method for dynamic compression of network transmission data according to an embodiment of the present disclosure is shown. The dynamic compression model in the dynamic compression method may include: an embedding layer, a bidirectional long short-term memory layer, a self-attention layer, and a fully connected layer connected in sequence.

[0053] like Figure 2 As shown, the process of the dynamic compression method for transmitting data over another network may include:

[0054] Step S201: determining target monitoring data based on transmission data of a target communication network in a first time period.

[0055] In this embodiment, the target monitoring data is the monitoring data generated for the transmission data of the above communication network and has a clear focus, and may include at least one of the following: transmission delay, packet loss rate, and compression efficiency. After obtaining the transmission data of the target communication network in the first time period, the above execution subject may determine the target monitoring data based on the transmission data of the communication network in the first time period. For details, please refer to Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0056] Step S202: when the target item in the target monitoring data meets the compression parameter adjustment condition, feedback data is generated according to the target monitoring data.

[0057] In this embodiment, the target item refers to at least one item in the target monitoring data, that is, the target item may be one, two or three items of the transmission delay, packet loss rate and compression efficiency included in the target monitoring data.

[0058] The compression parameter adjustment condition is a pre-set condition, which can be reflected by using existing technology or future technology to measure whether the measurement value of the above target item meets the threshold, meets the preset ratio or meets the preset distribution. Figure 1 Step S102 of the illustrated embodiment will not be described in detail here.

[0059] Step S203, using the embedding layer, taking the feedback data as input, and generating an initial embedding vector of the network state feature.

[0060] In this embodiment, the embedding layer is used to convert the feedback data into a vector representation of a fixed size, thereby converting the feedback data identifying the network state characteristics into an initial embedding vector.

[0061] Step S204, using a bidirectional long short-term memory layer, taking the initial embedding vector as input, and generating an intermediate embedding vector containing the forward and backward dependency information of the initial embedding vector.

[0062] In this embodiment, the Bi-LSTM layer is a special recurrent neural network (RNN) layer that can simultaneously consider the forward and backward information of the input sequence. When processing the initial embedding vector sequence data, the Bi-LSTM layer can capture richer context information of the initial embedding vector, thereby outputting an intermediate embedding vector that contains the forward and backward dependency information of the initial embedding vector.

[0063] Step S205, using the self-attention layer, taking the intermediate embedding vector as input, generates a target embedding vector including long-distance dependency information and dynamic weight adjustment information.

[0064] In this embodiment, the self-attention layer can capture the long-distance dependencies in the input intermediate embedding vector sequence, and optimize information extraction by dynamically adjusting weights, thereby outputting a target embedding vector that contains long-distance dependency information and dynamic weight adjustment information.

[0065] Step S206, using a fully connected layer, taking the target embedding vector as input, and generating target compression parameters of a compression algorithm corresponding to the transmitted data.

[0066] In this embodiment, in the fully connected layer, each neuron is connected to all neurons in the self-attention layer, and the target embedding vector extracted by the self-attention layer can be further processed and classified to obtain the estimated target compression parameter.

[0067] Step S207: Use the target compression parameters to update the parameter settings of the compression algorithm.

[0068] In this embodiment, after the target compression parameter is determined, the execution subject may adopt the target compression parameter to update the parameter setting of the local compression algorithm to obtain the compression algorithm with the updated parameter setting.

[0069] Step S208: Using the compression algorithm with updated parameter settings, process the transmission data of the target communication network in a second time period after the first time period.

[0070] In this embodiment, a compression algorithm with updated parameter settings can be used to process the transmission data of the communication network in the second time period after the first time period. That is, after adjusting the parameter settings of the compression algorithm according to the target monitoring data, the updated compression algorithm is used to process the subsequent data packets that need to be transmitted, thereby dynamically changing the compression parameters of the compression algorithm according to the network status, thereby improving the efficiency of network data transmission.

[0071] The dynamic compression method for network transmission data provided in this embodiment adopts a dynamic compression model including a sequentially connected embedding layer, a bidirectional long short-term memory layer, a self-attention layer and a fully connected layer, thereby introducing a Self-Attention mechanism on the traditional Bi-LSTM model architecture, which can more accurately capture and understand network state characteristics, and improve the accuracy of the target compression parameters used by the compression algorithm during the dynamic compression process, thereby improving the accuracy of the parameter adjustment of the dynamic compression algorithm.

[0072] In some optional implementations of this embodiment, the dynamic compression model is obtained based on a model generation step, and the model generation step includes:

[0073] The feedback data samples in the training sample set are taken as input, the compression parameter samples in the training sample set are taken as output, the compression efficiency of the feedback data samples in the training sample set and the transmission delay of the feedback data samples are minimized as the loss function, an adaptive learning rate optimization algorithm is used to iteratively optimize the parameters of the dynamic compression model, and a back propagation algorithm is used to update the gradient (the gradient of the loss relative to each parameter) at each iteration until the number of iterations reaches a preset number or the loss function converges;

[0074] In this implementation, the dynamic compression model is a machine learning model that dynamically adjusts compression parameters based on input feedback data. The generation goal of the model is to find the best model parameters to achieve a balance between compression efficiency and transmission delay.

[0075] The generation of the dynamic compression model can prepare a training sample set. The training sample set includes:

[0076] Feedback data samples in the training sample set: are the input part of the training sample set, usually composed of a series of historical data or simulated data, representing the actual or expected input that the model needs to process. Feedback data samples in the test sample set can be generated based on one or more of the transmission delay, packet loss rate and compression efficiency in the target monitoring data.

[0077] Compression parameter samples in the training sample set: are the output part of the training sample set, corresponding one-to-one to the feedback data samples in the training sample set. The compression parameter samples in the training sample set are known, optimal or near-optimal compression parameter configurations, which have been proven in previous experiments or practical applications to effectively balance compression efficiency and transmission delay.

[0078] The loss function used when training the dynamic compression model can measure the difference between the model output and the true output (or expected output, here the compression parameter sample corresponding to the input feedback data sample). In order to achieve better transmission efficiency of network transmission data, the loss function can be selected to minimize the compression efficiency and transmission delay of the feedback data sample. Specifically, minimizing the compression efficiency and transmission delay of the feedback data sample involves a trade-off between two goals: on the one hand, it is hoped that the compressed data is as small as possible (i.e., high compression efficiency) to reduce storage and transmission costs; on the other hand, it is hoped that the compression and decompression process is as fast as possible (i.e., low transmission delay) to ensure real-time performance or reduce processing time.

[0079] In some optional examples, minimizing a weighted sum of a first mean square error and a second mean square error may be used as a loss function, where the first mean square error is the mean square error of compression efficiency; and the second mean square error is the mean square error of transmission delay.

[0080] The adaptive learning rate optimization algorithm used when training the dynamic compression model can dynamically adjust the learning rate according to the performance during training, so as to adjust the model parameters during model training and minimize the loss function. The algorithm uses a larger learning rate in the early stage of training to converge quickly, and reduces the learning rate when approaching the optimal solution to avoid oscillation or missing the optimal solution. Common adaptive learning rate optimization algorithms include Adam, RMSprop, etc.

[0081] The back propagation algorithm used when training the dynamic compression model can calculate the gradient of the loss function relative to each model parameter. In each iteration, the algorithm first generates output data based on the current model parameters and input data, then calculates the value of the loss function based on the output data and the true value (or expected value), and then calculates the gradient of each parameter through the back propagation algorithm. These gradients can be used to update the model parameters to reduce the value of the loss function.

[0082] The iterative optimization process when training a dynamic compression model involves repeatedly performing the following steps: using the current model parameters and the training sample set to calculate the value of the loss function; using the backpropagation algorithm to calculate the gradient; using the adaptive learning rate optimization algorithm to update the model parameters. This process will continue until the number of iterations reaches the preset number (i.e. the maximum number of iterations) or the loss function converges (i.e. the value of the loss function changes little or no longer changes in multiple consecutive iterations).

[0083] After the dynamic compression model is trained, it is usually necessary to perform model validation and testing to ensure that the dynamic compression model can also perform well on unseen data. This can be achieved by applying the trained dynamic compression model to an independent validation set or test set, and evaluating its compression efficiency and transmission delay on these sets.

[0084] In the model generation step of this implementation, the Bi-LSTM model combined with Self-Attention can be used to generate compression parameters, and the back-propagation algorithm is used to calculate the gradient when training the model; the model parameters are updated using the adaptive learning rate optimization algorithm, which improves the efficiency and accuracy of the dynamic compression model in outputting the optimal compression parameters.

[0085] In some optional implementations, the above-mentioned model generation step may also include: using the dynamic compression model after iterative optimization of parameters, taking the feedback data samples in the test sample set as input, and generating predicted compression parameters corresponding to the feedback data in the test sample set, wherein the feedback data samples in the test sample set respectively correspond to the compression parameter samples in the test sample set; generating evaluation indicators based on the difference between the predicted compression parameters and the compression parameter samples in the test sample set; adjusting the hyperparameters of the dynamic compression model after iterative optimization of parameters according to the evaluation indicators, and using the dynamic compression model after adjusting the hyperparameters to execute the model generation step until the evaluation indicators meet the preset requirements (the evaluation indicators reach the extreme value point).

[0086] In this implementation, when adjusting the hyperparameters of the dynamic compression model, you can adjust the learning rate, embedding dimension, number of hidden units, increase or decrease the number of LSTM layers, change the batch size, etc. The evaluation index meets the preset requirements, which means that the evaluation index converges to the preset threshold range, or reaches the extreme point, or reaches the maximum number of iterations.

[0087] The test sample set used to evaluate the dynamic compression model includes feedback data samples and corresponding compression parameter samples.

[0088] Feedback data samples in the test sample set: can be feedback data samples of the same type as the feedback data samples in the training sample set. The feedback data samples in the test sample set and the feedback data samples in the training sample set can be feedback data samples from different batches. The feedback data samples in the test sample set can be generated based on one or more of the transmission delay, packet loss rate and compression efficiency in the target monitoring data.

[0089] Compression parameter samples in the test sample set: corresponding to the feedback data samples in the test sample set, are known, optimal or near-optimal compression parameter configurations, which have been proven in previous experiments or practical applications to effectively balance compression efficiency and transmission delay.

[0090] The dynamic compression model after iterative optimization of parameters is a dynamic compression model that has been preliminarily trained and adjusted by the above-mentioned optional implementation method. The feedback data samples in the test sample set can be input into the dynamic compression model after iterative optimization of parameters to obtain the output predicted compression parameters of the feedback data in the corresponding test sample set.

[0091] When calculating the difference between the predicted compression parameter of the feedback data corresponding to the test sample set and the actual compression parameter sample in the test sample set, it can be calculated by various measurement methods (such as mean square error, absolute error, relative error, etc.).

[0092] Based on the differences, one or more evaluation metrics are generated to quantify the predictive performance of the model. These metrics can be accuracy, recall, F1 score, mean square error (MSE), etc.

[0093] According to the evaluation indicators, some strategies (grid search, random search, Bayesian optimization, etc.) can be used to adjust the hyperparameters of the dynamic compression model after iterative optimization parameters, and the adjusted hyperparameters are used to perform the above model generation steps, that is, retrain the dynamic compression model, and iteratively optimize the parameters of the dynamic compression model until the evaluation indicators meet the preset requirements. For example, if the evaluation indicators show that the model is overfitting during the training process, you can try to reduce the learning rate or increase the regularization term; if the model performance is poor, you can try to increase the number of hidden units or the number of LSTM layers.

[0094] The model generation step in this implementation method adjusts the hyperparameters of the dynamic compression model after iterative optimization parameters through evaluation results, and retrains the dynamic compression model using the adjusted hyperparameters, thereby continuously improving the prediction performance of the dynamic compression model and making it more accurate and efficient when generating target compression parameters.

[0095] Please refer to Figure 3 , Figure 3 A flowchart showing an exemplary application scenario of a method for dynamic compression of network transmission data according to an embodiment of the present disclosure is shown.

[0096] In some specific application scenarios, the technical architecture including a data acquisition module 301, a real-time detection module 302, a logic judgment module 303, a feedback mechanism module 304, and a dynamic compression module 305 can be used to implement the above-mentioned dynamic compression method for data in network transmission.

[0097] Specifically, each module in the technical architecture works as follows:

[0098] The working mode of the data acquisition module 301 is as follows:

[0099] The data acquisition module uses the following specific methods and tools for data acquisition: Taking the CAN network as an example, CAN bus data recorders (such as CANalyzer, CANoe, etc.) are used for data acquisition. These tools can be connected to the vehicle's CAN network and collect all transmitted data frames in real time.

[0100] Specifically, the CAN bus data recorder can be configured first, and the acquisition parameters (such as sampling rate, data frame filtering rules, etc.) can be set; then, the data recorder can be connected to the vehicle's CAN network, through the standard interface of the on-board diagnostic system II (OBD-II) interface or directly connected to the CAN bus; then, the data recorder can be started to start real-time acquisition of data frames on the CAN bus and record all transmitted data; then, the collected data can be transmitted to the monitoring module for real-time processing.

[0101] The real-time monitoring module 302 works as follows:

[0102] The monitoring indicators and specific monitoring process of the real-time detection module are as follows:

[0103] Monitoring indicators include:

[0104] Transmission delay (Latency): The time interval from the sending to the receiving of a data packet.

[0105] Packet Loss Rate: The proportion of data packets lost during transmission.

[0106] Compression Ratio: The ratio of the compressed data size to the original data size.

[0107] The specific monitoring process includes:

[0108] Input: Real-time network transmission data collected by the data acquisition module, including the timestamp, ID, data length and data content of each data frame.

[0109] deal with:

[0110] Transmission delay calculation: Calculates the transmission delay of each data packet.

[0111] Packet loss rate calculation: Count the number of data packets lost per unit time.

[0112] Compression efficiency calculation: Calculate the ratio of data size before and after compression.

[0113] Specifically, transmission delay (Latency) refers to the time interval from the sending to the receiving of a data packet, which can be calculated by the following formula:

[0114] L i =T recv-T send ;

[0115] Among them, L i is the transmission delay; T recv is the receiving time of the data packet; T sen d is the time when the data packet is sent.

[0116] Packet Loss Rate is the ratio of packets lost during transmission, which can be expressed as P loss It means that the number of lost packets can be obtained by subtracting the number of received packets from the number of sent packets, and then the percentage of the number of lost packets to the number of sent packets is used as the packet loss rate. The above-mentioned execution subject can use a network monitoring tool or a protocol analyzer to help record these data.

[0117] Compression Ratio is the ratio of the compressed data size to the original data size of a data packet, which can be calculated using the following formula:

[0118]

[0119] Among them, C ratio is the compression efficiency, Size original is the original data size of the data packet, Size compressed is the compressed data size of the packet.

[0120] Output: Generates monitoring data including transmission delay, packet loss rate and compression efficiency.

[0121] The logic judgment module 303 works as follows:

[0122] The logic judgment module performs logic judgment based on the monitoring data, and if the conditions are met (such as the network load is higher than a certain threshold), the parameter adjustment is started.

[0123] Function: Make logical judgments based on monitoring data to determine whether compression parameters need to be adjusted.

[0124] Input: Monitoring data of real-time monitoring module.

[0125] Output: Judgment result, passed to the feedback mechanism module.

[0126] Judgment conditions: transmission delay is greater than a threshold; or, packet loss rate is greater than a threshold; or, compression efficiency is less than a compression efficiency threshold.

[0127] The feedback mechanism module 304 works as follows:

[0128] The feedback mechanism module feeds back the monitoring data and judgment results to the dynamic compression module.

[0129] Function: Feedback the judgment results and monitoring data to the dynamic compression adjustment module based on the AI ​​model.

[0130] Input: judgment results and monitoring data of the logic judgment module.

[0131] Output: Feedback data, passed to the dynamic compression adjustment module.

[0132] Process: Generate feedback data based on the logical judgment result: M(t)={L i ,P loss ,C ratio}, where M(t) is the feedback data, L i is the transmission delay; P loss is the packet loss rate, C ratio is the compression efficiency.

[0133] The dynamic compression module 305 works as follows:

[0134] The dynamic compression module calls the Bi-LSTM model and adjusts the compression algorithm parameters according to the feedback data.

[0135] Function: Dynamically adjust the parameters of the compression algorithm through the AI ​​model (Bi-LSTM).

[0136] Input: Feedback data from the feedback mechanism module.

[0137] Output: optimized compression parameters, adjusted compression algorithm, and generated compressed data packets.

[0138] The adjustment process includes:

[0139] 1. Input feedback data M(t) to the Bi-LSTM model.

[0140] 2. The Bi-LSTM model outputs the optimization parameter Popt.

[0141] 3. Adjust the compression algorithm and generate the data compression package Dcomp(t).

[0142] That is, in the above-mentioned dynamic compression method for network transmission data, the data acquisition module collects CAN network data D(t) in real time; the real-time monitoring module calculates the current transmission efficiency E(t) and compression performance C(t), and generates monitoring data M(t)={E(t), C(t)}; the logic judgment module: judges whether E(t) and C(t) meet the preset conditions, such as E(t)<threshold, and if so, starts parameter adjustment; the feedback mechanism module feeds back the monitoring data M(t) and the judgment result to the dynamic compression module; the dynamic compression module inputs the feedback data M(t) to the Bi-LSTM model, outputs the optimization parameter Popt, adjusts the parameters of the compression algorithm, and improves the data compression performance.

[0143] Please refer to Figure 4 , Figure 4 An exemplary architecture of a dynamic compression model of a dynamic compression method for network transmission data according to an embodiment of the present disclosure is shown.

[0144] Specifically, the architecture of the dynamic compression model disclosed in the present invention includes four layers:

[0145] Embedding Layer: Embed the input network state features into a vector representation (EmbededCharacter vector).

[0146] Bi-LSTM Layer: The embedding vector is processed by bi-directional LSTM to capture the previous and next dependency information.

[0147] Self-Attention Layer: Performs Self-Attention processing on the output of Bi-LSTM to capture long-distance dependencies and dynamic weight adjustments.

[0148] Fully Connected Layer: Pass the output of the Self-Attention layer to the fully connected layer and output the optimized parameter p of the compression algorithm.

[0149] At the same time, considering that traditional LSTM is prone to lose long-distance dependency information when processing long sequences, the Self-Attention mechanism can effectively capture long-distance dependencies and improve the expressiveness of the model. Therefore, the present invention introduces the Self-Attention mechanism on the traditional Bi-LSTM model architecture to better capture network state characteristics: Through the Self-Attention mechanism, the model can more accurately capture and understand network state characteristics and improve the accuracy of parameter adjustment.

[0150] Please refer to Figure 5 , Figure 5 A flowchart of the training and deployment process of the dynamic compression model of the dynamic compression method for network transmission data according to an embodiment of the present disclosure is shown.

[0151] like Figure 5 As shown in the figure, the training and deployment process of the dynamic compression model includes:

[0152] Step S501, collecting CAN network status data.

[0153] In this example, when training the dynamic compression model, a CAN bus data logger or data acquisition script can be used to collect a large amount of CAN network data in real time, including transmission delay, packet loss rate, compression efficiency, etc. under different network conditions.

[0154] During model training, the data formats and features processed are as follows:

[0155] Feature Name type describe Timestamp time The time when the packet was sent or received Packet ID Integer Identifier of CAN data packet Data length Integer The length of the packet in bytes Original data size Integer Data size before compression (bytes) Compressed data size Integer Compressed data size (bytes) Transmission delay Floating point numbers The time interval from sending to receiving the data packet (milliseconds) Packet loss rate Floating point numbers The percentage of lost packets

[0156] Step S502: pre-processing the collected data.

[0157] In this example, when preprocessing the collected data, the data can be cleaned, standardized, and feature extracted to extract the feature vector X(t) and the target compression parameter Y(t). 70% of the preprocessed data can be used as a training set, and the remaining 30% can be used as a validation set.

[0158] Step S503: construct a dynamic compression model.

[0159] In this example, the Embedding layer, the Bidirectional LSTM layer, and the Self-Attention layer are used to build a Bi-LSTM model, which inputs the network state feature X(t) and outputs the compression parameter P(t).

[0160] Step S504: training a dynamic compression model.

[0161] In this example, the loss function of the dynamic compression model can use the mean square error (MSE) as the loss function, and the optimization goal is to minimize the weighted sum of compression efficiency and transmission delay.

[0162] Loss = α.MSE(L i )+β.MSE(C ratio )

[0163] Among them, Loss is the loss function; α is the weight of the mean square error of the transmission delay; MSE(L i ) is the mean square error of transmission delay; β is the weight of the mean square error of compression efficiency; MSE(C ratio ) is the mean square error of compression efficiency.

[0164] The optimization algorithm of the dynamic compression model can use the adaptive learning rate optimization algorithm AdamW to optimize the model parameters and use the back propagation algorithm to update the gradient. The gradient update formula is as follows:

[0165]

[0166] Among them, θ t+1 ,θ tRepresent the model parameters at the t+1th and tth iterations respectively; η is the learning rate, a hyperparameter used to control the step size of parameter update; is the gradient of the loss function with respect to the model parameters θ at the tth iteration.

[0167] When training the dynamic compression model on the training set, a large amount of CAN network data can be used for training, and the model parameters can be iteratively optimized until the loss function converges.

[0168] Step S505: verify the dynamic compression model.

[0169] In this example, you can evaluate the model performance on the validation set and tune the model hyperparameters.

[0170] Step S506: deploy a dynamic compression model.

[0171] In this example, the trained Bi-LSTM model can be embedded in the dynamic compression adjustment module to monitor the CAN network status in real time and dynamically adjust the compression algorithm parameters through the Bi-LSTM model.

[0172] Step S507: transferring the real-time monitoring data to the dynamic compression model.

[0173] Step S508: adjust the compression parameters of the compression algorithm.

[0174] Step S509: CAN network transmits data.

[0175] In this example, when transmitting data in the CAN network, the adjusted compression algorithm can be used to compress the transmitted data packets.

[0176] The training and deployment process of the dynamic compression model in this example can evaluate the improvement of the dynamic compression model in compression efficiency and transmission performance through actual operation data, and realize the dynamic compression parameter adjustment of the Bi-LSTM model combined with Self-Attention in the CAN network, which significantly improves the efficiency and performance of data transmission.

[0177] The compression algorithm parameters that can be adjusted by the above dynamic compression model may include:

[0178] Compression Ratio: Adjust the ratio of compressed data. The compression ratio can be changed dynamically according to the network status to balance data transmission efficiency and compression effect.

[0179] Block Size: The size of data blocks affects the compression effect and transmission performance. By dynamically adjusting the block size, the transmission and decompression efficiency can be optimized.

[0180] Iteration Count: In the compression algorithm, increase or decrease the number of iterations to optimize the compression effect and computational overhead.

[0181] Dictionary Size: In dictionary compression algorithms, the dictionary size is adjusted to accommodate different data types and transmission requirements.

[0182] Hash Table Size: In the hash table compression algorithm, adjust the size of the hash table to improve compression and search efficiency.

[0183] Window Size: In the sliding window compression algorithm, the window size is dynamically adjusted to adapt to different transmission rates and data patterns.

[0184] The following are some examples:

[0185] Assuming that the LZ77 compression algorithm is used, the Bi-LSTM model can dynamically adjust the following parameters based on the real-time monitored network status:

[0186] Window size: In the case of high transmission delay or high packet loss rate, increase the window size to reduce the compression frequency and improve data transmission efficiency.

[0187] Dictionary size: When the data pattern changes greatly, increase the dictionary size to improve the compression ratio.

[0188] Block size: When the network status is stable, reduce the block size to increase the compression speed.

[0189] Based on the dynamic compression method of network transmission data in the above application scenario, the specific application implementation steps of the present disclosure in actual engineering are as follows:

[0190] 1. System integration: Integrate the data acquisition module, real-time monitoring module, logic judgment module, feedback mechanism module and dynamic compression module into the vehicle control system.

[0191] 2. Data collection: Start the vehicle system, the data collection module begins to collect CAN network data and inputs it into the real-time monitoring module.

[0192] 3. Real-time monitoring: The monitoring module calculates the transmission efficiency and compression performance of the current network status and generates monitoring data.

[0193] 4. Logical judgment: The logical judgment module determines whether the compression parameters need to be adjusted based on the monitoring data. If the conditions are met, the feedback mechanism is activated.

[0194] 5. Parameter adjustment: The feedback mechanism module sends the monitoring data and judgment results to the dynamic compression module, and the Bi-LSTM model optimizes the compression parameters according to the input data output.

[0195] 6. Compression processing: The dynamic compression module adjusts the compression algorithm parameters and generates compressed data packets to improve transmission efficiency and performance.

[0196] 7. Feedback and optimization: The system runs continuously and continuously optimizes the data compression and transmission process through the feedback mechanism and real-time adjustment of the Bi-LSTM model.

[0197] Through the above steps, the present invention realizes the dynamic adjustment of data compression in the CAN network, improves the efficiency and reliability of data transmission, and adapts to the real-time changing network environment.

[0198] In the present embodiment, a dynamic compression device for network transmission data is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and the descriptions thereof will not be repeated. As used below, the term "module" can implement a combination of software and / or hardware of a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and contemplated.

[0199] This embodiment provides a dynamic compression device for network transmission data, such as Figure 6 As shown, including:

[0200] The data determination module 601 is used to determine target monitoring data based on the transmission data of the target communication network in the first time period; the target monitoring data includes at least one of the following: transmission delay, packet loss rate and compression efficiency;

[0201] A data generating module 602 is used to generate feedback data according to the target monitoring data when the target item in the target monitoring data meets the compression parameter adjustment condition;

[0202] The parameter generation module 603 is used to use the dynamic compression model and the feedback data as input to generate target compression parameters of the compression algorithm used for transmitting data, wherein the target compression parameters include at least one of the following: compression rate, block size, and number of iterations;

[0203] A parameter updating module 604, used to update the parameter settings of the compression algorithm using the target compression parameters;

[0204] The data processing module 605 is used to process the transmission data of the target communication network in a second time period after the first time period by using the compression algorithm with updated parameter settings.

[0205] In some optional implementations of the present embodiment, the dynamic compression model in the parameter generation module 603 includes: an embedding layer, a bidirectional long short-term memory layer, a self-attention layer, and a fully connected layer; the embedding layer is used to take the feedback data as input to generate an initial embedding vector of the network state feature; the bidirectional long short-term memory layer is used to take the initial embedding vector as input to generate an intermediate embedding vector containing the front and back dependency information of the initial embedding vector; the self-attention layer is used to take the intermediate embedding vector as input to generate a target embedding vector including long-distance dependency information and dynamic weight adjustment information; the fully connected layer is used to take the target embedding vector as input to generate the target compression parameters of the compression algorithm corresponding to the transmitted data.

[0206] In some optional implementations of the present embodiment, the dynamic compression model in the parameter generation module 603 is obtained based on a model generation module (not shown in the figure), and the model generation module is used to: take the feedback data samples in the training sample set as input, take the compression parameter samples in the training sample set as output, take minimizing the compression efficiency of the feedback data samples in the training sample set and the transmission delay of the feedback data samples as the loss function, use an adaptive learning rate optimization algorithm to iteratively optimize the parameters of the dynamic compression model, and use a back propagation algorithm to perform gradient update in each iteration until the number of iterations reaches a preset number or the loss function converges; wherein the feedback data samples in the training sample set correspond to the compression parameter samples in the training sample set respectively.

[0207] In some optional implementations of the present embodiment, the model generation module is further used to: use the dynamic compression model after iterative optimization of parameters, take the feedback data samples in the test sample set as input, and generate predicted compression parameters corresponding to the feedback data in the test sample set, wherein the feedback data samples in the test sample set respectively correspond to the compression parameter samples in the test sample set; generate evaluation indicators based on the difference between the predicted compression parameters and the compression parameter samples in the test sample set; adjust the hyperparameters of the dynamic compression model after iterative optimization of parameters according to the evaluation indicators, and use the dynamic compression model after adjusting the hyperparameters to execute the model generation step until the evaluation indicators meet the preset requirements.

[0208] In some optional implementations of this embodiment, the model generation module uses minimizing compression efficiency and transmission delay as a loss function, including: minimizing the weighted sum of the first mean square error and the second mean square error as a loss function, the first mean square error is the mean square error of the compression efficiency; the second mean square error is the mean square error of the transmission delay.

[0209] In some optional implementations of this embodiment, the apparatus further includes: the target item in the target monitoring data in the data generation module 602 meets the compression parameter adjustment condition, including the first limitation; and / or the target compression parameter in the parameter generation module 603 includes the second limitation;

[0210] Among them, the first limitation includes at least one of the following: the central tendency measure value (mean, median, mode, etc.) of the transmission delay is greater than the transmission delay threshold; the central tendency measure value of the compression efficiency is less than the compression efficiency threshold; the packet loss rate is greater than the packet loss rate threshold; the second limitation includes at least one of the following: dictionary size, hash table size and window size.

[0211] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0212] The dynamic compression device for network transmission data in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above functions.

[0213] See also Figure 7 , Figure 7 : is a schematic diagram of the structure of a vehicle provided by an optional embodiment of the present disclosure, the vehicle comprising: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. The various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process instructions executed in the vehicle, including instructions stored in or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple vehicles can be connected, and each device provides some necessary operations (for example, as a vehicle network, or a multi-processor system).

[0214] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0215] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0216] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system and applications required for at least one function; the data storage area may store data created according to the use of the vehicle, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the vehicle 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.

[0217] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0218] The embodiments of the present disclosure also provide a computer-readable storage medium. The above-mentioned method according to the embodiments of the present disclosure can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium and downloaded through a network, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state drive, etc.; further, the storage medium can also include a combination of the above-mentioned types of memory. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0219] A part of the present disclosure may be applied as a computer program product, such as a computer program instruction, which, when executed by a computer, can call or provide the method and / or technical solution according to the present disclosure through the operation of the computer. Those skilled in the art should understand that the existence of computer program instructions in computer-readable media includes, but is not limited to, source files, executable files, installation package files, etc., and accordingly, the way in which computer program instructions are executed by a computer includes, but is not limited to: the computer directly executes the instruction, or the computer compiles the instruction and then executes the corresponding compiled program, or the computer reads and executes the instruction, or the computer reads and installs the instruction and then executes the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to the computer.

[0220] Although the embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for dynamic compression of network transmission data, characterized in that: The method comprises: Determine target monitoring data based on transmission data of the target communication network in the first time period; the target monitoring data includes at least one of the following: transmission delay, packet loss rate and compression efficiency; When the target item in the target monitoring data meets the compression parameter adjustment condition, generating feedback data according to the target monitoring data; Using a dynamic compression model, taking the feedback data as input, generating target compression parameters of a compression algorithm used for the transmission data, the target compression parameters including at least one of the following: compression rate, block size, and number of iterations; updating parameter settings of the compression algorithm using the target compression parameters, and The compression algorithm with the updated parameter setting is used to process the transmission data of the target communication network in a second time period after the first time period.

2. The method according to claim 1, characterized in that The dynamic compression model includes: an embedding layer, a bidirectional long short-term memory layer, a self-attention layer, and a fully connected layer; The using of the dynamic compression model, taking the feedback data as input, and generating target compression parameters of the compression algorithm corresponding to the transmission data, comprises: Using the embedding layer, taking the feedback data as input, to generate an initial embedding vector of network state features; Using the bidirectional long short-term memory layer, taking the initial embedding vector as input, generating an intermediate embedding vector containing forward and backward dependency information of the initial embedding vector; Using the self-attention layer, taking the intermediate embedding vector as input, generating a target embedding vector including long-range dependency information and dynamic weight adjustment information; Using the fully connected layer, the target embedding vector is taken as input to generate target compression parameters of the compression algorithm corresponding to the transmission data.

3. The method according to any one of claims 1 or 2, characterized in that: The dynamic compression model is obtained based on a model generation step, and the model generation step includes: Taking feedback data samples in a training sample set as input, taking compression parameter samples in the training sample set as output, minimizing the compression efficiency of feedback data samples in the training sample set and the transmission delay of the feedback data samples as a loss function, adopting an adaptive learning rate optimization algorithm to iteratively optimize the parameters of the dynamic compression model, and using a back propagation algorithm to perform gradient update at each iteration until the number of iterations reaches a preset number or the loss function converges; The feedback data samples in the training sample set respectively correspond to the compression parameter samples in the training sample set.

4. The method according to claim 3, characterized in that The model generation step further includes: Using the dynamic compression model after iterative parameter optimization, taking the feedback data samples in the test sample set as input, generating predicted compression parameters corresponding to the feedback data in the test sample set, wherein the feedback data samples in the test sample set respectively correspond to the compression parameter samples in the test sample set; generating an evaluation index based on a difference between the predicted compression parameter and a compression parameter sample in the test sample set; According to the evaluation index, the hyperparameters of the dynamic compression model after the iterative optimization parameters are adjusted, and the dynamic compression model after the hyperparameters are adjusted is used to execute the model generation step until the evaluation index meets the preset requirements.

5. The method according to claim 3, characterized in that: Taking minimizing the compression efficiency and the transmission delay as a loss function includes: taking minimizing a weighted sum of a first mean square error and a second mean square error as a loss function, wherein the first mean square error is the mean square error of the compression efficiency; and the second mean square error is the mean square error of the transmission delay.

6. The method according to claim 1, characterized in that The method further comprises: The target item in the target monitoring data meets the compression parameter adjustment condition, including the first limitation; and / or The target compression parameters include a second limitation; The first limitation includes at least one of the following: the central tendency metric value of the transmission delay is greater than the transmission delay threshold; the central tendency metric value of the compression efficiency is less than the compression efficiency threshold; the packet loss rate is greater than the packet loss rate threshold; The second limitation includes at least one of the following: dictionary size, hash table size and window size.

7. A dynamic compression device for network transmission data, characterized in that: The device comprises: A data determination module, configured to determine target monitoring data based on transmission data of a target communication network in a first time period; the target monitoring data includes at least one of the following: transmission delay, packet loss rate, and compression efficiency; A data generating module, configured to generate feedback data according to the target monitoring data when the target item in the target monitoring data meets the compression parameter adjustment condition; A parameter generation module, configured to use a dynamic compression model and the feedback data as input to generate a target compression parameter of a compression algorithm used for the transmission data, wherein the target compression parameter includes at least one of the following: compression rate, block size, and number of iterations; a parameter updating module, configured to update the parameter settings of the compression algorithm using the target compression parameters, and The data processing module is used to process the transmission data of the target communication network in a second time period after the first time period by using a compression algorithm with updated parameter settings.

8. A vehicle, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the dynamic compression method for network transmission data according to any one of claims 1 to 6 by executing the computer instructions.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the dynamic compression method for network transmission data according to any one of claims 1 to 6.

10. A computer program product, characterized in that The method comprises computer instructions, wherein the computer instructions are used to enable a computer to execute the dynamic compression method for network transmission data according to any one of claims 1 to 6.

Images