A method and system for redundant data retransmission in vehicle-to-everything (V2X) networks based on dynamic priority scheduling

Abstract

This application provides a method and system for redundant data retransmission in vehicle-to-everything (V2X) networks based on dynamic priority scheduling, relating to the field of V2X technology for new energy vehicles. The method includes: acquiring multimodal vehicle state time-series data; analyzing the multimodal vehicle state time-series data using an early warning prediction model to predict the early warning probability of the data; determining the retransmission priority of multimodal data based on the timeliness, data volume, and early warning probability of the multimodal data; performing layered compression processing on different types of data based on the retransmission priority to obtain compressed data; storing the compressed data in an emergency buffer and / or main storage area according to preset rules, and retransmitting it according to the retransmission priority after network recovery. This application solves the technical problems of high data redundancy overhead and low retransmission efficiency in existing technologies for redundant data retransmission in V2X networks based on dynamic priority scheduling.

Description

Technical Field

[0001] This application relates to the field of new energy vehicle Internet of Vehicles technology, and in particular to a method and system for supplementing redundant data transmission in the Internet of Vehicles based on dynamic priority scheduling. Background Technology

[0002] In complex traffic environments and under wireless communication conditions, vehicle-to-everything (V2X) data transmission is susceptible to factors such as channel fading, network congestion, and node switching, leading to data loss or transmission delays. To ensure system security and reliability, redundant data retransmission mechanisms are typically designed to retransmit lost data after network recovery, restoring data integrity and continuity. Existing V2X data retransmission technologies mostly employ static priority or fixed-rule scheduling strategies, i.e., retransmission is uniformly ordered based on data type or sampling time. These methods lack comprehensive consideration of differences in data importance, changes in data timeliness, and the dynamic nature of vehicle operation risks, easily leading to problems such as delayed retransmission of high-risk data or bandwidth consumption by low-value data, resulting in high system redundancy overhead and low retransmission efficiency. Therefore, a data retransmission method is urgently needed to solve the technical problems of high redundancy overhead and low retransmission efficiency in vehicle-mounted data retransmission. Summary of the Invention

[0003] This application provides a method and system for redundant data retransmission in vehicle networking based on dynamic priority scheduling, which solves the technical problems of large data redundancy overhead and low retransmission efficiency in existing technologies.

[0004] To achieve the above objectives, this application adopts the following technical solution:

[0005] Firstly, a method for redundant data retransmission in vehicle-to-everything (V2X) networks based on dynamic priority scheduling is provided, comprising: acquiring multimodal vehicle state time-series data, including: drive motor temperature, battery temperature, voltage, current, insulation alarm, tire pressure abnormality, door status, vehicle operating parameters, GPS data, and environmental data; wherein, vehicle operating parameters include: speed, acceleration, and load; analyzing the multimodal vehicle state time-series data through an early warning prediction model to predict the early warning probability of the data; the early warning prediction model is a hybrid prediction model constructed using a Transformer coding layer and a Long Short-Term Memory (LSTM) network, used to analyze the multimodal vehicle state time-series data. Feature extraction and temporal correlation modeling are performed to predict the early warning probability of the data. Based on the timeliness, data volume, and early warning probability of multimodal vehicle status time series data, the priority for retransmission of multimodal vehicle status time series data is determined. Based on the retransmission priority of multimodal vehicle status time series data, different types of data are subjected to layered compression processing to obtain compressed data. Among them, continuous data is processed using differential coding and Huffman compression algorithms, and discrete alarm data is compressed using bit-field merging coding. The compressed data is stored in the emergency buffer and / or main storage area according to preset rules, and is retransmitted according to the retransmission priority after the network is restored.

[0006] In conjunction with the first aspect mentioned above, one possible implementation involves analyzing multimodal vehicle state time-series data using an early warning prediction model to predict the early warning probability of the data. This includes: performing global dependency modeling on the multimodal vehicle state time-series data using the time-aware position encoding mechanism of the Transformer encoding layer in the early warning prediction model to obtain a high-dimensional representation vector containing multimodal time-series features; extracting the trend features of the high-dimensional representation vector's change in the time dimension through the LSTM network layer in the early warning prediction model to obtain time-series correlation features; and obtaining the early warning probability of the input data through the prediction layer in the early warning prediction model based on the time-series correlation features.

[0007] In conjunction with the first aspect mentioned above, in one possible implementation, the time-aware position encoding mechanism of the Transformer encoding layer in the early warning prediction model refers to adding time interval information to the standard sinusoidal position encoding. The time-aware position encoding mechanism satisfies the following formula:

[0008]

[0009]

[0010] in, The value encoded for the position of time step pos in 2i dimensions. Let be the positional encoding value of time step pos in dimension 2i+1, where pos is the index of the current sampling point in the time series data, and i is the index of the positional encoding dimension. For the hidden layer dimension, As a time-enhancing factor, and This is a time interval mapping function. The time interval between adjacent sampling;

[0011] In conjunction with the first aspect mentioned above, in one possible implementation, the high-dimensional representation vector... Satisfy the following formula:

[0012]

[0013] in, The data represents the time series of multimodal vehicle states at time t. The location code value at time t. This indicates the Transformer encoding layer.

[0014] In conjunction with the first aspect mentioned above, in one possible implementation, the priority for retransmitting multimodal vehicle status time-series data is determined based on the timeliness, data volume, and warning probability of the multimodal vehicle status time-series data. This includes: determining the warning type based on the warning probability using preset corresponding rules; calculating a comprehensive retransmission priority score for multimodal vehicle status time-series data based on its timeliness, data volume, warning probability, and warning type; the timeliness of the multimodal vehicle status time-series data includes the urgency weight and freshness of the data; and sorting the comprehensive retransmission priority scores to obtain the retransmission priority of the multimodal vehicle status time-series data.

[0015] In conjunction with the first aspect mentioned above, one possible implementation involves prioritizing the supplementary transmission of multimodal vehicle state timing data. Satisfy the following formula:

[0016]

[0017] in, Weights for data attribute items in the score. To score the urgency level, As the weight for the urgency score, Score the warning type. As the scoring weight for warning type, To score the freshness of the data, Assign weights to the data freshness score. Score the data volume. Weights for data volume scores. The early warning probability score is calculated as follows. The weighting is used for the early warning probability score.

[0018] In conjunction with the first aspect mentioned above, in one possible implementation, the early warning prediction model adopts a combination of offline training and online incremental learning; offline training refers to training the initial model using historical alarm datasets; online incremental learning refers to dynamically updating the model parameters based on real-time collected multimodal vehicle state time-series data, training the model to adapt to individual vehicle characteristics.

[0019] Secondly, a vehicle-to-everything (V2X) redundant data retransmission system based on dynamic priority scheduling is provided, including: an onboard data acquisition module, an early warning and prediction module, a priority scheduling module, a hierarchical compression module, and a redundant storage and retransmission module;

[0020] The system includes: an onboard data acquisition module for acquiring multimodal vehicle status time-series data, including drive motor temperature, battery temperature, voltage, current, insulation alarm, tire pressure anomaly, door status, vehicle operating parameters, GPS data, and environmental data; vehicle operating parameters include speed, acceleration, and load; a warning and prediction module for extracting multimodal features and modeling temporal correlations based on the multimodal vehicle status time-series data using a hybrid prediction model constructed with a Transformer encoding layer and a Long Short-Term Memory network, and outputting warning probabilities; and a priority scheduling module for prioritizing data timeliness. The system calculates the comprehensive retransmission priority score of multimodal vehicle status time-series data based on data volume and early warning probability, and determines the retransmission order according to the score. A hierarchical compression module is used to compress different types of multimodal vehicle status time-series data based on retransmission priority. Continuous data is compressed using differential encoding and Huffman compression algorithms, while discrete alarm data is compressed using bit-field merging encoding. A redundant storage and retransmission module stores the compressed data into an emergency buffer and / or main storage area according to preset rules, and executes data retransmission tasks sequentially according to retransmission priority after network recovery.

[0021] In conjunction with the second aspect above, in one possible implementation, the early warning prediction module includes: a Transformer encoding submodule, a long short-term memory submodule, and a prediction output submodule;

[0022] The Transformer encoding submodule is used to perform time-aware location encoding and global feature extraction on the input multimodal time-series data; the Long Short-Term Memory submodule is used to extract the temporal trend of global features; and the prediction output submodule is used to calculate the warning probability based on the trend features.

[0023] In conjunction with the second aspect mentioned above, in one possible implementation, the redundant storage and retransmission module includes a two-level structure of an emergency buffer and a main storage area; the emergency buffer is used to store high-priority or high-probability data; the main storage area is used to store ordinary data; after the network is restored, the system retrieves data from the emergency buffer and the main storage area in order of priority and performs retransmission.

[0024] In conjunction with the second aspect above, in one possible implementation, the hierarchical compression module includes a data type identification unit, which is used to adaptively select difference coding, Huffman compression, and bit field merging algorithms for processing based on the continuity characteristics, redundancy threshold, and real-time transmission bandwidth of the multimodal vehicle state time series data.

[0025] Thirdly, this application provides an electronic device, including: a processor and a storage medium; the storage medium includes instructions, and the processor is configured to execute the instructions to implement the methods described in the first aspect and any possible implementation thereof. This electronic device may be an electronic device or a chip within an electronic device.

[0026] Fourthly, this application provides a computer program product containing instructions that, when run on an electronic device, cause the electronic device to perform the methods described in the first aspect and any possible implementation thereof.

[0027] This application provides a method and system for redundant data retransmission in vehicle-to-everything (V2X) networks based on dynamic priority scheduling. It utilizes a prediction model combining a Transformer coding layer and a Long Short-Term Memory (LSTM) network to perform global feature extraction and temporal correlation modeling on multimodal time-series data such as vehicle operating status, battery information, and environmental parameters, thereby accurately predicting the early warning probability of the data. By incorporating data timeliness, data volume, and early warning probability into the priority calculation model, the retransmission order is dynamically and adaptively adjusted. Simultaneously, layered compression strategies such as differential encoding, Huffman compression, and bit-field merging encoding are designed for different types of data to effectively reduce transmission load. Combined with a dual-layer storage mechanism of an emergency buffer and a main storage area, high-risk and time-sensitive data can be prioritized for retransmission after network recovery, thereby improving the accuracy, real-time performance, and reliability of redundant data retransmission in V2X environments. This solves the technical problems of high data redundancy overhead and low retransmission efficiency in existing technologies for vehicle-mounted retransmission.

[0028] It should be understood that the descriptions of technical features, technical solutions, beneficial effects, or similar language in this application do not imply that all features and advantages can be achieved in any single embodiment. Rather, it is understood that the description of a feature or beneficial effect means that a specific technical feature, technical solution, or beneficial effect is included in at least one embodiment. Therefore, the descriptions of technical features, technical solutions, or beneficial effects in this specification do not necessarily refer to the same embodiment. Furthermore, the technical features, technical solutions, and beneficial effects described in this embodiment can be combined in any suitable manner. Those skilled in the art will understand that embodiments can be implemented without one or more specific technical features, technical solutions, or beneficial effects of a particular embodiment. In other embodiments, additional technical features and beneficial effects may be identified in specific embodiments that do not embody all embodiments. Attached Figure Description

[0029] Figure 1 A system architecture diagram of a vehicle-to-everything (V2X) redundant data retransmission system based on dynamic priority scheduling is provided in an embodiment of this application.

[0030] Figure 2 A flowchart illustrating a method for redundant data retransmission in a vehicle-to-everything (V2X) network based on dynamic priority scheduling, provided in an embodiment of this application.

[0031] Figure 3 A flowchart illustrating another method for redundant data retransmission in vehicle-to-everything (V2X) networks based on dynamic priority scheduling, provided in an embodiment of this application.

[0032] Figure 4 This is a flowchart illustrating another method for redundant data retransmission in vehicle networking based on dynamic priority scheduling, provided in an embodiment of this application. Detailed Implementation

[0033] In the description of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. The "and / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone. Furthermore, "at least one" means one or more, and "multiple" means two or more. The terms "first," "second," etc., do not limit the quantity or order of execution, and "first," "second," etc., do not necessarily imply differences.

[0034] It should be noted that, in this application, the terms "exemplary" or "for example" are used to indicate that something is being described as an example, illustration, or illustration. Any embodiment or design described as "exemplary" or "for example" in this application should not be construed as being more preferred or advantageous than other embodiments or design solutions. Specifically, the use of terms such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0035] The vehicle-to-everything (V2X) redundant data retransmission method based on dynamic priority scheduling provided in this application can be applied to, for example... Figure 1 In the vehicular network redundant data retransmission system shown, based on dynamic priority scheduling, as follows: Figure 1 As shown, the system includes: vehicle data acquisition module 101, early warning and prediction module 102, priority scheduling module 103, hierarchical compression module 104, and redundant storage and retransmission module 105.

[0036] The system includes: an onboard data acquisition module for acquiring multimodal vehicle status time-series data, including drive motor temperature, battery temperature, voltage, current, insulation alarm, tire pressure anomaly, door status, vehicle operating parameters, GPS data, and environmental data; vehicle operating parameters include speed, acceleration, and load; a warning and prediction module for extracting multimodal features and modeling temporal correlations based on the multimodal vehicle status time-series data using a hybrid prediction model constructed with a Transformer encoding layer and a Long Short-Term Memory network, and outputting warning probabilities; and a priority scheduling module for prioritizing data timeliness. The system calculates the comprehensive retransmission priority score of multimodal vehicle status time-series data based on data volume and early warning probability, and determines the retransmission order according to the score. A hierarchical compression module is used to compress different types of multimodal vehicle status time-series data based on retransmission priority. Continuous data is compressed using differential encoding and Huffman compression algorithms, while discrete alarm data is compressed using bit-field merging encoding. A redundant storage and retransmission module stores the compressed data into an emergency buffer and / or main storage area according to preset rules, and executes data retransmission tasks sequentially according to retransmission priority after network recovery.

[0037] In one possible implementation, the early warning prediction module includes: a Transformer encoding submodule, a long short-term memory submodule, and a prediction output submodule;

[0038] The Transformer encoding submodule is used to perform time-aware location encoding and global feature extraction on the input multimodal time-series data; the Long Short-Term Memory submodule is used to extract the temporal trend of global features; and the prediction output submodule is used to calculate the warning probability based on the trend features.

[0039] In one possible implementation, the redundant storage and retransmission module includes a two-level structure of an emergency buffer and a main storage area; the emergency buffer is used to store high-priority or high-probability data; the main storage area is used to store ordinary data; after the network is restored, the system retrieves data from the emergency buffer and the main storage area in order of priority and performs retransmission.

[0040] In one possible implementation, the hierarchical compression module includes a data type identification unit, which adaptively selects difference coding, Huffman compression, and bit field merging algorithms for processing based on the continuity characteristics, redundancy threshold, and real-time transmission bandwidth of the multimodal vehicle state time series data.

[0041] To address the technical problems of high redundancy overhead and low transmission efficiency in existing vehicle-mounted data retransmission technologies, this application provides a method for redundant data retransmission in vehicle-to-everything (V2X) networks based on dynamic priority scheduling. This method includes: acquiring multimodal vehicle state time-series data, including: drive motor temperature, battery temperature, voltage, current, insulation alarm, tire pressure anomaly, door status, vehicle operating parameters, GPS data, and environmental data; wherein, vehicle operating parameters include: speed, acceleration, and load; analyzing the multimodal vehicle state time-series data through a warning prediction model to predict the warning probability of the data; the warning prediction model is constructed using a Transformer encoding layer and a Long Short-Term Memory (LSTM) network. A hybrid prediction model is used to extract features and model temporal correlations of vehicle status time-series data to predict the early warning probability. Based on the timeliness, data volume, and early warning probability of multimodal vehicle status time-series data, the priority for retransmission of multimodal vehicle status time-series data is determined. Based on the retransmission priority of multimodal vehicle status time-series data, different types of data are subjected to layered compression processing to obtain compressed data. Among them, continuous data is processed using differential encoding and Huffman compression algorithms, and discrete alarm data is compressed using bit-field merging encoding. The compressed data is stored in the emergency buffer and / or main storage area according to preset rules, and is retransmitted according to the retransmission priority after network recovery.

[0042] Figure 2 This is a flowchart illustrating the method for redundant data retransmission in vehicle-to-everything (V2X) networks based on dynamic priority scheduling, as provided in the embodiments of this application. Figure 2 As shown, the method includes:

[0043] S201. Obtain multimodal vehicle state time series data.

[0044] The multimodal vehicle status time-series data includes, but is not limited to: drive motor temperature, battery temperature, voltage, current, insulation alarm, abnormal tire pressure, door status, vehicle operating parameters, GPS data, and environmental data; among which, vehicle operating parameters include: speed, acceleration, and load.

[0045] In one possible implementation, the vehicle data acquisition module polls or passively receives data reported by various sensors and ECUs via CAN bus, LIN, or direct I / O interface. It performs preliminary verification, noise reduction, and zero-value filtering on the raw measurement values ​​according to a preset sampling frequency to form standardized data, which is accompanied by a timestamp.

[0046] It should be noted that sampling may be uneven, therefore the system should save the sampling interval during recording. Furthermore, hardware interrupts or high-priority channels should be configured for critical signals to prevent loss.

[0047] As an example, in this embodiment, the multimodal vehicle state time-series data acquisition module is integrated into the vehicle's central control unit (VCU), communicating directly with the battery management system, environmental sensor module, and vehicle operation control unit. The sampling frequency is set to 10Hz–100Hz. The acquired data is preprocessed and input into the subsequent model in time-series format.

[0048] S202. Analyze multimodal vehicle status time series data through early warning prediction models to predict the early warning probability of the data.

[0049] Among them, the early warning prediction model is a hybrid prediction model constructed using a Transformer coding layer and a long short-term memory network. It is used to extract features and model the temporal correlation of vehicle status time series data to predict the early warning probability of the data.

[0050] In one possible implementation, the input sequence data is used to perform global dependency modeling on the multimodal vehicle state time series data through the time-aware location encoding mechanism of the Transformer encoding layer in the early warning prediction model, thereby obtaining a high-dimensional representation vector containing multimodal time series features. The LSTM network layer in the early warning prediction model is used to extract the trend features of the high-dimensional representation vector in the time dimension to obtain the time series correlation features. Based on the time series correlation features, the early warning probability of the input data is obtained through the prediction layer in the early warning prediction model.

[0051] It should be noted that when deploying models on vehicles, both computational overhead and accuracy should be considered. Lightweight deployment methods such as model pruning, quantization, or knowledge distillation can be adopted.

[0052] As an example, in this embodiment of the application, a hybrid model is trained using historical alarm datasets and exported in TensorFlow Lite format during the offline phase, and the vehicle-mounted device periodically loads the model; when a large number of new samples or distribution drift occurs, the online incremental update module is enabled, the parameter difference summary is uploaded to the cloud for federated or centralized updates, and then the latest weights are distributed to the vehicle-mounted device to maintain model adaptability.

[0053] S203. Based on the timeliness, data volume, and early warning probability of multimodal vehicle status time series data, determine the priority of retransmission of multimodal vehicle status time series data.

[0054] Timeliness includes the freshness and urgency of the data; data volume refers to the size of the data during transmission or the estimated bandwidth usage; and retransmission priority is a value used by the system to determine the retransmission order after sorting by comprehensive score.

[0055] In one possible implementation, the warning type is determined by a preset corresponding rule based on the warning probability; a comprehensive retransmission priority score for the multimodal vehicle status time series data is calculated based on the timeliness, data volume, warning probability, and the warning type of the multimodal vehicle status time series data; the timeliness of the multimodal vehicle status time series data includes the urgency weight and freshness of the multimodal vehicle status time series data; the comprehensive retransmission priority score is sorted to obtain the retransmission priority of the multimodal vehicle status time series data.

[0056] S204. Based on the priority of multimodal vehicle state time-series data retransmission, different types of data are subjected to layered compression processing to obtain compressed data.

[0057] Continuous data is processed using differential encoding and Huffman compression algorithms, while discrete alarm data is compressed using bit-field merging encoding.

[0058] In one possible implementation, for continuous data such as temperature, voltage, and current, a combination of differential encoding and Huffman compression is used to encode the data to remove redundancy and reduce information entropy; for discrete alarm or status data, bit-field merging encoding is used to compress multiple Boolean states into a single byte or bit stream to improve transmission efficiency.

[0059] It should be noted that this layered compression strategy takes into account both real-time performance and recoverability. The system records metadata indexes and restoration parameters during compression, which enables the original numerical sequence to be accurately restored during the data decompression stage.

[0060] As an example, in an embodiment of this application, the difference encoding process satisfies: ,in, This represents the difference at the current time step. The original values ​​of adjacent time steps are represented separately, and Huffman coding is applied to the difference sequence to achieve a compression ratio of 40%–60%. For alarm data, eight alarm states are compressed into one byte through bit field merging, further reducing bandwidth usage. The compression layer is divided into three layers—high, medium, and low—based on the retransmission priority. High-priority data uses a low compression ratio strategy to ensure accuracy, while low-priority data uses a high compression ratio to save resources.

[0061] S205. Store the compressed data in the emergency buffer and / or main storage area according to the preset rules, and retransmit it according to the retransmission priority after the network is restored.

[0062] The emergency buffer stores high-priority compressed data, while the main storage area stores low-priority or non-critical data. This partitioning mechanism ensures that critical data is prioritized for retransmission once communication is restored.

[0063] In one possible implementation, the system sets a storage threshold based on a calculated priority score. For example, data with a priority greater than 0.7 enters the emergency buffer, while data with a priority less than 0.7 is stored in the main storage area. When the communication link is restored, the system sends data packets sequentially from highest to lowest priority to achieve ordered retransmission.

[0064] It should be noted that, in order to prevent cache overflow, the system can adopt a FIFO strategy and a compression ratio self-adjustment mechanism to ensure that high-priority data is stored persistently, while low-priority data is delayed or discarded when space is insufficient.

[0065] As an example, the emergency buffer is implemented using high-performance flash memory modules with a write latency of less than 10ms; the main storage area is based on local SSD storage and is suitable for batch data caching.

[0066] As an example, in this embodiment of the application, the data retransmission order satisfies the following formula:

[0067]

[0068] in, Let be the overall priority score of the i-th data, and desc be the cache area.

[0069] This application's embodiments achieve comprehensive perception of vehicle operating status through multimodal vehicle state time-series data acquisition. By integrating multi-source information such as drive motor temperature, battery temperature, voltage, current, insulation alarm, tire pressure anomaly, door status, speed, acceleration, load, GPS, and external environmental data, the model can reflect the vehicle's physical, electronic, and operating environment status from multiple perspectives, thereby improving the coverage and robustness of fault warnings. By introducing a Transformer encoding layer to encode the time-series data, the advantages of the self-attention mechanism in long-range dependency modeling are fully utilized to capture the global dependencies between multidimensional state signals. Through a multi-head attention computation mechanism, the interactive correlations between different time steps and different modal features are explicitly modeled, thereby obtaining high-level semantic feature representations. A Long Short-Term Memory (LSTM) network is used to extract the temporal correlations of the Transformer output features. Its gating structure preserves long-term dependencies and filters short-term interference, further enhancing the model's ability to characterize dynamic evolution trends. The synergy of these two approaches makes the prediction results more consistent with real physical changes. Finally, by using partitioned data transfer, the efficiency of data transfer is improved, solving the technical problems of large data redundancy overhead and low transfer efficiency in existing vehicle-mounted data transfer technologies.

[0070] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 3 As shown, S202 above analyzes multimodal vehicle state time series data through an early warning prediction model. The early warning probability of the predicted data can be specifically achieved through S301, S302 and S303, which are explained in detail below:

[0071] S301. By using the time-aware location encoding mechanism of the Transformer encoding layer in the early warning prediction model, global dependency modeling is performed on the multimodal vehicle state time series data to obtain a high-dimensional representation vector containing multimodal time series features.

[0072] Among them, the time-aware position coding of this application is a mechanism that introduces time interval information on the basis of standard sinusoidal position coding, which is used to enhance the model's ability to model data with unequal sampling intervals.

[0073] In one possible implementation, a time-aware position vector is superimposed on the data at each time step before it is input into the Transformer to describe the temporal position information and time interval characteristics of the sampling points, thereby improving the model's ability to capture long-term dependencies. Standard sinusoidal coding assumes that time steps are equally spaced, but vehicle operation data often exhibits non-uniform sampling. Therefore, this invention introduces a time enhancement factor to correct this assumption. Based on this, the time-aware position encoding mechanism of the Transformer encoding layer in the early warning prediction model is used to perform global dependency modeling on multimodal vehicle state time-series data, obtaining a high-dimensional representation vector containing multimodal time-series features.

[0074] As an example, in this embodiment of the application, the time-aware position encoding mechanism of the Transformer encoding layer in the early warning prediction model refers to adding time interval information to the standard sinusoidal position encoding. The time-aware position encoding mechanism satisfies the following formula:

[0075]

[0076]

[0077] in, The value encoded for the position of time step pos in 2i dimensions. Let be the positional encoding value of time step pos in dimension 2i+1, where pos is the index of the current sampling point in the time series data, and i is the index of the positional encoding dimension. For the hidden layer dimension, As a time-enhancing factor, and This is a time interval mapping function. The time interval between adjacent sampling;

[0078] As an example, in an embodiment of this application, a high-dimensional representation vector is used. Satisfy the following formula:

[0079]

[0080] in, The data represents the time series of multimodal vehicle states at time t. The location code value at time t. This indicates the Transformer encoding layer.

[0081] S302. By using the LSTM network layer in the early warning prediction model, the changing trend features of the high-dimensional representation vector in the time dimension are extracted to obtain the time-series correlation features.

[0082] In one possible implementation, the high-dimensional feature sequence output by the Transformer is sequentially input into the LSTM layer. The LSTM network layer further models the dynamic evolution relationship of multimodal temporal features in the time dimension. The LSTM controls the information flow through a gating mechanism to retain long-term dependent features and suppress invalid information, extracting the trend features of the high-dimensional representation vector in the time dimension to obtain the temporal correlation features.

[0083] It should be noted that the LSTM layer and the Transformer layer are connected through a linear transformation interface to achieve dimension matching and information continuity.

[0084] As an example, an LSTM layer consists of two layers, each with 128 hidden units, employing tanh and sigmoid activation functions to achieve interactive modeling of long-term and short-term features. Hidden states in an LSTM... The update satisfies the following formula:

[0085]

[0086] in, For activation function, This is the weight matrix. For bias terms, It is a high-dimensional representation vector.

[0087] S303. Based on the temporal correlation characteristics, the early warning probability of the input data is obtained through the prediction layer in the early warning prediction model.

[0088] In one possible implementation, the prediction layer is configured as a fully connected network that maps temporal features to the alarm probability space. Specifically, the hidden state of the last time step output by the LSTM is processed by a linear layer and a softmax layer to obtain the probability distribution of various alarm events.

[0089] It should be noted that the model can output a single alarm probability or a multi-label probability to support the identification of multiple types of anomalies.

[0090] As an example, the prediction layer uses the Sigmoid function to output a continuous value between 0 and 1 as the probability of warning for the data. (Output warning probability) Satisfy the following formula:

[0091]

[0092] in, For the Sigmoid function, For output of the hidden layer, This is the weight matrix. This is a bias term.

[0093] In one possible implementation, the early warning prediction model constructed by S301-S303 above adopts a combination of offline training and online incremental learning; offline training refers to training the initial model using historical alarm datasets; online incremental learning refers to dynamically updating the model parameters based on real-time collected multimodal vehicle state time series data, training the model to adapt to individual vehicle characteristics.

[0094] This application's embodiments utilize a deep fusion structure of the Transformer encoding layer and a Long Short-Term Memory (LSTM) network to achieve high-dimensional feature extraction and dynamic trend analysis of multimodal vehicle state time-series data. Through a time-aware position encoding mechanism, the model introduces time interval information on top of standard sinusoidal position encoding. This allows the input sequence to not only contain position information but also reflect changes in the sampling time interval, enabling the Transformer to effectively model multimodal time-series data with unequal sampling intervals and enhancing its ability to express global dependencies. This time-aware mechanism allows the model to adaptively capture the temporal correlations between different modalities when processing multi-source vehicle network data with different sampling frequencies and asynchronicity, effectively improving the temporal continuity and robustness of feature representation. By further extracting temporal features from the high-dimensional representation vector output by the Transformer through the LSTM network layer, and effectively capturing the long-term dependence and short-term fluctuation features of state changes using its gating structure, dynamic trend modeling of vehicle operating status is realized. This avoids the problem of local time feature loss that may occur when relying solely on the Transformer for sequence modeling. It achieves a hierarchical modeling process from global dependency features to temporal dynamic features and then to risk probability output, thereby improving the early warning accuracy and response timeliness of intelligent analysis of vehicle network data.

[0095] In one possible implementation of the embodiments of this application, combined with Figure 2 ,like Figure 4 As shown, the above-mentioned S203 determines the priority of retransmission of multimodal vehicle status time series data based on the timeliness, data volume, and early warning probability of the multimodal vehicle status time series data. This can be specifically achieved through the following S401, S402, and S403, which are explained in detail below:

[0096] S401. Based on the early warning probability, the early warning type is determined by preset corresponding rules.

[0097] The warning type refers to the category of potential anomalies or risks predicted based on vehicle operating status data, used to differentiate data processing strategies for different risk levels. Warning types may include, but are not limited to: battery overheating, drive motor malfunction, voltage fluctuations, low tire pressure, insulation alarm, and communication interruption. Each type has a different priority weight in the retransmission strategy.

[0098] In one possible implementation, the system first performs a hierarchical mapping of different warning intervals based on the warning probability output by the warning prediction model and a set threshold interval. Through this hierarchical mechanism, continuous probability values ​​are discretized into warning type labels, and corresponding warning scores are obtained.

[0099] It should be noted that the warning type depends not only on the current warning probability, but can also be dynamically adjusted based on the probability trend within a time window. For example, when the average probability over several consecutive time steps has been rising above a certain threshold, the warning level can be raised in advance, thereby enabling early intervention for trend-based risks.

[0100] As an example, in this embodiment of the application, a certain warning type is set as follows: [0, 0.3) is the low-risk range, corresponding to the normal state; [0.3, 0.7) is the medium-risk range, corresponding to the minor abnormality type; [0.7, 1] is the high-risk range, corresponding to the serious abnormality or critical alarm type. If the vehicle battery temperature rises from 45°C to 65°C in the past 10 seconds, and the predicted probability is... =0.82, the system classifies the data as "battery overheating - high risk", which corresponds to the highest priority category.

[0101] S402. Based on the timeliness, data volume, early warning probability, and early warning type of multimodal vehicle status time series data, calculate the priority score for comprehensive retransmission of multimodal vehicle status time series data.

[0102] The timeliness of multimodal vehicle state time series data includes the urgency weight of the multimodal vehicle state time series data and the freshness of the multimodal vehicle state time series data.

[0103] In one possible implementation, several scoring factors are defined for the data, including but not limited to the timeliness of multimodal vehicle state time series data, data volume, warning probability, and warning type. Weights are assigned to each scoring factor according to its importance, and priority scores are calculated by weighted summation.

[0104] As an example, in this embodiment of the application, the priority score for comprehensive retransmission of multimodal vehicle state timing data is used. Satisfy the following formula:

[0105]

[0106] in, Weights for data attribute items in the score. To score the urgency level, As the weight for the urgency score, Score the warning type. As the scoring weight for warning type, To score the freshness of the data, Assign weights to the data freshness score. Score the data volume. Weights for data volume scores. The early warning probability score is calculated as follows. The weighting is used for the early warning probability score.

[0107] S403. Sort the comprehensive retransmission priority scores to obtain the retransmission priority of multimodal vehicle status time series data.

[0108] In one possible implementation, the system collects the scores of all data items to be retransmitted, sorts them in descending order using a sorting function, and obtains the retransmission execution sequence.

[0109] In another possible implementation, to prevent information distortion caused by long-term delays in low-priority data, the system can set a retransmission saturation factor. When the cumulative delay time exceeds a threshold, its priority is automatically increased, implementing a "dynamic priority compensation" mechanism. This mechanism can effectively prevent long-term data backlog or the omission of important information.

[0110] As an example, when communication is restored, if the system detects that the motor temperature data score is 0.92, the tire pressure data score is 0.85, and the door status data score is 0.45, then it will perform the retransmission task in the order of motor temperature, tire pressure, and door status to maximize the security benefits of transmission bandwidth.

[0111] This application embodiment maps the original probability output to a clear warning type by matching the warning probability with preset rules, realizing the transformation from numerical-level risk identification to semantic-level risk classification, enabling the system to more intuitively identify potential fault risks. Based on this, the timeliness, volume, warning probability, and warning type of multimodal vehicle status time-series data are comprehensively evaluated to form a dynamic priority scoring mechanism. This allows the system to prioritize the retransmission of high-risk, high-timeliness data when bandwidth is limited, improving the availability and integrity of critical data. The retransmission order is automatically generated based on the comprehensive score results, ensuring an optimal balance between urgency, transmission cost, and real-time performance for various types of data. Based on these methods, the system achieves a transformation from static strategy to dynamic scheduling, significantly improving the accuracy of data retransmission and resource utilization, providing stronger assurance capabilities for reliable communication in the vehicle-to-everything (V2X) environment.

[0112] The above primarily describes the solutions of the embodiments of this application from the perspective of device implementation. It is understood that each device, such as a vehicle-to-everything (V2X) redundant data retransmission system based on dynamic priority scheduling, includes at least one of the hardware structures and software modules corresponding to each function in order to achieve the above-mentioned functions. Those skilled in the art should readily recognize that, based on the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein, this application can be implemented in hardware or a combination of hardware and computer software. Whether a function is executed in hardware or by computer software driving hardware depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0113] This application embodiment can divide the vehicle-to-everything (V2X) redundant data retransmission system based on dynamic priority scheduling into functional units according to the above method example. For example, each function can be divided into its own functional unit, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or as a software functional unit. It should be noted that the unit division in this application embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0114] The processor in this application may include, but is not limited to, at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller unit (MCU), or an artificial intelligence processor, etc., which are various computing devices that run software. Each computing device may include one or more cores for executing software instructions to perform calculations or processing. The processor may be a separate semiconductor chip or integrated with other circuits into a single semiconductor chip. For example, it may be integrated with other circuits (such as encoding / decoding circuits, hardware acceleration circuits, or various bus and interface circuits) to form a SoC (System-on-a-Chip), or it may be integrated as a built-in processor within an ASIC. The ASIC with the integrated processor may be packaged separately or together with other circuits. In addition to the cores for executing software instructions to perform calculations or processing, the processor may further include necessary hardware accelerators, such as field-programmable gate arrays (FPGAs), PLDs (programmable logic devices), or logic circuits that implement dedicated logic operations.

[0115] The memory in the embodiments of this application may include at least one of the following types: read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions; random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions; or electrically erasable programmable-only memory (EEPROM). In some scenarios, the memory may also be a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto.

[0116] This application also provides a computer-readable storage medium including instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0117] This application also provides a computer program product containing instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0118] This application also provides a chip including a processor and an interface circuit. The interface circuit is coupled to the processor. The processor is used to run computer programs or instructions to implement the above-described method. The interface circuit is used to communicate with other modules outside the chip.

[0119] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0120] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings, disclosure, and appended claims, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude multiple instances. A single processor or other unit can implement several functions listed in the claims. While different dependent claims may recite certain measures, this does not mean that these measures cannot be combined to produce good results.

[0121] Although this application has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made thereto without departing from the spirit and scope of this application. Accordingly, this specification and drawings are merely exemplary illustrations of this application as defined by the appended claims, and are considered to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from the spirit and scope of this application. Thus, if such modifications and modifications of this application fall within the scope of the claims of this application and their equivalents, this application is also intended to include such modifications and modifications.

Claims

1. A method for redundancy data retransmission based on dynamic priority scheduling in vehicle-to-everything (V2X), characterized in that, The method comprises the following steps: acquiring multi-modal vehicle state time series data, including: drive motor temperature, battery temperature, voltage, current, insulation alarm, tire pressure anomaly, door status, vehicle running parameters, GPS data and environmental data; wherein the vehicle running parameters include: speed, acceleration, load; analyzing the multi-modal vehicle state time series data by a warning prediction model to predict the warning probability of the data; the warning prediction model is a hybrid prediction model constructed by using a Transformer encoding layer and a long short-term memory network, which is used for feature extraction and time series correlation modeling of the vehicle state time series data to predict the warning probability of the data; determining the multi-modal vehicle state time series data retransmission priority based on the timeliness, data volume and warning probability of the multi-modal vehicle state time series data; based on the multi-modal vehicle state time series data retransmission priority, performing hierarchical compression processing on different types of multi-modal vehicle state time series data to obtain compressed data, wherein continuous data is processed by difference encoding and Huffman compression algorithm, and discrete alarm data is compressed by bit field merging encoding method; storing the compressed data in an emergency cache area and / or a main storage area according to a preset rule, and retransmitting according to the retransmission priority after network recovery.

2. The dynamic priority scheduling based redundancy data retransmission method for V2X according to claim 1, wherein, The method comprises the following steps: using the time-aware position encoding mechanism of the Transformer encoding layer in the warning prediction model to perform global dependence modeling on the multi-modal vehicle state time series data to obtain a high-dimensional representation vector containing multi-modal time series features; using the LSTM network layer in the warning prediction model to extract the change trend features of the high-dimensional representation vector in the time dimension to obtain time series correlation features; based on the time series correlation features, using the prediction layer in the warning prediction model to obtain the warning probability of the input data.

3. The dynamic priority scheduling based redundancy data retransmission method for V2X according to claim 2, wherein, The time-aware position encoding mechanism of the Transformer encoding layer in the warning prediction model refers to adding time interval information on the basis of standard sinusoidal position encoding, and the time-aware position encoding mechanism satisfies the following formula: wherein, is a position encoding value at 2i dimensions for time step pos, is a position encoding value at 2i+1 dimensions for time step pos, pos is an index of a current sampling point in the time series data, i is a position encoding dimension index, is a hidden layer dimension, is a time enhancement factor, and is a time interval mapping function, is a neighboring sampling time interval; the high-dimensional representation vector satisfies the following equation: wherein, is the multi-modal vehicle state time series data at time t, is the position encoding value at time t, denotes a Transformer encoding layer.

4. The dynamic priority scheduling based redundancy data retransmission method for V2X according to claim 1, wherein, The method comprises the following steps: based on the warning probability, determining the warning type by a preset corresponding rule; based on the timeliness, data volume, warning probability and warning type of the multi-modal vehicle state time series data, calculating a comprehensive retransmission priority score of the multi-modal vehicle state time series data; the timeliness of the multi-modal vehicle state time series data includes the emergency degree weight of the multi-modal vehicle state time series data and the freshness of the multi-modal vehicle state time series data; sorting the comprehensive retransmission priority score to obtain the multi-modal vehicle state time series data retransmission priority.

5. The dynamic priority scheduling based redundancy data retransmission method for V2X according to claim 4, wherein, The multimodal vehicle state time series data integrated retransmission priority score satisfies the following equation: wherein, is a data property item score weight, is an urgency score, is an urgency score weight, is an early warning type score, is an early warning type score weight, is a data freshness score, is a data freshness score weight, is a data volume score, is a data volume score weight, is an early warning probability score, is an early warning probability score weight.

6. The dynamic priority scheduling based redundant data retransmission method for V2X according to claim 1, wherein, The early warning prediction model adopts a combination of offline training and online incremental learning; the offline training refers to training an initial model using a historical early warning data set; the online incremental learning refers to dynamically updating model parameters based on real-time collected multi-modal vehicle state time series data, and training the model to adapt to individual vehicle characteristics.

7. A dynamic priority scheduling based redundancy data retransmission system for vehicular internet of things, for implementing the dynamic priority scheduling based redundancy data retransmission method of any one of claims 1-6. Comprise: A vehicle-mounted data acquisition module, an early warning prediction module, a priority scheduling module, a hierarchical compression module, and a redundant storage and retransmission module; The vehicle-mounted data acquisition module is configured to acquire multi-modal vehicle state time series data, including driving motor temperature, battery temperature, voltage, current, insulation alarm, tire pressure anomaly, door status, vehicle operating parameters, GPS data, and environmental data; wherein the vehicle operating parameters include speed, acceleration, and load; The early warning prediction module is configured to extract multi-modal features and model time series correlation based on the multi-modal vehicle state time series data by using a hybrid prediction model constructed by a Transformer encoding layer and a long short-term memory network, and output an early warning probability; The priority scheduling module is configured to calculate a multi-modal vehicle state time series data comprehensive retransmission priority score based on the timeliness, data volume, and early warning probability of the multi-modal vehicle state time series data, and determine a retransmission order according to the score result; The hierarchical compression module is configured to compress different types of multi-modal vehicle state time series data based on the retransmission priority, wherein continuous data is compressed using difference encoding and Huffman compression algorithm, and discrete alarm data is compressed using bit field merging encoding method; The redundant storage and retransmission module is configured to store the compressed data in an emergency cache area and / or a main storage area according to a preset rule, and execute data retransmission tasks in order according to the retransmission priority after network recovery. 8.The dynamic priority scheduling based redundancy data retransmission system for V2X application according to claim 7, wherein, The early warning prediction module comprises a Transformer encoding submodule, a long short-term memory submodule, and a prediction output submodule; The Transformer encoding submodule is configured to perform time-aware position encoding and global feature extraction on the input multi-modal vehicle state time series data; The long short-term memory submodule is configured to extract the time variation trend of the global features; The prediction output submodule is configured to calculate an early warning probability based on the trend features. 9.The dynamic priority scheduling based redundant data retransmission system for V2X application according to claim 7, wherein, The redundant storage and retransmission module comprises two levels of structure, namely an emergency cache area and a main storage area; The emergency cache area is used to store data with high priority or high early warning probability; The main storage area is used to store ordinary data; After network recovery, the system extracts data from the emergency cache area and the main storage area in order according to the priority and executes retransmission.

10. The dynamic priority scheduling based redundant data retransmission system for vehicle networking of claim 7, wherein, The hierarchical compression module comprises a data type identification unit, which is configured to adaptively select difference encoding, Huffman compression, and bit field merging algorithm for processing based on the continuity features, redundancy threshold, and real-time transmission bandwidth of the multi-modal vehicle state time series data.

Images