Data processing method and device, vehicle and computer readable storage medium

Abstract

Embodiments of the present application provide a data processing method and device, a vehicle and a computer readable storage medium. The present application can divide the vehicle-end data into different levels to obtain data of different levels, wherein the vehicle-end data includes at least one of data obtained based on a vehicle-mounted bus, data obtained based on a vehicle-mounted sensor and data obtained based on vehicle networking; the network state is predicted to obtain a network state score; and a processing mode of the data is determined according to the data level and the network state score, wherein the processing mode includes at least one of edge node processing, local processing and cloud processing. The present application can adopt different data processing modes for data of different levels in combination with the network state, so that the data transmission can adapt to the network state fluctuation, and the risk that important data cannot be uploaded in real time due to sharing of bandwidth in the uploading process of data of different levels is avoided.

Description

Technical Field

[0001] This application relates to the field of vehicle network data transmission technology, and more specifically, to a data processing method, apparatus, vehicle, and computer-readable storage medium. Background Technology

[0002] With the development of vehicle-to-everything (V2X) and vehicle-to-cloud (V2X) technologies, the amount of data transmitted between vehicles and the cloud is increasing. Related technologies classify and transmit vehicle-side data, effectively improving transmission efficiency. However, different types of data share bandwidth during the upload process, which means that critical data cannot be uploaded in real time when the network fluctuates. Summary of the Invention

[0003] This application proposes a data processing method, apparatus, vehicle, and computer-readable storage medium that can adopt different data processing methods for different data levels based on two dimensions: network status and data level. This enables data transmission to adapt to network status fluctuations and avoids the risk that more important data cannot be uploaded in real time due to bandwidth sharing between different data levels during the upload process, thus solving the aforementioned technical problems.

[0004] In a first aspect, embodiments of this application provide a data processing method, which includes: classifying vehicle-side data into data levels to obtain data of different levels, wherein the vehicle-side data includes at least one of data obtained based on the vehicle bus, data obtained based on vehicle sensors, and data obtained based on the Internet of Vehicles; predicting the network status to obtain a network status score; and determining the data processing method based on the data level and the network status score, wherein the processing method includes at least one of edge node processing, local processing, and cloud processing.

[0005] Secondly, embodiments of this application provide a data processing apparatus, which includes: a classification module for classifying vehicle-side data into different levels to obtain data of different levels, wherein the vehicle-side data includes at least one of data acquired based on the vehicle bus, data acquired based on vehicle sensors, and data acquired based on the Internet of Vehicles; a network status scoring module for predicting the network status and obtaining a network status score; and a dynamic decision-making module for determining the data processing method based on the data level and the network status score, wherein the processing method includes at least one of edge node processing, local processing, and cloud processing.

[0006] Thirdly, embodiments of this application provide a vehicle, which includes a memory and a processor. The memory stores an application program that, when invoked by the processor, causes the processor to execute the method provided in embodiments of this application.

[0007] Fourthly, embodiments of this application provide a computer-readable storage medium storing program code, which, when invoked by a processor, causes the processor to execute the method provided in embodiments of this application.

[0008] Fifthly, embodiments of this application provide a computer program product, which, when invoked by a processor, causes the processor to execute the method provided in embodiments of this application.

[0009] The data processing method provided in this application has the following technical effects: Based on two dimensions, data level and network status score, the data processing method is dynamically determined, including at least one of edge node processing, local processing and cloud processing. It can combine network status and adopt different data processing methods for different levels of data, so that data transmission can adapt to network status fluctuations and avoid the risk that data of higher importance cannot be uploaded in real time due to the sharing of bandwidth during the upload process of data of different levels. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments and drawings obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0011] Figure 1 This is a schematic diagram of a vehicle-road-cloud integrated collaborative architecture provided in an embodiment of this application; Figure 2 This is a flowchart of a data processing system provided in an embodiment of this application; Figure 3 This is a flowchart of a data processing method provided in an embodiment of this application; Figure 4 This is an operational flowchart of a Long Short-Term Memory (LSTM) network provided in an embodiment of this application; Figure 5 This is a structural block diagram of a data processing apparatus provided in an embodiment of this application; Figure 6 This is a structural block diagram of a vehicle provided in one embodiment of this application. Detailed Implementation

[0012] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0013] It should be understood that the " / " in this application indicates an "or" relationship, and "first", "second", and "third" are used to distinguish different levels and do not have specific actual meanings.

[0014] See Figure 1 , Figure 1 This is a schematic diagram of a vehicle-road-cloud integrated collaborative architecture provided in an embodiment of this application. The vehicle-road-cloud integrated collaborative architecture includes a vehicle 1, a roadside edge layer 2, and a cloud 3. The vehicle 1, the roadside edge layer 2, and the cloud 3 can communicate with each other through a wireless network.

[0015] Vehicle 1 includes an embedded hierarchical classification module and a network probe. The embedded hierarchical classification module deploys a decision tree for classifying vehicle-side data. The network probe monitors network security and performance, and measures round-trip time (RTT) and packet loss rate. By capturing and analyzing network packets, it monitors network activity in real time to identify potential network threats and optimize network performance. Vehicle 1 includes local nodes, including in-vehicle edge nodes and internal nodes. Each node in Vehicle 1 can be an Electronic Control Unit (ECU) with data processing, storage, and forwarding capabilities. The in-vehicle edge nodes support Federated Learning (FL) local model training and can upload parameters to the cloud. FL is a distributed machine learning method that can utilize multi-party data for model training while protecting data privacy.

[0016] Roadside edge layer 2 includes Roadside Units (RSUs). RSUs deploy differential data processing units (e.g., extracting LiDAR point cloud features) and support multi-vehicle data aggregation (reducing redundancy by 90%). Edge computing platforms (e.g., ETSI MEC) dynamically allocate computing power and implement low-latency routing through Software-Defined Networking (SDN) switches. ETSI MEC refers to the Mobile Edge Computing (MEC) platform proposed by the European Telecommunications Standards Institute (ETSI). SDN switches separate network control from the physical network topology, using SDN controllers to manage traffic forwarding and network functions.

[0017] The cloud 3 is equipped with a global resource scheduler. As the "intelligent brain" of the vehicle-road-cloud integrated collaborative architecture, the global resource scheduler of the cloud 3 has the core function of forming a continuous closed loop from perception, decision-making to execution and optimization.

[0018] The data processing method in this application can be applied to a data processing device or a vehicle, and the data processing device can be deployed in a vehicle. The vehicle is equipped with sensing devices for sensing environmental information around the vehicle (e.g., road traffic information, weather information, obstacle information, movement trajectories of other road users, etc.). The sensing devices can include at least one of cameras, lidar, millimeter-wave radar, and ultrasonic radar. The vehicle can be, but is not limited to, a gasoline vehicle or a new energy vehicle. New energy vehicles can include electric vehicles, and electric vehicles can include, but are not limited to, pure electric vehicles, hybrid electric vehicles, or fuel cell vehicles.

[0019] See Figure 2 , Figure 2 This is a flowchart of a data processing system provided in an embodiment of this application. The data processing system 100 can be deployed in a vehicle. The data processing system 100 includes a data acquisition module 110, a data classification module 120, a network status prediction module 130, and a result aggregation and feedback module 140. The data acquisition module 110 is used to dynamically acquire vehicle-side data. The data classification module 120 is used to classify the vehicle-side data into three levels: first-level data (urgent data), second-level data (critical data), and third-level data (ordinary data). The network status prediction module 130 is used to predict the network status based on a long short-term memory network and determine the network status score. The result aggregation and feedback module 140 is used to combine the network status score and the data level to determine the data processing method, including at least one of edge node processing, local processing, and cloud processing. For the specific operation process of each module of the data processing system 100, please refer to the corresponding parts of steps S110-S130 below.

[0020] See Figure 3 , Figure 3 This is a flowchart of a data processing method provided in an embodiment of this application. The data processing method includes steps S110-S130.

[0021] Step S110: Divide the vehicle-side data into data levels to obtain data of different levels. The vehicle-side data includes at least one of the following: data obtained based on the vehicle bus, data obtained based on vehicle sensors, and data obtained based on the Internet of Vehicles.

[0022] The vehicle bus may include at least one of the following: Controller Area Network (CAN), Local Interconnect Network (LIN), and Service-Oriented Architecture (SOA).

[0023] The vehicle-mounted sensors may include at least one of the following: cameras, lidar, millimeter-wave radar, and ultrasonic radar.

[0024] Among them, Vehicle to Everything (V2X) is used to realize information interaction between vehicles (V2V), vehicles to infrastructure (V2I), vehicles to pedestrians (V2P), and vehicles to network (V2N).

[0025] In some embodiments, vehicle-side data can be classified into different levels (basic levels) based on the degree of impact of the data on vehicle operation, resulting in data of different levels (static classification).

[0026] In other embodiments, vehicle-side data can be categorized into basic levels based on its impact on vehicle operation, resulting in data of different basic levels (static categorization). The final level, or final priority score, of the data is then determined based on its basic level, the current driving environment, and the data's timeliness (dynamic categorization). Compared to simply statically categorizing data levels, this embodiment dynamically updates data levels based on the driving environment and data timeliness, improving the accuracy of data level determination. This allows for the efficient processing of data with higher importance in the current driving environment (e.g., direct transmission via edge nodes).

[0027] As one implementation method, vehicle-side data can be divided into three levels based on preset classification rules and the degree of impact of the data on vehicle operation: first-level data (emergency data), second-level data (critical data), and third-level data (ordinary data). The degree of impact of the first-level, second-level, and third-level data on vehicle operation safety decreases in that order.

[0028] Table 1 Among them, Level 1 data (emergency data) has a very high impact on vehicle operation, i.e., a very high safety level, such as collision warning signals and emergency braking signals (delay requirement <50 milliseconds). Level 1 data can be labeled in real time using a lightweight classifier and supports dynamic upgrades. As shown in Table 1 (Static Data Classification Table), the basic level of engine fault codes is Level 2, and the level of engine fault codes is upgraded to Level 1 when the fault code is P0xxx (high-risk fault code). The basic level of navigation map update requests is Level 3, and the level of navigation map update requests is upgraded to Level 2 when the vehicle is on a highway (i.e., the vehicle speed is greater than the speed threshold).

[0029] The second-level data (critical data) has a higher impact on vehicle operation, i.e., a higher safety level. Examples include high-precision map differential update data and traffic light status data (delay 100-200 milliseconds). Data compression techniques (such as JPEG2000 / differential encoding) can be used to compress the second-level data, thereby reducing the bandwidth occupied during the transmission of second-level data.

[0030] Level 3 data (general data) has a lower impact on vehicle operation, meaning it has a lower safety level; examples include in-vehicle entertainment logs and diagnostic reports. This application allows for batch uploading or edge node caching of Level 3 data.

[0031] As an example, pseudocode for a lightweight classifier implementation on the vehicle side is provided here.

[0032] As one implementation method, classifying vehicle-side data into different levels based on its impact on vehicle operation can include classifying the data level according to its source and content. As an example, referring to the pseudocode of a lightweight vehicle-side classifier (using a binary decision tree, with resource consumption <5% CPU), if the data comes from LiDAR and the vehicle speed is greater than 60 km / h, the data's (basic) level can be determined to be Level 1; if the data comes from an On-Board Diagnostics (OBD) system and the fault code is an emergency fault code, the data's (basic) level can be determined to be Level 1; for data in other cases, the (basic) level can be determined to be Level 2 or Level 3 based on the data source and content.

[0033] As one implementation method, the final grade of the data can be determined (dynamically classified) based on the basic grade of the data, the current driving environment, and the timeliness of the data, according to the following expression: Final Grade = Basic Grade × Environmental Coefficient + Timeliness Coefficient (Data Generation Time Decay Factor). For ease of calculation, the final grade of the data (i.e., the final priority score) can be rounded to the nearest integer.

[0034] The environmental factor is determined based on the current driving environment. For example, if the current driving environment includes nighttime / rainy / foggy weather, the environmental factor can be 1.5; if the current driving environment includes high-speed conditions, the environmental factor can be 1; if the current driving environment includes emergency braking conditions, the environmental factor can be 2.

[0035] Step S120: Predict the network state and obtain a network state score.

[0036] As one implementation method, network bandwidth within a preset future time period can be predicted using a Long Short-Term Memory (LSTM) network based on network and environmental indicators; a network status score is then determined based on the prediction results. In this implementation, predicting network bandwidth within a preset future time period using LSTM enables real-time scoring of network status, improving the accuracy of network prediction.

[0037] Among them, network metrics refer to parameters that can be used to reflect the network status. For example, network metrics include base station load status, round-trip time (RTT), and packet loss rate.

[0038] Environmental indicators refer to parameters that can affect the vehicle network status during vehicle operation. These include real-time vehicle location, vehicle speed, and environmental coefficients corresponding to the current driving environment. Real-time vehicle location refers to the vehicle's position obtained based on the Global Positioning System (GPS). Both roadside units and vehicles can determine the environmental coefficients corresponding to the current driving environment in real time based on factors such as weather and terrain. Optionally, the vehicle can obtain the environmental coefficients corresponding to the current driving environment from the roadside units, or it can determine the environmental coefficients based on factors such as weather and terrain in real time.

[0039] The preset time period can be defined in advance, and the value range of the preset time period can be 5-10 seconds.

[0040] This application uses an LSTM-based prediction model to predict network bandwidth within a preset time period based on network and environmental metrics. LSTM is a special type of Recurrent Neural Network (RNN) designed to address the vanishing and exploding gradient problems encountered by traditional RNNs when processing long sequences of data. The core innovation of LSTM lies in its selective memorization and forgetting of historical information through the introduction of four main components: a forget gate, an input gate, an output gate, and a cell state, fundamentally solving the problem of long-distance dependencies. The core idea of ​​LSTM is to control the flow of information through gating signals, ensuring that key information is stored long-term in the cell state while irrelevant information in the cell state is forgotten. LSTM includes three key gating units and one cell state. The computation process for each time step (t) is as follows: points 1-4.

[0041] 1. Forget gate (used to determine the cell state at the previous time step) (Information that needs to be retained or discarded) Among them, f t This is the output information of the forget gate; It is the weight matrix of the forget gate; It is the hidden state of the previous time step; This is the input for the current time step; σ is the bias vector of the forget gate; σ is the sigmoid function with an output range of [0,1].

[0042] Forgot Gate Input It is the hidden state of the previous time step. Input x for the current time step t The splicing, the forget gate output f t Each element in f represents the retention probability of the corresponding cell state dimension. t =1 indicates that it is completely retained, f t =0 indicates complete forgetting).

[0043] 2. Input gate (used to control the updating of the cell state by the current input information, i.e., to determine the extent to which the current input information is written into the cell state): Among them, i t σ is the input gate weight, which determines the proportion of candidate cell states retained; σ is the sigmoid function, with an output range of [0,1]; W i It is the first weight matrix of the input gate; h (t-1) It is the hidden state of the previous time step; x t This is the input for the current time step; b i It is the first bias vector of the input gate; These are candidate cell states, activated by the hyperbolic tangent function (Tanh); tanh is the hyperbolic tangent function; W C It is the second weight matrix of the input gate; b C It is the second bias vector of the input gate.

[0044] The input gate generates the input information writing ratio through the sigmoid function, and then activates the generated candidate cell state through the tanh function.

[0045] 3. Cell state update (updating cell state by combining the results of the forgetting gate and the input gate): Among them, C t It represents the updated cell state; f t This is the output information of the forget gate; It is the element-wise product symbol; C (t-1) It represents the cell state at the previous time step; i t It is the percentage of candidate cell states that are preserved; This indicates the state of the candidate cell.

[0046] 4. Output gate (used to determine the output information of the hidden state at the current time step): Among them, o t It is the output information of the hidden state at the current time step; W o This is the weight matrix of the output gate; b o It is the bias vector of the output gate; h t It is the hidden state at the current time step, and also the output of the LSTM; Element-wise product sign; tanh is the hyperbolic tangent function; C t It represents the updated cell state.

[0047] The output gate uses the sigmoid function to determine which parts of the cell state will be output. The cell state is activated by the tanh function and then multiplied by the value of the output gate to obtain the final hidden state output. This ensures that the hidden state only contains the cell state information required by the current task, avoiding redundant output.

[0048] For ease of understanding, see [link to relevant documentation]. Figure 4 , Figure 4 This is a flowchart illustrating the operation of an LSTM provided in one embodiment of this application. Figure 4As shown, each LSTM unit at each time step outputs a hidden state and a cell state. For each LSTM unit at each time step, the forget gate filters out unwanted old information from the cell state based on the hidden state of the previous time step and the input of the current time step; the input gate writes new information into the cell state to obtain the updated cell state; the output gate selectively outputs the information from the cell state to the hidden state of the current time step as the input of the next time step, thus completing the information transfer of the current time step.

[0049] In this embodiment, the LSTM input (sliding window length of 10 seconds) includes network metrics and environmental metrics, and the LSTM output includes bandwidth predictions and confidence levels for a future preset time period. By generating a dynamic scene dataset based on SUMO (Simulation of Urban Mobility) traffic simulation and labeling the network state for the future preset time period, training data for the LSTM can be obtained. SUMO traffic simulation is an open-source urban traffic simulation software capable of modeling and simulating various traffic modes, including road vehicles, public transportation, and pedestrians.

[0050] In one implementation, the prediction result may include the bandwidth prediction value and confidence level for a predetermined future time period. The network status score can be determined based on the prediction result using the following expression: normalized bandwidth prediction value × confidence level. For example, if the total network status score is 5 points, and the bandwidth prediction value is 50 Mbps with a confidence level of 0.9, then the normalized bandwidth prediction value could be 50 / 10 = 5, and the network status score would be 5 * 0.9 = 4.5 points. For ease of calculation, the network status score can be rounded to the nearest integer.

[0051] In some embodiments, the vehicle can perform rating drift detection, that is, if the actual network latency exceeds 20% of the LSTM prediction value, it triggers LSTM online learning to update the LSTM network state weights, thereby improving the accuracy of network state prediction.

[0052] Step S130: Determine the data processing method based on the data level and network status score, wherein the processing method includes at least one of edge node processing, local processing and cloud processing.

[0053] Edge nodes generally refer to computing entities located between data sources and cloud data centers in the Internet of Vehicles (IoV), possessing data processing, storage, and forwarding capabilities. In this application's embodiments, edge nodes include in-vehicle edge nodes (i.e., local edge nodes within the vehicle) and roadside edge nodes (e.g., roadside units, micro base stations, and other road infrastructure).

[0054] In a vehicle dynamic data hierarchical transmission system based on edge nodes, the combination of different network status scores and data levels determines the data processing method, including at least one of edge node processing, local processing, and cloud processing.

[0055] In some embodiments, when the vehicle's latency limit, bandwidth usage, and energy consumption limits meet certain conditions, the data processing method can be dynamically decided based on data level and network status score. For example, when the vehicle's safety data is less than 100 milliseconds, non-critical data is less than or equal to 20% of the total bandwidth, and the vehicle-side communication module power consumption is less than 5 watts (W), the data processing method can be dynamically decided based on data level and network status score.

[0056] In some embodiments, the data processing method can be determined based on a pre-defined decision matrix, according to the data level and network status score. Referring to Table 2, the decision matrix includes the mapping relationship between data level, network status score, and processing method.

[0057] Referring to Table 2, under the condition of high network status score (4-5 points), network bandwidth > 50Mbps, latency < 30ms, LSTM predicts network status with high stability, and can prioritize processing first and second level data through edge nodes, and batch process third level data.

[0058] Referring to Table 2, with a network status score of 3, the network bandwidth is between 20-50Mbps and the latency is between 30-100ms. The network status predicted by LSTM is subject to fluctuation risk. The first and second level data can be processed first by local and edge nodes in collaboration, while the third level data is processed later.

[0059] Referring to Table 2, in the case of low network status score (1-2 points), bandwidth <10Mbps, latency >100ms, or network status interruption predicted by LSTM (such as tunnel scenarios), all levels of data can be transmitted through cloud compression, and all levels of data can also be processed through cloud rollback and disaster recovery. That is, all levels of data in weak network environments need to be designed for disaster recovery.

[0060] Table 2 (Data Processing Decision Matrix) As one implementation method, if the network status score is greater than or equal to the first preset score (4 points) and the data level is the first level, then the processing method is determined to include processing the first level data (e.g., collision warning signal) through edge nodes (including vehicle edge nodes or roadside edge nodes), that is, directly transmitting the first level data through the edge nodes.

[0061] As an example, see Table 2. When the network status score is ≥4, the processing method for the first level data may include: (1) Data transmission path selection: The first level data is directly transmitted to the nearest edge node (RSU or vehicle edge ECU) through the 5G slice private network, and the end-to-end delay is controlled within 50ms; (2) Redundancy mechanism: V2V multi-hop relay is started in parallel to ensure that it can still be transmitted when the network is weak. Among them, the V2V multi-hop relay adopts the Dedicated Short Range Communications (DSRC) protocol; (3) Technical guarantee: Edge nodes reserve computing resources (Kubernetes dynamic scheduling) and prioritize the processing of high-level tasks. Among them, Kubernetes, abbreviated as K8s, is the container orchestration and management platform for vehicles.

[0062] As one implementation method, if the network status score is greater than or equal to the first preset score (4 points), and the data level is the second level, then the processing method is determined to include processing the second-level data collaboratively through edge nodes and the cloud. That is, the processing method includes extracting and / or compressing the second-level data through edge nodes, and transmitting the extracted and / or compressed data through the cloud. The second level is lower than the first level.

[0063] As an example, see Table 2. When the network status score is ≥4, the processing methods for the second level of data (e.g., high-precision map update data) may include: (1) edge preprocessing: performing feature extraction at the RSU node (e.g., extracting target detection boxes based on LiDAR point cloud), reducing the data volume by 80% before uploading to the cloud; (2) compression technology: using JPEG2000 or differential encoding compression, reducing bandwidth usage by 75%.

[0064] As one implementation method, if the network status score is greater than or equal to the first preset score, and the data level is the third level, then the processing method is determined to include processing the third-level data collaboratively through edge nodes and the cloud. That is, the third-level data is compressed in batches through edge nodes, and the batch-compressed data is transmitted through the cloud. The third level is lower than the second level.

[0065] As an example, see Table 2. When the network status score is ≥4, the processing method for Level 3 data (such as log files) can include batch caching: edge nodes merge data packets and asynchronously upload them to the cloud using idle bandwidth, saving 60% of traffic.

[0066] As one implementation method, if the network status score is greater than the second preset score (2 points) and less than the first preset score (4 points), the data level is the first level. Then the processing method is determined to include processing the data of the first level through local and edge nodes in collaboration. That is, the data of the first level is compressed through the local node and the compressed data is backed up to the nearby edge node.

[0067] As an example, see Table 2. When the network status score is 3 points, the processing methods for the first level of data may include: (1) Local backup: the vehicle ECU performs emergency actions (such as braking) and compresses the data to the nearby edge node (delay <100ms); (2) Energy consumption optimization: turn off unnecessary communication modules and reduce power consumption by 40%.

[0068] As one implementation method, if the network status score is greater than a second preset score but less than a first preset score, and the data level is the second level, then the processing method is determined to include processing the second-level data locally, that is, performing differential updates on the second-level data through local nodes. The second level is lower than the first level.

[0069] As an example, see Table 2. When the network status score is 3 points, the processing methods for the second level data can include: (1) local lightweight computing: the differential update is completed in the vehicle computing unit, and only the difference part (such as the map change area) is uploaded; (2) dynamic weight adjustment: the reinforcement learning model adjusts the α / β coefficients according to the real-time feedback (such as adjusting α from 0.7 to 0.8). The reinforcement learning model can be a deep Q network (DQN) model. α is the level weight. The larger the α value, the more the system values ​​the urgency of the data itself when making decisions. β is the network weight. The larger the β value, the more the decision tends to adapt to the current network conditions and is adaptively adjusted by the model.

[0070] As one implementation method, if the network status score is greater than the second preset score but less than the first preset score, and the data level is the third level, then the processing method is determined to include processing the third-level data locally, that is, storing the third-level data through a local node using an exponential backoff retransmission method. The third level is lower than the second level.

[0071] As an example, referring to Table 2, when the network status score is 3, for the third level of data, the processing method can include exponential backoff retransmission: an initial retry interval of 1 second, a maximum of 5 retries, to avoid network congestion. Here, exponential backoff retransmission means that whenever an operation fails and needs to be retried, it is not retried immediately, but waits for a period of time before retrying, and the waiting time between each retry increases exponentially (usually doubling). Each retry increases the waiting time (1x for the first time, 2x for the second, 3x for the third, and so on). Exponential backoff retransmission can reduce server pressure, avoid network congestion or service avalanche caused by frequent retries, and improve the robustness and fault tolerance of the system.

[0072] As one implementation method, if the network status score is less than or equal to the second preset score (2 points) and the data level is the first level, the processing method is determined to include processing the first level data through local, edge node and cloud collaborative processing, that is, compressing the first level data through local nodes, uploading the compressed data to the cloud through V2V multi-hop relay, and storing the compressed data and recording the complete event through the cloud.

[0073] As an example, see Table 2. When the network status score is ≤2 points, the processing methods for the first level of data may include: (1) Simplified processing: Perform the minimum actions locally (such as triggering the airbag), and upload the data to the cloud via V2V multi-hop relay after compression; (2) Disaster recovery backup: Store the complete event record in the cloud for accident analysis.

[0074] As one implementation method, if the network status score is less than or equal to the second preset score and the data level is the second level, then the processing method is determined to include processing the data of the second level through local, edge node and cloud collaborative processing, wherein the second level is lower than the first level.

[0075] As an example, see Table 2. When the network status score is ≤2 points, the processing methods for the second level data may include: (1) Local caching: The complete data is temporarily stored in the vehicle storage. After the network is restored, the cloud actively pulls the first level data through the interruption and resumption method of the Transmission Control Protocol (TCP); (2) Differential compression: The high-precision map is compressed by ZSTD (Zstandard, a fast lossless compression algorithm), and the data volume is reduced by 90%.

[0076] As one implementation method, if the network status score is less than or equal to the second preset score, and the data level is the third level, then the processing method is determined to include processing the third-level data locally and in the cloud. That is, the third-level data is stored locally and transmitted to the cloud after the network is restored. The third level is lower than the second level.

[0077] As an example, see Table 2. When the network status score is ≤2, the processing method for the third level data can include offline queuing: the data is marked as the lowest level, uploaded after a delay of several hours, and used for later analysis.

[0078] As an example, pseudocode for a dynamic routing algorithm is provided here.

[0079] As an example, refer to the pseudocode of the dynamic routing algorithm, which can determine the data processing method based on the data level and network status score. Specifically, (1) For emergency data of the first level: if the network status score is ≥4 or the distance between the vehicle and the nearest edge node (e.g., roadside unit) is <50 meters, it can be determined that the emergency data will be processed through the edge node first; if the network status score is <4 and the distance between the vehicle and the nearest edge node is ≥50 meters, the emergency data will be directly transmitted to the cloud under strong network conditions. (2) For data that is not of the first level: if the network status score is ≤2, non-critical data will be cached locally under weak network conditions; if the vehicle's remaining SOC is >30%, the data will be transmitted in the background when the battery is high; (3) In other cases, the data will be transferred to the edge node when the battery is low.

[0080] In some embodiments, the vehicle can perform edge load balancing, that is, if the CPU utilization rate of the edge node is greater than 80%, the processing tasks of the second level of data can be automatically migrated to the neighboring node or degraded to local processing.

[0081] In some embodiments, the vehicle can be designed with a safety fallback strategy, that is, regardless of the network status score, the vehicle retains emergency response capabilities (such as braking redundancy control) locally for all Level 1 data.

[0082] To facilitate understanding, this application provides two examples to illustrate the data processing method of this application. Example 1: The scenario is: a vehicle detects an obstacle in a tunnel (first-level data), and at this time the network state score predicted by LSTM drops sharply from 4 points to 1 point. The data processing method includes steps S1A-S3A.

[0083] Step S1A: Local immediate braking, delay <20ms.

[0084] Step S2A: Compressed data is transmitted to the tunnel entrance RSU via V2V relay with a delay of 150ms.

[0085] Step S3A: The RSU forwards the information to the cloud to generate a detour route, which is then broadcast to subsequent vehicles.

[0086] Example 2: If the scenario is an emergency braking scenario caused by collision risk, the data processing method includes steps S1B-S3B.

[0087] Step S1B: The vehicle detects a pedestrian collision risk (Level 1 data), and LSTM predicts the current network score as 4 (5G coverage).

[0088] Step S2B: Data is directly transmitted to the cloud, and the cloud sends braking commands to the edge nodes, which are then broadcast to surrounding vehicles via the RSU (end-to-end delay <50ms).

[0089] Step S3B: In the event of a weak network (score = 1), trigger V2V multi-hop relay to ensure that the command can be reached.

[0090] Steps S110-S130 have the following technical effects: Based on two dimensions, namely data level and network status score, the data processing method is dynamically determined, including at least one of edge node processing, local processing and cloud processing. It can combine network status to adopt different data processing methods for different levels of data, so that data transmission can adapt to network status fluctuations, avoid the risk that important data cannot be uploaded in real time due to the sharing of bandwidth during the upload process of different levels of data, improve the timeliness of data processing, and reduce the upload of redundant data.

[0091] See Figure 5 , Figure 5 This is a structural block diagram of a data processing apparatus provided in an embodiment of this application. The data processing apparatus 200 includes a data classification module 210, a network status scoring module 220, and a dynamic decision-making module 230. The data classification module 210 is used to classify vehicle-side data into different data levels, wherein the vehicle-side data includes at least one of the following: data acquired based on the vehicle bus, data acquired based on vehicle sensors, and data acquired based on the vehicle network. The network status scoring module 220 is used to predict the network status and obtain a network status score. The dynamic decision-making module 230 is used to determine the data processing method based on the data level and the network status score, wherein the processing method includes at least one of edge node processing, local processing, and cloud processing.

[0092] In some embodiments, the dynamic decision module 230 is further configured to determine the data processing method based on a pre-set decision matrix, according to the data level and network status score, wherein the decision matrix includes a mapping relationship between the data level, the network status score and the processing method.

[0093] In some embodiments, the dynamic decision module 230 is further configured to: if the network status score is greater than or equal to a first preset score and the data level is a first level, determine that the processing method includes processing the data of the first level through edge nodes; if the network status score is greater than or equal to the first preset score and the data level is a second level, determine that the processing method includes processing the data of the second level through collaborative processing of edge nodes and the cloud, wherein the second level is lower than the first level; if the network status score is greater than or equal to the first preset score and the data level is a third level, determine that the processing method includes processing the data of the third level through collaborative processing of edge nodes and the cloud, wherein the third level is lower than the second level.

[0094] In some embodiments, the dynamic decision module 230 is further configured to: if the network status score is greater than a second preset score and less than a first preset score, and the data level is a first level, then determine that the processing method includes processing the first level data through local and edge node collaboration; if the network status score is greater than a second preset score and less than a first preset score, and the data level is a second level, then determine that the processing method includes processing the second level data locally, wherein the second level is lower than the first level; if the network status score is greater than a second preset score and less than a first preset score, and the data level is a third level, then determine that the processing method includes processing the third level data locally, wherein the third level is lower than the second level.

[0095] In some embodiments, the dynamic decision module 230 is further configured to: if the network status score is less than or equal to a second preset score and the data level is a first level, determine that the processing method includes processing the first level data collaboratively through local, edge node, and cloud; if the network status score is less than or equal to a second preset score and the data level is a second level, determine that the processing method includes processing the second level data collaboratively through local, edge node, and cloud, wherein the second level is lower than the first level; if the network status score is less than or equal to a second preset score and the data level is a third level, determine that the processing method includes processing the third level data through local and cloud, wherein the third level is lower than the second level.

[0096] In some embodiments, the classification module 210 is further configured to classify the vehicle-side data into basic levels based on the degree of impact of the data on vehicle operation, thereby obtaining data of different basic levels; and to determine the final level of the data based on the basic level of the data, the current driving environment, and the timeliness of the data.

[0097] In some embodiments, the network status scoring module 220 is further configured to predict the network bandwidth within a preset future time period based on network indicators and environmental indicators through a long short-term memory network. The network indicators include base station load status, data round-trip latency, and packet loss rate, while the environmental indicators include real-time vehicle positioning, vehicle speed, and environmental coefficients corresponding to the current driving environment. Based on the prediction results, a network status score is determined.

[0098] Those skilled in the art will clearly understand that the apparatus provided in the embodiments of this application can implement the methods provided in the embodiments of this application. The specific working process of the described apparatus and modules can be found in the corresponding processes of the methods in the embodiments of this application, and will not be repeated here.

[0099] In the embodiments provided in this application, the coupling, direct coupling, or communication connection between the modules shown or discussed may be indirect coupling or communication coupling through some interfaces, devices, or modules, and may be electrical, mechanical, or other forms. The embodiments of this application do not impose specific limitations on this.

[0100] Furthermore, the functional modules in the embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0101] See Figure 6 , Figure 6 This is a structural block diagram of a vehicle provided in an embodiment of this application. The vehicle 300 may include a memory 310 and a processor 320. The memory 310 stores an application program that is configured to cause the processor 320 to execute the method provided in the embodiment of this application when invoked by the processor 320.

[0102] The processor 320 may include one or more processing cores. The processor 320 uses various interfaces and lines to connect to various parts of the vehicle 300, and is used to run or execute instructions, programs, code sets or instruction sets stored in the memory 310, as well as to call and run or execute data stored in the memory 310, and perform various functions and process data of the vehicle 300.

[0103] The processor 320 can be implemented using at least one of the following hardware forms: Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), and Programmable Logic Array (PLA). The processor 320 can integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the displayed content; and the modem handles wireless communication. It is understood that the modem can also be implemented separately as a communication chip, without being integrated into the processor 320.

[0104] The memory 310 may include random access memory (RAM) or read-only memory (ROM). The memory 310 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 310 may include a program storage area and a data storage area. The program storage area may store instructions for implementing an operating system, instructions for implementing at least one function, instructions for implementing the various method embodiments described above, etc. The data storage area may store data created by the vehicle 300 during use.

[0105] This application also provides a computer-readable storage medium. The computer-readable storage medium stores program code that, when invoked by a processor, causes the processor to execute the method provided in this application.

[0106] Computer-readable storage media can be electronic storage devices such as flash memory, electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), hard disk, or ROM.

[0107] In some embodiments, the computer-readable storage medium includes a non-volatile computer-readable storage medium (Non-TCRSM). The computer-readable storage medium has storage space for program code that performs any of the method steps described above. This program code can be read from or written to one or more computer program products. The program code may be compressed in an appropriate form.

[0108] This application also provides a computer program product, which includes a computer program that, when invoked by a processor, causes the processor to execute the method provided in the embodiments of this application.

[0109] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A data processing method, characterized in that, include: Vehicle-side data is classified into different data levels. The vehicle-side data includes at least one of the following: data obtained from the vehicle bus, data obtained from vehicle sensors, and data obtained from the Internet of Vehicles. Predict the network state and obtain a network state score; Based on the data level and network status score, the data processing method is determined, which includes at least one of edge node processing, local processing, and cloud processing.

2. The method according to claim 1, characterized in that, The steps for determining the data processing method based on data level and network status scoring include: Based on a pre-defined decision matrix, the data processing method is determined according to the data level and network status score. The decision matrix includes the mapping relationship between the data level, network status score, and processing method.

3. The method according to claim 1 or 2, characterized in that, The steps for determining the data processing method based on data level and network status scoring include: If the network status score is greater than or equal to the first preset score, and the data level is the first level, then the processing method is determined to include processing the first level data through edge nodes; If the network status score is greater than or equal to the first preset score, and the data level is the second level, then the processing method is determined to include processing the second level data through collaborative processing of edge nodes and the cloud, wherein the second level is lower than the first level; If the network status score is greater than or equal to the first preset score, and the data level is the third level, then the processing method is determined to include processing the third level data through collaborative processing of edge nodes and the cloud, wherein the third level is lower than the second level.

4. The method according to claim 1 or 2, characterized in that, The steps for determining the data processing method based on data level and network status scoring include: If the network status score is greater than the second preset score and less than the first preset score, the data level is the first level, and the processing method is determined to include processing the first level data through local and edge node collaboration; If the network status score is greater than the second preset score and less than the first preset score, and the data level is the second level, then the processing method is determined to include processing the second level data locally, wherein the second level is lower than the first level; If the network status score is greater than the second preset score but less than the first preset score, and the data level is the third level, then the processing method is determined to include processing the third-level data locally, where the third level is lower than the second level.

5. The method according to claim 1 or 2, characterized in that, The steps for determining the data processing method based on data level and network status scoring include: If the network status score is less than or equal to the second preset score, the data level is the first level, and the processing method is determined to include processing the first level data through local, edge node and cloud collaborative processing; If the network status score is less than or equal to the second preset score, the data level is the second level. The processing method is determined to include processing the data of the second level through local, edge node and cloud collaborative processing. The second level is lower than the first level. If the network status score is less than or equal to the second preset score, and the data level is the third level, then the processing method is determined to include processing the third-level data through local and cloud processing, where the third level is lower than the second level.

6. The method according to claim 1, characterized in that, The steps for classifying vehicle-side data by data level include: Based on the degree of impact of the data on vehicle operation, the vehicle-side data is divided into basic levels to obtain data of different basic levels. The final data level is determined based on the basic data level, the current driving environment, and the timeliness of the data.

7. The method according to claim 1, characterized in that, The steps for predicting the network state and obtaining a network state score include: Based on network and environmental metrics, the network bandwidth within a preset time period is predicted using a long short-term memory network. The network metrics include base station load status, data round-trip latency, and packet loss rate, while the environmental metrics include real-time vehicle location, vehicle speed, and environmental coefficients corresponding to the current driving environment. Based on the prediction results, determine the network state score.

8. A data processing apparatus, characterized in that, include: The data classification module is used to classify vehicle-side data into different levels to obtain data of different levels. The vehicle-side data includes at least one of the following: data obtained from the vehicle bus, data obtained from vehicle sensors, and data obtained from the Internet of Vehicles. The network state scoring module is used to predict the network state and obtain a network state score. The dynamic decision-making module is used to determine the data processing method based on the data level and network status score. The processing method includes at least one of edge node processing, local processing, and cloud processing.

9. A vehicle, characterized in that, include: A memory and a processor, wherein the memory stores an application program that, when invoked by the processor, causes the processor to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores program code, and the product includes a computer program, which, when invoked by a processor, causes the program code to cause the processor to perform the method as described in any one of claims 1-7.

Images