Dynamic cache management method, device and computer equipment for internet of vehicles

Abstract

This invention relates to the fields of vehicle networking and artificial intelligence technology, and particularly to a dynamic cache management method, device, and computer equipment for vehicle networking. The method generates cache management strategies based on state space data at each target time step, obtains cache management strategies for each target time step, constructs training samples for each target time step by combining the state space data and cache management strategies, and uses the obtained training samples for dynamic cache management network updates. This constructs an active cache update mechanism, adapts to the real-time dynamic changes in vehicle networking data, improves the accuracy of vehicle networking cache demand prediction, and achieves time-series and refined prediction of cache demand.

Description

Technical Field

[0001] This invention relates to the fields of vehicle networking and artificial intelligence technology, and in particular to a dynamic cache management method, device, and computer equipment for vehicle networking. Background Technology

[0002] With the development of intelligent driving technology, large models are widely used in vehicle-to-everything (V2X) scenarios for core tasks such as path planning, road condition analysis, and traffic regulation adaptation. The dependence of V2X on external knowledge bases has increased significantly. Caching technology, as a key means to reduce data response latency and improve the operating efficiency of vehicle terminals, has become a core supporting technology for the implementation of large models in V2X.

[0003] Current cache management technology in the vehicle-to-everything (V2X) scenario mainly relies on the cache augmentation generation (CAG) architecture. By building a key-value pair cache pool of historical queries and output results, new queries are prioritized to match cached data to bypass model forward computation. However, existing cache management solutions suffer from low accuracy in predicting cache demand and poor adaptability to sudden scenarios, resulting in insufficient timeliness of cached data and difficulty in meeting the real-time decision-making needs of intelligent driving. Summary of the Invention

[0004] Based on this, the purpose of this invention is to provide a dynamic cache management method, device, computer equipment, and storage medium for vehicle networks. The method generates cache management strategies based on state space data at each target time step, obtains cache management strategies for each target time step, constructs training samples for each target time step by combining the state space data and cache management strategies, and uses the obtained training samples for dynamic cache management network updates. This constructs an active cache update mechanism, adapts to the real-time dynamic changes in vehicle network data, improves the accuracy of vehicle network cache demand prediction, and achieves time-series and refined prediction of cache demand.

[0005] In a first aspect, embodiments of this application provide a dynamic cache management method for vehicle networks, comprising the following steps:

[0006] Obtain the state space data of the vehicle network at each target time step within a preset time period, wherein the state space data includes the percentage of remaining cache space and the demand priority label; the target time step is any time step within the preset time period; Based on the state space data of the vehicle network at each target time step and the preset dynamic cache management network, the cache management analysis data of each target time step is obtained, wherein the cache management analysis data includes priority probability distribution data; Based on the priority probability distribution data and demand priority labels of each target time step, cache decision action codes are generated to obtain cache decision action codes for each target time step; based on the cache decision action codes and the proportion of remaining cache space for each target time step, the action execution timing and resource consumption prediction for each target time step are obtained; the cache decision action codes, action execution timing, and resource consumption prediction for each target time step are combined to construct the cache management strategy for each target time step. Cache management is performed according to the cache management strategy of each target time step. Based on the state space data of each target time step and the cache management strategy, training samples for each target time step are constructed. The dynamic cache management network is updated based on the training samples at each target time step to obtain the target dynamic cache management network; cache management is performed based on the vehicle network state space data at the current time step after the preset time period and the target dynamic cache management network.

[0007] Secondly, embodiments of this application provide a dynamic cache management device for vehicle networks, comprising: The data acquisition module is used to acquire the state space data of the vehicle network at each target time step within a preset time period. The state space data includes the percentage of remaining cache space and the demand priority label. The target time step is any time step within the preset time period. The data analysis module is used to analyze the state space data of the vehicle network at each target time step and the preset dynamic cache management network to obtain cache management analysis data for each target time step, wherein the cache management analysis data includes priority probability distribution data. The strategy reasoning module is used to generate cache decision action codes based on the priority probability distribution data and demand priority labels of each target time step, and obtain the cache decision action codes for each target time step; based on the cache decision action codes and the proportion of remaining cache space for each target time step, it obtains the action execution timing and resource consumption prediction for each target time step; and combines the cache decision action codes, action execution timing, and resource consumption prediction for each target time step to construct the cache management strategy for each target time step. The data processing module is used to perform cache management according to the cache management strategy of each target time step, and to construct training samples for each target time step based on the state space data and cache management strategy of each target time step. The cache management module is used to update the dynamic cache management network based on the training samples of each target time step to obtain the target dynamic cache management network; and to perform cache management based on the state space data of the vehicle network at the current time step after the preset time period and the target dynamic cache management network.

[0008] Thirdly, embodiments of this application provide a computer device, including: a processor, a memory, and a computer program stored in the memory and executable on the processor; when the computer program is executed by the processor, it implements the steps of the dynamic cache management method for the Internet of Vehicles as described in the first aspect.

[0009] Fourthly, embodiments of this application provide a storage medium storing a computer program, which, when executed by a processor, implements the steps of the dynamic cache management method for the Internet of Vehicles as described in the first aspect.

[0010] In this application embodiment, a dynamic cache management method, device, computer equipment, and storage medium for vehicle-to-everything (V2X) are provided. Cache management strategies are generated based on state space data at each target time step to obtain cache management strategies for each target time step. Training samples for each target time step are constructed by combining the state space data and cache management strategies. The obtained training samples are used for dynamic cache management network updates, constructing an active cache update mechanism that adapts to the real-time dynamic changes in V2X data, improves the accuracy of V2X cache demand prediction, and achieves time-series and refined prediction of cache demand.

[0011] To better understand and implement this invention, the following detailed description is provided in conjunction with the accompanying drawings. Attached Figure Description

[0012] Figure 1 A flowchart illustrating a dynamic cache management method for a vehicle network provided in one embodiment of this application; Figure 2 This is a flowchart illustrating step S1 of a dynamic cache management method for a vehicle network provided in one embodiment of this application. Figure 3 This is a flowchart illustrating step S2 of a dynamic cache management method for a vehicle network provided in one embodiment of this application. Figure 4 This is a flowchart illustrating step S3 of a dynamic cache management method for a vehicle network provided in one embodiment of this application. Figure 5 This is a flowchart illustrating step S4 of a dynamic cache management method for a vehicle network provided in one embodiment of this application. Figure 6This is a flowchart illustrating step S5 of a dynamic cache management method for a vehicle network provided in one embodiment of this application. Figure 7 This is a flowchart illustrating step S6 of a dynamic cache management method for a vehicle network provided in one embodiment of this application. Figure 8 A schematic diagram of the structure of a dynamic cache management device for a vehicle network provided in one embodiment of this application; Figure 9 This is a schematic diagram of the structure of a computer device provided in one embodiment of this application. Detailed Implementation

[0013] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0014] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0015] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0016] Please see Figure 1 , Figure 1 The flowchart illustrates a dynamic cache management method for a vehicle-to-everything (V2X) network according to an embodiment of this application. The method includes the following steps: S1: Obtain the state space data of the vehicle network at each target time step within a preset time period.

[0017] The implementing entity of the dynamic cache management method for the Internet of Vehicles in this application is the management device for the dynamic cache management method of the Internet of Vehicles (hereinafter referred to as the management device). In an optional embodiment, the management device may be a computer device, a server, or a server cluster composed of multiple computer devices.

[0018] In this embodiment, the management device obtains the state space data of the vehicle network at each target time step within a preset time period, wherein the target time step is any time step within the preset time period.

[0019] Specifically, the state space data includes cache demand value, cache remaining space percentage, average access latency, bandwidth utilization, resource sufficiency, and demand priority label. Specifically, the average access latency is obtained by real-time calculation of data transmission round-trip time, and the cache remaining space percentage and bandwidth utilization are predicted based on the state space data of the previous target time step and the cache management strategy.

[0020] The demand priority label is constructed based on the cache demand value and a preset demand priority threshold range, reflecting the urgency of the cache demand.

[0021] For the aforementioned cache requirement value, please refer to Figure 2 , Figure 2 The flowchart of S1 in the dynamic cache management method of the Internet of Vehicles provided in one embodiment of this application includes steps S11 to S14, as follows: S11: Obtain the multi-source data of the vehicle network at each target time step and the preset first long short-term memory network; perform feature classification, weight setting and feature weighting according to the multi-source data of the vehicle network at each target time step, and construct the core layer weighted feature vector and the auxiliary layer weighted feature vector for each target time step.

[0022] In this embodiment, the management device obtains multi-source data of the vehicle network at each target time step and a preset first long short-term memory network, wherein the multi-source data of the vehicle network includes vehicle status data and environmental feature data.

[0023] Specifically, the vehicle status data includes vehicle acceleration and remaining battery power. The environmental characteristic data includes traffic density change rate, weather level data, and emergency event information data. The weather level data indicates the weather conditions at the current target time step; the emergency event information data indicates whether an emergency has occurred at the current target time step, such as receiving temporary traffic control / construction / disaster notifications.

[0024] The system resource data includes operating parameters from the vehicle-mounted cache management module and communication module, such as cache occupied space, remaining space, data access latency, and bandwidth utilization. The historical access logs include time-series log data from the knowledge base request time, request content, and cache hit status stored locally on the vehicle terminal.

[0025] The management device performs feature classification, weight setting, and feature weighting based on the multi-source data of the vehicle network at each target time step, and constructs the core layer weighted feature vector and the auxiliary layer weighted feature vector for each target time step.

[0026] Specifically, the management device uses frequency change verification and correlation verification methods to filter redundant data in the multi-source data of the vehicle network, directly discarding it and not participating in the modeling. Based on a preset threshold, the filtered multi-source data of the vehicle network is conditionally judged. If any emergency / sudden condition is met, it is marked as a core layer feature vector. Non-redundant data that is not marked as a core layer feature vector is concatenated into an auxiliary layer feature vector, retaining effective features.

[0027] The management device calculates vehicle anomaly values ​​based on the vehicle acceleration and remaining battery power in the vehicle status data, obtaining the vehicle anomaly values ​​for each target time step, as described below:

[0028] In the formula, For vehicle abnormal values, For acceleration weighting parameters, This refers to the acceleration anomaly coefficient, which is obtained based on the vehicle's acceleration. For power weighting parameters, The remaining power anomaly coefficient is obtained based on the remaining power. Update the weight parameters for the destination. An anomaly coefficient is updated for the destination, which is read by the management device through the vehicle navigation system.

[0029] The management equipment calculates environmental anomalies based on the environmental characteristic data, including traffic density change rate, weather level data, and emergency event information data, to obtain the environmental anomaly values ​​for each target time step, as described below:

[0030] In the formula, For vehicle abnormal values, This refers to the traffic anomaly coefficient, which is obtained based on the traffic density change rate. This refers to the weather anomaly coefficient, which is obtained based on the weather level data. The emergency event coefficient is obtained based on the emergency event information data. This is the function for finding the maximum value.

[0031] The management device performs a weighted summation of vehicle and environmental anomalies at the same target time step to obtain the environmental urgency at each target time step. Based on the environmental urgency at each target time step, the management device sets the core layer and auxiliary layer weight parameters, obtaining the core layer and auxiliary layer weight parameters for each target time step. The management device then performs weighted processing on the core layer feature vector, auxiliary layer feature vector, core layer weight parameters, and auxiliary layer weight parameters for each target time step, obtaining the core layer weighted feature vector and auxiliary layer weighted feature vector for each target time step. By quantifying vehicle and environmental anomalies, calculating the scene urgency, and dynamically adjusting the core layer and auxiliary layer feature weights based on the urgency, the modeling capability for sudden scene features is enhanced.

[0032] S12: Obtain the memory unit feature vector and hidden state feature vector of the previous time step of each target time step calculated by the first long short-term memory network; extract hidden state features based on the core layer weighted feature vector of each target time step, the memory unit feature vector of the previous time step of each target time step, the hidden state feature vector, and the first long short-term memory network to obtain the first hidden state feature vector of each target time step.

[0033] In this embodiment, the management device obtains the memory cell feature vector and hidden state feature vector of the previous time step of each target time step calculated by the first long short-term memory network.

[0034] The first long short-term memory network is a miniature LSTM network with multiple hidden units. The management device extracts hidden state features based on the core layer weighted feature vector of each target time step, the memory unit feature vector of the previous time step of each target time step, the hidden state feature vector, and the first long short-term memory network to obtain the first hidden state feature vector of each target time step.

[0035] S13: Obtain the feature vector of the historical log access segment; concatenate the first hidden state feature vector, the auxiliary layer weighted feature vector, and the historical log access segment feature vector of each target time step to obtain the concatenated feature vector of each target time step; input the concatenated feature vector of each target time step into the preset second long short-term memory network for hidden state feature extraction to obtain the second hidden state feature vector of each target time step.

[0036] In this embodiment, the management device obtains a feature vector of historical log access segments, wherein the feature vector of historical log access segments is obtained by feature extraction based on the original historical access logs.

[0037] The management device concatenates the first hidden state feature vector, the auxiliary layer weighted feature vector, and the historical log access fragment feature vector for each target time step to obtain the concatenated feature vector for each target time step.

[0038] The management device inputs the spliced ​​feature vectors of each target time step into a preset second long short-term memory network to extract hidden state features and obtain the second hidden state feature vectors of each target time step.

[0039] Specifically, the second Long Short-Term Memory network adopts a two-layer cascaded LSTM structure. The first LSTM layer is set to have 128 hidden units, and the activation function is... The second LSTM layer is set to have 64 hidden units, and the activation function is... .

[0040] The management device inputs the concatenated feature vectors of each target time step into the first LSTM layer to extract hidden state features, thereby obtaining the intermediate hidden state feature vectors of each target time step.

[0041] The management device inputs the intermediate hidden state feature vectors of each target time step into the second LSTM layer, and uses a zero initialization strategy to extract hidden state features to obtain the second hidden state feature vectors of each target time step.

[0042] A two-layer cascaded LSTM structure is adopted to accurately capture long-short-term temporal dependencies based on the concatenated feature vectors of each target time step. The core layer weighted features, auxiliary layer weighted features, and historical access logs are integrated to capture long-short-term temporal dependencies, filter irrelevant temporal noise, retain key information valuable for cache demand prediction, and output feature vectors after temporal modeling.

[0043] S14: Perform dimensional compression and nonlinear mapping based on the second hidden state feature vector of each target time step to obtain the global feature vector of each target time step; perform cache requirement prediction based on the global feature vector of each target time step to obtain the cache requirement value of each target time step.

[0044] In this embodiment, the management device performs dimensionality compression and nonlinear mapping on the second hidden state feature vector of each target time step to obtain the global feature vector of each target time step, extracts the most critical global features for predicting cache demand, and provides dimension-adapted input vectors, as described below:

[0045] In the formula, For global feature vectors, It is a non-linear activation function. This is the first weight matrix. This is the feature vector of the second hidden state. This is the first bias vector.

[0046] The management device predicts cache requirements based on the global feature vectors for each target time step, obtaining the cache requirement value for each target time step, as described below:

[0047] In the formula, To cache the required values, It is the sigmoid activation function. This is the second weight matrix. This is the second bias vector.

[0048] Targeting the multi-source heterogeneous data characteristics of the Internet of Vehicles (IoV), this paper captures the long-term and short-term temporal dependencies of IoV data through multi-source feature hierarchical processing, scenario-based emergency quantification, and dynamic weight allocation. It then performs temporal enhancement modeling to achieve the fusion of multi-source IoV data and accurate prediction of caching needs, providing precise demand guidance for caching decisions.

[0049] In an optional embodiment, the management device employs a mean squared error loss calculation method. Based on the vehicle network multi-source data and cache requirement label values ​​of several sample time steps under normal and emergency scenarios, the mean squared error loss value under normal scenarios and the mean squared error loss value under emergency scenarios are calculated. The first long short-term memory network and the second long short-term memory network are trained based on the obtained mean squared error loss values ​​under normal and emergency scenarios to enhance the prediction accuracy of the model for key scenarios such as temporary traffic control, extreme weather, and vehicle anomalies.

[0050] S2: Analyze the state space data of the vehicle network at each target time step and the preset dynamic cache management network to obtain cache management analysis data for each target time step.

[0051] In this embodiment, the management device analyzes the state space data of the vehicle network at each target time step and the preset dynamic cache management network to obtain cache management analysis data for each target time step.

[0052] In an optional embodiment, the management device performs data format conversion and normalization verification on the state space data to ensure that the input data meets subsequent network requirements.

[0053] The dynamic cache management network includes a shared feature extraction network, a priority classification network, and a policy network; the cache management analysis data includes priority probability distribution data and action probability distribution data. Please refer to [link / reference]. Figure 3 , Figure 3 The flowchart of S2 in the dynamic cache management method of the Internet of Vehicles provided in one embodiment of this application includes steps S21 to S23, as follows: S21: Input the state space data of each target time step into the shared feature extraction network for linear transformation, activation function mapping and regularization processing to obtain the feature extraction vector of each target time step.

[0054] In this embodiment, the management device inputs the state space data of each target time step into the shared feature extraction network for linear transformation, activation function mapping and regularization processing to obtain the feature extraction vector of each target time step.

[0055] Specifically, the shared feature extraction network adopts a standard feature extraction architecture consisting of two hidden fully connected layers, a ReLU activation layer, and a Dropout regularization layer. The management device performs linear transformation, activation function mapping, and regularization processing on the state space data of each target time step and the shared feature extraction network to extract primary features related to caching decisions. It introduces nonlinear transformation capabilities to enable the network to fit the state-feature nonlinear mapping relationship in complex vehicle networking scenarios and obtain the feature extraction vectors of each target time step.

[0056] S22: Input the feature extraction vectors of each target time step into the priority classification network to calculate the priority probability distribution and obtain the priority probability distribution data of each target time step.

[0057] In this embodiment, the management device inputs the feature extraction vectors of each target time step into the priority classification network to calculate the priority probability distribution, thereby obtaining the priority probability distribution data of each target time step. The priority probability distribution data includes a plurality of priority probability distribution vectors.

[0058] Specifically, the priority classification network adopts a standard classification network architecture consisting of one hidden fully connected layer, one output fully connected layer, and a Softmax activation layer. The management device inputs the feature extraction vectors for each target time step into the priority classification network, and sequentially performs linear transformation, nonlinear activation, linear transformation, and Softmax normalization on the feature extraction vectors to obtain the priority probability distribution data for each target time step, refine the requirement priority determination, and provide a basis for action space pruning.

[0059] S23: Input the feature extraction vectors of each target time step into the policy network to calculate the action probability distribution and obtain the action probability distribution data of each target time step.

[0060] In this embodiment, the management device inputs the feature extraction vectors of each target time step into the policy network to calculate the action probability distribution and obtain the action probability distribution data of each target time step. The action probability distribution data includes action probability distribution vectors of several dimensions.

[0061] Specifically, the policy network adopts a standard policy network architecture of "2 hidden fully connected layers, 1 output fully connected layer, and Softmax activation layer". The management device calculates the action probability distribution based on the feature extraction vectors of each target time step and the policy network to obtain the action probability distribution data of each target time step, which is used to generate the optimal cache decision action.

[0062] S3: Generate cache decision action codes based on the priority probability distribution data and demand priority labels of each target time step to obtain cache decision action codes for each target time step; based on the cache decision action codes and the percentage of remaining cache space for each target time step, obtain the action execution timing and resource consumption prediction for each target time step; combine the cache decision action codes, action execution timing, and resource consumption prediction for each target time step to construct the cache management strategy for each target time step.

[0063] In this embodiment, the management device generates cache decision action codes based on the priority probability distribution data and demand priority tags for each target time step, obtaining cache decision action codes for each target time step. These cache decision action codes include cache action codes, update action codes, delete action codes, and delayed cache codes. The cache action code indicates that the target data should be cached in the local vehicle knowledge base; the update action code indicates that expired or low-precision data in the local cache should be updated; the delete action code indicates that redundant, invalid, or low-demand data should be deleted to release resources; and the delayed cache code indicates that the cache action should be executed when the bandwidth utilization rate is ≤60%. The resource usage prediction indicates the current action's impact on cache and communication resources.

[0064] Specifically, the management device determines the priority corresponding to the maximum probability distribution vector based on the priority probability distribution data of each target time step and the demand priority label, while constraining that this priority cannot be lower than the demand priority label, and constructs a priority label for each target time step. The priority label is used to indicate the priority corresponding to the maximum probability distribution vector, and the priority label is as follows:

[0065] In the formula, For the first t Priority labels for each target time step. For the first t Priority tags for each target time step For the priority probability distribution data, the first i A probability distribution vector of each priority level. This is a function for finding the extreme values.

[0066] The management device filters available actions based on the priority tags of each target time step and a preset complete action space to obtain the available action space for each target time step. Specifically, the complete action space includes cached actions, update actions, delete actions, and delayed cached actions.

[0067] The management device selects action codes based on the action probability distribution data and available action space for each target time step, obtaining the cached decision action code for each target time step. The cached decision action code is as follows:

[0068] In the formula, For the first t Encoding of cache decision actions for each target time step. For the probability distribution data of actions, the first... j A vector of action probability distributions in 1 dimension For the first indicator function, For the first t The available action space for each target time step, where 0 represents a cached action, 1 represents an update action, 2 represents a delete action, and 3 represents a delayed cached action.

[0069] The management device obtains the action execution timing and resource consumption prediction for each target time step based on the cache decision action code and the percentage of remaining cache space. The resource consumption prediction includes cache space consumption prediction and bandwidth consumption estimation. The cache space consumption prediction indicates the incremental space used by the cache module after the action execution is completed. The bandwidth consumption estimation indicates the total bandwidth consumption during the action execution. The action execution timing is as follows:

[0070] In the formula, For the first t Timing of action execution for each target time step. For the current time step, For the first t The percentage of remaining cache space for each target time step. This is the delay parameter.

[0071] The management device combines the cache decision action encoding, action execution timing, and resource usage prediction for each target time step to construct a cache management strategy for each target time step. It then dynamically generates the optimal cache strategy by combining state space data including cache demand values, thereby achieving reasonable allocation and proactive updating of cache resources.

[0072] S4: Perform cache management according to the cache management strategy of each target time step, and construct training samples for each target time step based on the state space data and cache management strategy of each target time step.

[0073] In this embodiment, the management device performs cache management according to the cache management strategy of each target time step, and constructs training samples for each target time step based on the state space data of each target time step and the cache management strategy.

[0074] Please see Figure 4 , Figure 4 The flowchart of S4 in the dynamic cache management method of the Internet of Vehicles provided in one embodiment of this application includes steps S41 to S43, as follows: S41: Perform cache management and reward calculation according to the cache management strategy of each target time step to obtain the global reward value of each target time step.

[0075] In this embodiment, the management device performs cache management and reward calculation according to the cache management strategy of each target time step to obtain the global reward value of each target time step.

[0076] Please see Figure 5 , Figure 5 The flowchart of S41 in the dynamic cache management method of the Internet of Vehicles provided in one embodiment of this application includes steps S411 to S412, as follows: S411: Execute cache management according to the cache management strategy of each target time step, and obtain the cache management execution indicators of each target time step.

[0077] In this embodiment, the management device performs cache management according to the cache management strategy of each target time step and obtains the cache management execution indicators of each target time step. The cache management execution indicators include the overall cache hit rate, access latency, bandwidth consumption, and resource prediction deviation.

[0078] S412: Calculate the reward based on the comprehensive cache hit rate, access latency, bandwidth consumption, resource prediction deviation, and the preset global reward calculation algorithm for each target time step to obtain the global reward value for each target time step.

[0079] In this embodiment, the management device calculates the reward based on the comprehensive cache hit rate, access latency, bandwidth consumption, resource prediction deviation, and a preset global reward calculation algorithm for each target time step, thereby obtaining the global reward value for each target time step. The global reward calculation algorithm is as follows:

[0080] In the formula, For the first t The global reward value at each target time step. For the first t Resource sufficiency at each target time step For the first t The overall cache hit rate for each target time step. For the first t Access latency for each target time step For the first t Bandwidth consumption per target time step For the first t Resource prediction bias at each target time step For the first t The validity of the data at each target time step is determined by the management device through real-time statistics from the onboard decision module. , , These are the first, second, and third reward coefficients, respectively.

[0081] S42: Based on the state space data of each target time step and the cache management strategy, obtain the predicted state space data of the next time step of each target time step.

[0082] In this embodiment, the management device obtains the predicted state space data of the next time step of each target time step based on the state space data of each target time step and the cache management strategy.

[0083] Please see Figure 6 , Figure 6The flowchart of S42 in the dynamic cache management method of the Internet of Vehicles provided in one embodiment of this application includes steps S421 to S424, as follows: S421: Obtain the space for deleting low-priority cached data at each target time step after performing cache management.

[0084] In this embodiment, the management device obtains the deletion space for low-priority cached data at each target time step after performing cache management.

[0085] S422: Based on the deletion space of low-priority cached data, the percentage of remaining cache space, the cache decision action encoding, the cache space occupancy prediction, and the preset cache remaining space percentage update algorithm for each target time step, obtain the percentage of remaining cache space for the next time step of each target time step.

[0086] In this embodiment, the management device obtains the percentage of the remaining cache space in the next time step of each target time step based on the deletion space of low-priority cached data, the percentage of remaining cache space, the cache decision action code, the cache space occupancy prediction, and a preset algorithm for updating the percentage of remaining cache space. The algorithm for updating the percentage of remaining cache space is as follows:

[0087] In the formula, For the first t The percentage of remaining cache space for each target time step. For the first t The percentage of remaining cache space for each target time step. For cache space usage prediction, For the first t Total onboard cache space for each target time step Space for deleting low-priority cached data. This is the second indicator function.

[0088] S423: Based on the bandwidth utilization rate, bandwidth consumption estimate, and preset bandwidth utilization update algorithm of each target time step, obtain the bandwidth utilization rate of the next time step of each target time step.

[0089] In this embodiment, the management device obtains the bandwidth utilization of the next time step for each target time step based on the bandwidth utilization rate, bandwidth consumption estimate, and a preset bandwidth utilization update algorithm. The bandwidth utilization update algorithm is as follows:

[0090] In the formula, For the first tBandwidth utilization at each target time step For bandwidth consumption estimation, For maximum bandwidth consumption, This is the third indicator function.

[0091] S424: Based on the percentage of remaining cache space, bandwidth utilization, and preset resource sufficiency update algorithm of each target time step, obtain the resource sufficiency of the next time step of each target time step.

[0092] In this embodiment, the management device obtains the resource sufficiency of the next time step for each target time step based on the percentage of remaining cache space, bandwidth utilization, and a preset resource sufficiency update algorithm. The resource sufficiency update algorithm is as follows:

[0093] In the formula, For the first t Resource sufficiency at each target time step.

[0094] S43: Combine the state space data of the same target time step, the cache management strategy, the global reward value of the target time step, and the predicted state space data of the next time step of each target time step to construct training samples for each target time step.

[0095] In this embodiment, the management device combines the state space data of the same target time step, the cache management strategy, the global reward value of the target time step, and the predicted state space data of the next time step of each target time step to construct training samples for each target time step.

[0096] S5: Update the dynamic cache management network based on the training samples of each target time step to obtain the target dynamic cache management network; perform cache management based on the state space data of the vehicle network at the current time step after the preset time period and the target dynamic cache management network.

[0097] In this embodiment, the management device updates the dynamic cache management network based on the training samples of each target time step to obtain the target dynamic cache management network. By constructing an active cache update mechanism, it adapts to the real-time dynamic change characteristics of vehicle network data, improves the accuracy of vehicle network cache demand prediction, and realizes time-series and refined prediction of cache demand.

[0098] In an optional embodiment, the management device employs a weighted filtering method to screen the training samples for each target time step, filtering out low-quality samples, samples that have failed execution, severely exceeded resource limits, or have low data validity, thereby obtaining a number of filtered training samples for the target time steps. The management device then updates the dynamic cache management network based on these filtered training samples to obtain the target dynamic cache management network.

[0099] The management device performs cache management based on the vehicle network state space data at the current time step after the preset time period and the target dynamic cache management network.

[0100] The dynamic cache management network also includes a value estimation network. See [link / reference] Figure 7 , Figure 7 The flowchart of S5 in the dynamic cache management method of the Internet of Vehicles provided in one embodiment of this application includes steps S51 to S55, as follows: S51: Based on the state space data in the training samples of each target time step and the predicted state space data of the next time step of each target time step, obtain the feature extraction vector of each target time step and the feature extraction vector of the next time step of each target time step.

[0101] In this embodiment, the management device obtains the feature extraction vectors for each target time step and the feature extraction vectors for the next target time step based on the state space data in the training samples of each target time step and the predicted state space data of the next target time step. Specific implementation details can be found in step S21, and will not be repeated here.

[0102] S52: Input the feature extraction vectors of each target time step and the feature extraction vectors of the next time step of each target time step into the value estimation network to calculate the state value estimation tensor, thereby obtaining the state value estimation tensor of each target time step and the state value estimation tensor of the next time step of each target time step.

[0103] In this embodiment, the management device inputs the feature extraction vectors of each target time step and the feature extraction vectors of the next time step of each target time step into the value estimation network to calculate the state value estimation tensor, thereby obtaining the state value estimation tensor of each target time step and the state value estimation tensor of the next time step of each target time step.

[0104] Specifically, the value estimation network adopts a standard value network architecture with two hidden fully connected layers and one output fully connected layer. The management device performs linear transformation and activation function mapping based on the feature extraction vectors of each target time step, the feature extraction vector of the next time step of each target time step, and the value estimation network to obtain the state value estimation tensor of each target time step and the state value estimation tensor of the next time step of each target time step, providing the core basis for the calculation of the advantage function, policy loss, and value loss optimization.

[0105] S53: Calculate the generalized dominance estimation tensor based on the global reward value of each target time step, the state value estimation tensor of each target time step, and the state value estimation tensor of the next time step of each target time step, to obtain the generalized dominance estimation tensor of each target time step.

[0106] In this embodiment, the management device performs generalized dominance estimation tensor calculation based on the global reward value of each target time step, the state value estimation tensor of each target time step, and the state value estimation tensor of the next time step of each target time step, to obtain the generalized dominance estimation tensor for each target time step. By accumulating the TD error of all future time steps, the long-term dominance value of the current action pair is estimated more accurately. The generalized dominance estimation tensor is:

[0107] In the formula, For the first t The TD error at each target time step As a discount factor, For the first t The state value estimation tensor of the next time step of the target time step. For the first t State value estimation tensor for each target time step For the first t Generalized advantage estimation tensor for each target time step. For balancing parameters, k Index for future time steps.

[0108] S54: Calculate the pruning strategy loss value based on the cache management strategy, global reward value, generalized advantage estimation tensor, and preset pruning strategy loss function for each target time step; calculate the value loss value based on the global reward value, state value estimation tensor, state value estimation tensor of the next time step for each target time step, and preset value loss function; calculate the entropy loss value based on the action space data of each target time step and preset entropy loss function.

[0109] In this embodiment, the management device calculates the loss based on the cache management strategy, global reward value, generalized advantage estimation tensor, and preset pruning strategy loss function for each target time step to obtain the pruning strategy loss value. The pruning strategy loss function is:

[0110] In the formula, The loss value for the pruning strategy. For shearing function, To find the expected function, This is the clipping threshold.

[0111] The management device calculates the value loss based on the global reward value at each target time step, the state value estimation tensor at each target time step, the state value estimation tensor at the next time step of each target time step, and a preset value loss function. The value loss function is as follows:

[0112] In the formula, This represents the value loss.

[0113] The management device calculates the entropy loss value based on the motion space data of each target time step and a preset entropy loss function. The entropy loss function is:

[0114] In the formula, This is the entropy loss value. This provides a complete action space.

[0115] S55: Update the dynamic cache management network based on the pruning strategy loss value, value loss value, and entropy loss value to obtain the target dynamic cache management network.

[0116] In this embodiment, the management device updates the dynamic cache management network based on the pruning strategy loss value, value loss value, and entropy loss value to obtain the target dynamic cache management network. Specifically, the control device performs a weighted summation of the pruning strategy loss value, value loss value, and entropy loss value to obtain the total loss. Backpropagation is then performed based on the obtained total loss to update the dynamic cache management network, obtaining the target dynamic cache management network. Real-time collection of decision execution effects and feedback data is also performed. Based on the obtained feedback data, model parameters are updated through backpropagation to achieve closed-loop optimization of the "prediction-decision" process, continuously improving the overall performance of the target dynamic cache management network and enhancing the accuracy of cache management.

[0117] Please refer to Figure 8 , Figure 8 This is a schematic diagram of the structure of a dynamic cache management device for a vehicle network according to an embodiment of this application. The device can be implemented in whole or in part through software, hardware, or a combination of both. The dynamic cache management device 8 for the vehicle network includes: The data acquisition module 81 is used to acquire the state space data of the vehicle network at each target time step within a preset time period, wherein the state space data includes the percentage of remaining cache space and the demand priority label; the target time step is any time step within the preset time period. The data analysis module 82 is used to analyze the state space data of the vehicle network at each target time step and the preset dynamic cache management network to obtain cache management analysis data for each target time step, wherein the cache management analysis data includes priority probability distribution data. The strategy reasoning module 83 is used to generate cache decision action codes based on the priority probability distribution data and demand priority labels of each target time step, and obtain the cache decision action codes for each target time step; based on the cache decision action codes and the proportion of remaining cache space for each target time step, it obtains the action execution timing and resource consumption prediction for each target time step; and combines the cache decision action codes, action execution timing and resource consumption prediction for each target time step to construct the cache management strategy for each target time step. The data processing module 84 is used to perform cache management according to the cache management strategy of each target time step, and to construct training samples for each target time step based on the state space data and cache management strategy of each target time step. The cache management module 85 is used to update the dynamic cache management network based on the training samples of each target time step to obtain the target dynamic cache management network; and to perform cache management based on the state space data of the vehicle network at the current time step after the preset time period and the target dynamic cache management network.

[0118] In this embodiment, a data acquisition module obtains the state space data of the vehicle network at each target time step within a preset time period. The state space data includes the percentage of remaining cache space and demand priority tags. The target time step is any time step within the preset time period. A data analysis module analyzes the state space data of the vehicle network at each target time step and a preset dynamic cache management network to obtain cache management analysis data for each target time step. This cache management analysis data includes priority probability distribution data. A strategy reasoning module generates cache decision action codes based on the priority probability distribution data and demand priority tags for each target time step, obtaining cache decision action codes for each target time step. Based on the cache decision data for each target time step... The system uses action encoding and cache remaining space percentage to obtain action execution timing and resource usage prediction for each target time step. It combines the cache decision action encoding, action execution timing, and resource usage prediction for each target time step to construct a cache management strategy for each target time step. Through the data processing module, cache management is executed according to the cache management strategy for each target time step. Based on the state space data and cache management strategy for each target time step, training samples for each target time step are constructed. Through the cache management module, the dynamic cache management network is updated based on the training samples for each target time step to obtain the target dynamic cache management network. Finally, cache management is executed based on the vehicle network state space data for the current time step after the preset time period and the target dynamic cache management network. Cache management strategies are generated based on the state space data of each target time step, and the cache management strategies for each target time step are obtained. Training samples for each target time step are constructed by combining the state space data and cache management strategies of each target time step. The training samples obtained for each target time step are used for dynamic cache management network updates, constructing an active cache update mechanism, adapting to the real-time dynamic change characteristics of vehicle network data, improving the accuracy of vehicle network cache demand prediction, and realizing time-series and refined prediction of cache demand.

[0119] Please refer to Figure 9 , Figure 9 This is a schematic diagram of the structure of a computer device provided in one embodiment of this application. The computer device 9 includes: a processor 91, a memory 92, and a computer program 93 stored in the memory 92 and executable on the processor 91; the computer device can store multiple instructions, which are adapted to be loaded and executed by the processor 91. Figures 1 to 7 For the method steps and specific execution process, please refer to [link / reference]. Figures 1 to 7 Specific details will not be elaborated here.

[0120] The processor 91 may include one or more processing cores. The processor 91 connects to various parts of the server using various interfaces and lines, and executes various functions and processes data of the vehicle-to-everything (V2X) dynamic cache management device 8 by running or executing instructions, programs, code sets, or instruction sets stored in the memory 92, and by calling data from the memory 92. Optionally, the processor 91 may 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 91 may integrate one or more 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 content required for the touch screen; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor 91.

[0121] The memory 92 may include random access memory (RAM) or read-only memory. Optionally, the memory 92 may include a non-transitory computer-readable storage medium. The memory 92 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 92 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch instructions), instructions for implementing the various method embodiments described above, etc.; the data storage area may store data involved in the various method embodiments described above, etc. Optionally, the memory 92 may also be at least one storage device located remotely from the aforementioned processor 91.

[0122] This application embodiment also provides a storage medium that can store multiple instructions, which are adapted to be loaded and executed by a processor as described above. Figures 1 to 7 For the method steps and specific execution process, please refer to [link / reference]. Figures 1 to 7 Specific details will not be elaborated here.

[0123] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0124] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0125] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the algorithm. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0126] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0127] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0128] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0129] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms.

[0130] This invention is not limited to the above-described embodiments. If any modifications or variations to this invention do not depart from the spirit and scope of this invention, and if such modifications and variations fall within the scope of the claims and equivalent technologies of this invention, then this invention also intends to include such modifications and variations.

Claims

1. A dynamic cache management method for Internet of Vehicles, characterized in that, Includes the following steps: Obtain the state space data of the vehicle network at each target time step within a preset time period, wherein the state space data includes the percentage of remaining cache space and the demand priority label; the target time step is any time step within the preset time period; Based on the state space data of the vehicle network at each target time step and the preset dynamic cache management network, the cache management analysis data of each target time step is obtained, wherein the cache management analysis data includes priority probability distribution data; Based on the priority probability distribution data and demand priority labels of each target time step, cache decision action codes are generated to obtain cache decision action codes for each target time step; based on the cache decision action codes and the proportion of remaining cache space for each target time step, the action execution timing and resource consumption prediction for each target time step are obtained; the cache decision action codes, action execution timing, and resource consumption prediction for each target time step are combined to construct the cache management strategy for each target time step. Cache management is performed according to the cache management strategy of each target time step. Based on the state space data of each target time step and the cache management strategy, training samples for each target time step are constructed. The dynamic cache management network is updated based on the training samples at each target time step to obtain the target dynamic cache management network; cache management is performed based on the vehicle network state space data at the current time step after the preset time period and the target dynamic cache management network.

2. The dynamic cache management method for the Internet of Vehicles according to claim 1, characterized in that: The state space data also includes cache requirement values; The process of obtaining the state space data of the vehicle network at each target time step within a preset time period includes the following steps: Obtain multi-source data of the Internet of Vehicles at each target time step and a preset first long short-term memory network; perform feature classification, weight setting and feature weighting based on the multi-source data of the Internet of Vehicles at each target time step, and construct the core layer weighted feature vector and the auxiliary layer weighted feature vector for each target time step; Obtain the memory unit feature vector and hidden state feature vector of the previous time step of each target time step calculated by the first long short-term memory network; extract hidden state features based on the core layer weighted feature vector of each target time step, the memory unit feature vector of the previous time step of each target time step, the hidden state feature vector, and the first long short-term memory network to obtain the first hidden state feature vector of each target time step. Obtain the feature vector of the historical log access segment; concatenate the first hidden state feature vector, the auxiliary layer weighted feature vector, and the historical log access segment feature vector of each target time step to obtain the concatenated feature vector of each target time step; input the concatenated feature vector of each target time step into a preset second long short-term memory network for hidden state feature extraction to obtain the second hidden state feature vector of each target time step. Dimensional compression and nonlinear mapping are performed on the second hidden state feature vectors of each target time step to obtain the global feature vectors of each target time step; cache requirement prediction is performed on the global feature vectors of each target time step to obtain the cache requirement values ​​of each target time step.

3. The dynamic cache management method for vehicle networking according to claim 2, characterized in that: The dynamic cache management network includes a shared feature extraction network, a priority classification network, and a policy network; the cache management analysis data includes priority probability distribution data and action probability distribution data. The step of analyzing the state space data of the vehicle network at each target time step and the preset dynamic cache management network to obtain cache management analysis data for each target time step includes the following steps: The state space data of each target time step is input into the shared feature extraction network for linear transformation, activation function mapping and regularization processing to obtain the feature extraction vector of each target time step; The feature extraction vectors of each target time step are input into the priority classification network to calculate the priority probability distribution, thereby obtaining the priority probability distribution data of each target time step, wherein the priority probability distribution data includes a number of priority probability distribution vectors. The feature extraction vectors of each target time step are input into the policy network to calculate the action probability distribution, thereby obtaining the action probability distribution data of each target time step, wherein the action probability distribution data includes action probability distribution vectors of several dimensions.

4. The dynamic cache management method for vehicle networking according to claim 3, characterized in that, The step of performing cache management according to the cache management strategy for each target time step, and constructing training samples for each target time step based on the state space data and cache management strategy for each target time step, includes the following steps: Cache management and reward calculation are performed according to the cache management strategy for each target time step to obtain the global reward value for each target time step; Based on the state space data of each target time step and the cache management strategy, the predicted state space data of the next time step of each target time step is obtained. The state space data of the same target time step, the cache management strategy, the global reward value of the target time step, and the predicted state space data of the next time step of each target time step are combined to construct the training samples for each target time step.

5. The dynamic cache management method for vehicle networking according to claim 4, characterized in that, The step of performing cache management and reward calculation based on the cache management strategy for each target time step to obtain the global reward value for each target time step includes the following steps: Cache management is performed according to the cache management strategy for each target time step, and cache management execution metrics for each target time step are obtained. The cache management execution metrics include comprehensive cache hit rate, access latency, bandwidth consumption, and resource prediction deviation. The reward is calculated based on the overall cache hit rate, access latency, bandwidth consumption, resource prediction deviation, and the preset global reward calculation algorithm for each target time step, to obtain the global reward value for each target time step.

6. The dynamic cache management method for vehicle networking according to claim 5, characterized in that: The state space data also includes bandwidth utilization and resource sufficiency; the resource usage prediction includes cache space usage prediction and bandwidth consumption estimation. The step of obtaining the predicted state space data for the next time step of each target time step based on the state space data of each target time step and the cache management strategy includes the following steps: Obtain the space for deleting low-priority cached data at each target time step after executing cache management; Based on the deletion space of low-priority cached data at each target time step, the percentage of remaining cache space, the encoding of cache decision actions, the prediction of cache space occupancy, and the preset algorithm for updating the percentage of remaining cache space, the percentage of remaining cache space at the next time step of each target time step is obtained. Based on the bandwidth utilization rate, bandwidth consumption prediction, and preset bandwidth utilization update algorithm of each target time step, the bandwidth utilization rate of the next time step of each target time step is obtained. Based on the percentage of remaining cache space, bandwidth utilization, and a preset resource sufficiency update algorithm for the next time step of each target time step, the resource sufficiency of the next time step is obtained.

7. The dynamic cache management method for vehicle networking according to claim 6, characterized in that: The dynamic cache management network also includes a value estimation network; The step of updating the dynamic cache management network based on the training samples at each target time step to obtain the target dynamic cache management network includes the following steps: Based on the state space data in the training samples of each target time step and the predicted state space data of the next time step of each target time step, the feature extraction vector of each target time step and the feature extraction vector of the next time step of each target time step are obtained. The feature extraction vectors of each target time step and the feature extraction vector of the next time step of each target time step are respectively input into the value estimation network to calculate the state value estimation tensor, thereby obtaining the state value estimation tensor of each target time step and the state value estimation tensor of the next time step of each target time step. The generalized dominance estimation tensor is calculated based on the global reward value of each target time step, the state value estimation tensor of each target time step, and the state value estimation tensor of the next time step of each target time step, to obtain the generalized dominance estimation tensor of each target time step. The loss is calculated based on the cache management strategy, global reward value, generalized advantage estimation tensor, and preset pruning strategy loss function for each target time step to obtain the pruning strategy loss value; the loss is calculated based on the global reward value, state value estimation tensor, state value estimation tensor of the next time step for each target time step, and preset value loss function to obtain the value loss value; the loss is calculated based on the action space data of each target time step and preset entropy loss function to obtain the entropy loss value. The dynamic cache management network is updated based on the pruning strategy loss value, value loss value, and entropy loss value to obtain the target dynamic cache management network.

8. A dynamic cache management device for vehicle networking, characterized in that, include: The data acquisition module is used to acquire the state space data of the vehicle network at each target time step within a preset time period. The state space data includes the percentage of remaining cache space and the demand priority label. The target time step is any time step within the preset time period. The data analysis module is used to analyze the state space data of the vehicle network at each target time step and the preset dynamic cache management network to obtain cache management analysis data for each target time step, wherein the cache management analysis data includes priority probability distribution data. The strategy reasoning module is used to generate cache decision action codes based on the priority probability distribution data and demand priority labels of each target time step, and obtain the cache decision action codes for each target time step; based on the cache decision action codes and the proportion of remaining cache space for each target time step, it obtains the action execution timing and resource consumption prediction for each target time step; and combines the cache decision action codes, action execution timing, and resource consumption prediction for each target time step to construct the cache management strategy for each target time step. The data processing module is used to perform cache management according to the cache management strategy of each target time step, and to construct training samples for each target time step based on the state space data and cache management strategy of each target time step. The cache management module is used to update the dynamic cache management network based on the training samples of each target time step to obtain the target dynamic cache management network; and to perform cache management based on the state space data of the vehicle network at the current time step after the preset time period and the target dynamic cache management network.

9. A computer device, characterized in that, It includes a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the dynamic cache management method for the Internet of Vehicles as described in any one of claims 1 to 7.

10. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the steps of the dynamic cache management method for the Internet of Vehicles as described in any one of claims 1 to 7.

Images