A vehicle networking method based on edge computing

Abstract

The present invention relates to the technical field of vehicle networking, and specifically to a vehicle networking method based on edge computing, including: judging whether the edge computing server needs to perform task migration according to the real-time load difference between edge computing servers, obtaining the data-location similarity of different tasks between edge computing servers to be migrated according to the similarity between task-related data of different tasks and the similarity of driving positions of vehicles between tasks, obtaining the reward function of the migration action, thereby iteratively updating the Q matrix of the Q learning algorithm, obtaining the optimal migration edge computing server of the edge computing server to be migrated, and then migrating the edge computing tasks in the vehicle networking. The present invention reduces the energy consumption of edge computing servers, improves overall resource utilization and load balancing, and improves the overall performance and sustainability of the vehicle networking system.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle networking, and in particular to a vehicle networking method based on edge computing. Background Art

[0002] The Internet of Vehicles is a complex system that connects vehicles, roads, pedestrians, and the cloud through communication technology. Edge computing, by equipping the roadside unit (RSU) of the Internet of Vehicles with an edge computing server, can receive data from the vehicle-mounted and roadside sensors in real time, analyze it, and transmit the results to other connected vehicles in the vicinity with extremely low latency.

[0003] In the vehicle networking method based on edge computing, due to the load imbalance problem between edge computing servers, edge computing migration is required to achieve edge-edge collaboration to improve resource utilization. The Q learning algorithm is usually used to achieve edge computing migration. The dynamic nature of the Internet of Vehicles determines that the edge computing server must have high real-time performance, and the Q learning algorithm requires a lot of iterations and explorations to find the optimal strategy, which may lead to untimely response in the dynamically changing Internet of Vehicles environment. Summary of the invention

[0004] The present invention provides a vehicle networking method based on edge computing to solve the existing problems.

[0005] A vehicle networking method based on edge computing of the present invention adopts the following technical solution:

[0006] An embodiment of the present invention provides a vehicle networking method based on edge computing, the method comprising the following steps:

[0007] Obtaining the operation data of the edge computing server, task-related data, and the distance between the edge computing servers, wherein the operation data includes at least: CPU utilization and total memory usage, and the task-related data includes at least: the rate of processing data when executing a number of tasks, and the driving position of the vehicle corresponding to each time in the number of tasks;

[0008] Analyze the CPU utilization, total memory usage, and data processing rate of the edge computing server when executing tasks, and calculate the real-time load of the edge computing server; by analyzing the real-time load differences between edge computing servers, determine whether the edge computing server needs to migrate tasks, and then select and obtain Class I edge computing servers;

[0009] According to the similarity between task-related data of different tasks and the similarity between the driving positions of vehicles between tasks, the data-position similarity of different tasks between the I-type edge computing servers is obtained; according to the data-position similarity of different tasks and the distance and state difference between the I-type edge computing servers, the reward function of the migration action is obtained, so as to iteratively update the Q matrix of the Q learning algorithm and obtain the optimal migration edge computing server of the I-type edge computing server;

[0010] According to the optimal migration edge computing server of Class I edge computing server, the edge computing tasks in the Internet of Vehicles are migrated.

[0011] Furthermore, the CPU utilization, total memory usage, and data processing rate of the edge computing server when executing tasks are analyzed, including the following specific methods:

[0012] Obtain the task execution effect of the edge computing server;

[0013] Multiply the CPU utilization, total memory usage, and task execution effect of the edge computing server to obtain the real-time load of the edge computing server.

[0014] Furthermore, the specific method for obtaining the task execution effect of the edge computing server is:

[0015]

[0016] in, Indicates The task execution effect of the edge computing server, Indicates The total number of tasks for edge computing servers, Indicates the current The edge computing server executes The rate at which data is processed during each task. Indicates Each edge computing server executes The average rate at which data is processed when performing a task. is the absolute value function.

[0017] Furthermore, the method of analyzing the real-time load difference between edge computing servers to determine whether the edge computing server needs to perform task migration, and then screening and obtaining Class I edge computing servers, includes the following specific methods:

[0018] Using the Q learning algorithm, the real-time load of each edge computing server is used as the initialization state of the Q learning algorithm, and the learning parameter A is preset. The action of the Q learning algorithm is the task migration of the edge computing server. The training goal of the Q learning algorithm is to make all edge computing servers achieve load balancing, and the initial Q matrix is ​​set to a zero matrix;

[0019] The load balancing means that the real-time load difference between any edge computing servers is less than or equal to a preset first parameter B;

[0020] Subtract the average state of all edge computing servers from the state of the edge computing server to obtain the current evaluation value of the edge computing server;

[0021] The edge computing servers whose current evaluation values ​​are greater than zero are recorded as Class I edge computing servers.

[0022] Furthermore, the data-location similarity of different tasks between the type I edge computing servers is obtained based on the similarity between the task-related data of different tasks and the similarity between the driving positions of vehicles between tasks, including the specific method of:

[0023] Obtain the data set of each task and the vehicle driving position sequence;

[0024] Obtain the similarity of data sets of different tasks and the similarity of vehicle driving position sequences;

[0025] The data-position similarity of different tasks is obtained by multiplying the similarity of data sets of different tasks with the similarity of vehicle driving position sequences.

[0026] Furthermore, the specific method for obtaining the data set of each task and the vehicle driving position sequence is:

[0027] Task-related data also includes: input data format, input data value, data type, output data format, and output data value of the task uploaded to the server;

[0028] For task-related data, the input data format, input data value, data type, output data format, and output data value uploaded to the server for each task constitute the data set of each task; in each task, the vehicle's driving position is organized into a vehicle driving position sequence in chronological order.

[0029] Furthermore, the specific method for obtaining the similarity of the data sets of different tasks and the similarity of the vehicle driving position sequences is:

[0030] The Jaccard similarity coefficient is used to obtain the similarity of data sets of different tasks between Class I edge computing servers, and the dynamic time warping algorithm is used to obtain the similarity of vehicle driving position sequences of different tasks between Class I edge computing servers.

[0031] Furthermore, the reward function of the migration action is obtained according to the data-location similarity of different tasks, the distance and state difference between the type I edge computing servers, so as to iteratively update the Q matrix of the Q learning algorithm to obtain the optimal migration edge computing server of the type I edge computing server, including the specific method as follows:

[0032] Get the number of similar task pairs between edge computing servers;

[0033] The reward function of the migration action is obtained as follows:

[0034]

[0035] in, Indicates Edge computing servers to The reward function for the migration action of an edge computing server; Indicates The edge computing server and The number of similar task pairs of edge computing servers; Indicates Edge computing servers to The Euclidean distance of edge computing servers; Indicates The status of the edge computing servers; Indicates The status of the edge computing servers; is the absolute value function, is an exponential function with a natural constant as base;

[0036] The Q matrix in the Q learning algorithm is updated according to the improved reward function and iterated until convergence to obtain the optimal migration edge computing server.

[0037] Furthermore, the specific method for obtaining the number of similar task pairs between edge computing servers is:

[0038] Task pairs whose data-location similarity of different tasks between Class I edge computing servers is greater than or equal to the preset second parameter are recorded as similar task pairs, similar task pairs of any two edge computing servers among Class I edge computing servers are obtained, and the number of similar task pairs is recorded as the number of similar task pairs between edge computing servers.

[0039] Furthermore, the specific method for obtaining the optimal migration edge computing server of the type I edge computing server for migrating edge computing tasks in the Internet of Vehicles is as follows:

[0040] The optimal migration edge computing server of type I edge computing servers is obtained through the improved Q learning algorithm. The tasks in type I edge computing servers are transferred to the optimal migration edge computing server in sequence according to the order of task similarity in descending order until load balance is achieved among the edge computing servers.

[0041] The beneficial effect of the technical solution of the present invention is that in the vehicle networking method based on edge computing, by utilizing the commonality of different vehicle data and considering the sharing of calculation results between similar computing tasks, the data processing efficiency can be significantly improved and redundant calculations can be reduced. The Q learning algorithm can take into account the reuse value of historical calculation results when making migration decisions, thereby optimizing resource allocation and task scheduling, which not only speeds up data processing and reduces latency, but also reduces the energy consumption of edge computing servers, improves the overall resource utilization and load balancing of edge computing servers, and is of great significance to improving the overall performance and sustainability of vehicle networking systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 This is a flow chart of the steps of a vehicle networking method based on edge computing of the present invention.

[0044] Figure 2 This is a flow chart of a Q learning algorithm for a vehicle networking method based on edge computing of the present invention. DETAILED DESCRIPTION

[0045] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of a vehicle networking method based on edge computing proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0046] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0047] The following is a detailed description of a specific solution of a vehicle networking method based on edge computing provided by the present invention in conjunction with the accompanying drawings.

[0048] See also Figure 1 , which shows a flow chart of a vehicle networking method based on edge computing provided by an embodiment of the present invention, the method comprising the following steps:

[0049] Step S001: Obtain the operating data of the edge computing server, task-related data, and the distance between the edge computing servers.

[0050] Specifically, in order to implement a vehicle networking method based on edge computing proposed in this embodiment, it is first necessary to collect the operation data of the edge computing server, task-related data, and the distance between the edge computing servers. The specific process is as follows:

[0051] Obtain the running data of the edge computing server, including CPU utilization and memory usage; obtain the task-related data of the edge computing server, including the total number of tasks, the speed of processing data when executing each task, the input data format, input data value, data type, output data format, output data value of each task uploaded to the server, the driving position of the vehicle in each task and the corresponding time; obtain the Euclidean distance between edge computing servers.

[0052] At this point, the operation data of the edge computing server, task-related data, and the Euclidean distance between the edge computing servers are obtained through the above method.

[0053] Step S002: Analyze the CPU utilization, total memory usage, and data processing rate of the edge computing server when executing tasks, and calculate the real-time load of the edge computing server; by analyzing the real-time load differences between edge computing servers, determine whether the edge computing server needs to migrate tasks, and then screen and obtain Class I edge computing servers.

[0054] It should be noted that when the task volume of a single edge computing server is overloaded, it is necessary to analyze the real-time load of nearby edge computing servers and migrate tasks to idle servers when migrating task computing.

[0055] Specifically, in step 2.1, the CPU utilization, total memory usage, and data processing rate of the edge computing server when executing tasks are analyzed to calculate the real-time load of the edge computing server.

[0056] It should be noted that in the vehicle networking method based on edge computing, such as busy roads during rush hour in the morning and evening, a single edge computing server may be limited by computing power, storage and communication resources when processing large-scale vehicle data. When the edge computing server is overloaded, there will be huge delays, making it difficult to answer the requests of the on-board users in a timely manner. The changes in vehicle mobility data cause the answers received by the on-board users to become less accurate and meaningless, or enter the range of the next server to requeue, affecting the decision-making behaviors of the on-board users such as driving planning. However, due to the differences in the spatiotemporal distribution of services, other edge computing servers have a certain amount of idle resources and can perform task migration. Task migration needs to consider many factors, such as the direction of vehicle movement, the distance between the edge computing server and the task response time. The Q learning algorithm can automatically learn and optimize the task migration strategy in edge computing to minimize delays and energy consumption.

[0057] It should be noted that the prerequisite for migrating the request task of the in-vehicle user is that the real-time load of the current edge computing server is too high. Therefore, the real-time load of each edge computing server is obtained first.

[0058] First, the task execution effect of the edge computing server is obtained based on the difference between the processing rate of all current tasks of the edge computing server and the average processing rate in the historical records.

[0059] As an embodiment, a method for obtaining the task execution effect of the edge computing server is:

[0060]

[0061] in, Indicates The task execution effect of the edge computing server, Indicates The total number of tasks for edge computing servers, Indicates the current The edge computing server executes The rate at which data is processed during each task. Indicates Each edge computing server executes The average rate at which data is processed when performing a task. is the absolute value function.

[0062] It should be noted that the historical records are historical records in which the number of onboard users is less than half of the current number of onboard users.

[0063] It should be noted that Indicates The difference between the processing rate of all current tasks of an edge computing server and the average processing rate in the historical relatively idle state represents the ability of the edge server to process the tasks of vehicle users at that moment. The larger the difference, the more it means that the edge server cannot process the tasks in time, which will cause a large delay and the current load of the server is too large.

[0064] Secondly, based on the operating data of the edge computing server, the real-time load of the edge computing server is obtained.

[0065] It should be noted that the real-time load of the edge computing server is related to the operating data of the edge computing server, and the operating data includes CPU utilization, memory usage, and task execution effect. The higher the CPU utilization, the higher the load of the server; when the total memory usage is close to the total memory of the server, it may cause server performance to decline or system crash, the delay of the server processing tasks increases, and the greater the load of the server; the worse the task execution effect, the edge server cannot process the task in time, which will cause a large delay, and the current load of the server is too large. Therefore, the higher the CPU utilization, the more total memory usage, and the greater the task execution effect, the greater the real-time load of the edge computing server.

[0066] As an embodiment, a method for obtaining the real-time load of an edge computing server is as follows:

[0067]

[0068] Indicates The real-time load of the edge computing servers; Indicates CPU utilization of edge computing servers; Indicates The total memory usage of edge computing servers, Indicates The task execution effect of an edge computing server.

[0069] Step 2.2: By analyzing the real-time load differences between edge computing servers, determine whether the edge computing servers need to migrate tasks, and then select Class I edge computing servers.

[0070] First, the Q learning algorithm is used. The real-time load of each edge computing server is used as the initialization state of the Q learning algorithm. The learning parameter A is preset. The action of the Q learning algorithm is the task migration of the edge computing server. The training goal of the Q learning algorithm is to make all edge computing servers achieve load balancing. The initial Q matrix is ​​set to a zero matrix.

[0071] It should be noted that, because the vehicle networking system focuses on real-time performance, the larger the learning parameter, the more importance it places on new action results. Therefore, A is preset to 0.8 based on experience and can be adjusted based on actual conditions. This embodiment does not make any specific limitations.

[0072] It should be noted that the load balancing means that the real-time load difference between any edge computing servers is less than or equal to the preset first parameter B.

[0073] It should be noted that B is preset to 0.2 based on experience and can be adjusted according to actual conditions, and is not specifically limited in this embodiment.

[0074] Secondly, according to the status difference of the edge computing server, the current evaluation value of the edge computing server is obtained.

[0075] As an embodiment, a method for obtaining the current evaluation value of the edge computing server is:

[0076]

[0077] Indicates The current evaluation value of edge computing servers; Indicates The status of the edge computing servers; Indicates the average status of all edge computing servers.

[0078] It should be noted that the real-time load of the edge computing server is also the status of the edge computing server, and the average real-time load of the overall server of the Internet of Vehicles is also the average status of all edge computing servers.

[0079] Finally, the edge computing servers whose current evaluation values ​​are greater than zero are recorded as Class I edge computing servers.

[0080] It should be noted that the edge computing server whose current evaluation value is greater than zero indicates that subsequent processing is required, and the selected action is to migrate the task to other edge computing servers, and the number of migrated tasks satisfies the load balancing with the target edge computing server after migration.

[0081] At this point, a Class I edge computing server is obtained through the above method.

[0082] Step S003: According to the similarity between task-related data of different tasks and the similarity between the driving positions of vehicles between tasks, the data-position similarity of different tasks between Class I edge computing servers is obtained; according to the data-position similarity of different tasks and the distance and state difference between Class I edge computing servers, the reward function of the migration action is obtained, so as to iteratively update the Q matrix of the Q learning algorithm and obtain the optimal migration edge computing server of the Class I edge computing server.

[0083] It should be noted that due to the dynamic nature of the vehicle networking system, request tasks need to be responded to in a timely manner. Therefore, the optimal migration target edge computing server can be selected based on the similarity between the request tasks of different vehicle users, and the computing resource results can be shared to improve the response speed.

[0084] Specifically, in step 3.1, based on the similarity between task-related data of different tasks and the similarity between the driving positions of vehicles between tasks, the data-location similarity of different tasks between the type I edge computing servers is obtained.

[0085] It should be noted that when the edge computing server that needs to migrate tasks chooses to randomly migrate tasks to another edge computing server, the Q learning algorithm uses a greedy strategy to optimize the state of the edge computing server after migration and minimize the real-time load. However, the transmission delay and energy consumption of the migration task may result in a longer response time for processing the task requested by the vehicle user and failure to deliver it in time. Therefore, considering that the task requests put forward by the vehicle users have commonalities, when the computing tasks are similar, it is possible to consider sharing the task results to reduce delays.

[0086] It should be noted that the task similarity can be measured by the input data format, data type, output data format, vehicle driving route and activity area uploaded to the server by the task.

[0087] First, for task-related data, the input data format, input data value, data type, output data format, and output data value uploaded to the server for each task constitute the data set of each task. In each task, the vehicle's driving position is organized into a vehicle driving position sequence in chronological order.

[0088] Secondly, the Jaccard similarity coefficient is used to obtain the similarity of data sets of different tasks between Class I edge computing servers, and the dynamic time warping algorithm is used to obtain the similarity of vehicle driving position sequences of different tasks between Class I edge computing servers.

[0089] It should be noted that the Jaccard similarity coefficient is an existing indicator for comparing the similarity between two sets, so it is not described in detail in this embodiment.

[0090] It should be noted that the dynamic time warping algorithm is an existing method for measuring the similarity between two time series, so it is not described in detail in this embodiment.

[0091] Finally, the data-position similarity of different tasks is obtained by combining the similarity of data sets of different tasks with the similarity of vehicle driving position sequences.

[0092] As an embodiment, a method for obtaining data-location similarity of different tasks is:

[0093]

[0094] in, Indicates Tasks and The data-location similarity of the tasks, For the Tasks and The similarity of the data sets of the tasks, For the Tasks and The similarity of the vehicle driving position sequences of the tasks.

[0095] It should be noted that the tasks mentioned here refer only to tasks between different edge computing servers, and do not include tasks between the same edge computing server.

[0096] In step 3.2, according to the data-location similarity of different tasks, the distance and state difference between the type I edge computing servers, the reward function of the migration action is obtained, so as to iteratively update the Q matrix of the Q learning algorithm and obtain the optimal migration edge computing server of the type I edge computing server.

[0097] It should be noted that the higher the data-location similarity between the tasks of the edge computing server and the edge computing server that performs task migration, the shorter the processing time of the migration task will be. The higher the data-location similarity between tasks while ensuring the reduction of real-time load, the higher the real-time performance of edge computing can be guaranteed while taking into account the dynamics of vehicle movement. Therefore, the reward function is improved through the data-location similarity between tasks of different edge servers.

[0098] It should be noted that the reward function represents the quality of the migration action. The higher the reward function value, the better the choice of the target server for the edge computing server task migration.

[0099] First, task pairs whose data-location similarity of different tasks between Class I edge computing servers is greater than or equal to a preset second parameter are recorded as similar task pairs, similar task pairs of any two edge computing servers in Class I edge computing servers are obtained, and the number of similar task pairs is recorded as the number of similar task pairs between edge computing servers.

[0100] It should be noted that the second parameter is preset to 0.7 based on experience, and can be adjusted according to actual conditions, and is not specifically limited in this embodiment.

[0101] It should be noted that for edge computing servers, the quality of migration actions is reflected in the number of similar task pairs between the edge computing server and the edge computing server to be migrated, the distance between the edge computing servers, and the state difference. The higher the similarity between the tasks of the edge computing servers and the more similar task pairs there are, the greater the possibility of migration. When the task is migrated, the lower the delay of task processing due to high task similarity, the higher the reward function value; when the two edge servers are closer, the lower the delay of their transmission tasks, and the faster the task processing speed is, the higher the reward value; the greater the difference in real-time load between the two edge computing servers, it means that one of them is overloaded, and the lower the load of the other, the more migration should be performed, and the greater the reward function value. Therefore, the reward function of the migration action can be obtained through the task similarity between edge computing servers, the distance between edge computing servers, and the state difference.

[0102] As an embodiment, the method for obtaining the reward function of the migration action is:

[0103]

[0104] in, Indicates Edge computing servers to The reward function for the migration action of an edge computing server; Indicates The edge computing server and The number of similar task pairs of edge computing servers; Indicates Edge computing servers to The Euclidean distance of edge computing servers; Indicates The status of the edge computing servers; Indicates The status of the edge computing servers; is the absolute value function, is an exponential function with a natural constant as its base.

[0105] It should be noted that Indicates The edge computing server and The state difference of the edge computing servers, that is, the difference in real-time load, The larger it is, the greater the difference in real-time load between the two edge computing servers is. If one of them is overloaded, the lower the real-time load of the other one is, and it is more suitable as the target edge computing server for task migration, and the larger the reward function is.

[0106] Finally, the Q matrix in the Q learning algorithm is updated according to the improved reward function and iterated until convergence to obtain the optimal migration edge computing server.

[0107] It should be noted that the process of updating the Q learning algorithm is as follows Figure 2 shown.

[0108] At this point, the optimal migration edge computing server is obtained through the above method.

[0109] Step S004: Migrate the tasks of the edge computing servers in the Internet of Vehicles according to the optimal migration edge computing server of the Class I edge computing server.

[0110] Specifically, the optimal migration edge computing server of Class I edge computing servers is obtained through the improved Q learning algorithm, and the tasks in Class I edge computing servers are transferred to the optimal migration edge computing server in sequence according to the order of descending similarity of tasks until load balance is achieved between the edge computing servers.

[0111] At this point, this embodiment is completed.

[0112] It should be noted that the The model is only used to represent negative correlation and constrain the output of the model to be in In the specific implementation, it can be replaced by other models with the same purpose. This embodiment is only based on The model is described as an example without any specific limitation. is the input to the model.

[0113] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A vehicle networking method based on edge computing, characterized in that: The method comprises the following steps: Obtaining the operation data of the edge computing server, task-related data, and the distance between the edge computing servers, wherein the operation data includes at least: CPU utilization, total memory usage, and the task-related data includes the total number of tasks, the speed of processing data when executing each task, the input data format, input data value, data type, output data format, output data value, and the driving position and corresponding time of the vehicle in each task uploaded to the server; Analyze the CPU utilization, total memory usage, and data processing rate of the edge computing server when executing tasks, and calculate the real-time load of the edge computing server; by analyzing the real-time load differences between edge computing servers, determine whether the edge computing server needs to migrate tasks, and then select and obtain Class I edge computing servers; According to the similarity between task-related data of different tasks and the similarity between the driving positions of vehicles between tasks, the data-position similarity of different tasks between the I-type edge computing servers is obtained, and the tasks are only between tasks of different edge computing servers, and do not include tasks of the same edge computing server; according to the data-position similarity of different tasks and the distance and state difference between the I-type edge computing servers, the reward function of the migration action is obtained, so as to iteratively update the Q matrix of the Q learning algorithm and obtain the optimal migration edge computing server of the I-type edge computing server; Migrate edge computing tasks in the Internet of Vehicles based on the optimal migration edge computing server of Class I edge computing servers; The method of analyzing the real-time load difference between edge computing servers to determine whether the edge computing server needs to perform task migration, and then screening and obtaining a Class I edge computing server, includes the following specific methods: Using the Q learning algorithm, the real-time load of each edge computing server is used as the initialization state of the Q learning algorithm, and the learning parameter A is preset. The action of the Q learning algorithm is the task migration of the edge computing server. The training goal of the Q learning algorithm is to make all edge computing servers achieve load balancing, and the initial Q matrix is ​​set to a zero matrix; The load balancing means that the real-time load difference between any edge computing servers is less than or equal to a preset first parameter B; Subtract the average state of all edge computing servers from the state of the edge computing server to obtain the current evaluation value of the edge computing server; The edge computing server with a current evaluation value greater than zero is recorded as a Class I edge computing server; The method of obtaining the data-location similarity of different tasks between the type I edge computing servers according to the similarity between the task-related data of different tasks and the similarity between the driving positions of the vehicles between the tasks includes: Obtain the data set of each task and the vehicle driving position sequence; Obtain the similarity of data sets of different tasks and the similarity of vehicle driving position sequences; Multiply the similarity of data sets of different tasks by the similarity of vehicle driving position sequences to obtain the data-position similarity of different tasks; The reward function of the migration action is obtained according to the data-location similarity of different tasks, the distance and state difference between the type I edge computing servers, so as to iteratively update the Q matrix of the Q learning algorithm to obtain the optimal migration edge computing server of the type I edge computing server, including the specific method as follows: Get the number of similar task pairs between edge computing servers; The reward function of the migration action is obtained as follows: in, Indicates Edge computing servers to The reward function for the migration action of an edge computing server; Indicates The edge computing server and The number of similar task pairs of edge computing servers; Indicates Edge computing servers to The Euclidean distance of edge computing servers; Indicates The status of the edge computing servers; Indicates The status of the edge computing servers; is the absolute value function, is an exponential function with a natural constant as base; The model indicates a negative correlation and the output of the constraint model is in the interval (0,1). Represents the input of the model; Update the Q matrix in the Q learning algorithm according to the improved reward function and iterate until convergence to obtain the optimal migration edge computing server; The specific method for obtaining the optimal migration edge computing server of the type I edge computing server for migrating edge computing tasks in the Internet of Vehicles is as follows: The optimal migration edge computing server of type I edge computing servers is obtained through the improved Q learning algorithm. The tasks in type I edge computing servers are transferred to the optimal migration edge computing server in sequence according to the order of task similarity in descending order until load balance is achieved among the edge computing servers.

2. The vehicle networking method based on edge computing according to claim 1, characterized in that: The specific method for analyzing the CPU utilization, total memory usage, and data processing rate of the edge computing server when executing tasks is as follows: Obtain the task execution effect of the edge computing server; Multiply the CPU utilization, total memory usage, and task execution effect of the edge computing server to obtain the real-time load of the edge computing server.

3. A vehicle networking method based on edge computing according to claim 2, characterized in that: The specific method for obtaining the task execution effect of the edge computing server is: in, Indicates The task execution effect of the edge computing server, Indicates The total number of tasks for edge computing servers, Indicates the current The edge computing server executes The rate at which data is processed during each task. Indicates Each edge computing server executes The average rate at which data is processed during each task. is the absolute value function.

4. The vehicle networking method based on edge computing according to claim 1, characterized in that: The specific method for obtaining the data set of each task and the vehicle driving position sequence is as follows: Task-related data also includes: input data format, input data value, data type, output data format, and output data value of the task uploaded to the server; For task-related data, the input data format, input data value, data type, output data format, and output data value uploaded to the server for each task constitute the data set of each task; in each task, the vehicle's driving position is organized into a vehicle driving position sequence in chronological order.

5. The vehicle networking method based on edge computing according to claim 1, characterized in that: The specific method for obtaining the similarity of the data sets of different tasks and the similarity of the vehicle driving position sequences is: The Jaccard similarity coefficient is used to obtain the similarity of data sets of different tasks between Class I edge computing servers, and the dynamic time warping algorithm is used to obtain the similarity of vehicle driving position sequences of different tasks between Class I edge computing servers.

6. The vehicle networking method based on edge computing according to claim 1, characterized in that: The specific method for obtaining the number of similar task pairs between edge computing servers is: Task pairs whose data-location similarity of different tasks between Class I edge computing servers is greater than or equal to the preset second parameter are recorded as similar task pairs, similar task pairs of any two edge computing servers among Class I edge computing servers are obtained, and the number of similar task pairs is recorded as the number of similar task pairs between edge computing servers.

Images