Vehicle-dependent task offloading method and system for edge computing

Abstract

This invention relates to a method and system for offloading vehicle-dependent tasks in edge computing. The method includes the following steps: clustering volunteer vehicles within a defined communication range to obtain volunteer vehicle clusters, where each cluster represents a volunteer alliance, and vehicles within the same alliance share resources; prioritizing all subtasks based on their latest execution time to obtain a priority queue; initializing the offloading point of each subtask according to the obtained priority queue to obtain a task offloading strategy; and iteratively optimizing the task offloading strategy using a multi-objective ant colony optimization algorithm. By evaluating task offloading on the vehicle edge server using two metrics—task completion latency and task execution cost—and minimizing task completion latency and task execution cost within the constraints of edge device computing resources and cache space, the task offloading problem is solved.

Description

Technical Field

[0001] This invention relates to the field of vehicle edge computing technology, specifically to a method and system for offloading vehicle-dependent tasks in edge computing. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Vehicle cloud computing (VCC) is a technology that solves the problem of vehicle network communication. It can utilize resources such as GPS, storage, Internet and vehicle computing power to make real-time decisions and process tasks in the cloud. However, because cloud servers are far away from vehicles, the communication latency is large and cannot meet the low latency requirements of in-vehicle applications. Therefore, VCC faces the challenge of low latency and high bandwidth.

[0004] To address the aforementioned issues, mobile edge computing (MEC) is currently being used to move servers to the network edge, providing users with low latency and high reliability services. For example, vehicle ad hoc networks (VANET) are integrated into MEC to form vehicle edge computing (VEC).

[0005] However, compared to cloud servers, VEC servers have limited processing power and available resources. Faced with a large number of computationally intensive vehicle applications, VEC servers will be unable to meet users' Quality of Service (QoS). To address the resource limitations of VECs, some existing technologies treat vehicles as providers of computing services. This means that vehicles can act not only as task initiators but also as task servers, thereby extending the capabilities of edge servers and improving user QoS. Cooperative task scheduling can also be addressed through vehicle cloud (VC) computing offloading, where computational tasks can be offloaded from the VC to vehicles for collaborative execution. However, in practical applications, most tasks generated by vehicles are related, so the order of task execution must be considered when executing related tasks, and task dependencies must be taken into account during task offloading. Existing algorithms in this area are not ideal. Summary of the Invention

[0006] To address the technical problems mentioned above, this invention provides a method and system for offloading vehicle-dependent tasks using edge computing. The method employs a multi-target ant colony algorithm to optimize the task offloading strategy, thereby improving the speed of task offloading and reducing the cost of task processing.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] The first aspect of the present invention provides a method for offloading vehicle-dependent tasks in edge computing, comprising the following steps:

[0009] Volunteer vehicles within the defined communication range are clustered to obtain volunteer vehicle clusters. Each cluster represents a volunteer alliance, and vehicle resources are shared within the same alliance.

[0010] All subtasks are prioritized based on their latest execution time to obtain a priority queue.

[0011] Based on the obtained priority queue, initialize the unloading point of each subtask to obtain the task unloading strategy;

[0012] Based on the obtained task unloading strategy, the multi-objective ant colony optimization algorithm is invoked for iterative optimization.

[0013] Furthermore, the methods for unloading subtasks include three categories:

[0014] The first category is to execute locally. If the available resources on the user's local machine can meet the task processing requirements and latency requirements, then the task can be executed locally.

[0015] The second type is offloading to a roadside unit. If the available resources on the user's local machine cannot meet the task processing requirements and latency requirements, then the task will be offloaded to a roadside unit for processing.

[0016] The third category involves offloading to volunteer vehicles. If the available resources on the user's local machine cannot meet the task processing requirements and latency requirements, then the task will be offloaded to a volunteer vehicle for processing.

[0017] Furthermore, the goal of the multi-objective ant colony optimization algorithm is to minimize task completion latency and task execution cost.

[0018] Furthermore, the constraints for minimizing task completion latency and task execution cost include:

[0019] The completion time for each subtask must not exceed the maximum completion time;

[0020] Each subtask can only be executed on one edge device;

[0021] The reward received by an edge device cannot exceed the offer that the user can make for the requested vehicle.

[0022] The load level of each edge device must not exceed its maximum load capacity;

[0023] The bandwidth resources allocated to each downlink cannot exceed the downlink bandwidth resources of the mobile device;

[0024] The bandwidth resources allocated to each uplink cannot exceed the uplink bandwidth resources of the mobile device;

[0025] The computing resources allocated to each subtask cannot exceed the total computing resources of the vehicle.

[0026] The computing resources allocated to each subtask cannot exceed the computing resources of all roadside units.

[0027] Furthermore, the analytic hierarchy process (AHP) is used to initialize the pheromone concentration and heuristic information for the ant colony algorithm.

[0028] The criteria layer of the analytic hierarchy process (AHP) considers the distance between the user-requesting vehicle and the edge devices, as well as the load level of each edge device.

[0029] A second aspect of the present invention provides a system for implementing the above-described method, comprising:

[0030] The volunteer alliance partitioning module is configured to cluster vehicles in the system so that vehicles with achievable density are grouped into one class.

[0031] The priority queuing module is configured to prioritize all subtasks according to their latest execution time, thus creating a priority queue.

[0032] Ant colony algorithm initialization module: Based on the priority queue, the analytic hierarchy process (AHP) is used to initialize the pheromone concentration and heuristic information;

[0033] The optimization module is configured to perform iterative optimization using a multi-objective ant colony algorithm based on the pheromone concentration generated during initialization and heuristic information.

[0034] A third aspect of the present invention provides a computer-readable storage medium.

[0035] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps in the vehicle-dependent task offloading method for edge computing as described above.

[0036] A fourth aspect of the present invention provides a computer device.

[0037] A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the steps of the vehicle-dependent task offloading method for edge computing as described above.

[0038] Compared with existing technologies, one or more of the above technical solutions have the following beneficial effects:

[0039] 1. Two metrics, task completion latency and task execution cost, were designed to evaluate task offloading on the vehicle edge server. Under the constraints of edge device computing resources and cache space, the task offloading problem was solved by minimizing task completion latency and task execution cost.

[0040] 2. The ant colony algorithm was improved by utilizing the ideas of Pareto advantage and analytic hierarchy process (AHP), and a new outer set update strategy was designed to achieve the goal. After evaluation, it was proved that the algorithm has better performance, improves task completion time, and reduces task execution cost. Attached Figure Description

[0041] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0042] Figure 1 This is a flowchart of a vehicle-dependent task offloading method for edge computing provided in one or more embodiments of the present invention;

[0043] Figure 2 This is a schematic diagram of task allocation during the task unloading process provided by one or more embodiments of the present invention. Detailed Implementation

[0044] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0045] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, 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 pertains.

[0046] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0047] As described in the background section, in the field of vehicular edge computing (VEC), when a vehicle user offloads multiple dependent tasks from multiple access points to the vehicular edge computing platform, the computing tasks can be offloaded from the VC to the vehicle for collaborative execution. However, in practical applications, most tasks generated by the vehicle are related. Therefore, the order of task execution must be considered when executing related tasks, and the dependencies between tasks must be taken into account during task offloading. Existing algorithms are not ideal in this regard.

[0048] Therefore, the following embodiments provide a method and system for unloading vehicle-dependent tasks in edge computing. The task unloading strategy is optimized by using a multi-target ant colony algorithm, which improves the speed of task unloading and reduces the cost of task processing.

[0049] Example 1:

[0050] like Figure 1-2 As shown, the vehicle-dependent task offloading method for edge computing is proposed firstly, which is a multi-volunteer vehicle-assisted computing task model in vehicle edge computing.

[0051] In this scenario, there are requesting vehicles, VEC servers, volunteer vehicles, and volunteer alliances. When the resources of the requesting vehicle meet the task's computational requirements, the task will be executed locally. When the resources of the requesting vehicle are insufficient, the task will be offloaded to a volunteer vehicle, a nearby RSU (a type of edge device), or an adjacent volunteer alliance.

[0052] A multi-vehicle edge computing network has M RSUs and N vehicles, represented by sets S = {s1, s2, ..., s...} M} and V={v1,v2,...,v N The symbol indicates that both the vehicle and the RSU are equipped with antennas, enabling them to transmit wirelessly to each other.

[0053] Suppose there are Z requesting vehicles in the VEC network, and each requesting vehicle has a task to perform. The requesting vehicle task consists of U interdependent subtasks, and each subtask can be executed on a different edge device.

[0054] The request is for vehicle i to generate subtask j, which is composed of triples. S indicates that i,j R represents the size of the subtask. i,j This represents the total number of CPU cycles required to complete a subtask. This indicates a delay constraint for the subtask.

[0055] The requested vehicle unloading policy can be represented as "off". i,j ∈{0, 1}. off i,j =1 indicates that the subtask is offloaded to a nearby edge device for execution, off i,j =0 indicates that the subtask is executed locally.

[0056] For ease of expression, we introduce a task unloading strategy X = {X1, X2, ..., X}. i ,...,X Z}, where X i ={x i,1 ,x i,2 ,...,x i,U}, x i,j ={x i,j,1 ,x i,j,2 ,...,x i,j,k ....,x i,j,M+N}

[0057] x i,j,k =1 indicates that an edge device is selected to execute the subtask; otherwise, x i,j,k =0, when the subtask selects edge device k, it will be unable to select other edge devices.

[0058] B k =(f k ,ω k C k ) represents the edge device k, where f k ω represents the computing power of edge devices. k It is the load level of the edge device, C k It is the amount of computing resources available on the edge device.

[0059] The dependencies between subtasks can be represented using a directed acyclic graph (DAG). For example, subtask T... 1,5 Only in subtask T 1,2 Sub-task T 1,3 Execute upon completion. Similarly, subtask T 1,6 Sub-task T 1,7 Only in subtask T 1,3 Execution occurs after processing. Subtasks at the same level can be executed in parallel, while sub-threads at different levels can only execute after their predecessor subtask has been completed.

[0060] Therefore, based on the workflow of vehicle edge computing, an objective function is designed, and the task unloading problem is modeled, which is planned to minimize the total task completion delay and task execution cost.

[0061] To better evaluate the rationality of the uninstallation location, two objective functions were designed: minimizing the total task completion delay and the task execution cost, expressed as:

[0062]

[0063]

[0064] C1:

[0065] C2:

[0066] C3:

[0067] C4:

[0068] C5:

[0069] C6:

[0070] Among them, FC i Indicates task T i Duration, This represents the total cost of requesting vehicle i to complete its own task. Let represent the latency of processing a task on the k-th edge device. Constraints C1 and C2 ensure that each subtask is executed on only one edge device. Constraint C3 ensures that each subtask does not exceed the maximum allowed latency limit. Constraint C4 ensures that the available computation on each edge device is sufficient to complete the processing task. Constraint C5 guarantees that the reward of the edge device is no greater than the offer that the requesting vehicle can provide. Constraint C6 ensures that the load level of each edge device does not exceed its maximum load limit.

[0071] Among them, FC i The calculation method is as follows:

[0072]

[0073]

[0074]

[0075]

[0076]

[0077]

[0078]

[0079]

[0080]

[0081] CT i,j =ES i,j +TE i,j

[0082]

[0083] Among them, v v,k W represents the offloading channel rate. v,k p represents the bandwidth requested for the channel allocation between vehicle v and edge device k. v Let d represent the transmission power of vehicle v, d be the distance from the vehicle to the edge server, α be the path loss exponent, and σ be the path loss index. 2 Add Gaussian white noise power. This indicates the latency of processing subtasks locally. This indicates the latency required for the task to be offloaded to the RSU for processing. This indicates the service rate of the RSU for this task. This indicates the task completion rate. This indicates the transmission delay of the task to the RSU. This indicates the transmission delay of the task to the volunteer vehicle. This indicates processing delays on volunteer vehicles. TE represents the migration delay of the task to the new volunteer vehicle. i,j Subtask T i,j Total execution latency, ES i,j Subtask T i,j The earliest start time, MR i,j CT indicates the moment when resources are available. i,j Subtask T i,j The final moment of completion.

[0084] Among them, task execution cost The calculation method is as follows:

[0085]

[0086]

[0087]

[0088]

[0089] m k =ξf k +ηω k

[0090]

[0091]

[0092]

[0093]

[0094] in, E represents the energy consumed when requesting the vehicle to execute a subtask locally. unit This indicates the amount of resources consumed per CPU cycle. This represents the energy consumed when the task is offloaded to the RSU to execute the subtask. κ represents the energy consumed by the transport and migration tasks. i,j Indicates whether a subtask needs to be migrated; if a subtask needs to be migrated, then κ. i,j =1, otherwise κ i,j =0, This represents the energy consumed when the data is unloaded onto volunteer vehicles to perform sub-tasks. This represents the amount of resources consumed per CPU cycle on the volunteer vehicle, m. k Let ξ represent the unit CPU cycle price of edge device k, ξ represent the unit price coefficient related to computing power, and η represent the unit price coefficient related to load level. This represents a reward for edge devices that assist in performing tasks. Indicates the energy consumption during transmission. Subtask T i,j The total cost on edge device k, where γ represents the unit price related to energy consumption, α represents the reward ratio, and β represents the self-energy consumption ratio.

[0095] This embodiment provides a method for offloading dependent tasks for vehicle users in multi-access vehicle edge computing, such as... Figure 1 As shown, the specific steps include:

[0096] Step 1: Cluster the volunteer vehicles using the DBSCAN clustering algorithm to divide the volunteer alliance into neighborhoods. First, assume ε represents the neighborhood radius and M represents the neighborhood density threshold for clustering. The number of vehicles in the ε-neighborhood of vehicle i is defined by the following formula:

[0097] N ε (i)={v∈V|d(i,v)<M}

[0098]

[0099] Where V represents the set of vehicles, and d(i,v) represents the straight-line distance between vehicle i and vehicle v.

[0100] The density of vehicle i is then expressed as:

[0101] ρ(i)=|N ε (i)|

[0102] Then the DBSCAN algorithm is used to solve for the volunteer alliance.

[0103] Step 2: Use the Analytic Hierarchy Process (AHP) to initialize pheromone concentration and heuristic information.

[0104] When using the Analytic Hierarchy Process (AHP) to initialize pheromone concentration and heuristic information, the target layer is for selecting appropriate edge devices (RSUs or volunteer vehicles). The criterion layer for selecting appropriate edge devices considers two factors: the distance between each edge device and the load level of the edge device. Appropriate edge devices are selected based on probabilities calculated using the AHP.

[0105] First, the judgment matrix is ​​defined:

[0106]

[0107] Here, 'c' represents the relative importance of distance factors compared to load level factors in selecting suitable volunteer vehicles. The value of 'c' ranges from 1 to 9.

[0108] Design importance matrices for the two factors in the criterion layer respectively:

[0109]

[0110]

[0111] Among them, B1(Vv) i ,Vv j ) represents the edge device Vv i Compared to edge devices Vv j The importance of distance, Dist(Vq) k ,Vv i ) indicates a request for vehicle Vq k To volunteer vehicle Vv i The distance; B2(Vv i ,Vv j ) represents the edge device Vv i Compared to edge devices Vv j Importance of load level, Load(Vv) i () indicates the load level on the volunteer vehicle.

[0112] Then perform a consistency check:

[0113]

[0114] Where, λ max Let n represent the largest eigenvalue of the matrix, and n+m represent the dimension of the matrix. Then, the consistency index RI is obtained by looking up a table. Finally, the formula for calculating the consistency ratio is derived:

[0115]

[0116] When the consistency ratio CR < 0.1, the degree of inconsistency of the matrix is ​​considered to be within the acceptable range, and it has satisfactory consistency. It passes the consistency test and its normalized eigenvector can be used as the weight vector.

[0117] After performing a consistency check on matrices A, B1, and B2, the corresponding weight vectors w are obtained. A We then calculated the initial pheromone concentration on the edge device v, expressed by the following formula:

[0118]

[0119] Step 3: Calculate the unloading probability using the transition probability formula, and select the unloading position for each ant using a roulette wheel. The specific process is as follows:

[0120]

[0121]

[0122]

[0123]

[0124]

[0125]

[0126] Where ρ is the pheromone evaporation parameter. The current pheromone concentration, Let λi represent the change in pheromone concentration. λ1 and λ2 represent the effects of the current solution and the outer set on the change in pheromone concentration, respectively. The heuristic for placing subtask j of task i on edge device v for processing depends on latency and cost, as shown below:

[0127]

[0128] c represents the impact of system performance on heuristic information. k is a parameter that controls the heuristic information, preventing it from getting stuck in local optima when it is too large, and from getting stuck in random search when it is too small, thus failing to find the optimal solution.

[0129] In the (t+1)th iteration, the roulette wheel algorithm is used to select the unloading location for the ant. The probability of the unloading location for task i is randomly selected. If the probability falls between volunteer vehicle v-1 and volunteer vehicle v, then task i is selected to unload onto volunteer vehicle v for processing.

[0130] P i =randi([0,1])

[0131]

[0132] Let ES represent the set of volunteer vehicles available for processing task data in the current network state. This represents the probability that task i will be offloaded to edge device v for processing.

[0133]

[0134] Step 4: Update the current solution and find the non-dominated solution using the fitness evaluation function;

[0135] The fitness function is shown below:

[0136]

[0137]

[0138] For a solution a1, fitness T () is a fitness evaluation function related to time T, and its value is inversely proportional to the magnitude of time T; fitneSs cost () is a fitness evaluation function related to cost. Where θ, χ and φ are both penalty factors. Therefore, the fitness function describes the minimum value of the objective function when the constraints are satisfied.

[0139] Dominant solutions are eliminated based on their fitness evaluation values, and non-dominated solutions are added to the outer set. For example, given two solutions al... i and al j If al i <al j Then al i Dominate al j Otherwise al j Dominate al i If al i and al j If they do not dominate each other, then they will all be added to the outer set.

[0140] Step 5: External set update

[0141] If the capacity of the outer set is not limited, the number of solutions within the outer set increases with the number of iterations, slowing down the algorithm and reducing its efficiency. Furthermore, since the solutions within the outer set may include both non-dominated and dominated solutions, these dominated solutions affect the calculation of pheromone concentration and the transition probability in the next iteration, preventing the ant colony from moving towards the optimal solution under these disturbances. Therefore, this embodiment limits the capacity of the outer set and incorporates an outer set elimination strategy. (Using Cap...) ER Let represent the maximum capacity of the outer set, that is, the number of solutions it can hold.

[0142] If the capacity of the outer set exceeds Cap ER (ER stands for outer set), and there are the following two operations:

[0143] 1. Restrict the conditions for adding the current solution to the outer set. The restriction method is to calculate the average T and average cost of the current outer set, i.e., Avg. T and Avg cost If the current solution s i If the following expression is satisfied, then s i Add to the external set.

[0144] T(s i )≥Avg T (ER)||cost(s i )≥Avg cost (ER)

[0145] 2. Eliminate some solutions from the outer set. In this embodiment, a Fitness Deviation Ratio (FDR) is defined to sort all solutions in the outer set. The FDR is shown below:

[0146]

[0147] in and Let represent the minimum value of T and the minimum value of cost in the outer set, respectively. For a solution al in the outer set, and Let T and W represent the time deviation ratio and cost deviation ratio, respectively, and let T and W be the distances from the optimal solution and the cost, respectively. The FDR of all solutions in the outer set is calculated using the above formula, and then sorted according to the size of the FDR value. The larger the FDR, the greater the T or cost in the solution must be compared to the optimal T or cost. Therefore, in this embodiment, FDR is used as an indicator of the outer set elimination strategy. The number of solutions added in step (1) is determined by the number of solutions eliminated according to FDR, thus keeping the number of solutions in the outer set at Cap. ER about.

[0148] Step 6: Repeat steps 2-5 until the iteration ends. Select a suitable solution from the Pareto optimal set (outer set).

[0149] The above process designed two metrics, task completion latency and task execution cost, to evaluate task offloading on the vehicle edge server. Under the constraints of edge device computing resources and cache space, the task offloading problem is solved by minimizing task completion latency and task execution cost.

[0150] This problem is NP-hard. The ant colony algorithm is improved by utilizing the ideas of Pareto dominance and the Analytic Hierarchy Process (AHP), and a new outer set update strategy is designed to achieve the goal. After evaluation, it is proved that the algorithm has good performance, improves the task completion time, and reduces the task execution cost.

[0151] Example 2:

[0152] A system for implementing the above method includes:

[0153] The volunteer alliance partitioning module is configured to cluster vehicles in the system so that vehicles with achievable density are grouped into one class.

[0154] The priority queuing module is configured to prioritize all subtasks according to their latest execution time, thus creating a priority queue.

[0155] Ant colony algorithm initialization module: Based on the priority queue, the analytic hierarchy process (AHP) is used to initialize the pheromone concentration and heuristic information;

[0156] The optimization module is configured to perform iterative optimization using a multi-objective ant colony algorithm based on the pheromone concentration generated during initialization and heuristic information.

[0157] The algorithm implemented by the above system uses two metrics, task completion latency and task execution cost, to evaluate task offloading on the vehicle edge server. Under the constraints of edge device computing resources and cache space, the algorithm aims to minimize task completion latency and task execution cost, thereby solving the task offloading problem.

[0158] Example 3:

[0159] This embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in the edge computing vehicle-dependent task offloading method as described in Embodiment 1 above.

[0160] The proposed method uses two metrics—task completion latency and task execution cost—to evaluate task offloading on the vehicle edge server. Under the constraints of edge device computing resources and cache space, the method aims to minimize task completion latency and task execution cost, thereby solving the task offloading problem.

[0161] Example 4:

[0162] This embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the edge computing vehicle-dependent task offloading method described in Embodiment 1 above.

[0163] The proposed method uses two metrics—task completion latency and task execution cost—to evaluate task offloading on the vehicle edge server. Under the constraints of edge device computing resources and cache space, the method aims to minimize task completion latency and task execution cost, thereby solving the task offloading problem.

[0164] The steps or modules involved in Embodiments 2 to 4 above correspond to those in Embodiment 1. For specific implementation details, please refer to the relevant description section of Embodiment 1. The term "computer-readable storage medium" should be understood as a single medium or multiple media including one or more instruction sets; it should also be understood as including any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and enabling the processor to perform any of the methods in this invention.

[0165] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for vehicle-dependent task offloading of edge computing, characterized in that: Includes the following steps: Volunteer vehicles within the defined communication range are clustered to obtain volunteer vehicle clusters. Each cluster represents a volunteer alliance, and vehicle resources are shared within the same alliance. All subtasks are prioritized based on their latest execution time to obtain a priority queue. Based on the obtained priority queue, initialize the unloading point of each subtask to obtain the task unloading strategy; Based on the obtained task unloading strategy, the multi-objective ant colony optimization algorithm is invoked for iterative optimization. The objectives of the multi-objective ant colony optimization algorithm are: to minimize task completion delay and task execution cost; and to select a suitable solution using an external set update strategy. The pheromone concentration and heuristic information for the ant colony algorithm are initialized using the analytic hierarchy process (AHP). The criterion layer of the AHP considers the distance between the user-requesting vehicle and the edge devices, as well as the load level of each edge device. Specifically: First, the judgment matrix A is defined, which is the importance matrix of the two factors of the criterion layer and respectively; then the consistency test is performed; finally, the consistency ratio is obtained; After the consistency check on matrix A, and the corresponding weight vector is obtained, and then the initial pheromone concentration on the edge device is obtained; The offloading probability is calculated using a transition probability formula, and the offloading position of each ant is selected through roulette. The heuristic information for placing the subtask j of the task i on the edge device v for processing depends on the latency and cost. In the (t+1)th iteration, the roulette algorithm is used to select the offloading position of the ant, and the offloading position probability of the task is randomly selected. If the probability falls between the volunteer vehicle v-1 and the volunteer vehicle v, the first task is selected to be offloaded to the volunteer vehicle v for processing. The offloading probability is calculated using a transition probability formula, and the offloading position of each ant is selected through roulette. The heuristic information for placing the subtask j of the task i on the edge device v for processing depends on the latency and cost. In the (t+1)th iteration, the roulette algorithm is used to select the offloading position of the ant, and the offloading position probability of the task is randomly selected. If the probability falls between the volunteer vehicle v-1 and the volunteer vehicle v, the first task is selected to be offloaded to the volunteer vehicle v for processing. The offloading probability is calculated using a transition probability formula, and the offloading position of each ant is selected through roulette. The heuristic information for placing the subtask j of the task i on the updating the current solution, finding non-dominated solutions by the fitness evaluation function; fitness function is a fitness evaluation function related to time T, whose value is inversely proportional to the size of time T; fitness function is a fitness evaluation function related to cost cost; eliminating dominated solutions according to the fitness evaluation value, and putting non-dominated solutions into the external set; ER represents the maximum capacity of the outer set; if the capacity of the outer set exceeds... Perform the following two operations: Step 1: Impose restrictions on the conditions for adding the current solution to the outer set, and calculate the average T and average cost of the current outer set, i.e. and If the current solution If the following expression is satisfied, then... Add to external set; Step 2: Eliminate some solutions from the outer set; Define the Fitness Bias Ratio (FDR), and sort all solutions in the outer set using the FDR, as shown below: For a solution within the outer set , and Let T and cost represent the time deviation ratio and cost deviation ratio, respectively, and let T and cost represent the distances from the optimal solution and the distances from the optimal solution, respectively. Calculate the FDR of all solutions in the outer set and sort them by FDR value. A larger FDR indicates that the T or cost in the solution is much larger than the optimal T or cost. Use FDR as an indicator of the outer set elimination strategy. Eliminate the solutions added in step 1 to keep the number of solutions in the outer set at a certain level. Within.

2. The edge computing vehicle-dependent task offloading method as described in claim 1, characterized in that: The methods for unloading subtasks include: The first category is to execute locally. If the available resources on the user's local machine can meet the task processing requirements and latency requirements, then the task can be executed locally. The second type is offloading to a roadside unit. If the available resources on the user's local machine cannot meet the task processing requirements and latency requirements, then the task will be offloaded to a roadside unit for processing.

3. The edge computing vehicle-dependent task offloading method as described in claim 2, characterized in that: Subtask unloading methods also include: The third category involves offloading to volunteer vehicles. If the available resources on the user's local machine cannot meet the task processing requirements and latency requirements, then the task will be offloaded to a volunteer vehicle for processing.

4. The edge computing vehicle-dependent task offloading method as described in claim 1, characterized in that: Constraints that minimize task completion latency and task execution cost include: The completion time for each subtask shall not exceed the maximum completion time; Each subtask is executed on only one edge device. The reward received by the edge device shall not exceed the quote offered by the vehicle requested by the user. The load level of each edge device shall not exceed its maximum load capacity; The bandwidth resources allocated to each downlink shall not exceed the downlink bandwidth resources of the mobile device; The bandwidth resources allocated to each uplink shall not exceed the uplink bandwidth resources of the mobile device; The computing resources allocated to each subtask shall not exceed the total computing resources of the vehicle. The computing resources allocated to each subtask shall not exceed the computing resources of all roadside units.

5. An edge computing vehicle-dependent task offloading system, employing the edge computing vehicle-dependent task offloading method as described in any one of claims 1-4, characterized in that: include: The volunteer alliance partitioning module is configured to cluster vehicles in the system so that vehicles with achievable density are grouped into one class. The priority queuing module is configured to prioritize all subtasks according to their latest execution time, thus creating a priority queue. Ant colony algorithm initialization module: Based on the priority queue, the analytic hierarchy process (AHP) is used to initialize the pheromone concentration and heuristic information; The optimization module is configured to perform iterative optimization using a multi-objective ant colony algorithm based on the pheromone concentration generated during initialization and heuristic information.

6. A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the vehicle-dependent task offloading method for edge computing as described in any one of claims 1-4.

7. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the steps of the vehicle-dependent task offloading method for edge computing as described in any one of claims 1-4.

Images