A multi-access edge computing vehicle task offloading method and system

Abstract

The application provides a kind of multi-access edge computing vehicle task unloading method and system, comprising: obtaining service vehicle information, edge server information, according to service vehicle information, edge server information, obtain task unloading available device set;Task vehicle's task unloading position and unloading proportion are initialized based on task unloading available device set, and task unloading strategy is obtained;With the total time and total energy consumption of task completion as target, the task unloading strategy is optimized using multi-objective particle swarm optimization algorithm, until optimization iteration condition is reached, and the final task unloading strategy is output.

Description

Technical Field

[0001] This invention belongs to the field of task offloading technology for Internet of Vehicles (IOV) devices, and particularly relates to a vehicle task offloading method and system for multi-access 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] With the development of IoT technology, vehicle-to-everything (V2X) technology has also entered a new stage. As an application of IoT, V2X has attracted increasing attention in recent years. By connecting vehicles with smart sensors, cameras, and other devices, applications such as object detection, traffic management, augmented reality, and autonomous driving have been realized. However, these applications generate a large number of latency-sensitive and computationally intensive tasks, and the computing power of the vehicle itself is far from sufficient to meet the demands of these tasks. Therefore, a major problem in V2X is how to offload tasks, that is, where to offload the tasks for processing, and whether the tasks should be offloaded entirely or partially. In recent years, various paradigms have been proposed for task offloading in V2X.

[0004] Tasks that cannot be processed locally in the vehicle can be offloaded to the cloud. In this process, devices with fewer resources offload computing or storage tasks to the cloud, forming mobile cloud computing (MCC). Using wireless communication, mobile cloud computing can provide intensive computing resources on a remote cloud to assist the vehicle in processing tasks.

[0005] While the cloud offers abundant computing and storage resources, its distance from users leads to increased network load as the number of users grows. Tasks may require waiting for processing, and congestion can occur during task upload and return, impacting processing efficiency. Therefore, data processing can be performed at the network edge, closer to the user, to improve efficiency. This computational shift of data processing to the network edge is called edge computing. Multi-access edge computing (MEC) distributes functions and services initially located in vehicles to the edge of the mobile network. These mobile computing devices are called edge servers, including mobile phones, nearby vehicles, and base stations. MEC provides computing resources to assist vehicles in performing tasks, meeting requirements such as low latency. This short-distance offloading in edge computing also protects against potential attacks from traditional transmissions, improving data integrity and security. In vehicular network applications, MEC can be considered as vehicle edge computing (VEC).

[0006] In recent years, a series of studies have focused on the offloading problem in vehicular networks (VECs), which can be divided into two categories based on the task offloading method. The first is binary offloading. In this case, the entire task is either offloaded to a local vehicle or an RSU. Game theory has been used to address the optimization problem in VECs for binary offloading. The second is partial offloading, which divides the task into two parts: one part is offloaded to a local vehicle, and the other part is offloaded to an RSU. Alternatively, one part is offloaded to a local vehicle, and the other part is offloaded to other idle vehicles. While binary offloading is easier to implement, partial offloading is more flexible and can effectively reduce latency and energy consumption of local devices.

[0007] However, none of the aforementioned studies considered fully utilizing the vehicle's own resources, nor did they comprehensively utilize idle vehicle and MEC resources. Therefore, designing a suitable unloading calculation scheme to fully utilize the collaborative resources between the vehicle and MEC is of great significance. Summary of the Invention

[0008] To overcome the shortcomings of the prior art, this invention provides a vehicle task unloading method and system with multi-access edge computing. It uses a multi-objective particle swarm optimization algorithm to optimize the task unloading ratio and unloading location, thereby reducing the task completion time and energy consumption.

[0009] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solution: a vehicle task offloading method for multi-access edge computing, comprising:

[0010] Obtain service vehicle information and edge server information, and obtain the set of available devices for task unloading based on the service vehicle information and edge server information;

[0011] The task unloading strategy is obtained by initializing the task unloading location and unloading ratio of the task vehicle based on the set of available devices for task unloading;

[0012] With the goal of minimizing the total time and energy consumption for task completion, the task unloading strategy is optimized using a multi-objective particle swarm optimization algorithm until the optimization iteration conditions are met, and the final task unloading strategy is output.

[0013] A second aspect of the present invention provides a vehicle task offloading system with multi-access edge computing, comprising:

[0014] The acquisition module is configured to: acquire service vehicle information and edge server information, and obtain the set of available devices for task unloading based on the service vehicle information and edge server information;

[0015] The initialization module is configured to initialize the task unloading location and unloading ratio of the task vehicle based on the set of available task unloading devices, and obtain the task unloading strategy.

[0016] The optimization module is configured to optimize the task unloading strategy using a multi-objective particle swarm optimization algorithm with the goal of minimizing the total time and energy consumption of task completion, until the optimization iteration conditions are met, and then output the final task unloading strategy.

[0017] A third aspect of the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, perform the steps described in the above method.

[0018] A fourth aspect of the present invention provides an electronic device including a memory and a processor, and computer instructions stored in the memory and running on the processor, wherein the computer instructions, when executed by the processor, perform the steps described in the above method.

[0019] The above one or more technical solutions have the following beneficial effects:

[0020] This invention designs two metrics, latency and energy consumption, to evaluate task offloading on vehicles and edge servers. Under the constraints of computing resources and processing capabilities of vehicles and edge servers, a task offloading problem is proposed with the goal of minimizing the total time and energy consumption for task completion. A multi-objective particle swarm optimization algorithm is adopted for the task offloading problem. Pareto optimality relation is introduced in the optimization process to find the optimal solution, which reduces the time and energy consumption for task completion and improves the speed of task offloading.

[0021] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] 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.

[0023] Figure 1 This is a flowchart illustrating the unloading of vehicle tasks in multi-access edge computing in Embodiment 1 of the present invention. Detailed Implementation

[0024] 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.

[0025] It should be noted that the terminology used herein is for the purpose of describing particular implementations only and is not intended to limit the exemplary implementations of the present invention.

[0026] Where there is no conflict, the embodiments and features in the embodiments of the present invention can be combined with each other.

[0027] Terminology Explanation:

[0028] Particle Swarm Optimization (PSO) is an algorithm inspired by the foraging behavior of birds. Each bird represents a particle, the bird's flight space is considered the search space of the optimization problem, the bird's flight speed is considered the particle's movement speed, and the food source the bird searches for is the optimal solution to the optimization problem. In PSO, particles update their speed and position by finding local and global optima. The PSO algorithm is described as follows:

[0029]

[0030]

[0031] in, This represents the particle's velocity at the current moment. This represents the location of the local optimal solution for the particle. This represents the position of the particle's global optimal solution. ω represents the current position of the particle. c1 and c2 are non-negative constants, known as inertia factors, r1 and r2 are random numbers distributed in the interval [0,1], and ω is the inertia weight.

[0032] Example 1

[0033] This embodiment discloses a vehicle task offloading method for multi-access edge computing, including:

[0034] Step 1: Obtain service vehicle information and edge server information, and obtain the set of available devices for task unloading based on the service vehicle information and edge server information;

[0035] Step 2: Initialize the task unloading location and unloading ratio of the task vehicle based on the set of available devices for task unloading to obtain the task unloading strategy;

[0036] Step 3: With the goal of minimizing the total time and energy consumption for task completion, optimize the task unloading strategy using the multi-objective particle swarm optimization algorithm until the optimization iteration conditions are met, and output the final task unloading strategy.

[0037] In this embodiment, the tasks are divided into two categories: the first category can only be executed on the local device; the second category can be executed partly locally and partly on the service vehicle, or partly locally and partly on the edge server.

[0038] In this embodiment, the execution method of the task is determined;

[0039] If the task size is smaller than the vehicle's local processing capacity, it will be executed directly on the local device.

[0040] If the size exceeds the local processing capacity of the vehicle, the task will be partially processed locally and partially offloaded to other service vehicles. Alternatively, the task can be partially processed locally and partially offloaded to an edge server.

[0041] In this embodiment, an objective function is designed to model the task unloading problem, which is planned as minimizing the total time and total energy consumption for task completion.

[0042] To better evaluate the rationality of the unloading location and unloading ratio, two objective functions were designed: minimizing the total time and total energy consumption for task completion, denoted as T1 and T2 as follows:

[0043] T1 = minT (1)

[0044] T2=minE (2)

[0045] T = {T1, T2} (3)

[0046] Where T represents the total time required to process the task, and E represents the total energy required to process the task.

[0047] The established constraints are as follows:

[0048]

[0049]

[0050]

[0051]

[0052]

[0053]

[0054] C4:E i ≤E max ,1≤i≤n (10)

[0055]

[0056] C6:0≤p i ≤p max ,1≤i≤n (12)

[0057] Among them, constraint C1 in the objective function is a bandwidth constraint, indicating that the transmission power must be greater than the required minimum transmission power; R i,i' R represents the power transmitted between vehicles. i,jThis indicates the transmission power between the vehicle and the MEC server.

[0058] Constraint C2 in the objective function is a delay tolerance constraint, which means that the task completion time of the vehicle and MEC must meet the maximum allowable delay requirement; This indicates the maximum allowed time for the vehicle to complete the task. This indicates the maximum task completion time allowed by MEC.

[0059] Constraint 3 C3 in the objective function is the processing capacity constraint, which means that the vehicle and MEC cannot exceed their specified maximum processing capacity; Indicates the vehicle's processing capacity. This indicates the processing capacity of the MEC.

[0060] Constraint C4 in the objective function is an energy consumption constraint, meaning that the energy consumption of each task is less than the maximum available energy; E i This indicates the energy consumption of each task.

[0061] Constraint C5 in the objective function is a task allocation ratio constraint, indicating that the task allocation ratio is between 0 and 1; λ s This indicates the task allocation ratio.

[0062] Constraint 6C6 in the objective function is the mission offload power constraint, indicating that the mission transmit power has a threshold and is non-negative; p i This indicates the transmission power of each vehicle.

[0063] In this embodiment, the method for calculating the task delay T is as follows:

[0064]

[0065] Among them, R i,i' This represents the transmission rate between task vehicle i and service vehicle i'. B i p represents the transmission bandwidth between vehicles. i,i' Represents task P s Offloaded transmission power to other vehicles, h i,i' It is the channel gain of the link between vehicles, σ i,i' 2 This indicates the noise power between vehicles.

[0066]

[0067] in, This indicates the transfer time for the task to be unloaded and moved to other vehicles. s λ represents the size of the task data. s This indicates the proportion of tasks that are offloaded to other vehicles.

[0068]

[0069]

[0070] Among them, R i,j This indicates the transmission rate between the vehicle and the MEC. (B) j p represents the transmission bandwidth between the vehicle and the MEC. i,j h represents the transmission power that subtask Si offloads to the MEC server. i,j It is the channel gain of the link between the vehicle and the MEC, σ i,j 2 This indicates the noise power between the vehicle and the MEC server.

[0071]

[0072] in, This indicates the transfer time from task unloading to MEC. Input data size I s R i,j This refers to the transmission power between the vehicle and the MEC. When a task is executed locally, transmission latency and energy consumption do not need to be considered; only task execution time and energy consumption are taken into account.

[0073]

[0074] Among them, T l Indicates the local calculation time. s The cycle represents the size of the task data. il This is the CPU cycle required for vehicle i to process 1 bit of data locally. The CPU speed of local vehicle i.

[0075]

[0076]

[0077]

[0078]

[0079]

[0080] Because in the unload execution mode, local and vehicle, and local and MEC are executed in parallel, the parallel execution time is T. p The local computation time for uninstallation is T. lp The vehicle's calculation time during unloading is... This is the CPU rotation speed of service vehicle i. The total time for task offloading to other vehicles is... We use Φv The ∈(0,1) indicates whether a vehicle has a task currently being executed when the task is unloaded to it. If Φ v =0, indicating that when the task arrives at the vehicle, the vehicle does not have any tasks being processed and does not need to wait in the queue; if Φ v =1 indicates that when the task arrives at the vehicle, the vehicle already has other tasks being processed, so the task needs to wait in the queue. The computation time for the task to be unloaded to the MEC is... The number of cycles that the MEC CPU processes per unit of data. js MEC CPU cycle frequency λ s ×I s ×cycle js This indicates the computational load of the current task in MEC.

[0081]

[0082] The method for calculating the energy consumption E of task execution is as follows:

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090]

[0091]

[0092] in, This indicates the energy consumption of transferring the task to the service vehicle. For local execution, the energy consumption is local. This represents the energy consumed by vehicle i's CPU in one cycle. The local execution energy consumption in unloading mode is... This represents the energy consumed by the vehicle's local CPU within one cycle. The energy consumption of other vehicles during unloading mode is... cycle is This represents the number of CPU cycles for vehicle i to process each unit of data. The CPU cycle frequency of the vehicle, λs ×I s ×cycle is This indicates the computational load of the current task on the vehicle. This represents the energy consumed by the vehicle's CPU in one cycle. MEC calculation energy consumption is... This represents the energy consumed by the CPU of an MEC server in one cycle. The energy consumption of a vehicle in partial offload mode is E. v In partial offloading mode, the energy consumption of MEC is E. m The total energy consumption is E.

[0093] In this embodiment, the task unloading location and unloading ratio are initialized. Considering the distance between each task's vehicle and other service vehicles or between the task vehicle and its MEC server, and the load capacity of the service vehicles and MEC servers, the entropy weight method is used to initialize the unloading location. Specifically:

[0094] Before using the entropy weight method to select equipment, it is first necessary to consider the distance formula. Calculate the distances between the task and the vehicle, and between the task and the MEC (Multi-access Edge Computing) system. Determine the available devices within the coverage area of ​​each task vehicle, and establish a set of usable devices. If a device is within the task's coverage area, assign values ​​to two key performance indicators (KPIs) for that device; otherwise, do not assign values. Here, x1 and y1 represent the coordinates of the task vehicle, and x2 and y2 represent the coordinates of the service vehicle or MEC server.

[0095] We need to assign weights to two key performance indicators (KPIs) for each task to more accurately evaluate each device. The two KPIs are distance and the load capacity of the vehicle or MEC. Closer distances and higher load capacities of the vehicle or NEC are assigned higher weights.

[0096] There are α available devices within the coverage area of ​​the mission vehicle. Available devices are those on which computational tasks can be offloaded. Available devices can be either vehicles or MEC servers. The matrix consisting of the distances between the α available devices and the mission vehicle, and the load capacity of the available devices, is as follows:

[0097]

[0098] matrix The data in the matrix is ​​normalized (standardized). The purpose is to normalize the matrix. The absolute values ​​of the data are converted into relative values.

[0099] In this embodiment, β (β∈{1,2}) is used to represent the two indicators: the distance between available equipment and the mission vehicle, and the load capacity of available equipment. Let the elements in the standardized matrix be represented as u. αβ ', use formula (34) on the matrix Data normalization:

[0100]

[0101] Then, calculate the proportion of the αth available device under the conditions of distance and load capacity respectively:

[0102]

[0103] ∑O αβ =1 (35)

[0104] Calculate the information entropy value under the conditions of distance and load capacity:

[0105]

[0106] Calculate the weights for distance and load capacity separately:

[0107]

[0108] Finally, calculate the overall score for each available device:

[0109]

[0110] The final score is obtained by multiplying the distance and load capacity by a standardized matrix. The coefficients 0.998 and 0.002 are used to make u... αβ The value is greater than 0. Here, 0.998 can be changed to any number closer to 1, such as 0.999, 0.997, etc. αβ u represents the score of the αth available device under the conditions of distance and load capacity. αβ 'for u αβ The standardized value. αβ Let e ​​be the weight of the αth available device under the conditions of distance and load capacity. β The information entropy values ​​are distance and load capacity. τ is the weight for distance and load capacity. α This represents the overall score for all available devices. Γ represents the number of available devices.

[0111] Then, based on the final score, a roulette wheel algorithm is used to select either a vehicle or a MEC as the unloading location during initialization. The higher the score of a location, the greater the probability that the task will be unloaded onto that location during initialization. The specific process of the roulette wheel algorithm is as follows:

[0112] First, calculate the probability that task s selects a certain vehicle or MEC through formula (39), then calculate the cumulative probability of each device through formula (40), and sort the cumulative probabilities in ascending order. Finally, generate a random array r ∈ (0, 1). If r < q α , then select individual α; otherwise, select individual ι such that q[ι - 1] < r ≤ q[ι] holds.

[0113]

[0114]

[0115] After initializing the offloading ratio, offload the tasks to the selected vehicles or MECs according to the offloading ratio.

[0116] Among them, τ α is the comprehensive score of each available device, p α is the proportion of a certain available device in the roulette wheel, and q α is the cumulative probability of the device. The cumulative probability represents the probabilities of each available device with line segments of different lengths, and these line segments are combined into a straight line with a length of 1 (the sum of the probabilities of each individual). In this straight line, the longer a certain segment of the line is, the greater the probability that the device is selected. The cumulative probability of any device is the sum of the previous several data corresponding to the device.

[0117] For example Figure 1 As shown, in this implementation, aiming at minimizing the total time and total energy consumption for task completion, the multi-objective particle swarm optimization algorithm is used to optimize the task offloading strategy. Specifically:

[0118] Step 301: Initialize the positions and velocities of the particles in the population of the multi-objective particle swarm optimization algorithm. The position of the particle is the task offloading strategy. ]>

[0119] Specifically, a 2*K matrix y is designed to represent the offloading decision of the task:

[0120]

[0121] In the matrix y, B s = {0, 1, 2, … m, m + 1, …, m + n} represents the position of offloading to the vehicle or MEC. If B s = 0, it means that the s-th task is completely processed locally; if B s = 1, it means that the s-th task is partially offloaded to the first MEC; if B s = m + 2, it means that the s-th task is partially offloaded to the second vehicle, and so on. λ s = {λ1, λ2, λ3, …, λk} indicates the proportion of the task that will be unloaded from execution. If the task is processed entirely locally, then λ s =0 and B s =0; if part of the task is processed locally, then λ s ≠0 and B s ≠0.

[0122] Assume the population size is q. Each particle has two attributes: position and velocity.

[0123] The position of each particle in D-dimensional space:

[0124] χ i =(x i1 ,x i2 ,x i3 ,…,x id ,…,x iD (42)

[0125] Individual 1≤s≤k.

[0126] Overall, a single k×D matrix is ​​used.

[0127]

[0128] The velocity of each particle in D-dimensional space:

[0129]

[0130] The optimal individual position k×D of the particle:

[0131] P i =(p i1 ,p i2 ,p i3 ,…,p id ,…,p iD (45)

[0132] The global optimal position of the particle is 1×D:

[0133] P g =(p g1 ,p g2 ,p g3 ,…,p gd ,…,p gD (46)

[0134] Step 302: Design a fitness function based on task execution time and execution energy consumption.

[0135] As the total time and energy consumption for task completion increase, the fitness function value increases. Our goal is to minimize the total time and energy consumption for task completion, so we should look for solutions with smaller fitness function values.

[0136] Because our problem is a bi-objective problem, the optimization processes of the two objective functions may conflict during the search for a local optimum. Therefore, we designed a fitness function that comprehensively considers time delay and energy consumption to select the local optimum.

[0137] Specifically, the fitness function is:

[0138]

[0139]

[0140]

[0141] Among them, f T (sl) and f E (sl) is a function of execution time and execution energy consumption, inversely proportional to both time and energy consumption. Here, θ1 and θ2 are exponents of task execution time and task execution energy consumption, which can be set according to user needs. T1 and T2 are the two objective functions of the design, namely, minimizing the total time and total energy consumption for task completion.

[0142] Step 303: Calculate the objective function value for each particle, and compare the calculated objective function values ​​according to the Pareto dominance relation. Add the non-dominated solutions to the outer set.

[0143] Specifically, the objective function values ​​for the two different particles are T. 1i and T 1j T 2i and T 2j When T 1i >T 1j And T 2i ≥T 2j Or T 1i ≥T 1j And T 2i >T 2j If i dominates j, add i to the outer set; otherwise, add j to the outer set; if i and j do not dominate each other, add i and j together to the outer set.

[0144] Step 304: Find the local and global optima in the Pareto set based on the fitness function. Update the particle's position based on the particle's current solution, local optima, and global optima, recalculate the particle's fitness function and Pareto dominance, and add non-dominated solutions to the outer set.

[0145] Local optimal solution:

[0146] The initial fitness value is used as the local optimum for each particle. When updating the local optimum, each particle is iterated in the population, and the fitness value of the current solution is compared with the historical optimum of each particle. If the fitness function of the current solution is smaller, the local optimum is changed to the current function; otherwise, the local optimum remains unchanged.

[0147] Global optimal solution:

[0148] The optimal initial fitness value is taken as the current global optimum, and the position corresponding to the optimal fitness value is taken as the position of the global optimum. The solution with the minimum fitness function value in the outer set is the global optimum.

[0149] To prevent getting trapped in local optima, a probability transition formula is designed to optimize the particle update process. The probability transition formula is as follows:

[0150]

[0151]

[0152]

[0153]

[0154] In the formula, trp cur , and Let represent the probabilities of maintaining the previous execution method, the execution method close to the local optimum, and the execution method close to the global optimum, respectively. τ1 and τ2 represent the weighting coefficients for the completion time and execution energy consumption of the task in the decision-making process, respectively.

[0155] f T (x) and f E (x) is a correlation function of time and energy consumption, and is inversely proportional to time and energy consumption. T (x)+f E The larger the value of (x), the greater the probability of transitioning to that solution. For example, given a random probability trp x ∈[0,1], if trp x ∈[0,trp cur If ], then choose to move in the current particle's direction of movement; if Then choose to move towards the particle's local optimal position; if Then, the particle is moved to its globally optimal position.

[0156] In this embodiment, an outer set is used to store Pareto solutions, and the number of solutions in the outer set is capped. If necessary, one or more solutions in the outer set need to be removed to meet the threshold limit of the outer set's capacity.

[0157] When updating the outer set, the following cases apply: If the outer set is empty, the newly generated solution is directly added to the outer set; if the newly generated solution is dominated by any solution in the outer set, the outer set is not updated; if the newly generated solution dominates any solution in the outer set, the dominated old solution is removed from the outer set, and the newly generated solution is added; if the newly generated solution and a solution in the outer set are mutually non-dominant, it is determined whether the outer set is full: if the outer set is not full, it is directly added; if the outer set is full, an adaptive mesh method is used to delete solutions from the outer set, and the solution density value is calculated. If the density value of the newly generated solution is large, it is not added to the outer set; if the density value of a solution in the outer set is large, the solution with the largest density value in the outer set is deleted, and the newly generated solution is added to the outer set.

[0158] The steps of the adaptive mesh method are as follows: Calculate the boundary of the target space in the current iteration. and Where z is the iteration number;

[0159] Calculate the modulus of the mesh:

[0160]

[0161]

[0162] G = M × M (56)

[0163] G is the number of grid cells. M is the number of grid cells to be divided. These are the maximum and minimum values ​​of the first objective at the z-th iteration, respectively; similarly, These are the maximum and minimum values ​​of the second objective at the z-th iteration.

[0164] Iterate through all particles in the outer set and determine their positions in the grid. The grid number for a given particle is:

[0165]

[0166] Int[] is the floor function. K = 1, 2, 3, ..., PK is the number of members in the outer set in the current iteration. and For particles The values ​​of objective functions T1 and T2.

[0167] make This represents the r-th coordinate component of the k-th non-dominant solution. r = 1, 2, ..., R, where R is the number of objectives in the problem to be optimized. In this embodiment, r = 1 and 2.

[0168] Calculate the particle density values ​​in the mesh. Lower density values ​​(based on spatial distance) increase the probability of being selected. Higher density values ​​increase the probability of being deleted. Then, re-mesh the mesh.

[0169]

[0170] Let A be the individual density of any solution in the Pareto optimal solution of the external set; K (K = 1, 2, 3, ..., PK, where PK is the number of members in the outer set); A o It is an external collection of files that is different from A. K Any other non-dominant solution. PCD(A) K A o ) is A K With A o The distance between parallel grids can be calculated using the following formula:

[0171]

[0172] when With A o When all grid coordinate components are the same, it means that the two solutions are completely identical in the grid and share the corresponding M grids. Therefore, the parallel grid distance between them is assigned to 0.5.

[0173] Step 305: Repeat steps 301-304 until the iteration ends. Select the solution with the smallest fitness function value from the external set as the global optimal solution, and output the corresponding task unloading strategy.

[0174] Example 2

[0175] The purpose of this embodiment is to provide a vehicle task unloading system with multi-access edge computing, characterized by including:

[0176] The acquisition module is configured to: acquire service vehicle information and edge server information, and obtain a set of available devices for task unloading based on the service vehicle information and edge server information;

[0177] The initialization module is configured to initialize the task unloading location and unloading ratio of the task vehicle based on the set of available task unloading devices, and obtain the task unloading strategy.

[0178] The optimization module is configured to optimize the task unloading strategy using a multi-objective particle swarm optimization algorithm with the goal of minimizing the total time and energy consumption of task completion, until the optimization iteration conditions are met, and then output the final task unloading strategy.

[0179] Example 3

[0180] The purpose of this embodiment is to provide a computing device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the above-described method.

[0181] Example 4

[0182] The purpose of this embodiment is to provide a computer-readable storage medium.

[0183] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the steps of the above method.

[0184] The steps and methods involved in the apparatuses of Embodiments 2, 3, and 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.

[0185] Those skilled in the art will understand that the modules or steps of the present invention described above can be implemented using general-purpose computer devices. Optionally, they can be implemented using computer-executable program code, thereby allowing them to be stored in a storage device for execution by a computer device, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. The present invention is not limited to any particular combination of hardware and software.

[0186] While the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, this is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art without creative effort based on the technical solutions of the present invention are still within the scope of protection of the present invention.

Claims

1. A method for unloading vehicle tasks in multi-access edge computing, characterized in that, include: Obtain service vehicle information and edge server information, and obtain the set of available devices for task unloading based on the service vehicle information and edge server information; The task unloading strategy is obtained by initializing the task unloading location and unloading ratio of the task vehicle based on the set of available devices for task unloading; With the goal of minimizing the total time and energy consumption of task completion, the task unloading strategy is optimized using a multi-objective particle swarm optimization algorithm until the optimization iteration condition is met, and the final task unloading strategy is output. Design two objective functions with the goal of minimizing the total time and total energy consumption for task completion. T 1 and T 2, T 1 and T 2 is represented as: T 1= minT , T 2= minE, in ，T Indicates the total time required to process the task. ，E The total energy consumption required to process the task is represented by the multi-objective particle swarm optimization algorithm and the task unloading strategy described above. Specifically, the positions and velocities of particles in the multi-objective particle swarm optimization algorithm population are initialized, and the positions of the particles are the task unloading strategies; a fitness function is designed based on the total time and total energy consumption of task completion; the objective function value of each particle is calculated, and the calculated objective function values ​​are compared according to the Pareto dominance relationship, and non-dominated solutions are added to the outer set. Find local and global optima in the Pareto set based on the fitness function, including updating the particle position based on the particle's current solution, local optima, and global optima, recalculating the particle's fitness function and Pareto dominance, and adding non-dominated solutions to the outer set. The solution with the smallest fitness function value is selected from the external set as the target optimal solution, and the corresponding task unloading strategy is output. The particle update process is optimized using a probability transition formula, which is: in, , and Let represent the probabilities of maintaining the previous execution method, the execution method close to the local optimum, and the execution method close to the global optimum, respectively. , and These are the functions related to the total time and total energy consumption of the task, respectively, and are inversely proportional to the total time and total energy consumption. The fitness function is: in, and These are functions of the total time and total energy consumption for task completion, respectively, and are inversely proportional to the total time and total energy consumption. and This is an index representing the total time and total energy consumption required to complete the task.

2. The vehicle task offloading method for multi-access edge computing as described in claim 1, characterized in that, The task unloading location and unloading ratio of the task vehicle are initialized based on the set of available devices for task unloading, specifically as follows: Obtain the distances between the task vehicle and the service vehicle and the edge server, as well as the load status of the service vehicle and the task vehicle; The entropy weight method is used to initialize the unloading location of the task selected by the task vehicle to the service vehicle or edge server, and the set of available devices for task unloading is obtained. The probability of each task on the task vehicle selecting a service vehicle or edge server from the set of available devices for task unloading is calculated using the roulette wheel algorithm. Based on the probability of each task being on a certain service vehicle or edge server, the cumulative probability of task unloading on each service vehicle or edge server in the set of available devices is calculated and sorted, and the unloading ratio of tasks on the task vehicle is initialized.

3. The vehicle task offloading method for multi-access edge computing as described in claim 1, characterized in that, The constraints that must be satisfied to minimize the total time and total energy consumption for task completion include: bandwidth constraints, latency tolerance constraints, processing capacity constraints, energy consumption constraints, task allocation ratio constraints, and task offloading power constraints.

4. The vehicle task offloading method for multi-access edge computing as described in claim 1, characterized in that, When the outer set is full, the adaptive mesh method is used to delete solutions from the outer set and calculate the density of the solutions. If the density of the newly generated solution is large, it is not added to the outer set. If the density value of a certain solution in the outer set is large, the solution with the largest density value in the outer set is deleted and the newly generated solution is added to the outer set.

5. A vehicle task unloading system using multi-access edge computing as described in claim 1, characterized in that, include: The acquisition module is configured to: acquire service vehicle information and edge server information, and obtain a set of available devices for task unloading based on the service vehicle information and edge server information; The initialization module is configured to initialize the task unloading location and unloading ratio of the task vehicle based on the set of available task unloading devices, and obtain the task unloading strategy. The optimization module is configured to optimize the task unloading strategy using a multi-objective particle swarm optimization algorithm with the goal of minimizing the total time and energy consumption of task completion, until the optimization iteration conditions are met, and then output the final task unloading strategy.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the vehicle task offloading method for multi-access edge computing as described in any one of claims 1-4.

7. A processing apparatus, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the vehicle task offloading method for multi-access edge computing as described in any one of claims 1-4.

Images