Task offloading method for improving task execution efficiency and edge node resource utilization

Abstract

The application discloses a task offloading method for improving task execution efficiency and edge node resource utilization, and relates to the technical field of mobile communication, comprising the following steps: constructing a task offloading model under an edge cooperation environment, and acquiring total task execution time; taking minimizing the total task execution time as a target, jointly optimizing resource allocation decisions, user and edge server association decisions, task offloading proportions, and execution location decisions according to an improved teaching optimization algorithm, and acquiring the minimum total task execution time. The application jointly optimizes task offloading location decisions, task offloading proportions, computing resource allocation, and user and server association strategies, comprehensively optimizes multiple key decision variables, effectively reduces the complexity of the task offloading problem, and improves scheduling efficiency.

Description

Technical Field

[0001] This invention belongs to the field of mobile communication technology, specifically relating to a task offloading method that improves task execution efficiency and edge node resource utilization. Background Technology

[0002] With the development of mobile communication technology, smart applications such as smart homes, smart healthcare, facial recognition, and augmented reality / virtual reality are constantly emerging. Although mobile devices have improved in terms of hardware performance, due to resource limitations such as battery capacity and CPU computing power, there is a mismatch between the needs of many emerging applications and the device's performance.

[0003] To address this issue, Mobile Cloud Computing (MCC) technology emerged, aiming to use cloud data centers to provide computing support and decouple computing needs from mobile devices. However, with the development of 5G communication and the Internet of Things (IoT), emerging applications require lower latency and faster data transmission, leading to Mobile Edge Computing (MEC) becoming a popular solution.

[0004] In this context, mobile edge computing pushes computing resources to the network edge to reduce data transmission latency and improve performance. However, due to the wide distribution of users, limited network access range, and heterogeneity of applications, a single edge server struggles to meet the computing needs of multiple user devices, potentially leading to a decline in service quality. Therefore, there is an urgent need to provide a task offloading method to improve service quality. Summary of the Invention

[0005] To address the aforementioned problems in the existing technology, this invention provides a task offloading method that improves task execution efficiency and edge node resource utilization. The technical problem to be solved by this invention is achieved through the following technical solution:

[0006] In a first aspect, the present invention provides a task offloading method to improve task execution efficiency and edge node resource utilization, comprising:

[0007] Construct a task offloading model in an edge collaboration environment and obtain the total task execution time;

[0008] With the goal of minimizing the total task execution time, an improved teaching optimization algorithm is used to jointly optimize resource allocation decisions, user and edge server association decisions, task offloading ratios, and execution location decisions to obtain the minimum total task execution time.

[0009] The beneficial effects of this invention are:

[0010] This invention provides a task offloading method to improve task execution efficiency and edge node resource utilization. In task offloading scheduling, it comprehensively considers the user's location, the server's coverage area, and the communication relationship between servers, making task offloading decisions more intelligent, better adaptable to mobile edge computing environments, reducing task execution latency, and improving user experience. It fully considers the servers associated with task offloading and the allocated wireless communication bandwidth resources, thereby better utilizing edge server resources, achieving higher resource utilization, reducing resource waste, lowering costs, and improving system performance. It can jointly optimize task offloading location decisions, task offloading ratios, computing resource allocation, and user-server association strategies, comprehensively optimizing multiple key decision variables, effectively reducing the complexity of the task offloading problem and improving scheduling efficiency.

[0011] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0012] Figure 1 This is a flowchart of a task unloading method for improving task execution efficiency and edge node resource utilization provided by an embodiment of the present invention;

[0013] Figure 2 This is a schematic diagram of a task offloading model in an edge collaboration environment provided in an embodiment of the present invention;

[0014] Figure 3 This is a schematic diagram of an improved teaching optimization algorithm provided in an embodiment of the present invention. Detailed Implementation

[0015] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0016] The following research exists in related technologies:

[0017] Fan et al.'s approach to source and assistant server partitioning: Fan et al. proposed dividing servers into source and assistant servers; the load on the source server is offloaded to the assistant server via a wired link to reduce system latency and energy consumption. A heuristic optimization algorithm was developed to maximize the overall efficiency in terms of latency and energy consumption. An interior-point method was used to solve the scheduling strategy, which improved system performance.

[0018] Zhou et al.'s two-layer algorithm: Zhou et al. considered the collaboration between edge servers, paying particular attention to fairness among tasks. They proposed a two-layer algorithm where the upper layer uses an Artificial Bee Colony-Particle Swarm Optimization-Genetic Algorithm Task Binary Offloading Algorithm (APGTO) to search for the optimal offloading scheme. The lower layer is a fairness-aware resource allocation algorithm designed to fully utilize server resources and generate a resource allocation scheme that guarantees fairness among tasks. This approach effectively reduces the system's total energy consumption.

[0019] The relevant research has the following shortcomings:

[0020] The locational relationship between users and servers has been neglected: Most existing research focuses on server collaboration while ignoring user location and its relationship with servers. User location significantly impacts the latency and performance of task offloading scheduling, especially in mobile edge computing environments. Current research typically assumes users will always connect to the nearest edge server; however, in densely deployed networks, edge server coverage areas often overlap, meaning users can access multiple edge servers. Therefore, task offloading and scheduling need to comprehensively consider user location, offloading decisions, server associations, and communication bandwidth resource allocation to better utilize edge server resources and improve system performance.

[0021] There is a lack of research on partial task offloading methods: While partial offloading has proven effective for computational tasks involving code partitioning, existing research has not fully integrated the advantages of edge-to-edge horizontal collaboration and partial task offloading. This combination requires joint optimization of task offloading split ratios, execution server locations, task forwarding paths, and computational resource allocation ratios. This increases the complexity of the problem, necessitating the development of new models and algorithms to address it.

[0022] The lack of simulation experiments adapting to different environments is a significant issue: the effectiveness of edge-to-edge collaboration depends on environmental factors such as server topology, user distribution, task configuration parameters, and server performance. However, existing research often overlooks the heterogeneity of these factors and fails to consider different load conditions. This makes it difficult for existing algorithms to adapt to different application scenarios and environmental differences. Therefore, simulation experiments are needed to consider different parameter environments in order to better evaluate the advantages of MEC task scheduling models and strategies based on edge-to-edge collaboration.

[0023] In view of this, this invention provides a task offloading method to improve task execution efficiency and edge node resource utilization. It studies task offloading methods in multi-server, multi-user edge-to-edge collaborative environments, considering the network topology of servers and the locational relationship between users and servers. With the goal of minimizing task execution time, a task offloading model for edge-to-edge collaboration is constructed. Based on the non-convexity of the optimization problem, a fixed variable and sub-problem decomposition approach is adopted, combined with an improved teaching optimization algorithm, to propose an iterative optimization strategy for task offloading.

[0024] Please see Figure 1 and Figure 2 , Figure 1 This is a flowchart of a task offloading method for improving task execution efficiency and edge node resource utilization, provided by an embodiment of the present invention. Figure 2 This is a schematic diagram of a task offloading model in an edge collaboration environment provided by an embodiment of the present invention. The present invention provides a task offloading method for improving task execution efficiency and edge node resource utilization, comprising:

[0025] S101. Construct a task offloading model in an edge collaboration environment and obtain the total task execution time;

[0026] S102. With the goal of minimizing the total execution time of the task, the improved teaching optimization algorithm is used to jointly optimize resource allocation decisions, user and edge server association decisions, task offloading ratio, and execution location decisions to obtain the minimum total execution time of the task.

[0027] In an optional embodiment of the present invention, for task i, the expression for the total execution time T of task i is:

[0028]

[0029] Where i represents the index of the task, and N represents the set of tasks, N = {1, 2, ..., n}. This indicates the local execution time of task i. This indicates the total execution time for task i to be unloaded at the edge.

[0030] In an optional embodiment of the present invention, the expression for the total execution time of task i's edge unloading is:

[0031]

[0032] Where A represents the user's association decision with the edge server, A = {a ij},i=1,2,...,N;j=1,2,...,M,a ij ∈{0,1} represents the association decision variable between task i and edge server j, B represents the execution location decision, B={bik},i=1,2,...,N;j=1,2,...,M,b ik ∈{0,1} represents the execution location decision variable between task i and edge server k, δi represents the task offloading ratio of task i, and 0≤δ i ≤1, F represents resource allocation decision, This represents the computing resources allocated by edge server k to task i, M represents the set of edge servers, M = {1, 2, ..., m}, j represents the index of the edge server, and a ij The decision variable representing the association between task i and edge server j. This indicates the upload time of the data from task i to edge server j. u i Indicates the size of the task input data, r ij This represents the uplink wireless link transmission rate of the task data, and k represents the index of the edge server. b represents the forwarding time of task i between edge server j and edge server k. ik This represents the execution location decision variable for task i and edge server k. This indicates the execution time of task i on edge server k. s i This indicates the computational complexity of task i.

[0033] In an optional embodiment of the present invention, the expression for the local execution time of task i is:

[0034]

[0035] Among them, s i This indicates the amount of computation required for task i to be processed by the CPU. δ represents the local computing resources of task i. i This indicates the percentage of tasks that are unloaded from task i.

[0036] In an optional embodiment of the present invention, constructing a task offloading model in an edge collaboration environment includes:

[0037] Specifically, in this implementation, the set of edge servers is set as M = {1, 2, ..., m}. Each edge server can provide task offloading and execution services to mobile devices within its coverage area. The topology of the edge server network is represented as G = {V, E}, where V = {v j |j∈M} represents the edge server node and its location, E={e jk |j,k∈M} represents the wired connection between edge server j and edge server k, e jk Indicates the bandwidth of the wired connection, e jk=0 indicates that there is no connection between the two edge servers. Assuming the wired connection between the edge servers is bidirectional, when j≠k, e jk =e kj When j = k, e jk =+∞.

[0038] Let the task set be N = {1, 2, ..., n}, randomly distributed within the coverage area of ​​the edge server, u i This indicates the size of the input data for task i, in bits (s). i This indicates the amount of computation processed by the CPU for task i, measured in Gigacycles. Task i can be processed locally or partially offloaded to an edge server for execution.

[0039] Set δ i (0≤δ i ≤1) represents the proportion of tasks offloaded to edge servers for execution, where tasks are executed in parallel on both the local and edge servers, where (1-δ) i )s i Indicates execution locally, δ i s i This indicates that the execution will be offloaded to an edge server that can be directly accessed, or forwarded to an edge server outside the local access range through cooperation between edge servers.

[0040] Let A represent the set of decision variables associated with a user and an edge server, A = {a ij},i=1,2,...,N;j=1,2,...,M,a ij (i∈N,j∈M) represents the decision variables for the association between task i and edge server j, where N={1,2,…,n}, M={1,2,…,m}, a ij =1 indicates that task i uses the wireless channel of edge server j to upload data. This indicates that the task is executed locally; let B represent the set of decision variables for task execution on the edge server, B = {b ik},i=1,2,...,N;j=1,2,...,M,b ij (i∈N,j∈M) represents the decision variable for the final execution location of the task, b ij =1 indicates that the task is executed on edge server j; otherwise, b ij =0; Set Θ j ={i|i∈N,j∈M,a ij =1} represents the set of tasks that occupy the wireless channel j of the edge server for data transmission; the load on the edge server increases as the amount of offloaded tasks increases, and Φ is set. j ={i|i∈N,j∈M,b ij=1} represents the set of tasks that are ultimately executed on edge server j.

[0041] In an optional embodiment of the present invention, the expression for the local execution time of task i is:

[0042]

[0043] Among them, s i This indicates the amount of computation required for task i to be processed by the CPU. Represents the local computing resources of task i, (1-δ i )s i This represents the size of the task executed locally, 0 ≤ δ i ≤1.

[0044] In an optional embodiment of the present invention, the process of obtaining the total execution time of task edge unloading for task i includes:

[0045] The execution time for offloading a portion of the task to the edge server consists of three parts, including: task data upload time. Task data forwarding time between edge servers during collaborative execution and the edge computing time of the task Assuming that users connected to the same edge server do not interfere with each other during data transmission, and the transmission channel remains unchanged within the same offload execution cycle, orthogonal spectrum is allocated to users associated with the same edge server, r ij The wireless link transmission rate representing the uplink data transfer rate of the task indicates the data upload time of the task. It can be defined as:

[0046]

[0047] Based on the task offloading model in an edge collaboration environment, the shortest path Path(j,k) between edge servers j and k is defined. Combining the user-edge server association decision variables and the task execution location decision variables, the task forwarding path is associated with the shortest path between edge servers, and the path Ω forwarded by task i is defined. i for:

[0048] Ω i =a ij ×b ik ×Path(j,k) (3);

[0049] Assuming the bandwidth between edge servers remains constant during each decision-making period, the transmission rate between edge servers is high, the number of offloaded requests is small, and time conflicts caused by concurrent transmission tasks are negligible, and since collaboration occurs between adjacent edge servers, propagation latency is negligible, the task forwarding transmission time between edge servers is:

[0050]

[0051] Assume δ i The proportion of tasks is ultimately offloaded to edge server j for execution, and edge server j allocates computing resources for the tasks. This indicates the execution time of task i on the edge server. for:

[0052]

[0053] Combining formulas (2), (4), and (5), the total time for task unloading to edge execution is expressed as:

[0054]

[0055] The task is divided into a local execution portion and an edge server execution portion, which are executed in parallel. Local execution utilizes CPU resources, while data uplink offloading utilizes I / O resources. Communication and computation do not interfere with each other. The execution time T of task i is... i Maximum execution time for local execution and edge unloading:

[0056]

[0057] Define the execution time of all user tasks as a function T, with the following expression:

[0058]

[0059] The optimization problem P1, which aims to minimize the execution time of all user tasks, jointly optimizes the user-edge server association decision A, the task execution location selection decision B, the task offloading ratio decision δ, and the edge server resource allocation F, is defined as follows:

[0060]

[0061]

[0062]

[0063]

[0064]

[0065]

[0066]

[0067]

[0068] In an optional embodiment of the present invention, with the goal of minimizing the total execution time of the task, the minimum total execution time of the task is obtained by jointly optimizing resource allocation decisions, user and edge server association decisions, task offloading ratios, and execution location decisions based on an improved teaching optimization algorithm, including:

[0069] Initialize the set of students in the class P, let p = 1. Until all students in the class student set P are initialized; wherein, the class student set P = {z p}, p = 1, 2, ..., K, where each student is represented as i = 1, 2, ..., N The location where task i is executed can be decoded into execution location decision B. i , The unloading ratio of task i can be decoded as the task unloading ratio δ of task i. i

[0070] Initialize the number of iterations g = 1; where the upper limit of iterations is set to G;

[0071] Construct a fitness function, and based on the fitness function, obtain the student z with the best current fitness. best The fitness function is expressed as follows:

[0072]

[0073] The student z with the best current fitness best Updated to teacher, teacher=z best At the same time, update the average knowledge level of students in each subject (mean). i Its expression is:

[0074]

[0075] The new knowledge generated by each student learning from the teacher is z. newp Update the student z with the best current fitness. best ;

[0076] Each student chooses to learn from another student, updating the current best-fit student's knowledge (z). best ;

[0077] Until g = G, we obtain the optimal edge server computing resource allocation F, the decision A for associating with the edge server, the task offloading ratio δ, and the execution location decision B.

[0078] In an optional embodiment of the present invention, the new knowledge is z. newp The expression is:

[0079]

[0080] in, This represents a random number between (0,1). TF = round(1 + rand(0,1)) is a random number that takes the value 1 or 2.

[0081] In an optional embodiment of the present invention, a fitness function is constructed, and the student with the best current fitness is obtained based on the fitness function, including:

[0082] Let p = 1;

[0083] Based on the computing resource allocation strategy and the user-edge server association strategy, the edge server computing resource allocation F is obtained. p And user-edge server association decision A p Initialize A best =A p F best =F p ;

[0084] Let p = p + 1;

[0085] Based on the computing resource allocation strategy and the user-edge server association strategy, the edge server computing resource allocation F is obtained. p And user-edge server association decision A p ;

[0086] If f(z) best A best ,F best ) <f(z p A p ,F p Let z best =z p A best =A p F best =F p ;

[0087] If p < K, continue executing p = p + 1 until the iteration ends, and output the student z with the best fitness at the current time. best .

[0088] In an optional embodiment of the invention, the new knowledge that each student learns from the teacher is z. newp Update the student z with the best current fitness. best ,include:

[0089] Let p = 1;

[0090] q = round(rand(0,K));

[0091] Calculate the new knowledge z newp The calculation expression is:

[0092]

[0093] in,

[0094] Based on the computing resource allocation strategy and the user-edge server association strategy, the edge server computing resource allocation F is obtained. newp And user-edge server association decision A newp ;

[0095] If f(z) newp A newp ,F newp ) <f(z p A p ,F p Let z p =z newp A p =A newp F p =F newp ;

[0096] p = p + 1;

[0097] Until p = K, output the updated student z with the best current fitness. best .

[0098] In an optional embodiment of the present invention, resource allocation decisions are obtained through a resource allocation strategy, the expression of which is:

[0099]

[0100] Where, δ i This indicates the percentage of task unloading for task i. s represents the local computing resources of task i. i This represents the computational load of task i processed by the CPU, M represents the set of edge servers, j represents the index of the edge server, and r represents the value of the edge server. ijDenote the uplink wireless link transmission rate of task data as u i Denote the size of task input data Denote the forwarding time of task i between edge server j and edge server k, and its expression is:

[0101]

[0102] In an optional embodiment of the present invention, the user-edge server association strategy includes:

[0103] Initialize the iteration number g = 1, traverse all users, associate each user to the edge server with the largest channel gain, and obtain the initial association strategy A

[0104] Calculate the communication transmission time TR(A) of the current system. The expression of the communication time TR(A) is:

[0105]

[0106] Wherein Denote the data upload time of task i Denote the transmission time of task i between edge servers

[0107] Initialize i = 1

[0108] Select another server j', set a ij = 0, a ij' = 1, and obtain the association strategy A'

[0109] If TR(A') < TR(A), then update the association strategy, and let A = A' "

[0110] i = i + 1

[0111] If i ≤ N, then return to continue selecting the server

[0112] Until g = G, obtain the optimal association decision A for the edge server, where G represents the iteration upper limit

[0113] In the above embodiment, the computing resource allocation strategy

[0114] The above-described problem P1 is a large-scale mixed integer non-linear programming problem (Mixed Integer Non-Linear Programming, MINLP). Finding the optimal solution usually requires exponential time complexity and cannot obtain the optimal solution through a polynomial time algorithm. The problem will be solved by subproblem decomposition combined with an improved metaheuristic algorithm. The algorithm hierarchical structure is as Figure 3 shown Figure 3This is a schematic diagram of an improved teaching optimization algorithm provided in an embodiment of the present invention. Observing the optimization problem P1, it can be seen that constraints (9f) and (9g) are decoupled from other constraints. By first fixing the user-edge server association decision A, the task execution location decision B, and the task offloading ratio δ, the optimization problem P1 is transformed into:

[0115]

[0116]

[0117]

[0118]

[0119]

[0120]

[0121] Expanding formula (10), we get:

[0122]

[0123] For tasks whose local execution time is greater than their edge execution time, reduce the allocation of computing resources to make the edge execution time equal to the local execution time, and return the redundant computing resources to the resource pool of the edge server for reallocation. Assuming that the tasks initially allocated to edge server j all satisfy the condition that the local execution time is less than the edge execution time, use the Lagrange multiplier method to solve for the current optimal computing resource allocation, determine the relationship between the local execution time and the edge execution time, allocate the redundant resources to other tasks, and repeat the above process until there are no redundant resources, and obtain the optimal computing resource allocation under the current execution location decision, user and server association decision, and task offloading ratio. Redefine the optimization problem of (10) as P1.1, and its expression is:

[0124]

[0125]

[0126]

[0127] Observation shows that constraints (12a) and (12b) are convex constraints, and the objective function can be defined as:

[0128]

[0129] Taking the second derivative of χ(F), we get:

[0130]

[0131]

[0132] From formulas (14) and (15), we can see that the Hessian matrix of function (13) is a positive definite matrix, and χ(F) is a convex function. Combining the constraints (12a) and (12b), we define the Lagrangian function, whose expression is:

[0133]

[0134] At the optimal solution, the KKT (Karush-Kuhn-Tucker) conditions should be satisfied to obtain the amount of computing resources allocated to each task by the edge server. Its expression is:

[0135]

[0136] Therefore, resources are reallocated for tasks whose local execution time is longer than their edge execution time. Specifically, the operation is as follows: Get the updated amount of resources that should be allocated. The redundant resources are then returned to the edge server computing resource pool for reallocation. The updated allocated resource amount... The expression is:

[0137]

[0138] In the above embodiments, the user-edge server association strategy is as follows:

[0139] This invention considers the scenario of densely deployed edge servers. Based on the geographical relationship between users and edge servers, users in overlapping coverage areas can access multiple edge servers, while users in non-overlapping areas can only access one edge server; the association decision remains unchanged. Therefore, when formulating the user-edge server association decision, only users who can access multiple edge servers need to be considered, and the associated edge server is determined based on the task unloading time and forwarding time. This invention employs a greedy association strategy to minimize the sum of the global task upload time and forwarding time, obtaining a suboptimal solution for the user-edge server association strategy. The global task communication time is defined as TR(A), and its expression is:

[0140]

[0141] In the global task communication time TR(A), the task forwarding time is associated with the shortest path Path(j,k) between edge servers. Assuming the wired link bandwidth between edge servers in each time slot remains constant and data transmission concurrency is not considered, the problem is transformed into finding the shortest path between two points. The shortest path problem can be implemented using the Floyd algorithm, executed once during the algorithm's initialization phase. This invention proposes a greedy search algorithm based on minimizing the global communication time for user-edge server association decisions. The algorithm's pseudocode is shown in Table 1 below. Table 1 represents the greedy search algorithm for user-edge server association decisions.

[0142]

[0143]

[0144] In Algorithm 1, steps 1 to 6 iterate through all users and associate each user with the nearest edge server to obtain the initial association strategy A; step 7 calculates the total communication transmission time TR(A) of the current system; steps 8 to 16 find replaceable edge servers j′ for user i in user set U. If the communication transmission time of the updated decision is better, A is updated to A′ until the user association decision with the edge server no longer changes; step 17 returns the optimal set of user association decisions with the edge server.

[0145] In the above embodiments, the algorithm for searching the task unloading execution location and unloading ratio is as follows:

[0146] Encoding and Decoding: Assume the set of students in the class is P, and each student's knowledge operator... Defined as a two-dimensional variable, corresponding to the unloading execution location decision B and the unloading ratio δ. According to constraints (9b) and (9d), the task has only one execution location, so the execution location decision variable is encoded as a one-dimensional variable. Uninstall ratio when When j=0, it means that the task is executed entirely locally.

[0147] Fitness function design: In the task unloading scenario, decode to obtain the current unloading execution position decision and the task unloading ratio δ p The user-edge server association decision A is obtained by calculating the resource allocation strategy and the user-edge server association strategy. p And optimal computational resource allocation F p The task execution time is then calculated. The fitness function is expressed as follows:

[0148]

[0149] Update teacher and class average knowledge: based on fitness function f(z) p Ap ,F p Find the best student's score in the current class and update that student as the teacher: teacher = argminf(z p A p ,F p At the same time, calculate the average knowledge level of all students in each subject.

[0150] Teacher-led instruction stage: The new knowledge acquired by each student from the teacher is z. newp The generation of new knowledge is defined as:

[0151]

[0152] in, Let TF be a random number between (0,1), and let TF = round(1 + rand(0,1)) randomly select 1 or 2. Calculate the fitness function value f(z) according to formula (21). newp A newp ,F newp When f(z) newp A newp ,F newp ) <f(z p A p ,F p When the original solution is replaced by the new solution, the original solution will replace the new solution; otherwise, the original solution will remain unchanged.

[0153] Student peer learning phase: In this phase, student p1 randomly selects another student p2 to learn from, where p1 ≠ p2. The learning direction is determined by the fitness of students p1 and p2. The strategy function for student p1 to generate a new solution newp1 during the peer learning phase is defined as:

[0154]

[0155] in, Let represent a random number between (0,1). The new solution replacement strategy is the same as in the teacher teaching phase; after each student chooses to learn from another student, the mutual learning phase ends. This completes one round of the TLBO algorithm iteration process, which continues iteratively until convergence. The task scheduling algorithm based on the improved teaching optimization algorithm is described in Table 2. Table 2 shows the task scheduling algorithm based on the improved teaching optimization algorithm.

[0156]

[0157]

[0158] Step 1 first uses the Floyd algorithm to find the shortest transmission path between edge servers, initializing the number of iterations to 1. Steps 3-5 store the best individual knowledge from the previous generation, update the current global best solution to the teacher, and update the average value of each subject in the current class. Steps 6-8: Each student learns knowledge from the teacher, decodes the new solution, and replaces the old solution if the new solution is better. Steps 9-14: Each student randomly learns knowledge from another student, obtains a new solution, decodes it, and replaces the old solution if the new solution is better. Afterward, the teacher and class average knowledge are updated, and the current iteration ends. If the maximum number of iterations has been reached, the current optimal task scheduling decision is returned. Otherwise, return to Step 2 and start a new round of search operations. Step 19 returns the optimal scheduling decision.

[0159] In summary, the task offloading method provided by this invention, which improves task execution efficiency and edge node resource utilization, has the following beneficial effects:

[0160] 1. Considering User Location and Server Coverage: Zhou et al. only considered the collaboration process between servers, neglecting the user's task unloading process. Since edge server computing, communication resources, and coverage are limited, in mobile edge computing environments, the user-server association strategy based on user location and the number of users within server coverage significantly impacts task unloading scheduling latency. This invention comprehensively considers user location, server coverage, and communication relationships between servers in task unloading scheduling. This makes task unloading decisions more intelligent, better adaptable to mobile edge computing environments, reduces task execution latency, and improves user experience.

[0161] 2. Improved Resource Utilization: Zhou et al. assumed that users would always connect to the nearest edge server. However, in densely deployed network environments, the coverage areas of adjacent edge servers typically overlap to avoid blank areas not covered by any edge server. This means that users in overlapping areas can access any edge server covering them. This invention fully considers the servers associated with task offloading and the allocated wireless communication bandwidth resources, thereby better utilizing edge server resources. This achieves higher resource utilization, reduces resource waste, lowers costs, and improves system performance.

[0162] 3. Comprehensive Optimization: This invention proposes a task scheduling algorithm based on an improved teaching optimization algorithm, which can jointly optimize task unloading location decisions, task unloading ratios, computing resource allocation, and user-server association strategies. This comprehensive optimization considers multiple key decision variables, effectively reducing the complexity of the task unloading problem and improving scheduling efficiency.

[0163] 4. Adaptability to various environments: This invention has implemented various experimental simulation environments to verify the effectiveness of the model and algorithm. This means that the technical solution of this invention is not only applicable to a single environment, but can also adapt to different application scenarios and environmental conditions, thus improving its versatility.

[0164] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations are intended to cover non-exclusive inclusion, such that an article or device comprising a list of elements includes not only those elements but also other elements not expressly listed. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the article or device comprising said element. Terms such as "connected" or "linked" are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect. The orientations or positional relationships indicated by terms such as "upper," "lower," "left," and "right" are based on the orientations or positional relationships shown in the accompanying drawings and are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the invention.

[0165] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0166] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A task offloading method to improve task execution efficiency and edge node resource utilization, characterized in that, include: Construct a task offloading model in an edge collaboration environment and obtain the total task execution time; With the goal of minimizing the total execution time of the task, the improved teaching optimization algorithm is used to jointly optimize resource allocation decisions, user and edge server association decisions, task offloading ratio, and execution location decisions to obtain the minimum total execution time of the task. Among them, for the task The total execution time of the task The expression is: ； in, Indicates the index of the task. Represents a set of tasks. Indicates task Local execution time, Indicates task Total execution time for edge unloading; The task The expression for the total execution time of edge unloading is: ； in, This indicates the decision to associate users with edge servers. Indicates the execution of location decisions. Indicates task Task uninstallation ratio Indicates resource allocation decisions, Represents a set of edge servers. Indicates the index of the edge server. Indicates task With edge servers The associated decision variables, Indicates task Upload to edge server Data upload time, Indicates the index of the edge server. Indicates task On the edge server With edge servers The forwarding time between them Indicates task With edge servers The execution location decision variable, Indicates task On the edge server Execution time; The step of minimizing the total task execution time, based on an improved teaching optimization algorithm, jointly optimizes resource allocation decisions, user-edge server association decisions, task offloading ratios, and execution location decisions to obtain the minimum total task execution time, includes: Initialize the class student set ,make , , until all students in the class have gathered. All students in the class are initialized; wherein, the set of students in the class is... Each student is represented as , Indicates task The execution location can be decoded into an execution location decision. , Indicates task The uninstallation ratio can be decoded into task Task uninstallation ratio , Index representing students; Initialize the number of iterations The upper limit of the iteration is set as follows: ; Construct a fitness function, and based on the fitness function, obtain the student with the best current fitness. The fitness function is expressed as follows: , Indicates the allocation of computing resources on edge servers. This indicates the decision-making process for associating users with edge servers; The student with the best current fitness Updated to teacher, At the same time, update students' average knowledge level in each subject. Its expression is: ； Each student learns new knowledge from the teacher. Update the student with the best current fitness. ; Each student chooses to learn from another student, updating the list of students with the best current fitness level. ; Until To obtain the optimal allocation of edge server computing resources Used for decision-making related to edge servers Task uninstallation ratio and execution location decision ; The resource allocation decision is obtained through a resource allocation strategy, wherein the resource allocation strategy... The expression is: ； in, Indicates task Task uninstallation ratio Indicates task Local computing resources Indicates task The amount of computation processed by the CPU Represents a set of edge servers. Indicates the index of the edge server. This indicates the uplink wireless link transmission rate of the task data. Indicates the size of the task input data; The user-edge server association strategy includes: Initialize the number of iterations Iterate through all users and associate each user with the edge server that has the highest channel gain to obtain the initial association decision. ; Calculate the current system's communication transmission time The communication transmission time The expression is: ； in, Indicates task Data upload time, Indicates task Transmission time between edge servers; initialization ; Choose another server ,set up , To obtain related decisions ; if Then update the associated decision, making ; ； if If so, return to continue selecting a server; Until To obtain the optimal decision for associating with the edge server. ,in, This indicates the upper limit of iterations.

2. The task offloading method for improving task execution efficiency and edge node resource utilization according to claim 1, characterized in that, The task The expression for local execution time is: ； in, Indicates task The amount of computation processed by the CPU Indicates task Local computing resources Indicates task The percentage of tasks that are unloaded.

3. The task offloading method for improving task execution efficiency and edge node resource utilization according to claim 1, characterized in that, The new knowledge The expression is: ； in, Represents a random number between (0,1). The random number is either 1 or 2. It refers to a teacher.

4. The task offloading method for improving task execution efficiency and edge node resource utilization according to claim 1, characterized in that, The fitness function is constructed, and the student with the best current fitness is obtained based on the fitness function. ,include: make ; Based on the computing resource allocation strategy and the user-edge server association strategy, the edge server computing resource allocation is obtained. and user-edge server association decisions ,initialization , ; make ; Based on the computing resource allocation strategy and the user-edge server association strategy, the edge server computing resource allocation is obtained. and user-edge server association decisions ; if Then let , , ; if Then continue execution. Continue this process until the iteration ends, then output the student with the best current fitness. .

5. The task offloading method for improving task execution efficiency and edge node resource utilization according to claim 1, characterized in that, The new knowledge that each student learns from the teacher is Update the student with the best current fitness. ,include: make ; ； Calculate the new knowledge The calculation expression is: ； in, ; Based on the computing resource allocation strategy and the user-edge server association strategy, the edge server computing resource allocation is obtained. and user-edge server association decisions ; if Then let , , ; ； Until Output the updated student with the best current fitness. .

Images