Joint resource management and task offloading method and system for multi-service edge computing

Abstract

The application discloses a joint resource management and task offloading method and system of multi-service edge computing, and relates to the technical field of resource allocation. The method comprises the following steps: acquiring tasks uploaded by multiple users; establishing a multi-objective function with the maximum utility of all base stations and the maximum utility of each user as the target; decomposing the multi-objective function into a service provider resource allocation problem and a user and base station matching problem; solving the user and base station matching problem by using a user and base station matching algorithm based on coalition game theory to obtain a user and base station matching scheme; decomposing the service provider resource allocation problem into a sum of multiple independent sub-problems, solving the sum of multiple independent sub-problems by using a Lagrange dichotomy difference algorithm based on the user and base station matching scheme, and obtaining a task offloading scheme. In the multi-service scenario, the application combines the algorithms of resource allocation, task scheduling and coalition game task offloading, reduces the time delay and cost of users, and improves the benefits of base stations.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of resource allocation, and particularly relates to a joint resource management and task offloading method and system for multi-service edge computing. BACKGROUND

[0002] With the rapid development of artificial intelligence technology, the demand for computing power and performance of hardware by artificial intelligence applications has also rapidly grown. For example, the application of artificial intelligence in virtual reality, augmented reality, human-computer interaction, intelligent manufacturing, autonomous driving and other fields requires fast real-time computing power interaction. Generally, these applications require high-quality services, such as low latency, high access rate, high bandwidth and low energy consumption. In order to meet the growing demand for high-quality applications, mobile edge computing has become a new computing model, which establishes a base station close to the terminal application and extends the centralized cloud computing capability to the edge.

[0003] The base station deploys an edge server to meet the computing needs of the user terminal, and can access the cloud to meet higher computing and storage needs, forming an end-edge-cloud layered computing offloading architecture. Although computing offloading is an important part of many studies in mobile edge computing, it is often overlooked that the heterogeneity and diversity of mobile services on the edge server, and how to achieve multi-service edge resource management and multi-user task offloading strategies.

[0004] Due to limited computing and storage resources, an edge server in a base station only has a subset of services in the service set, rather than all services. The edge server can only handle service type requests hosted on it, where the type and format of data required by each service and the delivery method can also be different. In addition, when the edge server needs to provide services for a large number of users at the same time, continuous resource provisioning and computing offloading operations may occur, directly reducing the data rate of application service users and making it difficult to ensure the quality of service for delay-sensitive users. In addition, users need to pay for the service when enjoying the service, which is related to the quality of service provided by the base station. The existing technology only considers the user side benefit or the edge benefit, but in the multi-service edge computing scenario, both parties need to consider the benefits. SUMMARY

[0005] The embodiments of the present application provide a joint resource management and task offloading method and system for multi-service edge computing, to solve the problem that the user side benefit and the edge provider benefit are not considered at the same time in the prior art.

[0006] In one aspect, the embodiments of the present application provide a joint resource management and task offloading method for multi-service edge computing, comprising:

[0007] obtaining tasks uploaded by a plurality of users;

[0008] A multi-objective function is established with the maximum utility of all base stations and the maximum utility of each user as the target;

[0009] The multi-objective function is converted into a multi-objective optimization problem, and the multi-objective optimization problem is decomposed into a service provider resource allocation problem SPRA and a user and base station matching problem UBM;

[0010] A user and base station matching CG-UBM algorithm based on coalition game theory is used to solve the user and base station matching problem UBM, and a user and base station matching scheme is obtained, which maximizes the utility of each user;

[0011] The service provider resource allocation problem SPRA is decomposed into a sum of multiple independent sub-problems SSPRA, and the sum of multiple independent sub-problems SSPRA is solved based on the user and base station matching scheme by using a Lagrange interval difference LII algorithm to obtain a task offloading scheme;

[0012] According to the offloading scheme, the tasks are offloaded to the corresponding base stations or cloud servers.

[0013] On the other hand, the embodiments of the present application also provide a joint resource management and task offloading system for multi-service edge computing, comprising:

[0014] A task acquisition module is configured to acquire tasks uploaded by multiple users;

[0015] A target function establishment module is configured to establish a multi-objective function with the maximum utility of all base stations and the maximum utility of each user as the target;

[0016] A function conversion module is configured to convert the multi-objective function into a multi-objective optimization problem, and decompose the multi-objective optimization problem into a service provider resource allocation problem SPRA and a user and base station matching problem UBM;

[0017] A scheme matching module is configured to use a user and base station matching CG-UBM algorithm based on coalition game theory to solve the user and base station matching problem UBM, and obtain a user and base station matching scheme, which maximizes the utility of each user;

[0018] A scheme solving module is configured to decompose the service provider resource allocation problem SPRA into a sum of multiple independent sub-problems SSPRA, and solve the sum of multiple independent sub-problems SSPRA based on the user and base station matching scheme by using a Lagrange interval difference LII algorithm to obtain a task offloading scheme;

[0019] A scheme execution module is configured to offload the tasks to the corresponding base stations or cloud servers according to the offloading scheme.

[0020] The joint resource management and task offloading method and system for multi-service edge computing have the following advantages:

[0021] 1. A server benefit-aware resource allocation and task scheduling model is established in a multi-service edge scenario. Considering the limited computing, storage and network resources of the edge server, a subset of service types on the edge server, and multi-service interference, the application allocates computing resources and communication resources under the NOMA (Non-Orthogonal Multiple Access) model to each user and schedules tasks to the cloud server to improve the efficiency of the edge server and achieve efficient use of resources.

[0022] 2. The problem of high delay or high cost caused by the user's nearby or random selection of edge nodes is solved. On the basis of resource allocation and task scheduling, the application establishes a user-base station joint matching model, taking the low delay and low cost of each user as the starting point, freely selecting the group with the maximum benefit of itself according to the rules, providing user fairness selection and reducing the delay and cost of each user.

[0023] 3. The Lagrange solution method and the greedy pruning algorithm are used to solve the resource allocation and task scheduling problem of the edge server, a user-base station matching algorithm based on cooperative game is established, and its convergence is proved. The simulation results show that the algorithm obtains the equilibrium state solution, and its convergence speed is fast, and the performance is better than other algorithms. BRIEF DESCRIPTION OF DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed in the embodiment or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0025] Figure 1 The flowchart of the joint resource management and task offloading method of multi-service edge computing provided by the embodiment of the present application;

[0026] Figure 2 The composition schematic diagram of the joint resource management and task offloading system of multi-service edge computing provided by the embodiment of the present application;

[0027] Figure 3 The schematic diagram of the cost change of each user provided by the embodiment of the present application;

[0028] Figure 4 The schematic diagram of the influence of the number of users on the calculation overhead and the number of iterations provided by the embodiment of the present application;

[0029] Figure 5 The schematic diagram of the comparison of several algorithms about the average delay of users provided by the embodiment of the present application;

[0030] Figure 6 Several algorithms provided in the embodiments of the present application are illustrated in the following base station utility comparison diagram. DETAILED DESCRIPTION

[0031] The technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0032] Figure 1 The flowchart of the joint resource management and task offloading method of multi-service edge computing provided in the embodiments of the present application. The embodiments of the present application provide a joint resource management and task offloading method of multi-service edge computing, which comprises:

[0033] S100, acquiring tasks uploaded by a plurality of users.

[0034] Exemplarily, the tasks uploaded by the users can be distributed to a plurality of base stations, each of which has an edge server to undertake edge computing tasks. Taking the channel of the nth base station as an example, the signal-to-noise ratio of user m in the wireless channel of base station n is represented as S m，n In addition, for each base station n, the channel gain satisfies the condition This means that the user is the ith weakest instantaneous channel in the user set of the nth base station. In addition, the user set selected by the nth base station is represented as an index variable , and the ith order in the user set is user m. Therefore, the transmission rate of user m through the wireless channel of base station n is:

[0035]

[0036] where B n represents the channel bandwidth of the nth base station, p m represents the transmission power range of the mth user, G m,n represents the channel gain between the mth user and the nth base station, η0 represents the background noise, I m,n is the interference between the mth user and the nth base station, and:

[0037]

[0038] where m represents the number of users, the symbol ∈ represents belonging to, q is other users in addition to m, and l is the sequence number of the sequence from i to .

[0039] In addition, the task of user m can be further offloaded to the cloud service, and the transmission rate of the nth base station offloading the task of the mth user to the cloud server is and the transmission rate satisfies:

[0040]

[0041] denotes the set of users whose tasks are offloaded to the cloud server by the nth base station, denotes the transmission rate limit of the nth base station offloading the task to the cloud server.

[0042] Then, according to the edge server's processing speed of the tasks performed thereon and the subset of user tasks offloaded to the edge node offloaded to the cloud server, the edge server's income and the required delay and cost of each user to complete the service are obtained.

[0043] The cost of the user includes the processing task fee paid to the service provider, the service delay, and the energy consumption of the user itself. If the user m is a single entity and the service provided is not used for other purposes, the utility of the user m can be represented as:

[0044] u m =-(d m，k Pr(D m )+D m +E m ) (4)

[0045] where d m，k denotes the size of the task data uploaded by the mth user belonging to the kth application service, Pr(D m ) denotes the payment of the service delay of the mth user, D m denotes the service delay of the mth user, and Em denotes the energy consumption of the mth user. Generally, the smaller D m , the higher the payment of the user m. It can be assumed that the Pr function is a decreasing function of the variable D m , and Pr(D m ) = e(D m ) b , where e and b are related parameters in the unit price function Pr(D m ), and b < 0.

[0046] If the task of user m is processed on base station n, the delay of the nth base station for the task uploaded by the mth user belonging to the kth application service is:

[0047]

[0048] where r m，ndenotes the transmission rate between the nth base station and the mth user, ρ k denotes the workload requirement of the kth application service, denotes the processing rate of the nth base station for the tasks uploaded by the mth user belonging to the kth application service, and are the end-to-base station transmission delay and the computation delay performed at the base station, respectively.

[0049] If the tasks of the mth user are further offloaded to the cloud server for execution by the nth base station, the delay of the nth base station for offloading the tasks uploaded by the mth user belonging to the kth application service to the cloud server is:

[0050]

[0051] denotes the transmission rate of the tasks uploaded by the mth user belonging to the kth application service offloaded to the cloud server, denotes the processing rate of the nth base station for offloading the tasks uploaded by the mth user belonging to the kth application service to the cloud server. The base station obtains revenue by processing the tasks of the users, and the revenue function is the sum of the payments of the users. In addition to the revenue of the base station, the cost generated by the base station for processing the tasks of the users is also considered. The cost consists of two parts: the energy consumption generated by executing the tasks of the users and the payment fee for the cloud server to process part of the tasks. According to the energy consumption model, the energy consumption of the nth base station for processing the tasks uploaded by the mth user belonging to the kth application service is:

[0052]

[0053] where κ n is the energy coefficient determined by the structural characteristics of the edge server of the nth base station. The total energy consumption of the nth base station is:

[0054]

[0055] The cost of processing the tasks τ m，k uploaded by the mth user on the cloud server is where a is the computation price per unit load, denotes the set of users in the nth base station whose tasks are executed on the base station. The total cost of processing the tasks offloaded from the nth base station to the cloud server is The utility of all base stations is:

[0056]

[0057] denotes the set of base stations, U n denotes the revenue of the nth base station.

[0058] S110, a multi-objective function is established with the maximum utility of all base stations and the maximum utility of each user as the target.

[0059] Exemplarily, from the perspective of multiple service providers, the most important thing is how to configure multiple services on multiple base stations and how to select the tasks unloaded to the base stations for further execution on the cloud server to maximize their utility. From the perspective of each user, the most important thing is how to select the base station to upload their own tasks to optimize their own utility. The multi-objective function is represented as:

[0060]

[0061] max u m = -d m,k Pr(D m )-D m -E m (11)

[0062]

[0063] formula (1)-(9)

[0064] where max represents the maximum value, n represents the number of base stations, s.t. represents the constraint condition, a m represents the base station selected by the mth user, the symbol represents any, k represents the number of application service types, represents the set of application service types, S n，k represents the configuration indicator variable of base station n for the kth application service, C n，k represents the computing resource of the kth application service in the nth base station, || represents the case under the following conditional restriction. Constraint (12) indicates that user m needs to select one base station from the base station setto unload its tasks. Constraint (13) indicates the computing resource limit of base station n. Constraint (14) indicates the computing resource limit of the kth application in base station n. Formula (1)-(9) indicates the allocation of multiple business resources, the computing resource of each base station, and the transmission rate of the unloaded tasks to each base station or cloud server.

[0065] S120, the multi-objective function is converted into a multi-objective optimization problem, and the multi-objective optimization problem is decomposed into a service provider resource allocation problem SPRA and a user and base station matching problem UBM.

[0066] Exemplarily, the multi-objective function is essentially a multi-objective optimization problem, which is divided into two sub-problems: 1. service provider resource allocation problem SPRA, which includes multi-base station service allocation, base station / cloud computing resource allocation, and transmission rate allocation; 2. user and base station matching problem UBM. The specific steps are as follows: each user matches the optimal base station according to his own objective function; then, in order to maximize the utility of the service provider, the supplier dynamically adjusts the resource allocation, computing allocation, and further task allocation unloaded to the cloud server according to the user's matching selection and task demand, further, the user re-adjusts the selection of the base station, and iterates in turn, and finally reaches a balance.

[0067] In S130, a user and base station matching CG-UBM algorithm based on coalition game theory is used to solve the user and base station matching problem UBM, and a user and base station matching scheme is obtained, which maximizes the utility of each user.

[0068] Exemplarily, before solving the service provider resource allocation problem SPRA, it is necessary to determine the selection of users to base stations, so it is necessary to solve the user and base station matching problem UBM first.

[0069] From the selfish point of view of each user, he has the right to choose the base station that maximizes his utility. In order to solve this problem, the present application uses a coalition game method to show the interactive scenario of multiple participants aiming to maximize their respective benefits. Users who choose the same base station are regarded as a coalition, for example, the coalition How to find a stable coalition set to maximize the utility of each user and base station is the key to this problem. This game is called UBM coalition game.

[0070] The present application formulates the following irreversible transfer rule: coalition The subsets S1, S3 in the coalition S2 in the coalition Can join the disjoint coalition

[0071] If Then

[0072] If Then

[0073] If Then

[0074] Wherein, Is a comparison relation, that is, the utility of the left set is greater than that of the right set. Given two sets And They are the same set partitioned so that and or the coalitions in are disjoint, means there exists at least one subset S in such that for all m∈S, satisfies u m (S1)>u m (S2) and

[0075] For a set of initial selection of the nearest base station to form a coalition and iterative application of the user transfer rule, coalition formation game Has the optimal solution that satisfies individual rationality, group stability

[0076] The embodiments of the application adopt a user and base station matching CG-UBM algorithm based on coalition game theory to solve the base station matching problem UBM, and specifically include the following steps:

[0077] S131, initialize the user set to where N is the number of base stations, and represent the user set selecting the 1st, 2nd, nth and Nth base station respectively, and the initial value of each base station n and can be obtained by calling the SPRA algorithm;

[0078] S132, calculate the utility u m and the utility of all base stations

[0079] S133, start loop iteration, and determine whether the currently calculated is the same as the result obtained in the last iteration, if not, it means that it is uncertain whether the current coalition has reached the optimal state, then continue iteration;

[0080] S134, in the new iteration, each user compares the utility of its current state with the utility obtained after joining other coalitions to determine whether it should switch coalitions. If the utility u m and the utility of all base stations increases by joining other coalitions , update the user and base station matching scheme a m to the scheme n' after iteration;

[0081] S135, if the currently calculated The same as the result obtained in the last iteration, it is shown that the UBM based on coalition game has reached a stable equilibrium state, and the final user and base station matching scheme can be output.

[0082] In S140, the service provider resource allocation problem SPRA is decomposed into a sum of multiple independent sub-problems SSPRA, and the sum of multiple independent sub-problems SSPRA is solved based on the user and base station matching scheme by using a Lagrangian II algorithm to obtain a task offloading scheme.

[0083] Exemplarily, the embodiments of the present application focus on the resource allocation of the edge server, the transmission rate between the user and the base station, and the task processing selection of the base station. Considering the sufficiency of the cloud server resources, the transmission rate of the base station uploaded to the cloud server and the processing rate of the cloud server are respectively defined as r Cl and C Cl The SPRA problem can be modified as a sum of n independent sub-problems SSPRA.

[0084]

[0085]

[0086] (1)-(9), (13)-(14)

[0087] α represents a task offloading scheme of all users, and represents the total utility of the nth base station under the offloading scheme α. The embodiments of the present application convert the SSPRA problem into a problem with constraint conditions and .

[0088] The Lagrangian function of the sum of multiple independent sub-problems SSPRA is:

[0089]

[0090] In the formula, L(c, λ, μ) represents a Lagrangian function, λ represents a Lagrangian inequality operator, represents the total utility of the nth base station, μ m represents a Lagrangian equality operator, and c and μ are and μ m respectively.

[0091] The KKT (Karush-Kuhn-Tucker) condition of the Lagrangian function is:

[0092]

[0093] wherein, and e and b represent the relevant parameters in the unitary function Pr(Dm). d m,k The square of, express The square of.

[0094] The KKT conditions are solved using the Lagrange bisection interpolation (LII) algorithm, and the resulting solution is the unloading scheme. Specifically, the input during the solution process is... in, This is the user-base station matching scheme obtained from S130. λ0 is the initial value of the left boundary of λ in equation (18), and the initial value of the right boundary of λ is -ε, where ε and ε are two arbitrary decimals. After inputting the initial values, the solution is obtained using Lagrange bisection interpolation LII. Then the value C n Substitution The formula obtained The values ​​of λ and ε are compared to determine whether the KKT conditions are satisfied. After satisfying the KKT conditions, the left or right endpoint of λ is updated until the conditions λ(h) - λ(h-1) ≤ ε and ε are satisfied. That is, the difference between the values ​​of λ obtained in two consecutive iterations is small enough, and The difference between the values ​​obtained in two adjacent iterations is also small enough.

[0095] The Lagrange bisection interpolation algorithm LII is based on a given... This is a fact. To solve the SSPRA problem, it is necessary to determine the optimal... Make Maximize, then solve the optimal solution to the SPRA problem. Since the determination... In The SPRA problem is NP-hard, therefore this application proposes a heuristic algorithm to solve it. Since... Each addition or deletion of a user will affect and Therefore, this heuristic algorithm includes: in the user set Given a certain situation, the user set Set it to an empty set, and then collect it from the user each time. Add some users until the total utility is reached. No more additions, keep the current user set. As the optimal set of users.

[0096] Specifically, when the heuristic algorithm is executed, it first offloads the task to the user set of the cloud server via base station n. The initial value is set to an empty set, and then based on the current... The computational resource allocation for base station n is calculated using the Lagrange bisection interpolation algorithm (LII). and calculate the utility U n , according to the method, a loop iteration is performed. In the iteration, if the current U n (s) is greater than the last round U n (s-1), then another user is selected and added to , where is updated by M1. M1 is updated by selecting w from , so that U n (M n , M1∪{w}) is maximum. In each round of preparation for adding a user to M1, U is used as the objective function. If for any user w∈M2, there is always:

[0097]

[0098] then M1 is optimal.

[0099] In the Lagrange dichotomy interpolation algorithm LII, the overhead of each user obtaining the solution is O(β) when λ is determined. Therefore, the calculation overhead of the Lagrange dichotomy interpolation algorithm LII is In the SPRA problem, the calculation time of one loop iteration is Assuming that the number of loop iterations is H, the calculation overhead of the base station n is Let be the maximum value in M n , and the number of iterations of the algorithm CG-UBM is L, then the calculation overhead of the algorithm CG-UBM is

[0100] S150, according to the unloading scheme, the task is unloaded to the corresponding base station or cloud server.

[0101] The embodiments of the present application also provide a joint resource management and task unloading system for multi-service edge computing, as shown in Figure 2 , the system comprises:

[0102] A task acquisition module is configured to acquire tasks uploaded by a plurality of users;

[0103] An objective function establishment module is configured to establish a multi-objective function with the maximum utility of all base stations and the maximum utility of each user as the target;

[0104] A function conversion module is configured to convert the multi-objective function into a multi-objective optimization problem, and decompose the multi-objective optimization problem into a service provider resource allocation problem SPRA and a user and base station matching problem UBM;

[0105] A scheme matching module is configured to solve a base station matching problem UBM by using a user and base station matching algorithm based on coalition game theory CG-UBM, and obtain a user and base station matching scheme, which maximizes the utility of each user.

[0106] A scheme solving module is configured to decompose a service provider resource allocation problem SPRA into a sum of multiple independent sub-problems SSPRA, solve the sum of multiple independent sub-problems SSPRA by using a Lagrange interval difference algorithm LII based on the user and base station matching scheme, and obtain a task offloading scheme.

[0107] A scheme executing module is configured to offload tasks to corresponding base stations or cloud servers according to the offloading scheme.

[0108] Experimental demonstration

[0109] The rationality and effectiveness of the method proposed in the present application will be evaluated below. The main parameters are shown in Table 1. The experimental simulation results and analysis are as follows.

[0110] Table 1 Main parameters

[0111] Figure 2 is the overall framework diagram, that is, from the perspective of users and base stations, a multi-service type edge cloud resource allocation, task scheduling and UE task offloading model based on coalition game. Each user has the right to select a base station with the highest utility to offload its tasks. The base station determines whether the tasks offloaded to the base station are executed on the edge server or further offloaded to the cloud based on the best utility decision of the edge service provider, that is, even if the required service is placed on the base station.

[0112] The convergence of the CG-UBM algorithm under the scenario of 50 users and 5 base stations will be verified below. Figure 3 Part of the user cost change is shown. Under the condition of base station utility maximization strategy, after multiple rounds of mutual negotiation between users, the cost of 76% of the users shows a downward trend, and the cost reduction ratio is as high as 80%, and 24% of the users shows a small upward trend. With the increase of the number of iterations, the cost of each user reaches a stable value. In addition, the system reaches a system balance state after 6 iterations, which reflects the high efficiency of the proposed method.

[0113] In this experiment, not only the convergence and efficiency of the proposed method are to be verified, but also the rationality of the proposed method is to be evaluated. The performance of the algorithm CG-UBM is analyzed from the system computational overhead and the number of iterations. First, under the condition that the number of base stations is constant, how the system computational overhead and the number of iterations change with the increase of the number of users is analyzed. The number of base stations is 5, and the number of users is selected from the interval [40 80] with an increment of 5. The number of iterations of the algorithm CG-UBM is shown in Table 2. Figure 4In the figure, with the increase of the number of users, the blue and red bars represent the computing overhead and the number of iterations, respectively. In addition, a corresponding trend line is added to each column group. With the increase of the user scale, the computing overhead of the system increases in the form of a quadratic function, and the fitting curve degree is more than 0.99. In addition, with the increase of the number of users, the number of iterations of the system does not show an increasing trend, but always remains around 5 to 10.

[0114] The CG-UBM algorithm is compared with four baselines: the nearest task offloading (NTO) algorithm, the random task offloading (RTO) algorithm, the edge task scheduling (ETS) algorithm and the edge resource allocation (ERA) algorithm. Figure 5 and 6 The changes of the average latency of users and the utility of base stations with the increase of the number of users are shown, respectively. It can be seen from the figure that with the increase of the scale of the number of users, the average latency of users and the utility of base stations of each algorithm show an increasing trend. Compared with the ETS algorithm, the average latency of users of ETS is smaller when the number of users is 45 and 50, but in other cases, the CG-UBM algorithm reduces the average latency of users by 2%-8%. In Figure 6 In the comparison of the average cost of users between the CG-UBM and ETS algorithms, the CG-UBM has a slight advantage. But in the comparison of the utility of base stations, the CG-UBM algorithm has a 2%-10% improvement compared with the ETS algorithm. Compared with NTO, RTO and ERA, the CG-UBM reduces the average latency of users by 23%, 22% and 17%, respectively. The average cost of users is reduced by 6%, 4% and 4%, respectively. In addition, the utility of base stations of CG-UBM is increased by 5%, 10% and 7%, respectively.

[0115] In summary, in the multi-service scenario, the CG-UBM algorithm combining resource allocation, task scheduling and coalition game task offloading can not only reduce the latency and cost of users, but also improve the benefits of base stations through comprehensive task scheduling and resource allocation.

[0116] Although the preferred embodiments of the present application have been described, those skilled in the art can make additional changes and modifications to the embodiments once they know the basic inventive concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0117] Obviously, those skilled in the art can make various modifications and variations to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these modifications and variations.

Claims

1. A method for joint resource management and task offloading of multi-service edge computing, characterized in that, The method comprises the following steps: obtaining a plurality of tasks uploaded by users; establishing a multi-objective function aiming at maximizing the utility of all base stations and maximizing the utility of each user; converting the multi-objective function into a multi-objective optimization problem, and decomposing the multi-objective optimization problem into a service provider resource allocation problem (SPRA) and a user and base station matching problem (UBM); solving the user and base station matching problem (UBM) by using a user and base station matching algorithm based on coalition game theory (CG-UBM) to obtain a user and base station matching scheme, which maximizes the utility of each user; decomposing the service provider resource allocation problem (SPRA) into a sum of a plurality of independent sub-problems (SSPRA), and solving the sum of a plurality of independent sub-problems (SSPRA) by using a Lagrange interval difference algorithm (LII) based on the user and base station matching scheme to obtain an offloading scheme of the tasks; offloading the tasks to corresponding base stations or cloud servers according to the offloading scheme. The multi-objective function is expressed as: In the formula, max represents taking the maximum value. This represents the utility of all base stations. n This represents the base station number, with the symbol ∈ indicating that it belongs to. Represents a set of base stations. Indicates the first n Revenue per base station Indicates the first n Energy consumption of each base station Indicates the first n The total cost of offloading tasks to cloud servers for each base station. Indicates the first m The utility of an individual user Indicates the first m The user uploaded the [number]th [user name] belonging to the [number]th [user name]. k The size of task data for application services. Indicates the first m Payment delays for individual users' services. Indicates the first m Service delay for individual users Indicates the first m Energy consumption per user, where st represents the constraint condition. Indicates the first m The base station selected by each user m User ID, symbol Indicates any, Represents a set of users. k The number representing the application service type. Represents a set of application service types. Indicates base station n Regarding the first k Configuration indicator variables for application services, Indicates the first n The first base station k Computing resources for application services Indicates the first n The total computing resource limit for each base station Indicates the first n The base station for the first m The user uploaded the [number]th [user name] belonging to the [number]th [user name]. k The processing rate of tasks in application-like services. This indicates the situation under the following constraints. Indicates the first n Base station and the m Transmission rate between users Indicates the first n Channel bandwidth of each base station Indicates the first m Transmit power range for each user Indicates the first m The user and the n Channel gain between base stations Indicates background noise. Indicates the first m The user and the n Interference between base stations Indicates that the task is the first n A set of users whose data is offloaded from a base station to a cloud server. Indicates the first n The base station will be the first m The transfer rate at which a user's tasks are offloaded to the cloud server. Indicates the first n The transmission rate limit for each base station to offload tasks to the cloud server. Indicates the first n The base station for the first m The user uploaded the [number]th [user name] belonging to the [number]th [user name]. k The latency of tasks in application-like services Indicates the first k The workload requirements of application-like services Indicates the first n The first base station will be the first m The user uploaded the [number]th [user name] belonging to the [number]th [user name]. k The latency of offloading application-like service tasks to the cloud server. Indicates the first m The user uploaded the [number]th [user name] belonging to the [number]th [user name]. k The task of application services is offloaded to the cloud server's transmission rate. Indicates the first n The first base station will be the first m The user uploaded the [number]th [user name] belonging to the [number]th [user name]. k The processing speed of offloading application-like service tasks to cloud servers. Indicates the first n The first base station processes the first m The user uploaded the [number]th [user name] belonging to the [number]th [user name]. k Energy consumption caused by tasks of application services Indicates the first n The energy coefficient of each base station is determined by the structural characteristics of the edge server. Indicates the selection of the first n User set of each base station and user set The difference set; The method for solving the user and base station matching problem (UBM) by using a user and base station matching algorithm based on coalition game theory (CG-UBM) comprises the following steps: User set Initialize to , ,…, ,…, ,in N It refers to the number of base stations. , , and These represent selecting the 1st, 2nd, and 3rd options respectively. n The and the first N A set of users for each base station; Compute utility for each user and utility for all base stations ; Determine the result of the current calculation Is the result the same as the result obtained in the previous iteration? If not, continue iterating. In the next iteration, each user compares the utility of its current state with the utility of joining other coalitions if the utility of joining other coalitions is higher than the utility of its current state, the user updates its matching scheme with the base station to the one after the iteration ​​ If the current computation yields The final user-to-base station matching solution is output, as in the previous iteration. 2.The joint resource management and task offloading method of multi-service edge computing according to claim 1, characterized in that, The sum of a plurality of independent sub-problems (SSPRA) is expressed as: max represents taking the maximum value, represents the total utility of all users' tasks for the offloading scheme, represents the total utility of all users' tasks for the offloading scheme represents the total utility of all users' tasks for the offloading scheme n represents the total utility of all users' tasks for the offloading scheme represents the total utility of all users' tasks for the offloading scheme n represents the total utility of all users' tasks for the offloading scheme represents the total utility of all users' tasks for the offloading scheme n represents the total utility of all users' tasks for the offloading scheme represents the total utility of all users' tasks for the offloading scheme n represents the total utility of all users' tasks for the offloading scheme The Lagrange function of the sum of a plurality of independent sub-problems (SSPRA) is expressed as: wherein, denotes the Lagrangian function, denotes the Lagrangian inequality operator, denotes the total utility of the n th base station, m denotes the user's index, the symbol ∈ denotes belonging to, denotes the set of users selecting the n th base station, denotes the set of users whose tasks are offloaded by the n th base station to the cloud server, denotes the difference set of and , respectively, denotes the processing rate of the n th base station for the tasks uploaded by the m th user belonging to the k th application service, denotes the total computing resource limit of the n th base station, denotes the Lagrangian equality operator, and are the short hand of and , respectively; The KKT condition of the Lagrange function is: wherein the symbol represents any, , , represents the n energy coefficient determined by the structural characteristics of the edge server, represents the k workload requirement of the application service of the class, m represents the k size of the task data uploaded by the e user belonging to the application service of the b class, represents the relevant parameters in the unit price function represents the square, represents the n transmission rate between the m base station and the user, represents the square of The KKT condition is solved by using a Lagrange interval difference algorithm (LII) to obtain the offloading scheme. 3.The joint resource management and task offloading method of multi-service edge computing of claim 2, wherein, In the case that the user set is determined, the user set is set to the empty set, and then some users are added to the user set until the total utility no longer increases, the current user set is taken as the optimal user set.

4. The system of claim 1-3, wherein the joint resource management and task offloading method of multi-service edge computing is applied, characterized in that, The method comprises the following steps: a task acquisition module for obtaining a plurality of tasks uploaded by users; a target function establishment module for establishing a multi-objective function aiming at maximizing the utility of all base stations and maximizing the utility of each user; a function conversion module for converting the multi-objective function into a multi-objective optimization problem, and decomposing the multi-objective optimization problem into a service provider resource allocation problem (SPRA) and a user and base station matching problem (UBM); a scheme matching module for solving the user and base station matching problem (UBM) by using a user and base station matching algorithm based on coalition game theory (CG-UBM) to obtain a user and base station matching scheme, which maximizes the utility of each user; a scheme solving module for decomposing the service provider resource allocation problem (SPRA) into a sum of a plurality of independent sub-problems (SSPRA), and solving the sum of a plurality of independent sub-problems (SSPRA) by using a Lagrange interval difference algorithm (LII) based on the user and base station matching scheme to obtain an offloading scheme of the tasks; a scheme execution module for offloading the tasks to corresponding base stations or cloud servers according to the offloading scheme.

Images