阅读说明：本技术 一种移动边缘计算网络中联合任务卸载和资源分配方法 (Joint task unloading and resource allocation method in mobile edge computing network ) 是由 鲜永菊 李栋 于 2020-03-12 设计创作，主要内容包括：本发明涉及一种移动边缘计算网络中联合任务卸载和资源分配方法,属于无线通信和移动边缘计算领域。该方法包括：UE产生新的计算任务,向MEC服务器发送任务卸载请求；MEC服务器收集本时隙内所有用户端发送的计算卸载请求信息；将用户计算任务和MEC服务器资源进行首次匹配,构成初始卸载策略集,计算初始目标函数的值；求得最小的目标函数值和获得最优目标函数值的用户。更新所有用户的卸载决定,判断得到的最优的目标函数不小于上一次的目标函数值是否成立,成立则输出卸载决定,信道分配矩阵和最优计算资源。本发明降低了用户卸载的成本开销,节约了移动用户总的成本；能够接纳更多的计算卸载任务,提升了系统的任务执行效率。(The invention relates to a joint task unloading and resource allocation method in a mobile edge computing network, belonging to the field of wireless communication and mobile edge computing. The method comprises the following steps: UE generates a new calculation task and sends a task unloading request to an MEC server; the MEC server collects the calculation unloading request information sent by all the user terminals in the time slot; matching the user calculation task with the MEC server resource for the first time to form an initial unloading strategy set, and calculating the value of an initial objective function; the user that gets the smallest objective function value and the user that gets the best objective function value. And updating the unloading decisions of all users, judging whether the obtained optimal objective function is not less than the last objective function value, and if so, outputting the unloading decisions, the channel allocation matrix and the optimal computing resources. The invention reduces the cost overhead of user unloading and saves the total cost of the mobile user; more calculation unloading tasks can be accommodated, and the task execution efficiency of the system is improved.)

1. A method for joint task offloading and resource allocation in a mobile edge computing network is characterized by specifically comprising the following steps:

s1: a mobile user generates a new calculation task and sends a task unloading request to an MEC server;

s2: the MEC server collects the calculation unloading request information sent by all the user terminals in the time slot;

s3: the MEC server carries out first matching on the user computing task and the MEC server resource to form an initial unloading strategy set; according to a network scene, establishing a communication model, a mobile equipment local calculation model and an edge server calculation model to establish a system benefit maximization problem model, then calculating the cost of task unloading to each server, and solving the total user cost in an initial unloading state, namely calculating the value of an objective function;

s4: obtaining the unloading decision of all users by changing the unloading decision of a single user;

s5: assigning subchannels and computing resources to users according to the offloading decision of step S4;

s6: calculating objective function values for the subchannels and calculation resources allocated by the user according to step S5;

s7: finding the smallest objective function value and the user who obtained the optimum objective function value under the unloading decision of step S4;

s8: updating the offload decisions for all users;

s9: judging whether the end condition is satisfied, executing the step S10 if the end condition is satisfied, and returning to the step S4 if the end condition is not satisfied;

s10: and outputting the unloading decision, the channel allocation matrix and the optimal computing resource.

2. The method of claim 1, wherein the step S3 specifically includes the following steps:

s31: establishing a communication model;

definition C ═ C_nk]_N×KDenotes a subchannel allocation matrix, wherein c_nk1 denotes the assignment of a subchannel k to a user n, otherwise c_nk0; the transmission power p of a given user n_nWhen the user n uses the subchannel k in the Small Base Station (SBS) m, the Signal to Interference plus Noise Ratio (SINR) is:

wherein, V_mRepresents the set of users of the service within the scope of the SBSm,representing the channel gain from user n to SBSm on subchannel k,indicating the inter-cell interference, σ, superimposed by other cell co-channel users²Is additive white gaussian noise power;

the transmission rate of user n on subchannel k is:

the transmission rate of user n to the MEC server is expressed as:

s32: establishing a local calculation model;

order toRepresenting the computing power of user n, with user n tasks processing delays locallyAnd energy consumptionRespectively as follows:

wherein the coefficient v_nIn relation to the chip hardware architecture, the cost incurred by the local computation is solved according to equations (4) and (5)Comprises the following steps:

wherein the content of the first and second substances,weights representing delay and energy consumption, respectively, for performing a computational task at user n; all weight values are in the range of [0, 1%]And is prepared from

S33: establishing an edge server equipment calculation model;

let f_nThe computing capacity allocated to the user n by the server is represented, and the time delay generated by the unloading of the user n task to the MEC server comprises two parts of transmission delay and server execution delay which are respectively represented by symbolsAndshow, its transmission energy consumedRepresents; combining equation (3) yields:

wherein d is_nExpressed as the size of the input data, w_nIndicating the CPU cycles required to compute the task;

the remote computation overhead is expressed according to equations (7), (8) and (9) as:

s34: establishing a problem model of joint calculation unloading decision and resource allocation based on optimal time and energy consumption cost;

user n cost function Z_nExpressed as:

wherein the content of the first and second substances,the joint computation offload, subchannel and computation resource optimization problem is expressed as:

the user's offloading cost at the initial offloading decision is found by equation (12).

3. The method for joint task offloading and resource allocation in a mobile edge computing network according to claim 2, wherein the step S4 specifically includes: let A^l-1Representing the user's unload decision in iteration l-1, then in iteration l-1 user n changes all of the user's own unload decisionsUser offload decision A^l-1(n) is:

wherein the content of the first and second substances,is a modulo two addition method.

4. The method of claim 3, wherein in the step S5, the sub-channel is allocated to the user according to the offloading decision of the step S4, and the method specifically comprises the following steps:

s51: the users and the sub-channels are regarded as game parties, a rational decision meeting the benefit maximization of the users and the sub-channels is made, and preference functions of the game parties form a preference relation of selection, namely the users and the channels are subjected to matching preference selection by taking the maximized transmission rate as a target; the preference function for subscriber n for subchannel k is defined as:

U_n(k)＝R_nk(14)

the preference function of the MEC server for the subchannel k matched by the user n is:

wherein α represents the unit rate gain coefficient of user n transmitting task on sub-channel k, β represents the unit interference cost coefficient,for user n, the total interference from other offloaded users on subchannel k is defined as:

s52: each user n constructs the preference relation with the sub-channel according to the formula (14);

s53: each user n selects the best sub-channel k^*＝arg max_k∈KU_n(k) And a subchannel k^*Requesting a connection;

s54: the MEC server counts the requesting users of each sub-channel k;

s55: the MEC server selects the best user among its requesting users for each subchannel kAnd allocating the sub-channel to the user;

s56: updating the set of users which fail to match of each subchannel k, and then deleting the subchannel which is already allocated from the subchannel set;

s57: updating the set of users which fail to be matched;

s58: if there is no other request between the user and the sub-channel, the sub-channel allocation matrix is output, otherwise, the step S52 is returned to.

5. The method of claim 4, wherein in the step S5, the allocating computing resources for the user is determined according to the offloading of the step S4, and the method specifically comprises the following steps:

s61: computing the resource allocation problem according to equation (12) after the subchannel allocation is complete, as:

equation (17) satisfies condition ∑_n∈Nf_nF or less and condition F_n≥0,Wherein F represents the maximum computing resource of the MEC server, F_nRepresenting the computing power allocated to the user by the server; solving by adopting a convex optimization method;

s62: the Lagrange function expression under the inequality constraint of the calculation formula (17) is as follows:

wherein mu is more than or equal to 0 and is Lagrange multiplier, N_offIs a set of offload users;

s63: order toObtaining the optimal solution of formula (17)Comprises the following steps:

s64: substituting equation (19) for equation (18) yields a lagrangian dual function of equation (17) as:

s65: in equation (20), the first order partial derivative of μ is calculated and made equal to 0 to obtain the optimal Lagrangian multiplier μ^*Comprises the following steps:

s66: obtaining optimal computing resources according to equations (19) and (21)Comprises the following steps:

6. a stent according to claim 5The method for joint task offloading and resource allocation in a moving edge computing network is characterized in that step S6 specifically includes: unload decision A from step S4^l-1(n) the function value of the subchannel and calculation resource calculation formula (12) allocated to the user in step S5, and is represented as Q (A)^l-1(n))。

7. The method of claim 6, wherein the step S9 specifically includes: is judged to beIs true, wherein Q (A)^l) Denoted as being in the unload decision A^lThe value of equation (12); execution of step S10 is established, otherwise, return to step S4 is made.

Technical Field

The invention belongs to the field of wireless communication and mobile edge computing, and relates to a joint task unloading and resource allocation method in a mobile edge computing network.

Background

The advent of mobile edge computing provides an effective solution for resource constrained mobile devices. Mobile devices can offload all or part of the compute-intensive tasks over wireless channels to a Mobile Edge Computing server (MEC) with relatively rich Computing resources to reduce local processing task latency and energy consumption. However, extra transmission delay and energy consumption may be caused by performing computation offloading, and determining the optimal offloading decision is a research focus in the MEC network. Furthermore, both the radio resources and the computational capacity of the MEC server are limited resources, and in multi-User MEC systems radio and computational resources are provided in a shared manner to User Equipment (UE). Different users have different computing tasks and service quality requirements, so that different UEs have different radio and computing resource allocation modes, and joint computing offloading and server resource allocation is considered to be one of the most important problems for improving the performance of the MEC network.

In recent years, researchers have made extensive research around mobile edge computing networks, and an energy-saving resource management strategy of an asynchronous MEC system is researched in documents "You C, Zeng Y, Zhang R, et al. resource management for asynchronous mobile-edge computing and streaming [ C ]//2018 IEEE International Conference on Communications works (ICC works). IEEE, 2018", and an optimization problem is formulated to minimize total mobile user energy consumption under time sharing and computation delay constraints; the optimization problem is reduced to two consecutive problems to find the optimal scheduling order, and then jointly optimize data partitioning and time division given the optimal order. The document "Tran T X, Pompili D. Joint task offloading and resource allocation for multi-server mobile-edge computing networks [ J ]. IEEE Transactions on vehicle technology,2019,68(1):856 868" considers the problem of joint task offloading and resource allocation in a multi-cell wireless network scenario supporting MEC, the author decomposing the original problem into a Resource Allocation (RA) problem with fixed task offloading decisions and a Task Offloading (TO) problem with an optimal value function. The RA problem is solved through a convex and quasi-convex optimization technology, and a novel heuristic algorithm is provided for the TO problem, and the algorithm can realize suboptimal solution within polynomial time. The documents "Zhang J, Xia W, Cheng Z, et al, an evolution of the game for joint Wireless and closed resource allocation in mobile edge computing [ C ]// 20179 th international conference on Wireless Communications and Signal Processing (WCSP). IEEE,2017: 1-6" mainly study the joint resource allocation algorithm based on games and analyze the stability of games on the basis of considering the energy consumption and time delay of mobile terminals and the monetary cost in the mobile edge computing environment. The document ' Zhao Li Qiang, Lu Xiao Di, Liang Kai, Yang Ke, Song Feng Fei ' a network system and service method for providing mobile edge computing service, China 108494612[ P ] 2018.09.04 ' discloses a network system level service method for providing mobile edge computing service, which solves the MEC deployment problem on mobile communication network. The method comprises the steps that four modes of local MEC server calculation, multi-MEC server joint calculation, specific non-local MEC server calculation and cloud center calculation are respectively provided, and a controller judges which calculation mode is more reasonable according to the calculated amount and delay sensitivity of tasks. Chinese 108920279[ P ] 2018.11.30 ", in order to reduce the reaction delay and energy consumption of a mobile device, wherein the multi-user scenario is that a plurality of mobile devices are connected with an MEC server, each mobile device can select one of a plurality of channels between the mobile device and the MEC server to carry out communication, and the MEC server is connected with a central cloud through a backbone network. In particular to two stages: the first phase is to decide whether the task is executed on the local mobile device or off-loaded to the MEC server; the second phase is when the server resources are insufficient, determining whether the task continues to wait on the MEC server or is offloaded to a remote cloud center for execution.

However, in the prior art, in a heterogeneous network environment, the MEC server resource allocation has the following problems: in a multi-user single MEC server access scenario, the problem of optimizing energy consumption and delay of user equipment relates to joint allocation of wireless resources and computing resources. Or only the calculation resource allocation is considered and the influence of the wireless resource allocation on the optimization result is ignored; or the wireless resources are only allocated to each user in equal proportion in the form of sub-carriers although the joint allocation of the wireless resources and the computing resources is considered. This seems to guarantee per-user fairness, but for tasks with large input data volumes it may appear that resources are not sufficient, whereas for tasks with small transmission data resources are spare. The unreasonable resource allocation is caused, so the invention needs to design a finer-grained wireless resource allocation method according to the differentiation of specific attributes of tasks.

Disclosure of Invention

In view of this, the present invention provides a method for joint task offloading and resource allocation in a mobile edge computing network, which solves the problem of task offloading and resource allocation in a multi-user deployment scenario in a heterogeneous wireless network. The characteristics of each user are considered, the cost and expense for unloading the user are effectively reduced, and the total cost (time delay and energy consumption) of the mobile user is saved; more calculation unloading tasks can be accommodated, and the task execution efficiency of the system is improved.

In order to achieve the purpose, the invention provides the following technical scheme:

a method for joint task unloading and resource allocation in a mobile edge computing network specifically comprises the following steps:

s1: a mobile user generates a new calculation task and sends a task unloading request to an MEC server;

s2: the MEC server collects the calculation unloading request information sent by all the user terminals in the time slot;

s3: the MEC server carries out first matching on the user computing task and the MEC server resource to form an initial unloading strategy set; according to a network scene, establishing a communication model, a mobile equipment local calculation model and an edge server calculation model to establish a system benefit maximization problem model, then calculating the cost of task unloading to each server, and solving the total user cost in an initial unloading state, namely calculating the value of an objective function;

s4: obtaining the unloading decision of all users by changing the unloading decision of a single user;

s5: assigning subchannels and computing resources to users according to the offloading decision of step S4;

s6: calculating objective function values for the subchannels and calculation resources allocated by the user according to step S5;

s7: finding the smallest objective function value and the user who obtained the optimum objective function value under the unloading decision of step S4;

s8: updating the offload decisions for all users;

s9: judging whether the end condition is satisfied, executing the step S10 if the end condition is satisfied, and returning to the step S4 if the end condition is not satisfied;

s10: and outputting the unloading decision, the channel allocation matrix and the optimal computing resource.

Further, the step S3 specifically includes the following steps:

s31: establishing a communication model;

definition C ═ C_nk]_N×KDenotes a subchannel allocation matrix, wherein c_nk1 denotes the assignment of a subchannel k to a user n, otherwise c_nk0; the transmission power p of a given user n_nWhen the user n uses the subchannel k in the Small Base Station (SBS) m, the Signal to Interference plus Noise Ratio (SINR) is:

wherein, V_mRepresents the set of users served in the SBS m range,representing the channel gain from user n to SBS m on subchannel k,indicating the inter-cell interference, σ, superimposed by other cell co-channel users²Is additive white gaussian noise power;

the transmission rate of user n on subchannel k is:

the transmission rate of user n to the MEC server is expressed as:

s32: establishing a local calculation model;

order toRepresenting the computing power (in CPU cycles per second) of user n, whose task is locally delayedAnd energy consumptionRespectively as follows:

wherein the coefficient v_nRelating to chip hardware architecture, making upsilon according to existing measurement_n＝5×10^-27. Determining the cost of the local calculation according to equations (4) and (5)Comprises the following steps:

wherein the content of the first and second substances,weights representing delay and energy consumption, respectively, for performing a computational task at user n; all weight values are in the range of [0, 1%]And is prepared from

S33: establishing an edge server equipment calculation model;

let f_nExpressed as the computing power (in CPU cycles per second) allocated to user n by the server, the time delay generated by the unloading of user n tasks to the MEC server mainly comprises two parts of transmission delay and server execution delay which are respectively represented by symbolsAndshow, its transmission energy consumedRepresents; combining equation (3) yields:

wherein d is_nExpressed as the size of the input data, w_nIndicating the CPU cycles required to compute the task;

the remote computation overhead is expressed according to equations (7), (8) and (9) as:

s34: establishing a problem model of joint calculation unloading decision and resource allocation based on optimal time and energy consumption cost;

user n cost function Z_nExpressed as:

wherein the content of the first and second substances,the joint computation offload, subchannel and computation resource optimization problem is expressed as:

the user's offloading cost at the initial offloading decision is found by equation (12).

Further, the step S4 specifically includes: let A^l-1Representing the user's offload decision in iteration l-1, user n changes the offload decision a of all users under its own offload decision in iteration l-1^l-1(n) is:

wherein the content of the first and second substances,is a modulo two addition method.

Further, in the step S5, the step of determining to assign a sub-channel to the user according to the offloading in the step S4 specifically includes the following steps:

s51: the users and the sub-channels are regarded as game parties, a rational decision meeting the benefit maximization of the users and the sub-channels is made, and preference functions of the game parties form a preference relation of selection, namely the users and the channels are subjected to matching preference selection by taking the maximized transmission rate as a target; the preference function for subscriber n for subchannel k is defined as:

U_n(k)＝R_nk(14)

the preference function of the MEC server for the subchannel k matched by the user n is:

wherein α represents the unit rate gain coefficient of user n transmitting task on sub-channel k, β represents the unit interference cost coefficient,for user n, the total interference from other offloaded users on subchannel k is defined as:

s52: each user n constructs the preference relation with the sub-channel according to the formula (14);

s53: each user n selects the best sub-channel k^*＝arg max_k∈KU_n(k) And a subchannel k^*Requesting a connection;

s54: the MEC server counts the requesting users of each sub-channel k;

s55: the MEC server selects the best user among its requesting users for each subchannel kAnd allocating the sub-channel to the user;

s56: updating the set of users which fail to match of each subchannel k, and then deleting the subchannel which is already allocated from the subchannel set;

s57: updating the set of users which fail to be matched;

s58: if there is no other request between the user and the sub-channel, the sub-channel allocation matrix is output, otherwise, the step S52 is returned to.

Further, in the step S5, the step of allocating computing resources to the user according to the uninstallation decision in the step S4 specifically includes the following steps:

s61: computing the resource allocation problem according to equation (12) after the subchannel allocation is complete, as:

equation (17) satisfies the condition ∑_n∈Nf_nF or less and condition F_n≥0,Wherein F represents the maximum computing resource of the MEC server, F_nRepresenting the computing power allocated to user n by the server; due to the functionThe Hessian matrix ofAndand forming, wherein the Hessian matrix is a semi-positive matrix. It follows that the function is a convex function. And the feasible region of equation (17) is a convex set, the optimization problem of equation (17) is a convex optimization problem. Solving by adopting a convex optimization method;

s62: the Lagrange function expression under the inequality constraint of the calculation formula (17) is as follows:

wherein mu is more than or equal to 0 and is Lagrange multiplier, N_offIs a set of offload users; based on KKT condition, constraintIs relaxed and has been eliminated in equation (18).

S63: order toObtaining the optimal solution of formula (17)Comprises the following steps:

s64: substituting equation (19) for equation (18) yields a lagrangian dual function of equation (17) as:

s65: in equation (20), the first order partial derivative of μ is calculated and made equal to 0 to obtain the optimal Lagrangian multiplier μ^*Comprises the following steps:

s66: obtaining optimal computing resources according to equations (19) and (21)Comprises the following steps:

further, the step S6 specifically includes: unload decision A from step S4^l-1(n) the function value of the subchannel and calculation resource calculation formula (12) allocated to the user in step S5, and is represented as Q (A)^l-1(n))。

Further, the step S9 specifically includes: is judged to beIs true, wherein Q (A)^l) Denoted as being in the unload decision A^lThe value of equation (12); execution of step S10 is established, otherwise, return to step S4 is made.

The invention has the beneficial effects that: aiming at a mobile edge system under a heterogeneous network, the method combines user computing unloading, sub-channel allocation and allocation of MEC server computing resources in consideration of different resource requirements of different users. The unloading proportion of the user tasks in the whole system is improved, and the time delay and the energy consumption overhead of the total users are reduced.

Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention may be realized and attained by the means of the instrumentalities and combinations particularly pointed out hereinafter.

Drawings

For the purposes of promoting a better understanding of the objects, aspects and advantages of the invention, reference will now be made to the following detailed description taken in conjunction with the accompanying drawings in which:

fig. 1 is a schematic diagram of a deployment scenario of an MEC server in a heterogeneous network;

FIG. 2 is a flowchart of a joint optimization method for computing task offloading, sub-channel and computing resources according to the present invention.

Detailed Description

The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention in a schematic way, and the features in the following embodiments and examples may be combined with each other without conflict.

Referring to fig. 1 to 2, as shown in fig. 1, the heterogeneous network is composed of a Macro Base Station (MBS) and a Small Base Station (SBS), and the ues are randomly distributed in the whole area. The MEC server is a small data center deployed by telecommunication operators at the edge of heterogeneous networks, and is connected to MBS to provide computing resources for mobile users and communication resources for users through heterogeneous networks. The present invention represents a set of MBS and SBS as M ═ {0,1, 2.., M }, and a set of user compositions as N ═ 1, 2.., N }. It is assumed that there are K orthogonal sub-channels,the set of subchannels is denoted by K ═ 1, 2. Definition a_nWith {0,1} representing the n-th user's offload decision, 0 represents the user's choice to execute their task locally, and 1 represents the user's choice to offload the task to MEC execution. With A ═ a₁,a₂,...,a_N]Indicating all users' offload decisions.

As shown in fig. 2, a preferred method for jointly optimizing resource allocation and computation offload in a mobile edge computing network in the present invention includes the following steps:

step S1, when the mobile user generates a new computing task, a computing offload request may be sent to the MEC server over the control link. The compute offload request message contains information about the volume compute task, which may be denoted as { d }_n,w_n}. Wherein d is_nExpressed as the size of the input data, w_nIndicating the CPU cycles required to compute the task.

In step S2, the MEC server collects the computation offload request information sent by the user end every other time slot. And combining the number of the sub-channels and the self-computing resource residual situation to prepare for resource allocation.

Step S3, the MEC server matches the user computing task with the MEC server resource for the first time to form an initial offloading policy set, and assumes that all users offload in the initial state. With A⁰＝[1,1,...,1]Indicating the offload decisions of all users in the initial state. Before resource allocation is executed, according to a network scene, a communication model, a mobile equipment local calculation model and an edge server calculation model are established, and a system benefit maximization problem model is formulated. And the MEC server calculates the cost of task unloading to each server according to the established mathematical model, and then calculates the total user cost in the initial unloading state.

Step S3 specifically includes the following steps:

step S31, a communication model is established.

Definition C ═ C_nk]_N×KRepresenting a subchannel assignment matrix. Wherein c is_nk1 denotes the assignment of a subchannel k to a user n, otherwise c_nk0. The transmit power of a given user np_nThe Signal to interference plus Noise Ratio (SINR) when the user n uses the subchannel k in the SBSm is:

wherein, V_mRepresented as a set of users of SBSm-wide services.Is the channel gain from user n to SBSm on subchannel k. The second term of denominator in formula (1) is expressed as the inter-cell interference, σ, superimposed by other cell co-channel users²Is additive white gaussian noise power.

The transmission rate of user n on subchannel k is:

the transmission rate of user n to the MEC server can be expressed as:

and step S32, establishing a local calculation model.

By usingRepresenting the computing power of user n (in CPU cycles per second). UEn task local processing delayAnd energy consumption

Wherein the coefficient v_nRelating to chip hardware architecture, making upsilon according to existing measurement_n＝5×10^-27. The costs associated with the local calculation can be determined from equations (4) and (5)

Wherein the content of the first and second substances,representing the weight of the delay and energy consumption, respectively, of performing a computational task at the UEn. All weight values are in the range of [0, 1%]And is prepared from

Step S33, a server device calculation model is established.

Let f_nExpressed as the computing power (in CPU cycles per second) allocated to user n by the server. The time delay generated by unloading UEn tasks to the MEC server mainly comprises two parts of transmission delay and server execution delay which are respectively formed by symbolsAndshow, its transmission energy consumedAnd (4) showing. In conjunction with equation (3) one can obtain:

the remote computation overhead according to equations (7), (8) and (9) is expressed as

And step S34, establishing a problem model for joint calculation unloading decision and resource allocation based on optimal time and energy consumption cost.

Based on the above discussion, the UEn cost function Z_nCan be expressed as:

wherein the content of the first and second substances,the joint computation offload, subchannel and computation resource optimization problem can be expressed as:

the offloading cost of the user in the initial offloading decision can be found by equation (12).

Step S4, use A^l-1Indicating the user's unload decision at iteration l-1. With A^l-1(n) denotes the offload decision for all users under which user n changes its own offload decision in the l-1 th iteration, where l denotes the number of iterations. Computing offload decisions A for all users that change individual user's own offload decisions during each iteration^l-1(n) is:

wherein the content of the first and second substances,denoted as modulo two addition.

In step S5, a sub-channel is assigned to the user based on the offload decision of step S4. The method specifically comprises the following steps:

and step S51, the users and the sub-channels are regarded as game parties, a rational decision meeting the benefit maximization of the users and the sub-channels is made, and preference functions of the game parties form a preference relation of selection, namely the users and the channels are subjected to matching preference selection by taking the maximized transmission rate as a target. Defining the preference function of user n for subchannel k as

U_n(k)＝R_nk(14)

The preference function of the MEC server for the subchannel k matched by the user n is:

where α represents the unit rate gain coefficient for user n transmitting tasks on channel k, β is the unit interference cost coefficient,for user n, the total interference from other offloaded users on subchannel k is defined as

In step S52, each user n constructs its preference relationship with the sub-channel according to equation (14).

Step S53, each user n selects the best sub-channel k^*＝arg max_k∈KU_n(k) In that respect And pair sub-channel k^*A connection is requested.

In step S54, the MEC server counts the requesting users of each sub-channel k.

At step S55, the MEC server selects the best user among its requesting users for each subchannel kAnd assigns the subchannel to the user.

Step S56, updating the set of users failing to match for each subchannel k, and then deleting the subchannel that has been allocated from the subchannel set.

And step S57, finally updating the user set which fails to be matched.

In step S58, if there is no other request between the user and the sub-channel, the sub-channel allocation matrix is output, otherwise, the process returns to step S52.

Step S6, determining to allocate computing resources to the user according to the offloading of step S4, specifically comprising the steps of:

in step S61, the problem of calculating resource allocation according to equation (12) after the sub-channel allocation is completed can be expressed as

Wherein condition ∑ is satisfied_n∈Nf_nF or less and condition F_n≥0,Where F represents the maximum computing resource of the MEC server. Due to the functionThe Hessian matrix ofAndand forming, wherein the Hessian matrix is a semi-positive matrix. It follows that the function is a convex function. And the feasible region of equation (17) is a convex set, the optimization problem of equation (17) is a convex optimization problem. Solving by convex optimization method

In step S62, the lagrangian function expression under the inequality constraint of the calculation formula (17) is:

wherein mu is more than or equal to 0 and is Lagrange multiplier. Constraint f based on KKT condition_n≥0,Is relaxed and has been eliminated in equation (18).

Step S63, letAn optimal solution of equation (17) can be obtainedIs composed of

Step S64, substituting (19) into (18), we can obtain the lagrangian dual function of equation (17) as

In step S65, by calculating the first order partial derivative of μ and making it equal to 0 in equation (20), the optimal lagrange multiplier μ can be obtained^*Is composed of

In step S66, the optimal calculation resource f can be obtained according to the equations (19) and (21)_n ^*Is composed of

Step S7, determining A according to the unloading of step S4^l-1(n), step S5Sub-channels allocated to the user and step S6 allocates the user with a function value of the computational resource table calculation formula (12), which is expressed as Q (a)^l-1(n))。

In step S8, the unload decision A in step S4 is determined^l-1(n) optimal objective function valueAnd the user of the optimal objective function value

Step S9, update all users' offload decisions

In step S10, the judgment is yesIs true, wherein Q (A)^l) Denoted as being in the unload decision A^lThe value of equation (12) at time. Step S11 is executed if it is established, and the process returns to step S4 if it is not established.

And step S11, outputting the optimal unloading decision, the optimal channel allocation matrix and the optimal computing resource.

Finally, the above embodiments are only intended to illustrate the technical solutions of the present invention and not to limit the present invention, and although the present invention has been described in detail with reference to the preferred embodiments, it will be understood by those skilled in the art that modifications or equivalent substitutions may be made on the technical solutions of the present invention without departing from the spirit and scope of the technical solutions, and all of them should be covered by the claims of the present invention.