CN110740473A - management method for mobile edge calculation and edge server - Google Patents

Abstract

Before the edge server provides auxiliary calculation, the present invention comprehensively considers the influence of the uplink transmission delay, the calculation delay and the downlink transmission delay on the total delay, and optimizes the allocation of the current limited resources of the system for each mobile terminal, so as to reduce the execution of all The total latency of user tasks. Especially in the current situation where the size of downlink data in some scenarios is too large to be ignored due to technological development, the present invention can minimize the total delay of all tasks corresponding to the task execution request, and for the case where the size of downlink data is small , the present invention can also minimize the total delay of all tasks corresponding to the task request, and can efficiently meet the user's low-latency requirements in different scenarios, thereby improving user experience.

Description

A management method and edge server for mobile edge computing

technical field

The present invention relates to the field of wireless communication, in particular to the joint optimization of computing offloading and resource allocation in mobile edge computing, and more particularly, to a management method and an edge server for mobile edge computing.

Background technique

With the rapid development of mobile communications and the rapid popularization of intelligent mobile terminals, many new applications such as virtual reality, augmented reality and autonomous driving have emerged. Such applications with low latency and high reliability communication requirements place higher requirements on the computing power of the mobile terminal. Since mobile terminals with limited computing power will generate high application processing delay when processing such applications and affect the service experience of end users, how to reduce the application processing delay and improve the service experience of end users is an urgent need to solve. one of the key questions.

In response to the above problems, mobile cloud computing (MCC) technology is proposed. The goal of MCC is to extend the abundant computing resources of the cloud to resource-constrained mobile terminals, thereby enhancing the potential computing capabilities of mobile terminals and reducing application processing delays. To achieve this goal, mobile terminals need to migrate computationally intensive tasks to cloud servers through wireless access. Although this method can reduce the load of the mobile terminal, it also has obvious defects, that is, the long distance between the mobile terminal and the cloud server and a large number of terminal service requests will increase the network delay, thereby reducing the service experience of the terminal user.

The European Telecommunications Standards Institute (ETSI) proposes a new technology: mobile edge computing (MEC). ) to reduce the communication load and network delay of users within its coverage area. However, due to cost constraints, edge servers usually have relatively limited computing resources compared to cloud servers. Therefore, for edge servers, too many tasks may bring extra load to service nodes and affect network latency.

At present, there are mainly three types of methods to reduce the task delay of the mobile terminal in the MEC network:

The first approach is to design and optimize the offloading scheme for mobile users. This type of method assumes that the base station can obtain the basic information of the user, such as the transmission distance, the type of task requested, the current number of users in the cell, the location information of the user in the cell in the cellular network, and the computing power of the edge server. Mobile users make offload selections by comparing the execution delay of tasks on the edge server and the local execution delay, thereby completing the optimization of the offload strategy and enabling end users to obtain the lowest delay. In this type of method, the maximum computing power of the edge server is dynamically changed according to the total number of users in the cell, and the resources are evenly allocated to each requesting user, and the allocation of downlink bandwidth resources is ignored. Therefore, there are at least two defects in this method. One is that adaptive computing resources cannot be allocated to users according to the size of the user's task, resulting in the computing resources not being fully utilized or not meeting the requirements; the other is that downlink transmission is ignored. Delay. When there are users involved in application scenarios such as AR and VR, the size of the data in the calculation result cannot be ignored, and the downlink transmission delay may be large, which greatly affects the user experience.

The second type of method is to design a joint optimization scheme of offloading strategy and resource allocation. Compared with the first type of method, this type of method further limits the computing power and system communication resources of the edge server. When there is a user task request, the edge server will perform uplink resources and computing for the user through the obtained task request. Optimal allocation of resources to minimize latency. Then, the delay obtained according to the optimized allocation is compared with the local computing delay, and then the optimized offloading strategy and the minimum delay are obtained. In this method, the allocation of computing resources is allocated according to needs, that is to say, the more computing resources required by the user's task request, the more points are allocated, and the less computing resources required, the less points are allocated. Finally, under the constraints of total computing resources, the The minimum total delay for all users is the goal, and resources are allocated optimally as needed. Although the second type of method jointly optimizes communication resources, computing resources and offloading strategies when multiple users initiate task requests to minimize the execution delay of user tasks, the second type of method still does not consider the downlink. The impact of limited road resources on the total task delay.

The third method is to add cloud servers. The cloud server is assumed to be a server with unlimited resources. Users can perform tasks in the local, edge server, and cloud server, and complete the calculation offloading with the goal of minimizing the application processing delay. . However, the disadvantage of the third type of method is that the distance is long, which increases the communication delay of users, and when multiple users make service requests, it is easy to cause network congestion, which further increases the application processing delay. Moreover, the third type of method does not consider The effect of downlink transmission delay on the total task delay.

It can be seen that the existing calculation offloading and resource optimization scheme optimizes the task delay of the mobile terminal under the condition of ignoring the influence of downlink resources. Since the task consists of input data, output data, and CPU cycles required for the input data (affected by computing resources), the task delay of the mobile terminal will be affected by the following aspects: The first is the transmission time of the uplink. After the user enters data, it is very important to allocate the optimal bandwidth resources to the user to meet the user's needs; the second is the task delay, how to allocate the optimal computing resources for a task to meet the user's delay needs is also very important. The third important thing is the transmission delay of the downlink. After the calculation of the input data is completed, when the output data is sent to the user, how to allocate the optimal bandwidth resources to the user to meet the transmission requirements of the output data greatly affects the user experience.

In addition to the above three main methods, there are some other methods, which cannot be exhaustive. For example, the random offloading algorithm (ROC) randomly assigns offloading decisions to each user when a user has a task request, thereby randomly allowing some users' tasks to be executed locally, while other users' tasks are offloaded to the edge server for execution.

All the above methods assume that the output data is small, thus ignoring the downlink transmission delay. However, in the actual buffer communication application scenarios, the downlink data in some cases is large, such as AR, VR and remote monitoring, the downlink data will seriously affect the user delay, so it cannot be ignored, especially in some special application scenarios , there is a great demand for low user latency. For example: in driverless technology, the user needs to monitor the vehicle in real time through the mobile terminal. However, during this process, the monitoring information of the vehicle will be calculated and processed by the edge server and the calculation result will be returned to the user's mobile terminal through the downlink. Therefore, in this application scenario, the optimization of downlink resources is very important to reduce the task delay.

To sum up, the applicable scenarios of the existing solutions are limited. In some scenarios, the size of downlink data is too large and cannot be ignored due to technological development, and it cannot efficiently meet the low latency requirements of users in different scenarios. Therefore, it is necessary to improve the existing technology.

SUMMARY OF THE INVENTION

Therefore, the purpose of the present invention is to overcome the above-mentioned defects of the prior art, and to provide a mobile edge computing device for mobile edge computing that comprehensively considers the influence of the uplink transmission delay, calculation delay, and downlink transmission delay of executing user tasks on the total task delay. A management method and an edge server.

According to an aspect of the present invention, the present invention provides a management method for mobile edge computing, which is used for edge-assisted computing of a system composed of a base station, a mobile terminal and an edge server within the coverage of the base station, and system resource optimization before the auxiliary computing. Allocate, for each base station coverage area, perform the following steps:

S1, in response to the task requests of all mobile terminals at the current moment, initialize the unloading decision of each mobile terminal, and randomly set the initial unloading decision of each mobile terminal as the first decision or the second decision;

S2. Based on the mobile terminal information, the current resources of the system, and the initial unloading decision set in step S1, the tasks of calculating all mobile terminals when executed according to the initial decision include the uplink transmission delay, the calculation delay, and the downlink transmission delay at The total delay in

S3. Calculate in turn the total delay when the unloading decision of each mobile terminal is adjusted to a decision opposite to the current unloading decision, wherein, if a certain mobile terminal adjusts the current unloading decision, the total delay will be reduced, then adjust the mobile terminal. The uninstallation decision of the terminal, otherwise it will not be adjusted;

S4. After one round of calculation for each mobile terminal, it is determined that no mobile terminal has adjusted its unloading decision during the round of adjustment, and the adjustment process is ended to obtain the final unloading decision and final system resource allocation scheme of each mobile terminal;

S5, independently allocate the current uplink and downlink communication resources and the computing resources of the edge server this time for assisting the calculation of the task required by the mobile terminal for the final unloading decision according to the system resource allocation scheme;

S6. Execute the task request of each mobile terminal according to the final uninstallation decision of each mobile terminal; and/or

S7, enter the next task cycle after completing the task requests of all mobile terminals at the current moment, and re-execute steps S1-S7;

The first decision indicates that the mobile terminal offloads the task corresponding to its task request to the edge server for calculation, the second decision indicates that the mobile terminal places the task corresponding to its task request for local computing, and the uplink or downlink communication resources are calculated according to the The bandwidth resources are allocated in percentage to the mobile terminal whose final unloading decision is the first decision, and the computing resources of the edge server are allocated in percentage to the mobile terminal whose final unloading decision is the first decision. The percentage of resources occupied by a mobile terminal whose final unloading decision is the first decision is the percentage ratio of its task request demand to the total task request demand corresponding to all mobile terminals whose final unloading decision is the first decision.

Wherein, the step S3 includes:

S31. Calculate in turn the total delay when the unloading decision of each mobile terminal is adjusted to a decision opposite to the current unloading decision, and the total delay is adapted to the unloading decisions of all mobile terminals after being adjusted based on the current mobile terminal each time System resource allocation plan is calculated; and/or

S32. If a mobile terminal adjusts the current offloading decision so that the total time delay is reduced, adjust the offloading decision and resource allocation scheme of the mobile terminal, otherwise, do not adjust.

Preferably, the downlink transmission delay is the ratio of the size of the calculation result data to the downlink transmission rate. When calculating the downlink transmission delay, the size of the corresponding calculation result data can be provided according to one of the following methods: In the case that the edge server has calculated the same task in the early stage, according to the historical data of the edge server, provide the data obtained in the early stage. The size of the calculation result data is used as the size of the calculation result data of the corresponding task; if the edge server has not calculated the same but the same type of task in the early stage, a reference range is formed according to the historical data of the edge server and random within the reference range Provide the first random value as the size of the calculation result of the corresponding task; and if the edge server has neither calculated the same nor the same type of task in the early stage, the edge server randomly provides a size smaller than or equal to the size of the task. The second random value is used as the size of the corresponding calculation result.

According to another aspect of the present invention, the present invention provides an edge server for providing services including at least auxiliary computing for mobile terminals in an area covered by a base station associated therewith, the edge server comprising: one or more processors and a memory, wherein the memory is used to store executable instructions; the one or more processors are configured to perform the aforementioned method by executing the executable instructions.

Compared with the prior art, the advantages of the present invention are: the present invention fully considers the influence of the uplink transmission delay, calculation delay, and downlink transmission delay of executing all user tasks on the total task delay, and before the edge server provides auxiliary computing, Considering the influence of uplink transmission delay, calculation delay and downlink transmission delay on the total delay, the current limited resources of the system are optimally allocated for each mobile terminal to reduce the total delay of executing all user tasks. Especially in the current situation where the size of downlink data in some scenarios is too large to be ignored due to technological development, the present invention can minimize the total delay of all tasks corresponding to the task execution request, and for the case where the size of downlink data is small , the present invention can also minimize the total delay of all tasks corresponding to the task request, and can efficiently meet the user's low-latency requirements in different scenarios, thereby improving user experience.

Description of drawings

The embodiments of the present invention will be further described below with reference to the accompanying drawings, wherein:

1 is a schematic diagram of a conventional architecture of a base station system according to an embodiment of the present invention;

2 is a schematic diagram illustrating a comparison of total delays corresponding to a management method for mobile edge computing and three existing methods under different numbers of mobile terminals according to an embodiment of the present invention;

3 is a schematic diagram illustrating a comparison of total delays corresponding to a management method for mobile edge computing and three existing methods under different input data sizes according to an embodiment of the present invention;

FIG. 4 is a schematic diagram showing a comparison of total delays corresponding to a management method for mobile edge computing according to an embodiment of the present invention and three existing methods under different data sizes of calculation results.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings through specific embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

First, the background of the present invention is described.

When the inventor is conducting research on reducing user delay, by studying the factors affecting mobile user delay in the existing mobile edge computing network and related solutions, it is found that among the delay factors affecting mobile users, the transmission delay of downlink data is very important and cannot be ignored. Therefore, it is necessary to optimize the task delay of the mobile terminal by integrating the uplink data transmission and computing resources on the basis of considering the downlink data transmission. In the optimization process, due to the different computing capabilities of different users in the area covered by a single base station, the tasks performed by different users also have different characteristics, that is, the input data size, output data size and the number of CPU cycles required for the task are different, that is The uplink network bandwidth resources, downlink network bandwidth resources, and edge server computing resources required by different user task requests are different. In addition, in a single base station multi-user scenario, the computing power of the edge server and the spectrum resources of the base station are limited. Therefore, it is an important issue how to make the mobile terminals corresponding to all users obtain the lowest total delay under the request of multiple users. The solution proposed in the present application can significantly reduce the task delay of the mobile user in the scenario of considering the uplink and downlink communication resources, thereby improving the user service experience.

Secondly, some terms used in the present invention are defined as follows:

An edge server refers to a server deployed at the edge of the network, such as near a base station, to provide services to corresponding users. In this application, a service includes at least an auxiliary computing service, and a user may refer to a mobile terminal.

CPU cycles refer to the number of times an instruction is executed per second, and the corresponding English name is CPU cycles.

FIG. 1 shows a schematic diagram of a conventional architecture of a base station system according to an embodiment of the present invention. The system includes a base station, an edge server, and a mobile terminal within the coverage of the base station. The edge server is deployed at the base station at the edge of the network, and provides services for mobile terminals within the coverage area of the base station. In this application, each mobile terminal corresponds to a user of the base station. The base station is responsible for providing communication services for mobile terminals, and the edge server is responsible for providing computing services for mobile terminals. The mobile terminal will issue a task request to seek the edge server to provide auxiliary computing services. However, in a system composed of base stations, edge servers and mobile terminals, the current resources of the system are limited, and it is impossible to satisfy the task request of each mobile terminal. Therefore, before the edge server provides auxiliary computing, how to optimally allocate system resources for each mobile terminal needs to comprehensively consider the uplink transmission delay, calculation delay and downlink transmission delay. In the current situation that the size of downlink data in some scenarios is too large to be ignored due to technological development, the total delay of all tasks corresponding to the task execution request can be minimized, which can efficiently meet the user's low-latency requirements in different scenarios. Improve user experience.

According to an embodiment of the present invention, the present invention provides a management method for mobile edge computing, which is used for edge-assisted computing of a system composed of a base station, a mobile terminal within the coverage of the base station, and an edge server, and system resource optimization before the auxiliary calculation. Allocate, for each base station coverage area, perform the following main steps:

S1, in response to the task requests of all mobile terminals at the current moment, initialize the unloading decision of each mobile terminal, and randomly set the initial unloading decision of each mobile terminal as the first decision or the second decision;

S2. Based on the mobile terminal information, the current resources of the system, and the initial unloading decision set in step S1, the tasks of calculating all mobile terminals when executed according to the initial decision include the uplink transmission delay, the calculation delay, and the downlink transmission delay at The total delay within;

S3. Calculate in turn the total delay when the unloading decision of each mobile terminal is adjusted to a decision opposite to the current unloading decision, wherein, if a certain mobile terminal adjusts the current unloading decision, the total delay will be reduced, then adjust the mobile terminal. The uninstallation decision of the terminal, otherwise it will not be adjusted;

S4. After one round of calculation for each mobile terminal, it is determined that no mobile terminal has adjusted its unloading decision during the round of adjustment, and the adjustment process is ended to obtain the final unloading decision and final system resource allocation scheme of each mobile terminal;

S5, independently allocate the current uplink and downlink communication resources and the computing resources of the edge server this time for assisting the calculation of the task required by the mobile terminal for the final unloading decision according to the system resource allocation scheme;

S6. Execute the task request of each mobile terminal according to the final uninstallation decision of each mobile terminal; and/or

S7, enter the next task cycle after completing the task requests of all mobile terminals at the current moment, and re-execute steps S1-S7;

The first decision indicates that the mobile terminal offloads the task corresponding to its task request to the edge server for calculation, the second decision indicates that the mobile terminal places the task corresponding to its task request for local computing, and the uplink or downlink communication resources are calculated according to the The bandwidth resources are allocated in percentage to the mobile terminal whose final unloading decision is the first decision, and the computing resources of the edge server are allocated in percentage to the mobile terminal whose final unloading decision is the first decision. The percentage of resources occupied by a mobile terminal whose final unloading decision is the first decision is the percentage ratio of its task request demand to the total task request demand corresponding to all mobile terminals whose final unloading decision is the first decision.

Preferably, the method may further include: each time before step S1 is performed, acquiring the mobile terminal information of all mobile terminals in the coverage area of the base station and refreshing the user information table recording the mobile terminal information, so that each time based on the updated information in the user information table The mobile terminal information calculates the total delay this time. The mobile terminal information includes location information, computing capability and requested specific task information of the mobile terminal. The mobile terminal information may also include the transmit power of the mobile terminal.

Preferably, for step S1, the edge server may complete a round of task requests of users within the coverage area and perform auxiliary calculation on the received task unloaded by the mobile terminal, and then send the calculation result to the corresponding mobile terminal before starting. Respond to task requests of all mobile terminals at the current moment in a new round.

For better understanding of the present invention, each step is described in detail below with reference to specific embodiments.

S1. In response to the task requests of all mobile terminals at the current moment, initialize the uninstallation decision of each mobile terminal, and randomly set the initial uninstallation decision of each mobile terminal as the first decision or the second decision. For example, assuming that there are three user mobile terminals in the coverage area of the current base station that have sent task requests, at this time, the task requests of all mobile terminals at the current moment are the task requests sent by these three users, and the edge server responds to all mobile terminals at the current moment. For task request, the initial uninstallation decision corresponding to the three users is randomly set to 0 or 1. For example, the initial uninstallation decision corresponding to the three users is randomly set to 0, 1, and 1 respectively, where 1 represents the first decision and 0 represents the second decision.

S2. Based on the mobile terminal information, the current resources of the system, and the initial unloading decision set in step S1, the tasks of calculating all mobile terminals when executed according to the initial decision include the uplink transmission delay, the calculation delay, and the downlink transmission delay at total delay in . After the initial uninstallation decision of each user is randomly set, system resources will be allocated correspondingly according to the user's initial uninstallation decision to form an initial system resource allocation scheme corresponding to the initial uninstallation decision, so as to calculate the The corresponding total delay derived from the initial system resource allocation scheme. According to an embodiment of the present invention, it is assumed that each mobile terminal only requests to execute one task w _i ={s _i ,d _i , _ci }, where s _i represents the size of the input data corresponding to the task, and d _i represents the calculation result data The size of _ci represents the total amount of CPU cycles required for a task. The set of uninstallation decisions of all mobile terminals is an uninstallation policy.

其中，

表示上行传输数据的速率，可以表示为：Therefore, the uplink transmission delay can be expressed as

in,

Indicates the rate of uplink data transmission, which can be expressed as:

Among them, _ki represents the percentage of uplink bandwidth occupied by mobile terminal i uploading its tasks, P _i,s represents the transmit power of mobile terminal i, hi _,B represents the channel fading coefficient from mobile terminal i to the base station, and d represents mobile terminal i The distance from the base station, r represents the path loss, and σ ² represents the noise power of the channel.

其中，

表示下行传输数据的速率，可以表示为：The downlink transmission delay can be expressed as

in,

Represents the rate of downlink data transmission, which can be expressed as:

Among them, ξ _i represents the bandwidth percentage occupied by mobile terminal i when receiving downlink data, h _B,i represents the channel fading coefficient from the base station to the mobile terminal i, and P _B represents the transmit power of the base station.

There are two types of computing delays, one is the computing delay when the task is executed on the edge server, and the other is the computing delay when the task is executed locally.

The computing delay of tasks executed on the edge server can be expressed as: f _i,m represents the computing resources allocated by the edge server to the mobile terminal i.

The computational delay of the task executing locally can be expressed as: f _i,l represents the local computing capability of the mobile terminal i.

Therefore, the delay caused by offloading the task of the mobile terminal i to the edge server for execution can be expressed as: U _i,m =t _i,u +t _i,d +t _i,m . The time delay caused by the local execution of the task of the mobile terminal i can be expressed as: U _i,l =t _i,l . It can be seen that the delay of offloading the task from the mobile terminal to the edge server includes the uplink transmission delay, the calculation delay and the downlink transmission delay. The delay of the mobile terminal executing the task locally includes the calculation delay. Since there is no process of uploading the task and receiving the calculation result data, the delay of the mobile terminal executing the task locally has no uplink transmission delay and downlink transmission delay.

Preferably, for step S2, the factors affecting the total delay can be formally expressed by establishing a basic model of the delay problem, and the calculation method of the total delay can be limited. In other words, the basic model of the delay problem can be used to formally express the factors affecting the total delay based on the current resources of the system and the information of the mobile terminal. The time delay problem model proposed by the embodiment of the present invention is a model for multi-user task requests in an area covered by a single base station. In the basic model of the delay problem, the computing power of each mobile terminal can be different, each task can be different, and the bandwidth resources can be designed according to percentages; in the basic model of the delay problem, each mobile terminal can request one or more at a time. multiple tasks. Edge servers have concurrent processing capabilities, that is, they can process multiple tasks at the same time. Therefore, in the initial stage, the base station needs to collect the mobile terminal information of the users in the system. The mobile terminal information may include the computing capability of each mobile terminal, location information relative to the base station, and/or the transmit power of the mobile terminal. These mobile terminal information can be recorded in the user information table and refreshed periodically.

Preferably, the basic model of the delay problem can be expressed as:

f_i,m表示边缘服务器为移动终端i分配的计算资源,f_m表示边缘服务器可提供的总的计算能力，

k_i表示移动终端i上传其任务所占上行带宽百分比，

ξ_i表示移动终端i接收下行数据时所占的带宽百分比，a表示具体的一组卸载决策，a＝{a₁,a₂,...a_n}，a_n中的n是当前实际的用户总数，即实际的移动终端的总个数，U_i,m表示移动终端i将其任务卸载至边缘服务器计算产生的总时延，U_i,l表示移动终端i将任务放置在本地计算产生的计算时延，限制条件C1表示分配给至少两个移动终端的计算资源的和要小于等于服务器的计算能力，限制条件C2表示为分配给至少两个移动终端的上行频谱资源总和要小于等于系统的总上行带宽，限制条件C3表示为分配给至少两个移动终端的下行频谱资源总和要小于等于系统的总下行带宽，限制条件C4表示移动终端i的本地计算资源是非负的，限制条件C5表示为移动终端i的卸载决策是第一决策或者第二决策。Among them, A represents the uninstall strategy, A={a _i } _i∈L , a _i represents the uninstall decision of the mobile terminal i, a _i ∈{0,1}, when a _i =1, it represents the uninstall decision of the mobile terminal i For the first decision, mobile terminal i unloads the task corresponding to its task request to the edge server for calculation; when a _i = 0, it means that the unloading decision of mobile terminal i is the second decision, and mobile terminal i corresponds to its task request. The tasks are placed in local computing, L represents the set of all mobile terminals, L={i:i=1,2,...N},

f _i,m represents the computing resources allocated by the edge server to the mobile terminal i, f _m represents the total computing power that the edge server can provide,

k _i represents the percentage of uplink bandwidth occupied by mobile terminal i uploading its tasks,

ξ _i represents the bandwidth percentage occupied by mobile terminal i when receiving downlink data, a represents a specific set of offloading decisions, a={a ₁ , a ₂ ,...a _n }, _n in an is the current actual The total number of users, that is, the actual total number of mobile terminals, U _i,m represents the total delay caused by the mobile terminal i offloading its tasks to the edge server for computing, U _i,l represents the mobile terminal i places the tasks on the local computing generation. The calculation delay, the restriction condition C1 indicates that the sum of the computing resources allocated to at least two mobile terminals should be less than or equal to the computing capability of the server, and the restriction condition C2 means that the sum of the uplink spectrum resources allocated to the at least two mobile terminals should be less than or equal to the system Constraint C3 means that the sum of downlink spectrum resources allocated to at least two mobile terminals is less than or equal to the total downlink bandwidth of the system, Constraint C4 means that the local computing resource of mobile terminal i is non-negative, Constraint C5 means The offloading decision for mobile terminal i is the first decision or the second decision.

S3. Calculate in turn the total delay when the unloading decision of each mobile terminal is adjusted to a decision opposite to the current unloading decision, wherein, if a certain mobile terminal adjusts the current unloading decision, the total delay will be reduced, then adjust the mobile terminal. The uninstallation decision of the terminal, otherwise it will not be adjusted;

S4. After one round of calculation for each mobile terminal, it is determined that no mobile terminal adjusts its unloading decision during this round of adjustment, and the adjustment process is ended to obtain the final unloading decision and final system resource allocation scheme of each mobile terminal. .

According to an embodiment of the present invention, in step S3, in the adjustment process of calculating the total delay when the unloading decision of each mobile terminal is adjusted to a decision opposite to the current unloading decision in turn, the number of unloading decisions adjusted each time may be one or more. For example, calculating in turn the total time delay when the unloading decision of each mobile terminal is adjusted to a decision opposite to the current unloading decision may be adjusted each time only the unloading decision of a single mobile terminal and the unloading decisions of other mobile terminals are not adjusted this time. adjustment, so as to calculate the total delay when the unloading decision of the mobile terminal is adjusted to a decision opposite to the current unloading decision. Specifically, if the current unloading decision of a mobile terminal is adjusted to 1, it is adjusted to 0, and the unloading decisions of other mobile terminals are not adjusted this time, so as to calculate the adjusted total delay. Or, the total time delay when the unloading decision of each mobile terminal is adjusted to a decision opposite to the current unloading decision is calculated in turn. No adjustment is made, so as to calculate the total delay when the unloading decision of the two mobile terminals is adjusted to a decision opposite to the current unloading decision. Specifically, if it is currently adjusted to some two mobile terminals, and the current uninstallation decisions of these two mobile terminals are 1 and 0, respectively, they are adjusted to 0 and 1 respectively, and the uninstallation decisions of the remaining mobile terminals are not adjusted this time, so that This calculates the adjusted total delay.

According to an embodiment of the present invention, for steps S3 to S4, the final unloading decision and system resource allocation scheme can be analyzed and determined by constructing a delay problem solving model. The delay problem solving model proposed in this embodiment adopts a game model , but solving this problem is not limited to game models. The reason for constructing the model for solving the delay problem is that the calculation time for solving the basic model of the above-mentioned delay problem may be long or it is difficult to obtain a direct solution. Build time-delay problem solving models to reduce solution time. The delay problem solving model can be used to analyze and obtain the final offloading strategy and the final resource allocation scheme that can minimize the total delay required to complete the tasks corresponding to all task requests. It should be understood that the final uninstall policy is a set of final uninstall decisions of each mobile terminal.

Preferably, the delay problem solving model can be expressed as:

G={L, (A _i ) _i∈L , u _i };

表示至少两个移动终端中除了移动终端i以外的其他移动终端j的卸载决策。Among them, L represents the set of all mobile terminals, A _i represents the unloading strategy of mobile terminal i, _ui represents the delay utility function of mobile terminal i, and the delay of mobile terminal i is expressed as

represents the uninstallation decision of the other mobile terminal j except the mobile terminal i among the at least two mobile terminals.

According to an embodiment of the present invention, the method of the present application can allow each mobile terminal to obtain a final resource allocation scheme including an optimal solution through a game through a time delay problem solving model, which is expressed as:

Optimal solution of computing resources allocated by edge server to mobile terminal i

The optimal solution for the percentage of uplink bandwidth that mobile terminal i uploads its tasks

The optimal solution for the percentage of bandwidth occupied by mobile terminal i receiving downlink data

B表示总频谱带宽，h_i,B表示从移动终端i到基站的信道衰落系数，P_i,s表示为移动终端i的发送功率，d表示移动终端i与基站的距离，r表示路径损耗，σ²表示信道的噪声功率，h_B,i表示从基站到移动终端i的信道衰落系数，P_B表示基站的发射功率，μ、λ和θ分别为第一、第二和第三优化因子。第一优化因子μ、第二优化因子λ和第三优化因子θ在迭代调整的过程中每次相应移动终端的卸载决策发生调整时则对应地进行优化更新，优化更新所采用的公式分别为：in,

B represents the total spectrum bandwidth, hi _,B represents the channel fading coefficient from the mobile terminal i to the base station, Pi _,s represents the transmit power of the mobile terminal i, d represents the distance between the mobile terminal i and the base station, r represents the path loss, σ ² represents the noise power of the channel, h _B,i represents the channel fading coefficient from the base station to the mobile terminal i, P _B represents the transmit power of the base station, μ, λ and θ are the first, second and third optimization factors, respectively. In the iterative adjustment process, the first optimization factor μ, the second optimization factor λ and the third optimization factor θ are optimized and updated accordingly every time the unloading decision of the corresponding mobile terminal is adjusted, and the formulas used for the optimization and update are:

为迭代步长，迭代步长的取值范围可以为10^-5～10^-7，t为迭代次数且t≥0，μ(t+1)、λ(t+1)和θ(t+1)分别为μ、λ和θ在本次调整后的数值，μ(t)、λ(t)和λ(t)分别为μ、λ和θ在本次调整前的数值。在一种情况下，虽然μ(t)-μ(t+1)≤10^-6、λ(t)-λ(t+1)≤10^-6和θ(t)-θ(t+1)≤10^-6的情况并未完全满足，但是在调整过程中，在从某一个移动终端开始经历一轮调整过程再次回到该移动终端的整个过程中，均没有任何一个移动终端在该轮调整过程中调整其卸载决策，则结束调整过程,得到最小化总时延对应的每个移动终端的最终的卸载决策及最终的系统资源分配方案。在另一种情况下，某次相应移动终端的卸载决策发生调整且在本次调整后μ(t)-μ(t+1)≤10^-6、λ(t)-λ(t+1)≤10^-6和θ(t)-θ(t+1)≤10^-6的情况下确认得到最优解。优选的，迭代次数可以是预设的。比如，迭代次数可以是由人为设定的或者由边缘服务器根据经验设定和调整的。迭代次数比如可以设定为10000次～20000次中的某个值。迭代次数是指迭代的最大次数。实际情况下，可能还未达到最大次数已经得到了最终的卸载决策及最终的系统资源分配方案。比如，迭代次数设置为15000次，实际调整过程中，可能8000次时已经出现了上述两种情况之一，则调整过程结束。in,

is the iteration step size, the value range of the iteration step size can be 10 ^-5 ~ 10 ^-7 , t is the number of iterations and t≥0, μ(t+1), λ(t+1) and θ(t+1 ) are the values of μ, λ and θ after this adjustment, respectively, and μ(t), λ(t) and λ(t) are the values of μ, λ and θ before this adjustment, respectively. In one case, although μ(t)-μ(t+1)≤10 ^-6 , λ(t)-λ(t+1)≤10 ^-6 and θ(t)-θ(t+1) The condition of ≤10 ^-6 is not completely satisfied, but during the adjustment process, during the whole process of going through a round of adjustment process from a certain mobile terminal and returning to the mobile terminal, none of the mobile terminals are in this round of adjustment. If the unloading decision is adjusted during the process, the adjustment process is ended, and the final unloading decision and the final system resource allocation scheme of each mobile terminal corresponding to the minimized total delay are obtained. In another case, the unloading decision of a corresponding mobile terminal is adjusted and after this adjustment μ(t)-μ(t+1)≤10 ^-6 , λ(t)-λ(t+1) In the case of ≤10 ^-6 and θ(t)-θ(t+1)≤10 ^-6 , it was confirmed that the optimal solution was obtained. Preferably, the number of iterations may be preset. For example, the number of iterations can be set manually or set and adjusted by the edge server based on experience. For example, the number of iterations can be set to a certain value from 10,000 times to 20,000 times. The number of iterations refers to the maximum number of iterations. In actual situations, the final unloading decision and the final system resource allocation scheme may not have been reached yet. For example, if the number of iterations is set to 15,000 times, in the actual adjustment process, one of the above two situations may have occurred at 8,000 times, and the adjustment process ends.

It can be seen from the above that the system resource allocation scheme includes the first, second and third optimal solutions. The first optimal solution is the computing resource allocated for the corresponding mobile terminal. The second optimal solution is the percentage of uplink bandwidth occupied by the uploading task assigned to the corresponding mobile terminal. The third optimal solution is the bandwidth percentage allocated for the corresponding mobile terminal when receiving downlink data. The first, second and third optimal solutions correspond to the first, second and third optimization factors, respectively. Preferably, in order to prevent the iteration time from being too long, according to an embodiment of the present invention, the method further includes: in the process of executing steps S3 to S4, if one of the following conditions occurs, the adjustment process is ended: T1, used for optimization The reduction values of the first, second and third optimization factors of the system resource allocation scheme are all smaller than the set threshold; and T2, the number of actual iterative adjustments reaches the set maximum number of iterations. In this case, after the adjustment process is finished this time, the unloading decision and system resource allocation scheme of each mobile terminal at the end of the adjustment process are taken as the final unloading decision and the final system resource allocation scheme of each mobile terminal. The reduction values of the first, second and third optimization factors are defined by μ(t)-μ(t+1), λ(t)-λ(t+1) and θ(t)-θ(t +1) to the calculated value. After obtaining the final uninstall policy, the final uninstall policy can be written into the user information table. In this case, although the optimal unloading decision and system resource allocation scheme are not obtained, it is avoided that the long-term solution causes the solution time to be too long and affects the user experience.

According to the above analysis, a final system resource allocation scheme is obtained to allocate system resources to each mobile terminal, that is, uplink communication resources, computing resources and downlink communication resources. Subsequently, the base station and the edge server receive the task unloaded by the mobile terminal whose final unloading decision is the first decision under the condition of the currently allocated uplink communication resources and complete the auxiliary calculation of the received task, and then use the currently allocated downlink communication resources Under certain conditions, the calculation result corresponding to the task is transmitted to the corresponding mobile terminal. Therefore, in this way, the total time delay of executing all tasks is minimized, and the user experience is improved.

Preferably, in step S4, the requirement for ending the adjustment process is: in a round of adjustment process from a certain mobile terminal, no mobile terminal among all mobile terminals adjusts its unloading decision during this round of adjustment process. For example, assuming that a total of 5 mobile terminals have issued task requests, when the initial uninstallation decisions of the five mobile terminals are randomly set as the first decision or the second decision, the initial uninstallation decisions of the five mobile terminals are 1, 0, 1, 0, 0, after multiple adjustments, the unloading decisions of the five mobile terminals become 0, 0, 1, 1, 0 respectively. After the previous adjustment, in the next round, from the first mobile terminal to the fifth mobile terminal After the end of the terminal, its unloading decision has not been adjusted. After the round of adjustment, the unloading decisions of the five mobile terminals are still 0, 0, 1, 1, and 0, and the adjustment process is ended.

In actual scenarios, before the task is calculated to obtain the calculation result data, the size of the result data is difficult to predict. Therefore, preferably, according to an embodiment of the present invention, the method of the present invention may further include: for a calculated task, the information of the task and the size of the corresponding calculation result data are associated with each other, so that the corresponding task is based on historical data. The size of the calculation result data provides a reference.

According to an embodiment of the present invention, in order to use the size of the calculation result data to perform downlink communication resource allocation in advance, the size of the corresponding calculation result data can be provided according to one of the following methods:

The first way: In the case that the edge server has calculated the same task in the early stage, the size of the calculation result data obtained in the previous stage is provided according to the historical data of the edge server as the size of the calculation result data of the corresponding task; for example, a task edge The server has been calculated in the early stage, and the edge server retains historical data. The historical data records the information of the task and the size of the corresponding calculation result data, so that the size of the calculation result data obtained in the previous stage can be used as the calculation of the corresponding task. The size of the resulting data. The same task may refer to the same specific computing task, and correspondingly, the corresponding computing result data is also the same.

The second method: In the case that the edge server has not calculated the same task but has calculated the same type of task in the early stage, a reference range is formed according to the historical data of the edge server, and the first random value is randomly provided within the reference range as the calculation of the corresponding task The size of the result; for example, for a task, which is a certain data format in the AR field, the edge server has not calculated the same task in the previous stage, but has calculated a task in the same type of data format in the previous stage, and the historical data indicates that the data The ratio of the size of the calculation result data in the format to the size of the task is 20%, then the size of the task is multiplied by 10% to 30% to form a reference range. The ratio of the size of the calculation result data in this data format to the size of the task is 20% and 40% respectively, then multiply the size of the task by 20% to 40% to form a reference range. In the three calculations, the ratio of the size of the calculation result data in this data format to the size of the task is 10%, 20%, and 40% respectively, then multiply the size of the task by 10% to 40% to form a reference range, and then use the The random value within the reference range is used as the size of the calculation result of the corresponding task. A task of the same type may refer to a task with the same data format.

The third way: when the edge server has neither calculated the same nor the same type of task in the early stage, the edge server randomly provides a second random value less than or equal to the size of the task as the corresponding calculation result. size.

Preferably, the mobile terminal may include at least one of a mobile phone, a tablet computer, a notebook computer, a vehicle-mounted computer, VR glasses, AR glasses and a smart watch. Preferably, the on-board computer can be, for example, a trip computer of an automobile.

According to an example of the present invention, the proposed method of the present invention is simulated and verified through the following simulation parameters. The simulation parameters used in the simulation verification process are given in Table 1.

Table 1 Simulation parameters

In Figs. 2, 3 and 4, the CORAG curve corresponds to the method adopted in the present application, which is an abbreviation of a computation offloading and resource allocation game (CORAG) based on game theory. The LOC curve corresponds to the local offloading completely (LOC) algorithm. In this algorithm, all tasks are not offloaded to the edge server for execution, but are executed locally on each mobile terminal. The ROC curve corresponds to the random offloading completely (ROC) algorithm. The COURG curve corresponds to calculation offloading and uplink spectrum resource allocation (computation offloading and uplink resource game, COURG), and the second method of the background art adopts this algorithm.

FIG. 2 is a schematic diagram showing a method for assisting edge computing according to an embodiment of the present invention and a schematic diagram of the total delay corresponding to each of the three existing methods under different numbers of mobile terminals. It can be seen from FIG. 2 that the performance of the method proposed by the present invention is better than that of the LOC, ROC and COURG algorithms. With the increase of the number of mobile terminals, the advantages of CORAG algorithm are more obvious. This is because under the condition of limited computing and communication resources, the CORAG algorithm can better allocate optimal computing and communication resources for different mobile terminals, so the total delay of all mobile terminals is the lowest. The input parameters of FIG. 2 adopt the parameters listed in Table 1, wherein some of the input parameters are varied. For example, the last four parameters in Table 1 are random values given randomly by the computer within the normal distribution range described above.

FIG. 3 is a schematic diagram showing a management method for mobile edge computing according to an embodiment of the present invention and a comparison diagram of total delay corresponding to each of three existing methods under different input data sizes. As can be seen from Figure 3, with the increase of uplink input data, the total delay of the CORAG algorithm is lower than that of the LOC, ROC, and COURG algorithms. This is because as the input data increases, the CORAG algorithm can allocate the optimal uplink and downlink bandwidth resources and computing resources according to the size of the input data, so as to minimize the total delay of all terminal devices.

FIG. 4 shows a management method for mobile edge computing according to an embodiment of the present invention and a schematic diagram of a comparison of total delays corresponding to each of three existing methods under different data sizes of calculation results. As can be seen from Figure 4, with the gradual increase of the calculation result data, the total delay of the CORAG algorithm is lower than the three algorithms of LOC, ROC and COURG. The performance of the CORAG algorithm is obviously better than the other three algorithms, and with the gradual increase of the calculation result data, the performance advantage of the CORAG algorithm is more obvious.

In yet another embodiment of the present invention, an edge server is also provided, which can be used to provide a service including at least auxiliary computing for mobile terminals in an area covered by a base station associated therewith. The edge server includes: one or more processors; and memory, wherein the memory is used to store executable instructions; the one or more processors are configured to perform any of the preceding embodiments by executing the executable instructions. The technical solutions presented and/or the methods of the present application.

In yet another embodiment of the present invention, there is also provided an electronic device, more specifically, it can be referred to as a communication device, including an edge server and a base station that are communicatively connected to each other. The edge server can provide services including at least auxiliary computing for mobile terminals in the area covered by the base station. The edge server includes: one or more processors; and memory, wherein the memory is used to store executable instructions; the one or more processors are configured to perform the techniques described in any of the preceding embodiments by executing the executable instructions Protocols and/or methods of the present application.

It should be noted that although the steps are described above in a specific order, it does not mean that the steps must be executed in the above-mentioned specific order. In fact, some of these steps can be executed concurrently, or even change the order, as long as it can be achieved The required function can be.

The present invention may be a system, method and/or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for causing a processor to implement various aspects of the present invention.

A computer-readable storage medium may be a tangible device that retains and stores instructions for use by the instruction execution device. Computer-readable storage media may include, but are not limited to, electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing, for example. More specific examples (non-exhaustive list) of computer readable storage media include: portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable programmable read only memory (EPROM) or flash memory), static random access memory (SRAM), portable compact disk read only memory (CD-ROM), digital versatile disk (DVD), memory sticks, floppy disks, mechanically coded devices, such as printers with instructions stored thereon Hole cards or raised structures in grooves, and any suitable combination of the above.

Various embodiments of the present invention have been described above, and the foregoing descriptions are exemplary, not exhaustive, and not limiting of the disclosed embodiments. Numerous modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein was chosen to best explain the principles of the embodiments, the practical application or technical improvement in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.

Claims (10)

1. a management method for mobile edge computing, for the edge auxiliary computing of the system formed by the base station and the mobile terminal in the base station coverage, the edge server and the optimal allocation of system resources before the auxiliary computing, it is characterized in that, for each For a base station coverage area, perform the following steps: S1, in response to the task requests of all mobile terminals at the current moment, initialize the unloading decision of each mobile terminal, and randomly set the initial unloading decision of each mobile terminal as the first decision or the second decision; S2. Based on the mobile terminal information, the current resources of the system, and the initial unloading decision set in step S1, the tasks of calculating all mobile terminals when executed according to the initial decision include the uplink transmission delay, the calculation delay, and the downlink transmission delay at The total delay in S3. Calculate in turn the total delay when the unloading decision of each mobile terminal is adjusted to a decision opposite to the current unloading decision, wherein, if a certain mobile terminal adjusts the current unloading decision, the total delay will be reduced, then adjust the mobile terminal. The uninstallation decision of the terminal, otherwise it will not be adjusted; S4. After one round of calculation for each mobile terminal, it is determined that no mobile terminal adjusts its unloading decision during this round of adjustment, and the adjustment process is ended to obtain the final unloading decision and final system resource allocation scheme of each mobile terminal. 2. A management method for mobile edge computing according to claim 1, wherein the first decision represents that the mobile terminal offloads the task corresponding to its task request to the edge server for calculation, and the second decision Indicates that the mobile terminal places the task corresponding to its task request in the local computing. 3. A kind of management method for mobile edge computing according to claim 1 or 2, is characterized in that, also comprises: S5, independently allocate the current uplink and downlink communication resources and the computing resources of the edge server this time for assisting the calculation of the task required by the mobile terminal for the final unloading decision according to the system resource allocation scheme; S6. Execute the task request of each mobile terminal according to the final uninstallation decision of each mobile terminal. 4. a kind of management method for mobile edge computing according to claim 3, is characterized in that, uplink or downlink communication resource is according to the mobile terminal of the first decision that bandwidth resource is allocated to final offloading decision in percentage form , the computing resources of the edge server are allocated according to the percentage of the computing resources of the edge server to the mobile terminal whose final unloading decision is the first decision, and the percentage of resources occupied by each mobile terminal whose final unloading decision is the first decision is its The percentage of task request demand in the total task request demand corresponding to all mobile terminals whose final unloading decision is the first decision. 5. a kind of management method for mobile edge computing according to claim 3, is characterized in that, the task request that the system completes all mobile terminals at the current moment is a task cycle, and described method also comprises: S7. After completing the task requests of all mobile terminals at the current moment, enter the next task cycle, and perform steps S1-S6 again. 6 . The management method for mobile edge computing according to claim 5 , wherein the method further comprises: acquiring mobile terminal information of all mobile terminals in the coverage area of the base station each time before step S1 is performed. 7 . And refresh the user information table that records the mobile terminal information, and use the updated mobile terminal information each time to calculate the total delay this time; Wherein, the mobile terminal information includes location information, computing capability and specific task information requested of the mobile terminal. 7. The management method for mobile edge computing according to claim 5, wherein the step S3 comprises: S31. Calculate in turn the total delay when the unloading decision of each mobile terminal is adjusted to a decision opposite to the current unloading decision, and the total delay is adapted to the unloading decisions of all mobile terminals after being adjusted based on the current mobile terminal each time The system resource allocation plan is calculated; S32. If a mobile terminal adjusts the current offloading decision so that the total time delay is reduced, adjust the offloading decision and resource allocation scheme of the mobile terminal, otherwise, do not adjust. 8. A management method for mobile edge computing according to claim 1 or 2, wherein the downlink transmission delay is the ratio of the size of the calculation result data to the downlink transmission rate, wherein, when calculating the downlink transmission Delay, according to one of the following methods to provide the size of the corresponding calculation result data: In the case that the edge server has calculated the same task in the early stage, the size of the calculation result data obtained in the previous stage is provided according to the historical data of the edge server as the size of the calculation result data of the corresponding task; In the case that the edge server has not calculated the same task but has calculated the same type of task in the early stage, a reference range is formed according to the historical data of the edge server, and a first random value is randomly provided within the reference range as the size of the calculation result of the corresponding task; and In the case that the edge server has neither calculated the same nor the same type of task in the early stage, the edge server randomly provides a second random value less than or equal to the size of the task as the size of the corresponding calculation result. 9 . A computer-readable storage medium, characterized in that a computer program is contained thereon, and the computer program can be executed by a processor to perform the method of any one of claims 1 to 8 . 10. An edge server, configured to provide a service including at least auxiliary computing for mobile terminals in an area covered by a base station related thereto, wherein the edge server comprises: one or more processors; and memory, wherein the memory is used to store executable instructions; The one or more processors are configured to perform the method of any of claims 1-8 by executing the executable instructions.

Images