An edge computing network edge computing method and related apparatus

Abstract

The application discloses an edge computing network terminal computing method and related devices, and the application adopts a clustering optimization model to obtain an optimal cluster for processing a computing task, and unloads the residual computing task to each cluster point for processing in the case that the base station coverage range is reduced, the corresponding edge server computing capacity is weakened, and the residual computing task exists, thereby effectively solving the problem that the base station cannot complete the node computing request.

Description

Technical Field

[0001] This invention relates to an edge computing method and related apparatus for edge computing networks, belonging to the field of communication technology. Background Technology

[0002] Edge computing (MEC) involves deploying edge servers near base stations or access points close to user terminals (hereinafter referred to as "nodes") to provide users with computing and network resources to fulfill various user service requests. Users transmit data through base stations, offloading computing tasks to edge servers, thus achieving edge computing. With the rise of the Internet of Things (IoT), the interconnection of everything has driven the early development of edge computing technology. Edge computing offers lower latency and lower energy consumption, and 5G technology has further brought unprecedented attention to edge computing globally.

[0003] When faced with special circumstances such as war or natural disasters, base stations in edge computing networks may experience partial functional damage (i.e., abnormal operation risk), which often leads to a reduction in the coverage area of ​​the base station and a weakening of the computing power of the corresponding edge server. In such cases, the base station may be unable to complete the computing requests of the nodes. Summary of the Invention

[0004] This invention provides an edge computing method and related apparatus for edge computing networks, which solves the problems disclosed in the background art.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] An edge computing method for edge computing networks includes:

[0007] If the base station is not operating normally at the current moment, a node within the current coverage area is selected as the cluster head based on the information of the nodes within the current coverage area. Among them, the coverage area of ​​the base station that is not operating normally is smaller, the computing power of the corresponding edge server is weaker, and there are remaining computing tasks.

[0008] The system sends a pre-defined clustering scheme to the cluster head and receives clusters from the cluster head. The cluster head obtains candidate clusters based on the pre-defined clustering scheme and selects the preferred cluster from the candidate clusters based on the pre-defined clustering optimization model.

[0009] Based on the cluster feedback from the cluster head, the remaining computing tasks are allocated. The sub-computing tasks are then offloaded to the corresponding cluster nodes through the cluster head, and the results of the sub-computing tasks are received from the cluster nodes through the cluster head.

[0010] Based on the information of the nodes within the current coverage area, select one node within the current coverage area as the cluster head, including:

[0011] If a fixed node exists within the current coverage area, select the fixed node as the cluster head;

[0012] If there are no fixed nodes within the current coverage area, calculate the first weight of the nodes based on the node information, and select the node with the largest first weight as the cluster head.

[0013] The information of nodes within the current coverage area includes the node's remaining computing power, the node's location, the node's movement speed, and the sum of the remaining computing power of all neighboring nodes; among them, neighboring nodes are selected from the nodes within the normal coverage area of ​​the base station.

[0014] The formula for calculating the first weight of a node is:

[0015] ;

[0016] In the formula, Let the first weight of the i-th node be... Let be the remaining computing power of the i-th node. Let be the distance between the i-th node and the base station at the next moment, calculated based on the current position and movement speed of the i-th node. Let be the stability index of the i-th node. , They are respectively The weights, , K is the number of neighboring nodes of the i-th node. Let be the distance between the i-th node and its k-th neighbor node at the current time. Let be the distance between the i-th node and its k-th neighbor node at the next time step. Let be the mobility index of the i-th node relative to the k-th neighbor node, and the neighbor nodes are selected from the nodes within the normal coverage area of ​​the base station.

[0017] Clustering schemes include:

[0018] Option 1: Select cluster points within a one-hop range centered on the cluster head to construct a cluster; where the one-hop range is the range that can directly communicate with the cluster head; where the cluster points are selected from nodes within the normal coverage area of ​​the base station.

[0019] Option 2: Select cluster points within a two-hop range centered on the cluster head to construct a cluster; where the two-hop range is the range within which a node can communicate with the cluster head through one node.

[0020] The cluster selection process includes:

[0021] If Scheme 1 is adopted, the second weight of the nodes within one hop range is calculated based on the node information, and the nodes with the second weight greater than the threshold are regarded as cluster points;

[0022] If Scheme 2 is adopted, the second weight of the nodes within the two-hop range is calculated based on the node information, and the nodes with the second weight greater than the threshold are taken as cluster points.

[0023] The formula for calculating the second weight is:

[0024] ;

[0025] In the formula, The second weight of the j-th node. The remaining computing power of the j-th node. Let be the stability index of the j-th node. , They are respectively The weights, , N is the number of neighboring nodes of the j-th node. Let be the distance between the j-th node and its n-th neighbor node at the current time. Let be the distance between the j-th node and its n-th neighbor node at the next time step. Let be the mobility index of the j-th node relative to the n-th neighbor node, and the neighbor node is selected from the nodes within the normal coverage area of ​​the base station.

[0026] Clustering optimization models include:

[0027] Objective function:

[0028] ;

[0029] In the formula, The system performance index is defined as a system consisting of a base station and nodes in a cluster. F represents the total remaining computing power of all nodes participating in the computational task. , F and respectively The weights are given by T, where T is the total latency of processing and transmitting computational tasks, and E is the total energy consumption of processing and transmitting computational tasks. , The weights for T and E are respectively. Let m be the channel bandwidth of the m-th node in the system. Let m be the transmit power from the upstream node to the m-th node in the system.

[0030] The constraints are:

[0031] ;

[0032] ;

[0033] ;

[0034] ;

[0035] ;

[0036] In the formula, The upper limit of T, for The upper limit, for The upper limit, where M is the total number of nodes in the system. This represents the upper limit of the channel bandwidth utilized by the system. The remaining computing power of the m-th node. for The upper limit.

[0037] An edge computing device for an edge computing network includes:

[0038] The cluster head selection module selects a node within the current coverage area as the cluster head if the base station is not operating normally at the current time, based on the information of the nodes within the current coverage area. Among them, the coverage area of ​​the base station that is not operating normally is smaller, the computing power of the corresponding edge server is weaker, and there are remaining computing tasks.

[0039] The cluster acquisition module sends a preset clustering scheme to the cluster head and receives clusters from the cluster head; the cluster head obtains candidate clusters according to the preset clustering scheme and selects the preferred clusters from the candidate clusters according to the preset clustering optimization model.

[0040] The task processing module allocates the remaining computing tasks based on the clusters fed back by the cluster head. It then unloads the allocated sub-computing tasks to the corresponding clusters through the cluster head and receives the results of the clusters processing the sub-computing tasks through the cluster head.

[0041] A computer-readable storage medium storing one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform an edge computing method of an edge computing network.

[0042] The beneficial effects achieved by this invention are as follows: When the coverage area of ​​the base station is reduced, the computing power of the corresponding edge server is weakened, and there are remaining computing tasks, this invention adopts a cluster optimization model to obtain the preferred clusters for processing computing tasks, and offloads the remaining computing tasks to each cluster point for processing, which effectively solves the problem that the base station cannot complete the node computing request. Attached Figure Description

[0043] Figure 1 A flowchart illustrating the end-computing method in an edge computing network;

[0044] Figure 2 This is a clustering diagram when the active node is the cluster head.

[0045] Figure 3 This is a schematic diagram when the fixed node is the cluster head. Detailed Implementation

[0046] The present invention will be further described below with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and should not be used to limit the scope of protection of the present invention.

[0047] like Figure 1 As shown, an edge computing method for an edge computing network, implemented on the base station side, includes the following steps:

[0048] Step 1: If the base station is not operating normally at the current time, select a node within the current coverage area as the cluster head based on the information of the nodes within the current coverage area; where the coverage area of ​​the base station is not operating normally is smaller, the computing power of the corresponding edge server is weaker, and there are remaining computing tasks.

[0049] Step 2: Send the preset clustering scheme to the cluster head and receive the clusters fed back by the cluster head; wherein, the cluster head obtains candidate clusters according to the preset clustering scheme, and selects the preferred clusters from the candidate clusters according to the preset clustering optimization model.

[0050] Step 3: Based on the cluster feedback from the cluster head, allocate the remaining computing tasks, unload the allocated sub-computing tasks to the corresponding cluster points through the cluster head, and receive the results of the cluster points processing the sub-computing tasks through the cluster head.

[0051] The above method, when the base station coverage area shrinks, the corresponding edge server computing power weakens, and there are remaining computing tasks, adopts a cluster optimization model to obtain the preferred clusters for processing computing tasks, and offloads the remaining computing tasks to each cluster point for processing, effectively solving the problem that the base station cannot complete the node computing requests.

[0052] When faced with special circumstances such as war or natural disasters, there is a risk of partial functional damage to the base station side in the edge computing network, that is, the risk of abnormal operation of the base station. The computing tasks originally processed by the base station cannot be completed, that is, there are remaining computing tasks. Since each node has remaining computing power, based on computing power sharing, the remaining computing tasks can be offloaded to the nodes for computing.

[0053] Based on the above analysis, we can first select the node clusters to handle the remaining computing tasks, specifically by first selecting the cluster head and then selecting the cluster nodes.

[0054] Within the coverage area of ​​the base station, in most cases all nodes are mobile. In this case, in step 1 above, as an example of the present invention, it is necessary to calculate the first weight of the node based on the node information and take the node with the largest first weight as the cluster head.

[0055] Nodes within the current coverage area of ​​the base station will interact with the base station in real time. Each node sends its basic information to the base station, including its serial number, remaining computing power, location, movement speed, and the sum of the remaining computing power of all its neighboring nodes. Neighboring nodes are selected from those within the base station's normal coverage area. Node information can be expressed by the formula: Where id is the node's serial number, f is the node's remaining computing power, (x, y) is the node's current position (the Cartesian coordinate system in which this position is located is centered at the base station), and v is the node's movement speed. The direction of node movement. It is the sum of the remaining computing power of all neighboring nodes.

[0056] When calculating the first weight, the stability of the node, its remaining computing power, and its distance from the base station need to be considered. Specifically, this can be expressed as:

[0057] Assume the current position of the i-th node is Then the position of this node at the next moment can be represented as: , Let be the velocity of the i-th node. Then, the distance between the node and the base station at the next moment can be expressed as: ;

[0058] The remaining computing power of a node can be directly obtained from the node information; the stability of a node can be considered as its mobility, defined as the distance between the i-th node and its k-th neighbor node at the current time. At the next moment, the distance between the i-th node and its k-th neighbor node is... Then the mobility index of the i-th node relative to the k-th neighbor node , can be represented as:

[0059] ;

[0060] if The larger the value, the further apart the two nodes are; conversely, the smaller the value, the closer they are to each other.

[0061] The mobility of the i-th node is represented by the average of the local values ​​of its relative mobility with all its neighboring nodes, and can be expressed by the formula:

[0062] ;

[0063] In the formula, K is the number of neighboring nodes of the i-th node. The smaller the value, the lower the mobility of the node relative to its neighboring nodes, which means the higher the stability.

[0064] Therefore, the specific formula for the first weight can be:

[0065] ;

[0066] In the formula, Let the first weight of the i-th node be... Let be the remaining computing power of the i-th node. Let be the distance between the i-th node and the base station at the next moment, calculated based on the current position and movement speed of the i-th node. Let be the stability index of the i-th node. , They are respectively The weights.

[0067] Since nodes are mobile, this introduces a lot of uncertainty and requires repeated calculations of the first weight to determine the cluster head. If there is a fixed node within the current coverage area of ​​the base station, this fixed node could be an inactive terminal or a small-scale light pole base station, etc. In order to simplify the method and reduce the calculation, the fixed node can be preferred as the cluster head.

[0068] After the cluster head is determined, the base station sends the preset clustering scheme to the cluster head. The cluster head will obtain candidate clusters according to the preset clustering scheme and select the preferred cluster from the candidate clusters according to the preset clustering optimization model.

[0069] As one embodiment of the present invention, see Figure 2 and 3 This invention provides two clustering schemes:

[0070] Option 1: Select cluster points within a one-hop range centered on the cluster head to construct a cluster; where the one-hop range is the range that can directly communicate with the cluster head; where the cluster points are selected from nodes within the normal coverage area of ​​the base station; and where all cluster points communicate with the cluster head.

[0071] Option 2: Select cluster points within a two-hop range centered on the cluster head to construct a cluster; where the two-hop range is the range that can communicate with the cluster head through a single node; the cluster points in the two hops communicate with the cluster head through the cluster points in the first hop.

[0072] If Scheme 1 is adopted, the second weight of the nodes within one hop range needs to be calculated based on the node information, and the nodes with the second weight greater than the threshold are taken as cluster points. If Scheme 2 is adopted, the second weight of the nodes within two hop range also needs to be calculated based on the node information, and the nodes with the second weight greater than the threshold are taken as cluster points.

[0073] The second weight is calculated based on the node's remaining computing power and stability, and can be expressed by the formula:

[0074] ;

[0075] In the formula, The second weight of the j-th node. The remaining computing power of the j-th node. Let be the stability index of the j-th node. , They are respectively The weights, , N is the number of neighboring nodes of the j-th node. Let be the distance between the j-th node and its n-th neighbor node at the current time. Let be the distance between the j-th node and its n-th neighbor node at the next time step. Let be the mobility index of the j-th node relative to its n-th neighbor node.

[0076] Based on the above scheme, at least two types of clusters can be obtained. In order to obtain the optimal cluster, it is necessary to perform calculations through a cluster optimization model. A system consisting of a base station and nodes in the cluster is defined, and optimization calculations are performed with the goal of optimizing the system performance. This model considers the total remaining computing power of the system, the total latency of processing and transmitting computing tasks, and the total energy consumption of processing and transmitting computing tasks.

[0077] When using Scheme 1, the latency required for the computation task includes transmission latency and computation latency; whereby the computation latency can be expressed as:

[0078] ;

[0079] In the formula, M represents the total number of nodes in the system. Let be the computation delay for the m-th node. Let C be the remaining computing power of the m-th node, C be the CPU cycles required to process 1 bit of data, and m=1 represent the cluster head;

[0080] The delay at which the base station sends the computation task to the cluster head can be expressed as:

[0081] ;

[0082] In the formula, The latency for the base station to send computing tasks to the cluster head. Let L be the data transmission rate from the base station to the cluster head, and L be the total system loss, which is the weighted sum of energy consumption and latency.

[0083] The transmission delay of the cluster head sending the computation task to a one-hop node (cluster node) can be expressed as:

[0084] ;

[0085] In the formula, The transmission delay for the cluster head to send the computation task to the m1-th node;

[0086] Let m be the data transfer rate from the cluster head to the m1-th node. , Let m1 be the channel bandwidth of the m1-th node. For noise power, Let m be the transmit power from the upstream node (i.e., the cluster head) to the m1th node. Let be the channel gain of the m1-th node. The channel gain can be determined by considering the effects of distance-dependent path loss, multipath fading, and shadowing on the time slot. , Let m1 be the distance between the m1-th node and the base station. This refers to the path loss factor.

[0087] Therefore, the total delay T is taken as the maximum value among all concurrent delays. .

[0088] The energy consumption required for a computational task includes transmission energy consumption and computation energy consumption; among which, the node computation energy consumption can be expressed as:

[0089] ;

[0090] In the formula, Let m be the computational energy consumption of the m-th node. For effective switching of capacitors;

[0091] Based on the transmission delay mentioned earlier, the transmission loss from the base station to the cluster head can be defined as:

[0092] ;

[0093] In the formula, For the transmission loss from the base station to the cluster head, The energy consumption for transmitting signals to the current base station.

[0094] Similarly, the transmission loss from the cluster head to a one-hop node (cluster point) can be defined as:

[0095] ;

[0096] In the formula, This is the transmission loss from the cluster head to the first-hop node (cluster point);

[0097] Therefore, the energy consumption required for the computation task can be expressed as: .

[0098] When using Scheme 2, the computational task requires an additional two-hop transmission delay on top of the aforementioned delay (i.e., the delay in Scheme 1). The total delay T is the sum of the transmission delay from the cluster head to the first-hop node and the delay from the first-hop node to the second-hop node; therefore, the total delay T is the maximum value among all concurrent delays. The same energy consumption will increase the energy consumption of the second-hop transmission.

[0099] Based on the above analysis, the clustering optimization model can be expressed as:

[0100] Objective function:

[0101] ;

[0102] In the formula, The system performance index is defined as a system consisting of a base station and nodes in a cluster. F represents the total remaining computing power of all nodes participating in the computational task. , F and respectively The weights are given by T, where T is the total latency of processing and transmitting computational tasks, and E is the total energy consumption of processing and transmitting computational tasks. , The weights for T and E are respectively. Let m be the channel bandwidth of the m-th node in the system. Let m be the transmit power from the upstream node to the m-th node in the system.

[0103] The constraints are:

[0104] 1. Latency constraint: The total latency of the computation task must not exceed the maximum latency, which can be expressed by the formula: ;

[0105] 2. Transmit power constraint, which can be expressed by the formula: ;

[0106] 3. System bandwidth constraints, which can be expressed by the formula: ;

[0107] 4. Channel bandwidth constraint, which can be expressed by the formula:

[0108] 5. The residual computational force constraint can be expressed by the formula:

[0109] In the formula, The upper limit of T, for The upper limit, for The upper limit, This represents the upper limit of the channel bandwidth utilized by the system. for The upper limit.

[0110] The above model maximizes system performance through the rational allocation of computing and channel resources, and takes the remaining computing power as a feature to achieve maximum computing power sharing.

[0111] The optimization of the above model is a mixed-integer nonlinear programming problem. The cluster head side can solve the optimization problem through a constrained optimization objective. First, the initial original population space is generated. The z-th particle is represented as a two-dimensional vector. ( Let represent the optimal solution for the z-th particle. This represents the individual value of the z-th particle. Let represent the resources required by the z-th particle, and represent the resource allocation scheme. Using the system's performance index as the fitness value, the fitness value for each particle can be calculated. The velocity of the z-th particle is... ( It is the evolution rate of a particle in its current state, determined by the states of its two adjacent particles. (formed by coupling), similarly, the individual extreme value of the z-th particle is (i.e., the current individual extreme value of the particle) And the extreme values ​​of the states of the chain particles adjacent to itself. (formed by coupling), the collective population value (Ultimately, it can be achieved through at least two individuals) Value, to obtain the collective population value of this population. ).

[0112] In each iteration, the particle updates its velocity and position through individual extreme values ​​and group extreme values, that is:

[0113] ;

[0114] ;

[0115] In the formula, These are weighting coefficients. , s is the current iteration number, c1 and c2 are non-negative constants called acceleration factors, and r1 and r2 are random numbers distributed in the interval [0,1]. This represents the velocity of the device at iterations s and s+1. This indicates the refresh rate after each iteration. Let represent the individual extreme values ​​of the z-th particle and the g-th particle at iteration number s. This represents the current optimal solution for the z-th particle at iterations s and s+1.

[0116] To prevent particles from blindly searching, it is generally recommended to limit their position and velocity to a certain range. Conditions 2 and 4 in the optimization conditions are conditions that limit the initial position and velocity of each particle, obtaining individual extreme values ​​and group extreme values. Through continuous iteration, particles update their own velocity and position through individual extreme values ​​and group extreme values, and finally obtain the optimal fitness value.

[0117] The cluster head feeds back the selected clusters to the base station, specifically by sending the maintained routing table to the base station. The base station allocates computing tasks based on load balancing and sends the allocated scheme to the cluster head. The cluster head then sends the sub-computation tasks in the scheme to the corresponding cluster points. The cluster head receives the results of the cluster points processing the sub-computation tasks and feeds the processing results back to the base station, thus realizing end-to-end computing in the edge computing network.

[0118] This invention proposes an end-to-end computing scheme that can utilize the computing power of a node itself to complete computing tasks under special circumstances, thereby achieving computing power sharing.

[0119] Based on the same technical solution, this invention also discloses a virtual device for the above method, such as software. This virtual device is mainly installed on the base station side, specifically an edge computing device for an edge computing network, comprising:

[0120] The cluster head selection module selects a node within the current coverage area as the cluster head if the base station is not operating normally at the current time, based on the information of the nodes within the current coverage area. Among them, the coverage area of ​​the base station that is not operating normally is smaller, the computing power of the corresponding edge server is weaker, and there are remaining computing tasks.

[0121] The cluster acquisition module sends a preset clustering scheme to the cluster head and receives clusters from the cluster head. The cluster head obtains candidate clusters according to the preset clustering scheme and selects the preferred clusters from the candidate clusters according to the preset clustering optimization model.

[0122] The task processing module allocates the remaining computing tasks based on the clusters fed back by the cluster head. It then unloads the allocated sub-computing tasks to the corresponding clusters through the cluster head and receives the results of the clusters processing the sub-computing tasks through the cluster head.

[0123] The data processing flow and methods of each module in the above device are consistent, and will not be described again here.

[0124] Based on the same technical solution, the present invention also discloses a computer-readable storage medium that stores one or more programs, the one or more programs including instructions that, when executed by a computing device, cause the computing device to perform an edge computing method of an edge computing network.

[0125] Based on the same technical solution, the present invention also discloses a computer device, including one or more processors and one or more memories, wherein one or more programs are stored in the one or more memories and configured to be executed by the one or more processors, and the one or more programs include instructions for performing edge computing of an edge computing network.

[0126] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0127] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0128] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0129] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0130] The above are merely embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of the claims of the present invention pending approval.

Claims

1. An edge computing network end computing method, characterized in that, The method comprises the following steps: If the base station is not normally running at the current time and there is a fixed node in the current coverage range, the fixed node is selected as the cluster head; if the base station is not normally running at the current time and there is no fixed node in the current coverage range, the first weight of the node is calculated according to the information of the node, and the node with the largest first weight is selected as the cluster head; wherein the coverage range of the base station in abnormal operation is smaller, and the corresponding edge server computing capacity is weaker, and there is a remaining computing task; the formula for calculating the first weight of the node is: ； In the formula, is the first weight of the ith node, is the remaining computing power of the ith node, is the distance between the ith node and the base station at the next time according to the current position and the movement speed of the ith node, is the stability index of the ith node, , is the weight of , , K is the number of neighbor nodes of the ith node, is the distance between the ith node and the kth neighbor node at the current time, is the distance between the ith node and the kth neighbor node at the next time, is the mobility index of the ith node relative to the kth neighbor node, and the neighbor node is selected from the nodes in the normal coverage range of the base station. The preset clustering scheme is sent to the cluster head, and the cluster fed back by the cluster head is received; wherein the cluster head obtains a candidate cluster according to the preset clustering scheme, and obtains an optimal cluster from the candidate cluster according to the preset clustering optimization model; According to the cluster fed back by the cluster head, the remaining computing task is distributed, the sub-computing task obtained by distribution is unloaded to the corresponding cluster point through the cluster head, and the result of processing the sub-computing task by the cluster point is received through the cluster head.

2. The edge computing network's end computing method according to claim 1, wherein, The information of the node in the current coverage range includes the remaining computing power of the node, the position of the node, the motion speed of the node and the sum of the remaining computing power of all neighbor nodes; wherein the neighbor nodes are selected from the nodes in the normal coverage range of the base station.

3. The edge computing network's end computing method of claim 1, wherein, The clustering scheme comprises: Scheme one: selecting cluster points within one-hop range with the cluster head as the center to build a cluster; wherein the one-hop range is the range that can directly communicate with the cluster head; wherein the cluster points are selected from the nodes in the normal coverage range of the base station; Scheme two: selecting cluster points within two-hop range with the cluster head as the center to build a cluster; wherein the two-hop range is the range that can communicate with the cluster head through a node.

4. The edge computing network's end computing method according to claim 3, wherein, The cluster point selection process comprises: If scheme one is adopted, the second weight of the node within the one-hop range is calculated according to the information of the node, and the node with the second weight greater than the threshold value is selected as the cluster point; If scheme two is adopted, the second weight of the node within the two-hop range is calculated according to the information of the node, and the node with the second weight greater than the threshold value is selected as the cluster point.

5. The edge computing network's end computing method according to claim 4, wherein, The formula for calculating the second weight is: ； In the formula, is the second weight of the jth node, is the remaining computing power of the jth node, is the stability index of the jth node, , respectively, is the weight of the jth node, , N is the number of neighbor nodes of the jth node, is the distance between the jth node and the nth neighbor node at the current moment, is the distance between the jth node and the nth neighbor node at the next moment, is the mobility index of the jth node relative to the nth neighbor node, and the neighbor node is selected from the nodes in the normal coverage range of the base station.

6. The edge computing network's end computing method according to claim 1, wherein, The clustering optimization model comprises: Objective function: ； In the formula, is the performance index of the system, the system is a system composed of base stations and nodes in clusters, F is the total residual computing power of all nodes participating in processing computing tasks, , are the weights of F and respectively, T is the total delay of processing and transmitting computing tasks, E is the total energy consumption of processing and transmitting computing tasks, , are the weights of T and E respectively, is the channel bandwidth of the mth node in the system, is the transmission power of the upstream node to the mth node in the system; The constraint condition is: ； ； ； ； ； In the formula, is an upper limit of T, is an upper limit of is an upper limit of is an upper limit of is an upper limit of M, M is the total number of nodes in the system, is an upper limit of the channel bandwidth used by the system, is the residual computing power of the mth node, is an upper limit of is an upper limit of 7. An edge computing network's end computing device, characterized by, The cluster head selection module, if the base station is not normally running at the current time and there is a fixed node in the current coverage range, the fixed node is selected as the cluster head; if the base station is not normally running at the current time and there is no fixed node in the current coverage range, the first weight of the node is calculated according to the information of the node, and the node with the largest first weight is selected as the cluster head; wherein the coverage range of the base station in abnormal operation is smaller, and the corresponding edge server computing capacity is weaker, and there is a remaining computing task; the formula for calculating the first weight of the node is: The cluster acquisition module, the preset clustering scheme is sent to the cluster head, and the cluster fed back by the cluster head is received; wherein the cluster head obtains a candidate cluster according to the preset clustering scheme, and obtains an optimal cluster from the candidate cluster according to the preset clustering optimization model; ； In the formula, is the first weight of the ith node, is the remaining computing power of the ith node, is the distance between the ith node and the base station at the next moment, which is calculated according to the current position and the motion speed of the ith node, is the stability index of the ith node, , is the weight of , , K is the number of neighbor nodes of the ith node, is the distance between the ith node and the kth neighbor node at the current moment, is the distance between the ith node and the kth neighbor node at the next moment, is the mobility index of the ith node relative to the kth neighbor node, and the neighbor node is selected from the nodes in the normal coverage range of the base station. The task processing module, according to the cluster fed back by the cluster head, the remaining computing task is distributed, the sub-computing task obtained by distribution is unloaded to the corresponding cluster point through the cluster head, and the result of processing the sub-computing task by the cluster point is received through the cluster head. ​ 8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, the one or more programs comprising instructions, which when executed by a computing device, cause the computing device to perform any of the methods of claims 1-6.

Images