Edge Server Configuration Method and System Based on AP Clustering Algorithm and Multi-Objective Optimization Algorithm

Abstract

The present invention relates to an edge server configuration method and system based on AP clustering algorithm and multi-objective optimization algorithm, and the configuration method includes the following steps: S1, modeling bus stops and edge server variables; S2, constructing the number of servers and server coverage based on Affinity Propagation clustering algorithm; S3, constructing an edge server configuration model under a smart bus scenario; S4, modeling the delay, task load balancing, high-quality service request ratio, constraints and total flow in the edge network of the edge server configuration; S5, under the premise of meeting the constraints, based on the NSGAII genetic algorithm, the delay is minimized, the task load balancing is minimized, the total flow is minimized, and the high-quality service request ratio is maximized to determine the configuration of the edge server. The present invention realizes the automatic configuration of the edge server under the smart bus scenario, and obtains a better edge server configuration strategy.

Description

Technical Field

[0001] The present invention belongs to the technical field of edge server configuration, and particularly relates to an edge server configuration method and system based on an AP clustering algorithm and a multi-objective optimization algorithm. Background Art

[0002] Currently, the urban intelligent bus system needs to comprehensively use technologies such as positioning systems, wireless communication, and image and video processing to achieve the intelligence of bus vehicle operation scheduling, the informatization and visualization of bus vehicle operation, and to provide a perfect information service for the public passengers. The realization of these functions requires a network connecting bus stops and vehicles to obtain relevant information in real time, so as to finally achieve intelligent scheduling and intelligent monitoring. At the same time, it can also provide users with more accurate specific information of bus vehicles, thereby improving the service quality of the bus system. However, the traditional approach is to transmit this information through the network to the cloud center for relevant processing. This approach will bring two harms: one is the impact of excessive latency, resulting in the loss of timeliness of the information for scheduling and transmission to users, and reducing the scheduling quality and service quality; the other harm is that a large amount of data transmission may lead to an excessive network traffic load, ultimately resulting in network congestion and affecting this intelligent bus system.

[0003] In recent years, the rise of edge computing has brought a new solution. The core idea of edge computing is to provide the nearest service directly near the data source. Since edge computing is initiated at the edge side of the data source, it reduces the process of a large amount of raw data being transmitted over the network, so the network services it provides are also faster. Therefore, it has great advantages in some real-time services, application intelligence, etc. The first step in edge computing is to reasonably deploy edge servers to achieve better results as much as possible.

[0004] At present, many scholars have conducted relevant research on the deployment of edge servers and the deployment of similar micro-clouds, and have achieved good results. Regarding the deployment of edge servers, the research scenarios of scholars mainly fall into two categories. One is to consider the configuration of edge servers in the mobile edge computing scenario; the other is to consider the configuration of edge servers in the Internet of Things scenario combined with specific services. In the process of deploying edge servers in the mobile edge computing scenario, the objective functions that usually need to be considered mainly include delay, delay balance, load balance, and cost. In the mobile edge computing scenario, because it involves users' private information, in some specific scenarios, the privacy issues of users also need to be considered. Because edge computing itself has strong distributed characteristics, some scholars also consider the robustness of the network in the process of deploying edge servers, that is, in the case where some edge servers may fail, as few users as possible are affected. In specific networking scenarios, it is mainly to configure edge servers in combination with specific service scenarios. Some of the scenarios currently being studied are in the fields of smart farms and smart manufacturing. The configuration of micro-clouds has many similarities with the configuration of edge servers. In the process of configuring micro-clouds, in addition to the objective functions considered in edge server configuration, some scholars also consider the problem of network traffic in the configuration process and optimize it. Since the objective functions that need to be optimized in the process of edge server configuration are not single, the algorithms for solving the edge server configuration problem currently mainly include heuristic algorithms and integer programming algorithms. In the process of configuring edge servers, most of the research directly assumes the number of edge servers to be configured and optimizes on this basis. At the same time, few people consider the network traffic in the process of edge server configuration.

[0005] On the basis of considering both load and geographical space, in order to better determine the optimal number of servers and the coverage radius R, the Affinity Propagation clustering algorithm is used to determine the optimal number of clusters and the server coverage radius. In order to find a better server configuration strategy, the edge server configuration problem is reduced to a multi-objective optimization problem of delay, load balance, traffic, and high-quality service request ratio, and the NSGAII genetic algorithm is used to solve it. Summary of the Invention

[0006] Aiming at the above problems existing in the construction of a high-quality edge server network in the smart bus scenario in the prior art, the purpose of the present invention is to provide an edge server configuration method and system based on the AP clustering algorithm and the multi-objective optimization algorithm, model the optimization objectives of minimizing delay, load balance, minimizing traffic, and maximizing the high-quality service request ratio, and determine the optimal edge server configuration location.

[0007] In order to achieve the above purpose, the present invention provides the following technical solutions:

[0008] Edge server configuration method based on AP clustering algorithm and multi-objective optimization algorithm, comprising the following steps:

[0009] S1. Model the bus stops and edge server variables;

[0010] S2. Construct the number of servers and the server coverage range based on the Affinity Propagation clustering algorithm;

[0011] S3. Construct an edge server configuration model in the intelligent bus scenario;

[0012] S4. Model the delay, task load balance, high-quality service request ratio, constraint conditions, and total traffic in the edge network for the edge server configuration;

[0013] S5. On the premise of meeting the constraint conditions, minimize the delay, minimize the task load balance, minimize the total traffic, and maximize the high-quality service request ratio based on the NSGAII genetic algorithm to determine the configuration of the edge server.

[0014] As a preferred solution, the step S1 includes:

[0015] The edge network is represented by an undirected graph G = {V, E}, where V represents the set of bus stops and servers, and E represents the network connection between the bus stops and the servers;

[0016] Assume that there are n bus stops and m edge servers in the edge network, and m << n; the data of the bus stops is represented as a set BS = {bs 1 , bs 2 , …, bs n}, where bs i represents the i-th bus stop, 1 ≤ i ≤ n; the data of the edge servers is represented as a set ES = {es 1 , es 2 , …, es m}, where es j represents the j-th edge server, 1 ≤ j ≤ m.

[0017] As a preferred solution, the step S2 includes:

[0018] Cluster the bus stops that require edge servers based on the Affinity Propagation clustering algorithm, obtain the number of central points CN and the bus stops included in each cluster, calculate the distance from each bus stop to its corresponding cluster center and the total task load of each cluster set; sort all the distance sets and task load sets, and select the cluster sets suitable for the coverage radius R of the edge server and low load according to the 80 / 20 principle, and subtract the number of the low-load cluster sets from CN to obtain the number K of the required edge servers.

[0019] As a preferred solution, the step S3 includes:

[0020] Based on the number K of the required edge servers, select K positions from BS = {bs 1 , bs 2 , …, bs n} to configure edge servers, and the remaining bus stops are loaded by the bus stop closest to the configured edge server.

[0021] As a preferred solution, in the step S4, the constraint conditions include: First, the task load of each bus stop is shared by only one edge server. Second, at most one edge server can be configured for each bus stop;

[0022] In the step S4, the construction process of the delay includes:

[0023] Calculate the distance between the bus stop and the edge server in the edge network through the Haversine formula:

[0024]

[0025] Among them, and represent the latitudes of bs and es in radians respectively, and λ 1 and λ 2 represent the longitudes of bs and es in radians respectively; use this distance to represent the transmission delay;

[0026] Let x ij = {0, 1} represent whether the bus stop bs i is served by the edge server es j ; if x ij = 1, it means that the bus stop numbered i is served by the edge server j; otherwise, it means that the bus stop numbered i is not served by the edge server j;

[0027] The transmission delay of the bus stop numbered i and served by the edge server j is

[0028] The waiting delay of the bus stop numbered i and served by the edge server j is

[0029] The delay of the bus stop numbered i and served by the edge server j is:

[0030] The total delay in the edge network is: The average transmission delay of all bus stops in the edge network is

[0031] As a preferred solution, in the step S4, the construction process of task load balancing includes:

[0032] Let be the task arrival rate of each bus stop;

[0033] The edge server es numbered j j The total task load is expressed as:

[0034] The average load of each edge server is

[0035] The task load balance of the edge server is defined as σ WB :

[0036]

[0037] As a preferred solution, in the step S4, the construction process of the total traffic in the edge network includes:

[0038] Let represent the number of relays required for the bus stop numbered i to the edge server j that serves it;

[0039] The traffic generated in the edge network is represented by multiplying the task arrival rate by the number of relays. The traffic generated by the bus stop numbered i in the edge network is:

[0040] The total traffic controlled by the edge server numbered j is defined as

[0041] The total traffic in the entire edge network is defined as T total is:

[0042] As a preferred solution, in the step S4, the construction process of the high-quality service request ratio includes:

[0043] The total number of task loads in the entire edge network is WB total :

[0044] Define Indicates whether the distance between bus stop i and edge server j is no greater than R; if so, the value is 1, indicating a high-quality service request; if not, it indicates a low-quality service request;

[0045] Define the total number of high-quality service requests in the edge network as WB good :

[0046] The high-quality service request ratio is

[0047] As a preferred solution, the step S5 includes:

[0048] Among all bus stops, based on the NSGAII genetic algorithm, obtain the configuration location L = {l 1 , l 2 , …, l K} of the edge server, satisfying Min(σ WB ), Min(T total ), Max(gr), and also satisfying the following conditions:

[0049] x ij = {0, 1}

[0050]

[0051] The present invention also provides an edge server configuration system based on the AP clustering algorithm and the multi-objective optimization algorithm, including:

[0052] A modeling module, used for modeling bus stop and edge server variables, and also for constructing the number of servers and the server coverage range based on the AffinityPropagation clustering algorithm, and also for constructing an edge server configuration model in the intelligent bus scenario; also used for modeling the delay, task load balancing, high-quality service request ratio, constraint conditions, and total traffic in the edge network;

[0053] A configuration module, used for minimizing the delay, minimizing the task load balance, minimizing the total traffic, and maximizing the high-quality service ratio based on the NSGAII genetic algorithm under the premise of satisfying the constraint conditions, and determining the configuration of the edge server.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] The present invention realizes the automatic configuration of edge servers in the intelligent bus scenario, without the need for manual determination of the number of edge servers. By using a clustering algorithm, it simultaneously considers the characteristics of load and site geographical distribution. The configuration strategy of edge servers is attributed to a multi-objective optimization problem of delay, task load balancing, total traffic, and high-quality service ratio, and the NSGAII genetic algorithm is used to optimize it to obtain a better edge server configuration strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 is a model diagram of an edge computing framework in the intelligent bus scenario according to an embodiment of the present invention;

[0057] Figure 2 is a flowchart of a method for configuring an edge server based on the AP clustering algorithm and a multi-objective optimization algorithm according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] To more clearly illustrate the embodiments of the present invention, the specific implementation manners of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings, and other implementation manners can be obtained.

[0059] The method for configuring an edge server based on the AP clustering algorithm and a multi-objective optimization algorithm of the present invention finds a better edge server configuration strategy by modeling the minimization of delay, the minimization of edge server load balancing, the minimization of the total edge network traffic, and the maximization of the high-quality request ratio as optimization objectives.

[0060] Specifically, as Figure 1 shown, there are multiple candidate locations for edge servers in the network. By selecting some better configuration strategies, the minimization of delay, edge server load balancing, and total edge network traffic are achieved while maximizing the high-quality service request ratio.

[0061] As Figure 2 shown, the method for configuring an edge server based on the AP clustering algorithm and a multi-objective optimization algorithm according to an embodiment of the present invention includes the following steps:

[0062] S1. Model the variables of bus stops (i.e., bus stations) and edge servers;

[0063] S2. Construct a solution process for the number of servers and the server coverage range R based on the Affinity Propagation clustering algorithm;

[0064] S3. Construct an edge server configuration model in the intelligent bus scenario;

[0065] S4. Model the latency of the edge server configuration;

[0066] S5. Model the task load balancing of the edge server configuration;

[0067] S6. Model the total traffic in the edge network;

[0068] S7. Model the high-quality service request ratio of the edge server configuration;

[0069] S8. Model the constraint conditions of the edge server configuration;

[0070] S9. Based on the NSGAII genetic algorithm, minimize the latency, task load balancing, total traffic, and maximize the high-quality service request ratio to determine the edge server configuration strategy.

[0071] Among them, step S1 includes the following steps: The entire edge network can be represented by an undirected graph G = {V, E}, where V represents the set of bus stops and the set of edge servers, and E represents the network connection between the bus stops and the edge servers.

[0072] Suppose there are n bus stops and m (m << n) edge servers in the edge network. The data of the bus stops is represented as the set BS = {bs 1 , bs 2 , …, bs n}, where bs i represents the i-th bus stop, 1 ≤ i ≤ n. The data of the edge servers is defined as the set ES = {es 1 , es 2 , …, es m}, where es j represents the j-th edge server, 1 ≤ j ≤ m.

[0073] The above step S2 includes the following steps: Based on the Affinity Propagation clustering algorithm, cluster the sites that require edge servers to obtain the number of center points CN and which sites are included in each cluster. Calculate the distance from each site to the corresponding cluster center and the total task load of each cluster set. Sort the calculated distance set and load set, and select the cluster set suitable for the edge server coverage radius R and low load according to the 80 / 20 principle. Subtract the number of the low-load cluster set from CN to obtain the number K of the required edge servers.

[0074] After obtaining the number of edge servers required in step S2 above, we need to select from BS = {bs 1 , bs 2 , …, bs nSelect K positions to deploy edge servers, and the computing requirements generated by the remaining sites are loaded by the site with the edge server deployed closest to them. Finally, minimize the delay parameter, load balancing parameter, and total traffic load parameter in the edge network, while maximizing the high-quality service request ratio.

[0075] The above step S4 includes the following steps:

[0076] S41. Calculate the distance between the bus stop and the edge server in the edge network through the Haversine formula, where, and represent the latitudes of bs and es in radians respectively, λ 1 and λ 2 represent the longitudes of bs and es in radians respectively. Use this distance to represent the transmission delay.

[0077] S42. Let x ij ={0,1} represent whether the site bs i is served by the edge server es j x ij =1 means that the bus stop numbered i is served by the edge server j, otherwise, it means that the site is not served by the edge server j.

[0078] S43. The transmission delay of the bus stop numbered i and served by the edge server j is

[0079] S44. The waiting delay of the bus stop numbered i and served by the edge server j is DT i queue , and solve it using the queuing model.

[0080] S45. The delay of the bus stop numbered i and served by the edge server j is DT i =DT i trans +DT i queue .

[0081] S46. The total delay in the entire edge service network can be expressed as The average transmission delay of all sites in the entire edge network is expressed as

[0082] The above step S5 includes the following steps:

[0083] S51. Let be the task arrival rate of each bus stop, 1 < i < n;

[0084] S52. The total task load of the edge server es numbered j j can be expressed as

[0085] S53. The average load of each edge server is

[0086] S54. The index definition of edge server load balancing is σ WB , and the calculation formula is as follows:

[0087] The above step S6 includes the following steps:

[0088] S61. Let represent the number of relays required from the bus stop numbered i to the edge server j that serves it. R is the value obtained in step S2;

[0089] S62. The traffic generated in the edge network is represented by multiplying the task arrival rate by the number of relays. The specific calculation method of the traffic generated by the bus stop numbered i in the edge network is

[0090] S63. The total traffic controlled by the edge server numbered j is defined as The calculation formula is as follows:

[0091] S64. The total traffic in the entire edge server network is defined as T total , and the calculation formula is as follows

[0092] The above step S7 includes the following steps:

[0093] S71. The total number of task loads in the entire edge network is represented as WB total , and the calculation formula is as follows

[0094] S72. Define to represent whether the distance from site i to edge server j is less than or equal to R; if it is less than or equal to, the value is 1, indicating that this is a high-quality service request; otherwise, it is 0, indicating that this is not a high-quality request, that is, a low-quality service request.

[0095] S73. Define the total number of high-quality service requests in the edge network as WB good , and the calculation formula is as follows:

[0096] S74. The high-quality service request ratio is represented as

[0097] The above step S8 includes the following steps: The entire edge server configuration process should also meet the following two constraints: 1. The task load of each bus stop is not shared by multiple edge servers; 2. Only one edge server can be configured for each bus stop.

[0098] The above step S9 includes the following steps: Under the constraints of step S8, based on the NSGAII genetic algorithm, determine the final edge server configuration strategy to minimize the average delay, minimize the load balancing parameter, minimize the traffic, and maximize the high-quality service request ratio.

[0099] That is, among all the bus stops, find a set of configuration locations L = {l 1 , l 2 , …, l K}, which can achieve Min(σ WB ), Min(T total ), Max(gr), and at the same time also meet the following conditions:

[0100] x ij = {0, 1}

[0101]

[0102] In the embodiment of the present invention, according to the number K of edge servers and the coverage radius R, the NSGAII algorithm is used to optimize the above four objectives, mainly including initializing the initial population by combining specific site information, iterating the algorithm at the same time, and finally obtaining a set of relatively optimal edge server configuration strategies. At the same time, in each round of iteration, a clustering algorithm is introduced to add a part of relatively good solutions to enhance the optimization process of NSGAII.

[0103] In addition, the embodiment of the present invention also provides an edge server configuration system based on the AP clustering algorithm and the multi-objective optimization algorithm, corresponding to the above configuration method. Specifically, the edge server configuration system includes a modeling module and a configuration module.

[0104] Among them, the modeling module is used to model the bus stop and edge server variables, and is also used to construct the number of servers and the server coverage range based on the AffinityPropagation clustering algorithm, and is also used to construct an edge server configuration model in the intelligent bus scenario; it is also used to model the delay, task load balancing, high-quality service request ratio, constraint conditions, and total traffic in the edge network of the edge server configuration.

[0105] Among them, the process of modeling bus stop and edge server variables includes: The entire edge network can be represented by an undirected graph G = {V, E}, where V represents the set of bus stops and edge servers, and E represents the network connection between bus stops and edge servers.

[0106] Suppose there are n bus stops and m (m << n) edge servers in the edge network. The data of bus stops is represented as the set BS = {bs 1 , bs 2 , …, bs n}, where bs i represents the i-th bus stop, 1 ≤ i ≤ n. The data of edge servers is defined as the set ES = {es 1 , es 2 , …, es m}, where es j represents the j-th edge server, 1 ≤ j ≤ m.

[0107] The process of constructing the number of servers and the coverage range of servers based on the Affinity Propagation clustering algorithm includes: clustering the sites that need edge servers based on the Affinity Propagation clustering algorithm to obtain the number of central points CN and which sites are included in each cluster. Calculate the distance from each site to the corresponding cluster center and the total task load of each cluster set. Sort the calculated distance set and load set, and select the cluster sets suitable for the edge server coverage radius R and low load according to the 80 / 20 principle. Subtract the number of cluster sets with low load from CN to obtain the number K of edge servers required.

[0108] The process of constructing the edge server configuration model in the intelligent bus scenario includes: After obtaining the number of edge servers required, we need to select K positions from BS = {bs 1 , bs 2 , …, bs n} to configure edge servers, and the computing requirements generated by the remaining sites are loaded by the site with the nearest deployed edge server. Finally, minimize the delay parameter, load balancing parameter, and total traffic load parameter in the edge network, and maximize the high-quality service request ratio at the same time.

[0109] In addition, the following details the process of modeling the delay, task load balancing, high-quality service request ratio, constraint conditions, and total traffic in the edge network for edge server configuration.

[0110] (1) The process of constructing the delay includes:

[0111] Calculate the distance between the bus stop and the edge server in the edge network through the Haversine formula:

[0112]

[0113] Among them, and represent the latitudes of bs and es in radians respectively, and λ 1 and λ 2 represent the longitudes of bs and es in radians respectively.

[0114] Use this distance to represent the transmission delay;

[0115] Let x ij ={0,1} represent whether the site bs i is served by the edge server es j ; x ij =1 means that the bus stop numbered i is served by the edge server j, otherwise, it means that the site is not served by the edge server j;

[0116] The transmission delay of the bus stop numbered i and served by the edge server j is

[0117] The waiting delay of the bus stop numbered i and served by the edge server j is DT i queue , and solve it using the queuing model.

[0118] The delay of the bus stop numbered i and served by the edge server j is DT i =DT i trans +DT i queue ;

[0119] The total delay in the entire edge service network can be expressed as The average transmission delay of all sites in the entire edge network is expressed as

[0120] (2) The construction process of task load balancing includes:

[0121] Let be the task arrival rate of each bus stop, 1 < i < n;

[0122] The total task load of the edge server es numbered j j can be expressed as

[0123] The average load of each edge server is

[0124] The metric of edge server load balancing is defined as σ WB , and the calculation formula is as follows:

[0125] (3) The construction process of the total traffic in the edge network includes:

[0126] Let represent the number of relays required from the bus stop numbered i to the edge server j that serves it. R is the value obtained in step S2;

[0127] The traffic generated in the edge network is represented by multiplying the task arrival rate by the number of relays. The specific calculation method of the traffic generated by the bus stop numbered i in the edge network is

[0128] The total traffic controlled by the edge server numbered j is defined as The calculation formula is as follows:

[0129] The total traffic in the entire edge network is defined as T total , and the calculation formula is as follows

[0130] (4) The construction process of the high-quality service request ratio includes:

[0131] The total task load in the entire edge network is represented as WB total , and the calculation formula is as follows

[0132] Define to represent whether the distance between site i and edge server j is less than or equal to R; if it is less than or equal to, the value is 1, indicating that this is a high-quality service request; otherwise, it is 0, indicating that this is not a high-quality request, that is, a low-quality service request.

[0133] Define the total number of high-quality service requests in the edge network as WB good , and the calculation formula is as follows:

[0134] The high-quality service request ratio is represented as

[0135] (5) The edge server configuration process should also meet the following two constraints: 1. The task load of each bus stop is not shared by multiple edge servers; 2. Only one edge server can be configured for each bus stop.

[0136] On the basis of satisfying the above two constraints, the configuration module of the embodiment of the present invention determines the final configuration strategy of the edge server based on the NSGAII genetic algorithm, realizing the minimization of the average delay, the minimization of the load balancing parameter, the minimization of the traffic, and the maximization of the high-quality service request ratio.

[0137] Specifically, among all bus stops, find a set of configuration locations L = {l 1 , l 2 , …, l K}, which can achieve Min(σ WB ), Min(T total ), Max(gr), and at the same time satisfy the following conditions:

[0138] x ij = {0, 1}

[0139]

[0140] represents the fixed stop number i. Summing up all x ij (traversing all edge servers, and the sum is 1, thus indicating that a stop can only be served by one edge server) can only be 1, which means that a stop is only served by one edge server.

[0141] The present invention realizes the automatic configuration of the edge server in the intelligent bus scenario, without the need to manually determine the number of edge servers served, and the obtained edge server configuration strategy is relatively optimal.

[0142] The above is only a detailed description of the preferred embodiments and principles of the present invention. For those of ordinary skill in the art, according to the idea provided by the present invention, there will be changes in the specific implementation manners, and these changes should also be regarded as the protection scope of the present invention.

Claims

1. Edge server configuration method based on AP clustering algorithm and multi-objective optimization algorithm, Characterized in that, It includes the following steps: S1. Model bus stops and edge server variables; S2. Construct the number of servers and the coverage range of servers based on the Affinity Propagation clustering algorithm; S3. Construct an edge server configuration model in the intelligent bus scenario; S4. Model the delay, task load balance, high-quality service request ratio, constraint conditions, and total traffic in the edge network for edge server configuration; S5. On the premise of meeting the constraint conditions, based on the NSGAII genetic algorithm, minimize the delay, minimize the task load balance, minimize the total traffic, and maximize the high-quality service request ratio to determine the configuration of the edge server; The step S1 includes: The edge network is represented by an undirected graph G = {V, E}, where V represents the set of bus stops and servers, and E represents the network connection between bus stops and servers; Assume that there are n bus stops and m edge servers in the edge network, where m << n; the data of the bus stops is represented as the set BS = {bs 1 , bs 2 , …, bs n}, where bs i represents the i-th bus stop, 1 ≤ i ≤ n; the data of the edge servers is represented as the set ES = {es 1 , es 2 , …, es m}, where es j represents the j-th edge server, 1 ≤ j ≤ m; The step S2 includes: Cluster the bus stops that need edge servers based on the Affinity Propagation clustering algorithm to obtain the number of central points CN and the bus stops included in each cluster. Calculate the distance from each bus stop to its corresponding cluster center and the total task load of each cluster set. Sort all the distance sets and task load sets, and select the cluster sets suitable for the edge server coverage radius R and low load according to the 80 / 20 principle. Subtract the number of low-load cluster sets from CN to obtain the number K of required edge servers; The step S3 includes: Based on the required number K of edge servers, select K positions from BS = {bs 1 , bs 2 , …, bs n} to configure edge servers, and the remaining bus stops are loaded by the bus stops that configure the edge servers closest to them; In the step S4, the constraint conditions include: First, the task load of each bus stop is shared by only one edge server. Second, each bus stop can be configured with at most one edge server; In the step S4, the construction process of the delay includes: Calculate the distance between bus stops and edge servers in the edge network through the Haversine formula: Wherein, and represent the latitudes of bs and es in radians respectively, and λ 1 and λ 2 represent the longitudes of bs and es in radians respectively; the transmission delay is represented by using this distance; Let x ij ={0, 1} represent whether the bus stop bs i is served by the edge server es j ; if x ij = 1, it means that the bus stop numbered i is served by the edge server j; otherwise, it means that the bus stop numbered i is not served by the edge server j; The transmission delay of the bus stop numbered i and served by the edge server j is The waiting delay of the bus stop numbered i and served by the edge server j is DT i queue ; The latency of the bus stop numbered i and served by the edge server j is: DT i = DT i trans + DT i queue ; The total delay in the edge network is as follows: The average transmission delay of all bus stops in the edge network is 2. The edge server configuration method based on the AP clustering algorithm and multi-objective optimization algorithm according to claim 1, Characterized in that, In the step S4, the construction process of the task load balance includes: Let be the task achievement rate of each bus stop; The edge server es numbered j j has the total task load expressed as: The average load of each edge server is The task load balancing of the edge server is defined as σ WB :

3. The edge server configuration method based on the AP clustering algorithm and multi-objective optimization algorithm according to claim 2, Characterized in that, In the step S4, the construction process of the total traffic in the edge network includes: Let represent the number of relays required from the bus stop numbered i to the edge server j that serves it; The traffic generated in the edge network is represented by the task arrival rate multiplied by the number of relay times. The traffic generated by the bus stop numbered i in the edge network is as follows: The total traffic controlled by the edge server numbered j is defined as The total traffic in the entire edge network is defined as T total It is:

4. The edge server configuration method based on the AP clustering algorithm and multi-objective optimization algorithm according to claim 3, Characterized in that, In the step S4, the construction process of the high-quality service request ratio includes: The total number of task loads in the entire edge network is WB total : Definition Indicates whether the distance between bus stop i and edge server j is no greater than R; if so, the value is 1, indicating a high-quality service request; if not, it indicates a low-quality service request. Define the total number of high-quality service requests in the edge network as WB good : The high-quality service request ratio is 5. The edge server configuration method based on the AP clustering algorithm and multi-objective optimization algorithm according to claim 4, Characterized in that, The step S5 includes: Among all bus stops, the configuration locations L = {l 1 , l 2 , …, l K} of the edge servers obtained based on the NSGAII genetic algorithm satisfy Min(σ WB ), Min(T total ), Max(gr).

6. An edge server configuration system based on the AP clustering algorithm and multi-objective optimization algorithm, applying the edge server configuration method according to any one of claims 1-5, Characterized in that, The edge server configuration system includes: The modeling module is used to model bus stop and edge server variables, and is also used to construct the number of servers and the server coverage range based on the Affinity Propagation clustering algorithm. It is also used to construct an edge server configuration model in the intelligent bus scenario. It is also used to model the delay, task load balancing, high-quality service request ratio, constraint conditions, and total traffic in the edge network. The configuration module is used to minimize the delay, minimize the task load balancing, minimize the total traffic, and maximize the high-quality service request ratio based on the NSGAII genetic algorithm on the premise of meeting the constraint conditions, and determine the configuration of the edge server.

Images