Parameter configuration model training method, parameter configuration method, device and equipment

Abstract

The invention discloses a parameter configuration model training method, a parameter configuration method, a parameter configuration device and equipment, and the method comprises the steps: carrying out the environment scanning of a first access point of a to-be-configured parameter, and obtaining the intensity of a signal which is transmitted by a second access point and is received by the first access point; determining a plurality of adjacent access points of the first access point from the second access points according to the signal intensity and a preset signal intensity threshold value; determining a feature vector of the first access point according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels; obtaining parameter configuration labels of a plurality of first access points; and performing iterative training on the initial parameter configuration model according to the feature vectors of the plurality of first access points and the parameter configuration labels to obtain a trained parameter configuration model, and performing parameter configuration on the first access points through the trained parameter configuration model to realize full-automatic wireless network parameter configuration. And the accuracy of wireless network parameter configuration is improved.

Description

Technical Field

[0001] The present application relates to the field of wireless communication technology, and in particular to a training method, a parameter configuration method, a device and equipment for a parameter configuration model. Background Art

[0002] As the scale of wireless network usage increases, network topology tends to be denser, and the mutual influence between APs (Access Points) is also increasing. For example, in dense wireless usage scenarios, due to the overlapping conflict areas between APs on the same communication frequency, the co-channel interference between APs affects the overall wireless performance of the wireless network. Therefore, how to configure the best wireless network parameters such as channel, bandwidth, power and frequency band has become an important factor in improving wireless performance.

[0003] At present, wireless network parameters are generally predicted through machine learning methods. Specifically, the network model is trained through parameters related to the wireless network environment, and then the trained network model is used to directly predict the network parameters.

[0004] However, the above method cannot perform adaptive and fully automatic wireless network parameter configuration at the network topology level. Summary of the invention

[0005] The main purpose of the present application is to provide a parameter configuration model training method, parameter configuration method, device and equipment, which can perform adaptive and fully automatic wireless network parameter configuration at the network topology level.

[0006] In order to achieve the above-mentioned purpose, in a first aspect, the present application provides a method for training a parameter configuration model, including: performing an environmental scan on a first access point to be configured with parameters, and obtaining the strength of a signal sent by a second access point and received by the first access point; determining multiple adjacent access points of the first access point from the second access point according to the strength of the signal and a preset signal strength threshold; determining a feature vector of the first access point according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels; obtaining parameter configuration labels of the multiple first access points; iteratively training an initial parameter configuration model according to the feature vectors and parameter configuration labels of the multiple first access points to obtain a trained parameter configuration model, so as to perform parameter configuration on the first access point through the trained parameter configuration model.

[0007] In one embodiment, the method further includes: mapping the wireless network where the first access point is located to obtain a topological structure corresponding to the wireless network; wherein the wireless network topological structure includes m vertices and n edges, m is an integer greater than or equal to 1, and n is an integer greater than or equal to 0; the vertices of the topological structure indicate the m access points in the wireless network, the edges in the topological structure indicate the connection relationship between any two access points in the topological structure, and the weight of the edge indicates the strength of the signal.

[0008] In an embodiment, determining the feature vector of the first access point according to the scanning results of the first access point on different channels and the scanning results of the neighboring access points on different channels includes: acquiring the scanning results of the first access point on different channels and the scanning results of the neighboring access points on different channels; determining the first feature vectors of the m vertices according to the scanning results of the first access point on different channels and the scanning results of the neighboring access points on different channels; determining the second feature vectors of the n edges according to the scanning results of the first access point on different channels and the scanning results of the neighboring access points on different channels; and determining the feature vector of the first access point according to the first feature vectors of the m vertices and the second feature vectors of the n edges.

[0009] In an embodiment, the determining the first feature vectors of the m vertices according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels includes: determining the first initial feature vectors of the m vertices according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels; denoising the first initial feature vectors of the m vertices to obtain the denoised first initial feature vectors; for each vertex of the m vertices, performing neighbor sampling on each vertex to obtain the neighbor nodes of each vertex; aggregating the first feature vectors of the neighbor nodes of each vertex to obtain the aggregated features of each vertex; determining the first feature vector of each vertex based on the aggregated features of each vertex and the denoised first initial feature vector; summarizing the first feature vectors of each vertex to obtain the first feature vectors of the m vertices.

[0010] In an embodiment, determining the second eigenvectors of the n edges according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels includes: determining the second initial eigenvectors of the n adjacency matrices corresponding to the n edges according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels; denoising the second initial eigenvectors of the n adjacency matrices corresponding to the n edges to obtain the denoised second initial eigenvectors of the n adjacency matrices, wherein the n edges correspond to the n adjacency matrices one-to-one; for each of the n edges, obtaining a denoised eigenvector matching the each edge from the denoised second initial eigenvectors of the n adjacency matrices; using the denoised eigenvector matching the each edge as the second eigenvector of each edge; and summarizing the second eigenvectors of each edge to obtain the second eigenvectors of the n edges.

[0011] In one embodiment, the iterative training of the initial parameter configuration model according to the feature vectors and parameter configuration labels of the multiple first access points to obtain the trained parameter configuration model includes: iterative training of the initial parameter configuration model according to the feature vectors and parameter configuration labels of the multiple first access points, and obtaining the trained parameter configuration model when the loss function satisfies a preset condition.

[0012] In one embodiment, the loss function includes a feature mapping function and a feature edge mapping function; the initial parameter configuration model is iteratively trained based on the feature vectors and parameter configuration labels of the multiple first access points, and the trained parameter configuration model is obtained when the loss function meets a preset condition, including: based on the first feature vectors of the m vertices and the feature mapping function, obtaining the updated first feature vectors corresponding to the m vertices; based on the second feature vectors of the n edges and the feature edge mapping function, obtaining the updated second feature vectors corresponding to the n edges; based on the updated first feature vectors corresponding to the m vertices, the updated second feature vectors corresponding to the n edges, and the parameter configuration label, the initial parameter configuration model is trained, and the trained parameter configuration model is obtained when the feature mapping function and the feature edge mapping function meet a preset number of iterations.

[0013] In an embodiment, if the parameter configuration tag is a tag of a radio frequency switch state, the trained parameter configuration model is used to perform frequency band parameter configuration on the first access point.

[0014] In an embodiment, if the parameter configuration label is a label of channel deployment, the trained parameter configuration model is used to perform channel parameter configuration on the first access point.

[0015] In one embodiment, if the parameter configuration label is a transmit power label, the trained parameter configuration model is used to perform power parameter configuration on the first access point, wherein the transmit power is between a first transmit power and a second transmit power, the first transmit power represents the minimum transmit power supported by the first access point, and the second transmit power represents the maximum transmit power supported by the first access point.

[0016] In an embodiment, if the parameter configuration tag is a broadband coverage tag, the trained parameter configuration model is used to perform bandwidth parameter configuration on the first access point.

[0017] In a second aspect, an embodiment of the present application provides a parameter configuration method, including: performing an environmental scan on a first access point to be configured with parameters, and obtaining the strength of a signal sent by a second access point and received by the first access point; determining multiple adjacent access points of the first access point from the second access point according to the strength of the signal and a preset signal strength threshold; determining a feature vector of the first access point according to a scanning result of the first access point on different channels and a scanning result of the adjacent access points on different channels; performing parameter configuration processing on the feature vector of the first access point based on a trained parameter configuration model to obtain parameter configuration information corresponding to the first access point; wherein the trained parameter configuration model is trained according to the training method of the parameter configuration model described in any of the above embodiments.

[0018] In a third aspect, an embodiment of the present application provides a training device for a parameter configuration model, including:

[0019] A first scanning unit, configured to perform an environment scan on a first access point to be configured with parameters, to obtain a strength of a signal sent by a second access point and received by the first access point;

[0020] A first determining unit, configured to determine a plurality of neighboring access points of the first access point from the second access point according to the strength of the signal and a preset signal strength threshold;

[0021] A second determining unit, configured to determine a feature vector of the first access point according to a scanning result of the first access point on different channels and a scanning result of the adjacent access point on different channels;

[0022] An acquiring unit, configured to acquire parameter configuration tags of a plurality of the first access points;

[0023] A training unit is used to iteratively train an initial parameter configuration model according to the feature vectors and parameter configuration labels of the plurality of first access points to obtain a trained parameter configuration model, so as to perform parameter configuration on the first access points through the trained parameter configuration model.

[0024] In a fourth aspect, an embodiment of the present application provides a parameter configuration device, including:

[0025] A second scanning unit, configured to perform an environment scan on a first access point to be configured with parameters, to obtain a strength of a signal sent by a second access point and received by the first access point;

[0026] A third determining unit, configured to determine a plurality of neighboring access points of the first access point from the second access point according to the signal strength and a preset signal strength threshold;

[0027] a fourth determining unit, configured to determine a feature vector of the first access point according to a scanning result of the first access point on different channels and a scanning result of the adjacent access point on different channels;

[0028] a configuration unit, configured to perform parameter configuration processing on the feature vector of the first access point based on the trained parameter configuration model to obtain parameter configuration information corresponding to the first access point;

[0029] The trained parameter configuration model is obtained by training according to the parameter configuration model training method described in any of the above embodiments.

[0030] In a fifth aspect, an embodiment of the present application provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the training method of the parameter configuration model as described in any of the above embodiments or the steps of the parameter configuration method as described in any of the above embodiments are implemented.

[0031] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the training method of the parameter configuration model as described in any of the above embodiments or the steps of the parameter configuration method as described in any of the above embodiments.

[0032] The embodiment of the present application provides a training method, parameter configuration method, device and equipment for a parameter configuration model, including: performing an environmental scan on a first access point to be configured with parameters, obtaining the strength of a signal sent by a second access point received by the first access point; determining multiple adjacent access points of the first access point from the second access point according to the signal strength and a preset signal strength threshold; determining a feature vector of the first access point according to the scanning results of the first access point on different channels and the scanning results of adjacent access points on different channels; obtaining parameter configuration labels of multiple first access points; iteratively training an initial parameter configuration model according to the feature vectors and parameter configuration labels of multiple first access points to obtain a trained parameter configuration model, so as to configure the parameters of the first access point through the trained parameter configuration model. The present application improves the accuracy of wireless network parameter configuration by mapping the wireless network to the corresponding wireless network topology structure, so as to perform adaptive and fully automatic wireless network parameter configuration at the topology level from the network topology level. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The drawings constituting a part of this application are used to provide a further understanding of this application, so that other features, purposes and advantages of this application become more obvious. The schematic embodiment drawings and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0034] Figure 1 It is a flow chart of a training method of a parameter configuration model provided in an embodiment of the present application;

[0035] Figure 2 It is a flow chart of a parameter configuration method provided in an embodiment of the present application;

[0036] Figure 3 It is a structural schematic diagram of a training device for a parameter configuration model provided in an embodiment of the present application;

[0037] Figure 4 It is a structural schematic diagram of a parameter configuration device provided in an embodiment of the present application;

[0038] Figure 5 It is a schematic diagram of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0040] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the application described herein can be implemented in an order other than those illustrated or described herein.

[0041] It should be understood that in the various embodiments of the present application, the size of the serial number of each process does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0042] It should be understood that in the present application, "include" and "have" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or apparatus comprising a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0043] It should be understood that in the present application, "plurality" refers to two or more than two. "And / or" is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "Contains A, B and C", "Contains A, B, C" means that A, B, and C are all included, "Contains A, B or C" means that one of A, B, and C is included, and "Contains A, B and / or C" means that any one, any two, or any three of A, B, and C are included.

[0044] It should be understood that in this application, "B corresponding to A", "B corresponding to A", "A corresponds to B" or "B corresponds to A" means that B is associated with A and B can be determined based on A. Determining B based on A does not mean determining B based only on A, but B can also be determined based on A and / or other information. A and B match when the similarity between A and B is greater than or equal to a preset threshold.

[0045] Depending on the context, "if" as used herein may be interpreted as "when" or "when" or "in response to determining" or "in response to detecting."

[0046] The data involved in this application may be data authorized by the tester or fully authorized by all parties. The collection, dissemination, and use of the data shall comply with the relevant laws, regulations, and standards of the relevant countries and regions. The implementation methods / examples of this application may be combined with each other.

[0047] The embodiments of the present application can be applied to various scenarios such as video, voice interaction, and microphone connection.

[0048] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0049] Next, the present application scheme will be described through specific embodiments in conjunction with the accompanying drawings.

[0050] See also Figure 1 , Figure 1 A flow chart of a training method for a parameter configuration model provided in an embodiment of the present application. Figure 1 As shown, the following steps are included:

[0051] Step S101: Performing an environmental scan on a first access point to be configured with parameters, and obtaining the strength of a signal sent by a second access point and received by the first access point.

[0052] In step S101, a comprehensive environment scan is first performed on the first access point (AP) whose parameters are to be configured. This scanning process usually includes receiving wireless signals from other access points (i.e., the second access point) in the surrounding environment and measuring the strength of these signals. The strength of the signal is an important indicator for evaluating the distance and interference between two access points. By using a dedicated scanning tool (such as a Wi-Fi scanner) or software, these signal strength data can be captured and recorded in real time. These data will serve as the basis for subsequent steps to determine the network environment of the first access point.

[0053] For example, the scanning process is not limited to detecting Wi-Fi signals, but may also include other types of wireless signals, such as Bluetooth, Zigbee, etc.

[0054] The signal strength of each detected second access point (i.e., other surrounding access points) is measured by the received signal strength indication (RSSI). The higher the RSSI value, the stronger the signal, and vice versa.

[0055] Step S102: determining a plurality of neighboring access points of the first access point from the second access points according to the signal strength and a preset signal strength threshold.

[0056] In step S102, according to the signal strength data collected in step S101, the second access point adjacent to the first access point can be further screened out. This screening process is usually based on a preset signal strength threshold. The signal strength threshold is a pre-set critical value for measuring signal strength, and the signal strength threshold is usually determined based on a variety of factors such as network planning, device performance, environmental factors, and actual business needs. In actual applications, the signal strength threshold may need to be adjusted and optimized according to changes in the network environment, updates to device performance, and changes in business needs. By continuously adjusting and optimizing the signal strength threshold, it can be ensured that the determination of adjacent access points is more accurate. If the strength of a signal emitted by a second access point measured at the first access point exceeds the signal strength threshold, the second access point is considered to be adjacent to the first access point. In this way, multiple adjacent access points of the first access point can be accurately determined. These adjacent access points will have an important impact on the parameter configuration of the first access point.

[0057] In one embodiment, the method further includes:

[0058] Mapping the wireless network where the first access point is located to obtain a topological structure corresponding to the wireless network;

[0059] Among them, the wireless networking topology structure includes m vertices and n edges, m is an integer greater than or equal to 1, and n is an integer greater than or equal to 0; the vertices of the topology structure indicate the m access points in the wireless networking, the edges in the topology structure indicate the connection relationship between any two access points in the topology structure, and the weight of the edge indicates the strength of the signal.

[0060] Each vertex in the topology corresponds to an access point in the wireless network. These vertices represent physical or logical entities in the network and are nodes for information transmission and exchange.

[0061] The edge in the topology structure represents the connection relationship between any two access points in the topology structure. This connection relationship can be a physical connection (such as direct connection through a wired or wireless link) or a logical connection (such as indirect reachability through a routing protocol). The existence and number of edges directly reflect the complexity of the network structure and the degree of interconnection between access points.

[0062] In order to more accurately describe the connection relationship represented by the edge, a weight value can be assigned to each edge. The weight of the edge is usually used to indicate the strength of the signal, reflecting the communication quality or signal coverage strength between two access points. Signal strength is an important indicator for measuring the quality of wireless links, which directly affects the rate and reliability of data transmission. Therefore, using signal strength as the weight of the edge can make the topology structure closer to the actual wireless network environment and provide a more accurate basis for subsequent parameter configuration.

[0063] For example, the weight of the edge may be determined according to the strength of the signal measured in step S101 , for example, the higher the signal strength, the greater the weight value.

[0064] In an embodiment, the step of “mapping the wireless network where the first access point is located to obtain a topology structure corresponding to the wireless network” can be implemented by the following steps S1001 to S1002, specifically:

[0065] Step S1001: Acquire the wireless network where the first access point is located.

[0066] Among them, wireless networking, also known as wireless distributed system, is the process of connecting multiple wireless APs into a network through wireless communication technology to achieve data exchange and resource sharing.

[0067] In this wireless networking system, the following devices may exist:

[0068] Gateway: A gateway is a device that connects the LAN to the external network and is responsible for forwarding data between the LAN and the external network. It can enable communication between devices in the LAN and external networks (such as the Internet).

[0069] Switch: A switch is a device used to expand the scale of a network. It can connect multiple network devices, such as servers, computers, printers, etc., to establish a local area network. The switch can forward data packets based on the MAC address (physical address) to provide high-speed and stable local area network connections.

[0070] Wired AP (Access Point): Wired AP is connected to a network switch via a wired connection (such as Ethernet) and can provide wireless services to wireless clients. It is usually used to provide coverage in areas with a small range, such as offices, conference rooms, etc.

[0071] Wireless AP (Access Point): Wireless AP is also called wireless access point. It establishes a connection with wireless clients through wireless signals and provides wireless network services. Wireless AP is more flexible to deploy and can cover a wider range. It is suitable for outdoor scenes, large buildings, campuses and other areas that require wide coverage.

[0072] The basic principle of the wireless networking system is to achieve seamless coverage of the entire area through the combination of multiple wireless APs. These wireless APs can be connected to switches and gateways through wired or wireless means to form a unified network architecture. Wireless clients can connect to the nearest wireless AP and communicate with the external network through switches and gateways.

[0073] Step S1001 involves the identification and collection of the actually deployed wireless distributed system. At this stage, information of all wireless access points AP participating in the networking is collected through specific detection means or management systems. These APs may be devices that support different IEEE 802.11 standard versions (such as 11be, 11ax, 11n, etc.), and they work together to achieve wireless resource sharing and communication. Each AP plays the role of a data transmission node in the physical space and has the ability to provide wireless coverage independently or jointly.

[0074] Step S1002: Map the wireless network to obtain a wireless network topology structure corresponding to the wireless network.

[0075] When m wireless APs are configured in a wireless network, the wireless network is mapped to obtain a wireless network topology structure corresponding to the wireless network. The m wireless APs in the wireless network are mainly configured as m vertices in the wireless network topology structure, wherein the m wireless APs correspond one-to-one to the m vertices, and m is an integer greater than or equal to 1. Then, any two vertices among the m vertices are connected to form n edges, wherein n is an integer greater than or equal to 0, so that the wireless network topology structure is composed of m vertices and n edges.

[0076] Step S1002 further converts the specific layout and interaction relationship of the actual wireless network into a mathematically easy-to-handle topological structure model. In this model, each wireless AP is regarded as a vertex, and the m vertices represent various physical devices in the entire network. When there are direct or indirect communication links between the m wireless APs, n edges will be formed accordingly in the topological structure.

[0077] For example, in a simple network composed of three wireless APs, if each AP can establish two-way communication with any other AP, then a fully connected triangle topology will be formed, that is, there are two interconnected edges between the three vertices. For example, taking m=3 as an example, there are three wireless APs configured in the wireless network, namely wireless AP1, wireless AP2 and wireless AP3. When mapping this wireless network to a wireless networking topology, wireless AP1, wireless AP2 and wireless AP3 are respectively used as vertex 1, vertex 2 and vertex 3 of the wireless networking topology. Since wireless AP1, wireless AP2 and wireless AP3 communicate with each other, it can be known that vertex 1, vertex 2 and vertex 3 in the wireless networking topology are connected to form three edges, namely, the first edge formed by the connection between vertex 1 and vertex 2, the second edge formed by the connection between vertex 2 and vertex 3, and the third edge formed by the connection between vertex 1 and vertex 3. Therefore, the wireless networking topology formed is a triangle formed by the connection between vertex 1, vertex 2 and vertex 3.

[0078] Step S103: determining a feature vector of the first access point according to the scanning result of the first access point on different channels and the scanning result of the adjacent access point on different channels.

[0079] In step S103, in order to more comprehensively describe the network environment of the first access point, it is necessary to analyze its scanning results on different channels. At the same time, it is also necessary to consider the scanning results of adjacent access points on different channels. These scanning results will provide key information about channel interference, channel occupancy, etc. By analyzing this information, a feature vector of the first access point can be constructed. The feature vector usually contains multiple dimensions, such as signal strength, channel interference level, channel occupancy, etc., to fully reflect the network environment of the first access point. These feature vectors will serve as input data for subsequent training parameter configuration models.

[0080] In one embodiment, step 103 may be implemented through the following steps S1031 to S1034, specifically:

[0081] Step S1031: obtaining scanning results of the first access point on different channels and scanning results of adjacent access points on different channels.

[0082] For example, firstly, the networking information corresponding to the wireless networking is acquired, and then the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels are obtained according to the networking information.

[0083] The networking information may include detailed configuration information and real-time network status scanning information.

[0084] For example, configuration information may include quality of service settings, bandwidth allocation (such as traffic demand), and transmission power control for each wireless AP. For example, quality of service settings define how different types of traffic are prioritized, including priority, queuing strategies, etc. For example, bandwidth allocation specifies the maximum bandwidth allocated to each application or user to ensure that important applications and services can obtain sufficient resources. For example, transmission power control is used to adjust the strength of the signal sent by each wireless access point (AP) to reduce interference and optimize coverage.

[0085] For example, network status scanning information can be used to reflect the real-time environmental conditions of network operation. Network status scanning information may include channel utilization, background noise signals and external interference signals on the scanning channel, as well as neighbor scanning strength between wireless APs, spacing parameters between wireless APs, etc. For example, channel utilization reflects the ratio of the bandwidth actually used on the channel to the total available bandwidth. For example, the background noise signal represents the background noise level, which is usually composed of electromagnetic noise generated by electronic devices. For example, external interference signals may come from other wireless networks, microwave ovens, Bluetooth devices, etc. For example, neighbor scanning strength represents the measurement of signal strength between adjacent APs to help identify potential sources of interference. For example, the spacing parameter between wireless APs represents the physical distance between APs, which affects signal propagation and coverage.

[0086] Then, the scanning target, that is, which APs (including the first access point and its adjacent APs) need to be scanned and the channel range that may be affected by the scanning target can be determined according to information such as AP layout and transmission power control in the networking information.

[0087] Then, based on the network status scanning information such as channel utilization and noise floor level, the parameters of the scanning tool, such as scanning frequency, scanning time, scanning channel, etc., can be adjusted to ensure that key information can be accurately captured.

[0088] For example, for scanning the first access point: select the first access point as the scanning target; configure the scanning tool, specify the channel range to be scanned (such as channels within the 2.4GHz or 5GHz frequency band); perform the scanning operation, and record key indicators such as signal strength, channel utilization, background noise signal, external interference signal, etc. on each channel to obtain the scanning results of the first access point on different channels.

[0089] For example, for scanning adjacent access points: select multiple adjacent access points of the first access point as scanning targets; configure the scanning tool to specify the channel range to be scanned (such as channels within the 2.4 GHz or 5 GHz frequency band); perform a scanning operation to scan each adjacent access point of the first access point to obtain scanning results of each adjacent access point on different channels.

[0090] Step S1032: determining first eigenvectors of m vertices according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels.

[0091] For example, a first feature vector of m vertices is constructed according to the scanning result. The first feature vector includes the performance of each access point itself.

[0092] In one embodiment, determining first feature vectors of m vertices according to scanning results of the first access point on different channels and scanning results of adjacent access points on different channels includes:

[0093] Determine the first initial feature vectors of the m vertices according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels;

[0094] De-noising the first initial eigenvectors of the m vertices to obtain the de-noised first initial eigenvectors;

[0095] For each of the m vertices, perform neighbor sampling on each vertex to obtain the neighbor nodes of each vertex;

[0096] Aggregate the feature vectors of the neighboring nodes of each vertex to obtain the aggregated features of each vertex;

[0097] Determine a first eigenvector of each vertex based on the aggregated features of each vertex and the denoised first initial eigenvector;

[0098] The first eigenvector of each vertex is summed up to obtain the first eigenvectors of m vertices.

[0099] In the above embodiment, after m vertices are obtained, a vertex set can be formed by the m vertices, and the vertex set is recorded as V. Each vertex v among the m vertices is in V.

[0100] The first initial eigenvector for each vertex v is given by X V Indicates that X V It mainly includes the configuration information of the wireless AP (such as traffic requirements) and network status scanning information (such as channel utilization, background noise and external interference on the scanning channel).

[0101] For example, assuming there are p non-overlapping channels and q feature parameters, the first initial feature vector is X V ∈R (p+1)×qIt can be seen that the p+1 layer has q-dimensional features, that is, each channel has a corresponding feature value. In addition to the p-layer features, the additional layer serves as the channel mask of the current wireless AP. Among them, the generation and status update of the channel mask only depends on the current main channel and bandwidth settings of the wireless AP. The corresponding mask of the current coverage channel of the wireless AP is all set to 1, and the mask of other uncovered channels is all set to 0, which is used to assist the feature extraction of main channel interference and adjacent channel interference.

[0102] If the flow demand is considered, the first initial eigenvector X V The values ​​of the traffic demand dimension on each channel are kept consistent, and external interference can be represented by comprehensive parameters such as the signal energy and strength received from other APs outside the network.

[0103] During the wireless communication transmission process, due to the volatility of the wireless communication transmission process, the collected parameters may be interfered by noise. After the first initial eigenvectors of the m vertices are determined, the first initial eigenvectors of the m vertices need to be denoised to obtain the denoised first initial eigenvectors.

[0104] For example, for the first initial eigenvector of m vertices, the average value of the corresponding parameters of each vertex in the eigenvectors of all its neighboring nodes can be calculated, and this average value can be used as the parameter value of the first initial eigenvector of the vertex after denoising. For example, if the first initial eigenvector of a vertex contains the parameter of signal strength, the average value of the signal strength parameters in the eigenvectors of all its neighboring vertices can be calculated as the denoising result of the signal strength of the vertex. Repeat this process until all parameters have been denoised. Among them, the parameter values ​​between adjacent nodes should be similar under ideal conditions, while noise is often random and independent. Therefore, the similarity of the local network environment can be used to smooth and correct the parameter value of the first initial eigenvector of the vertex through the information of the neighboring nodes. By calculating the average value of the corresponding parameters in the eigenvector of the neighboring nodes, the influence of noise can be smoothed to a certain extent, thereby obtaining a more stable and accurate first initial eigenvector.

[0105] In addition, the present application also needs to perform neighbor sampling on each vertex among the m vertices to obtain the neighbor nodes of each vertex.

[0106] Before sampling each vertex, you need to set the number of samples L and the maximum number of samples S. After setting, traverse from the 1st order to the Lth order subgraph of each vertex, and sample at most S neighbor vertices in each layer of subgraph.

[0107] Specifically, when sampling each vertex, this vertex is taken as the central node, and then the neighbors connected to this central node with direct edges are taken as the first-order neighbor subgraph (i.e., the first layer, the number of hops is 1), and the k-th order neighbor means that it has an adjacent edge with the k-1-th order neighbor (i.e., the k-th layer, the number of hops is k).

[0108] The sampling method for other vertices is the same as above and will not be repeated here.

[0109] When the number of neighbor nodes is insufficient, the above sampling method can adopt sampling with replacement or other supplementary methods, including but not limited to uniform random sampling, importance sampling, Monte Carlo sampling, etc.

[0110] It should be noted that the sampling hop count and the maximum sampling count need to be determined according to the actual application scenario or based on the final test results of the overall network. If you want to consider the impact of a wider range of nodes when adjusting the channel, you need to set a higher sampling hop count, and 2 is generally sufficient in general scenarios; the maximum sampling count needs to be considered based on the actual wireless AP deployment density. The higher the density, the higher the recommended sampling count.

[0111] After the feature vectors of the neighboring nodes of each vertex are obtained through sampling, the feature vectors of the neighboring nodes of each vertex need to be aggregated to obtain the aggregated features of each vertex.

[0112] Assume that the number of sampling hops is L. When performing neighbor aggregation, we traverse from the Lth order to the 1st order subgraph and aggregate the subgraph features of each layer step by step.

[0113] Specifically, let the first-layer node feature of vertex vi be It depends on the l-1th layer node features of neighboring node N(vi)

[0114] Assume that the feature mapping relationship of the neighbor node aggregation of vertex vi is expressed as: Among them, aggr() represents the aggregation operation, and average aggregation, pooling aggregation, etc. can be used.

[0115] The aggregation features of each vertex are calculated by the above formula Then, the aggregated features of each vertex are merged with the first initial feature vector of the vertex after denoising, and the first feature vector of each vertex can be obtained.

[0116] Step S1033: determining second eigenvectors of n edges according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels.

[0117] For example, a second eigenvector of each edge is constructed according to the scanning result, and the second eigenvector represents the interaction between two vertices on each channel.

[0118] In one embodiment, determining the second feature vectors of n edges according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels includes:

[0119] Determine the second initial eigenvectors of n adjacency matrices corresponding to the n edges according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels;

[0120] De-noising the second initial eigenvectors of the n adjacency matrices corresponding to the n edges to obtain the de-noised second initial eigenvectors of the n adjacency matrices, wherein the n edges correspond to the n adjacency matrices one-to-one;

[0121] For each of the n edges, obtain a denoised eigenvector matching each edge from the denoised second initial eigenvectors of the n adjacency matrices;

[0122] The denoised feature vector matched with each edge is used as the second feature vector of each edge;

[0123] The second eigenvector of each edge is summed up to obtain the second eigenvectors of n edges.

[0124] In the above embodiment, the first eigenvectors of m vertices are obtained, and the n edges of the wireless networking topology structure can be determined through the m vertices. The n edges are summarized into the edge set E. Each edge e(v i ,v j )∈E, where e(v i ,v j ) represents the connecting edge between vertex i and vertex j, and its corresponding adjacency matrix is ​​recorded as A∈R |V|×|V| , element a in the adjacency matrix i,j represents the neighbor scanning intensity of vertex i and vertex j, N(v i ) represents the neighbor set of vertex i.

[0125] Among them, N(v i ) needs to be generated in combination with the vertex channel mask. Let Xv i and X vj The corresponding channel mask feature is W vi and W vj , v j ∈N(v i ) if and only if W vi and W vj The dot product is not 0, which means that the coverage channels of the two wireless APs overlap.

[0126] For the convenience of processing, the adjacency matrix can add a fixed value to the scanning intensity to make it in the positive range. If there is no scanning relationship between two vertices, the corresponding value is set to 0. In addition, in addition to describing the scanning intensity between wireless APs, the elements in the adjacency matrix can also be used to describe the parameters of the spacing between wireless APs, such as the actual distance between wireless APs.

[0127] This application uses the elements in the adjacency matrix as edge input features. Therefore, it is necessary to first obtain the second initial eigenvectors of the n adjacency matrices corresponding to the n edges, and then denoise the second initial eigenvectors of the n adjacency matrices to obtain the denoised second initial eigenvectors of the n adjacency matrices. In addition, the average value of the two vertices corresponding to the edge can also be calculated as the denoised second initial eigenvector of the adjacency matrix.

[0128] After denoising, the denoised eigenvector matching each edge can be directly obtained from the denoised second initial eigenvectors of the n adjacency matrices, and then the denoised eigenvector matching each edge is used as the second eigenvector of each edge. Finally, the second eigenvector of each edge is summarized to obtain the second eigenvectors of n edges.

[0129] Step S1034: Determine the eigenvector of the first access point according to the first eigenvectors of the m vertices and the second eigenvectors of the n edges.

[0130] For example, the first feature vectors of the m vertices and the second feature vectors of the n edges are integrated to form a comprehensive feature vector of the first access point, which integrates the performance of the access point itself and its interaction with neighboring access points.

[0131] After obtaining the first eigenvectors of the m vertices and the second eigenvectors of the n edges, the first eigenvectors of the m vertices and the second eigenvectors of the n edges can be used as input features of the model in the initialization stage.

[0132] The first eigenvectors of the m vertices and the second eigenvectors of the n edges obtained through the above embodiment constitute a training data set.

[0133] Step S104: Acquire parameter configuration tags of multiple first access points.

[0134] In step S104, it is necessary to collect parameter configuration tags of the first access point with multiple known parameter configurations. These tags are usually the optimal parameter configuration schemes obtained through expert experience or experimental tests. These tags may include multiple types such as RF switch status, channel deployment, transmission power, and bandwidth coverage, each of which corresponds to different aspects of network performance. These tags will serve as supervision information for the training parameter configuration model to guide the training process of the model. By comparing the differences between the parameter configuration scheme output by the model and the tags, the performance of the model can be continuously optimized.

[0135] In an embodiment, the parameter configuration tag may be a tag of a radio frequency switch state, that is, a tag indicating whether the radio frequency is turned off or on. The tag of the radio frequency switch state is used as an input parameter of an initial parameter configuration model for configuring frequency band parameters for the first access point.

[0136] For example, the parameter configuration tag of the current vertex is It directly determines whether the current vertex participates in wireless communication. In the label, c i =0 means the current vertex turns off the radio frequency and does not participate in communication; c i = 0 means that the current vertex turns on the radio frequency and is ready for data transmission. This binary label simplifies the learning process of radio frequency control, allowing the model to quickly learn when to turn on or off the radio frequency.

[0137] In an embodiment, the parameter configuration tag may be a channel deployment tag, that is, a tag characterizing a deployment target channel or an undeployed target channel, and is used as an input parameter of an initial parameter configuration model for performing channel parameter configuration on the first access point.

[0138] For example, the parameter configuration label of the current vertex is c∈R n , reflects the deployment of the current vertex on multiple channels. n represents the number of deployed channels, c i =1 means the current vertex is deployed to the target channel i; c i = 0 means that the current vertex is not deployed to the target channel i. This multi-valued label enables the model to learn how to allocate resources among multiple channels to optimize the overall performance of the network. By comparing the differences between the channel deployment plan output by the model and the label, the model can continuously adjust its parameter configuration strategy to get closer to the optimal solution.

[0139] In one embodiment, the parameter configuration tag can be a transmission power tag, which serves as an input parameter of an initial parameter configuration model for performing power parameter configuration on a first access point, wherein the transmission power is between a first transmission power and a second transmission power, the first transmission power represents the minimum transmission power supported by the first access point, and the second transmission power represents the maximum transmission power supported by the first access point.

[0140] For example, the parameter configuration label of the current vertex is c∈R, which provides the model with a specific value of the AP transmit power. i Indicates the transmission power of the current vertex, the transmission power c i In (p min ,p max ) range, p min is the minimum transmit power supported by the AP, p max The maximum transmit power supported by the AP. By learning these tags, the model can learn how to adjust the AP's transmit power according to the network environment and business requirements to achieve the best communication effect and coverage.

[0141] In an embodiment, the parameter configuration tag may be a broadband coverage tag, a tag representing a covered target channel or an uncovered target channel, and is used as an input parameter of an initial parameter configuration model for performing bandwidth parameter configuration on the first access point.

[0142] The parameter configuration label of the current vertex is c∈R n , indicating the coverage of AP on all channels. n indicates the number of all channels. c i =1 means the current vertex has a coverage target channel i; c i =0 means that the current vertex does not cover the target channel i. This label allows the model to learn how to allocate bandwidth and optimize coverage on multiple channels. When processing this multi-label output, that is, there can be multiple elements of 1 in c, and common threshold segmentation, Random k-Labelsets, label sorting method, classifier chain method, etc. can be used for conversion and processing. Finally, by parsing the output of the model, the specific deployment bandwidth and main channel selection scheme of the current vertex can be obtained. For example, when obtaining the final deployment result, the channel specified by the channel mask of the vertex is used as the starting point, and the connected domain containing the most consecutive "1" elements is selected as the deployment bandwidth. If the main channel needs to be determined at the same time, it is selected according to the output weight.

[0143] Step S105: iteratively training an initial parameter configuration model according to the feature vectors and parameter configuration labels of the plurality of first access points to obtain a trained parameter configuration model, and performing parameter configuration on the first access points through the trained parameter configuration model.

[0144] In step S105, the feature vector obtained in step S103 and the parameter configuration label obtained in step S104 are used as training data to iteratively train the initial parameter configuration model. The training process usually adopts supervised learning algorithms in machine learning methods, such as neural networks, decision trees, etc. By continuously optimizing the parameters and structure of the model, the model can accurately predict the optimal parameter configuration solution based on the input feature vector. The trained parameter configuration model finally obtained will be able to automatically configure the parameters of the new first access point without manual intervention.

[0145] In one embodiment, the initial parameter configuration model is iteratively trained according to the feature vectors and parameter configuration tags of the plurality of first access points to obtain a trained parameter configuration model, including:

[0146] According to the feature vectors and parameter configuration labels of the plurality of first access points, the initial parameter configuration model is iteratively trained, and when the loss function satisfies a preset condition, a trained parameter configuration model is obtained.

[0147] During training, the reward gain is the inverse or negated relationship of the overall interference, throughput performance or negotiation rate, etc. Therefore, the loss function can be the interference loss, as follows:

[0148]

[0149] Among them, the deployment channel of vertex i is j, cl i,j ,nf ij ,inf ij , L represent the interference-related parameters such as channel utilization, background noise, internal interference, etc. scanned on the deployment channel of vertex i on j.

[0150] Among them, the loss function includes a feature mapping function and a feature edge mapping function;

[0151] According to the feature vectors and parameter configuration labels of the plurality of first access points, the initial parameter configuration model is iteratively trained, and when the loss function satisfies a preset condition, a trained parameter configuration model is obtained, including:

[0152] Based on the first eigenvectors of the m vertices and the eigenmapping function, obtaining updated first eigenvectors corresponding to the m vertices;

[0153] Based on the second eigenvectors of the n edges and the eigenedge mapping function, obtaining updated second eigenvectors corresponding to the n edges;

[0154] Based on the updated first feature vectors corresponding to the m vertices, the updated second feature vectors corresponding to the n edges, and the parameter configuration labels, the initial parameter configuration model is trained, and the trained parameter configuration model is obtained when the feature mapping function and the feature edge mapping function meet the preset number of iterations.

[0155] During the iterative training process, the feature states of vertices and edges need to be continuously updated.

[0156] In the initialization phase, for each vertex, the first initial eigenvector of the vertex is obtained by X V As the vertex input feature (ie, the first eigenvector of the vertex), for each edge, the second initial eigenvector of the adjacency matrix corresponding to the edge is the edge input feature (ie, the second eigenvector of the edge).

[0157] In the state update phase, the first eigenvector of the vertex and the second eigenvector of the edge need to update the scanning information and preprocessing results in time according to the last deployment result, so as to obtain the updated first eigenvector and the updated second eigenvector.

[0158] For example, as the state is updated, the updated first eigenvectors corresponding to the m vertices and the updated second eigenvectors corresponding to the n edges can be obtained based on the first eigenvectors of the m vertices and the second eigenvectors of the n edges, respectively. Specifically, based on the first eigenvectors of the m vertices and the characteristic mapping function, the updated first eigenvectors corresponding to the m vertices are obtained; based on the second eigenvectors of the n edges and the characteristic edge mapping function, the updated second eigenvectors corresponding to the n edges are obtained.

[0159] The feature mapping function can be used to process the first feature vectors of m vertices and generate updated first feature vectors corresponding to the m vertices. These updated first feature vectors will better reflect the attributes of the vertex and the relationship with other vertices, thereby helping to improve the prediction performance of the model. During the training process, the feature mapping function will continuously adjust according to the feedback of the loss function to optimize the feature representation of the vertex.

[0160] Among them, the feature mapping function is expressed by the following formula:

[0161]

[0162] Among them, concat() represents the common concatenation operation in neural networks. Represents feature mapping relationships, which can be expressed through other networks, such as multi-layer artificial neural networks or convolutional neural networks. represents the aggregated features of the neighboring nodes of vertex vi at the lth layer, represents the node features of the l-1th layer of vertex vi, Represents the updated second eigenvector of the edge at the lth level of vertex vi.

[0163] Among them, the feature edge mapping function can be used to process the second feature vectors of n edges and generate updated second feature vectors corresponding to the n edges. These updated second feature vectors will more accurately describe the properties of the edges and the relationship between the vertices connected by the edges. Similarly, the feature edge mapping function will also be adjusted according to the feedback of the loss function during the training process to optimize the feature representation of the edges.

[0164] Among them, the characteristic edge mapping function is expressed by the following formula;

[0165]

[0166] in, The feature edge mapping relationship can be expressed through other networks, such as multi-layer artificial neural networks or convolutional neural networks or aggregation operations. represents the node feature of the l-1th layer of vertex vi, represents the aggregated features of the neighboring nodes of the vertex vi at the l-1th layer, Represents the updated second eigenvector of the edge at the l-1th level of vertex vi.

[0167] During the training process, the initial parameter configuration model is trained based on the updated feature vectors corresponding to the m vertices and the updated feature vectors corresponding to the n edges. As the training progresses, the feature mapping function and the feature edge mapping function are continuously optimized, so that the prediction performance of the model is gradually improved. When the loss function meets the preset conditions (such as reaching the preset number of iterations or the loss value is less than a certain threshold), the model can be considered to have converged, and the model obtained at this time is the trained parameter configuration model.

[0168] After obtaining the trained parameter configuration model, the first eigenvector of the vertices and the second eigenvector of the edges in any wireless networking topology can be extracted according to any wireless networking topology, and then the first eigenvector of the vertices and the second eigenvector of the edges are input into the trained parameter configuration model, and the wireless network parameters of this wireless network can be output. The wireless network parameters include but are not limited to frequency band, bandwidth, power, channel, etc.

[0169] In one embodiment, if the parameter configuration tag is a tag of a radio frequency switch state, the trained parameter configuration model is used to perform frequency band parameter configuration on the first access point.

[0170] Among them, the parameter configuration label can be a label of the RF switch state, that is, a label that represents whether the RF is turned off or on. For example, the configuration of the frequency band is used to determine the optimal operating frequency range for each device or link to achieve the best spectrum utilization and minimum interference. In the process of frequency band configuration, the AP's own characteristics (the first eigenvector of the vertex) and its neighbor's influence characteristics (the second eigenvector of the edge) are fully considered, so as to provide a more balanced and reasonable frequency band configuration result.

[0171] During the training process, the model gradually mastered the ability to predict the optimal RF switch state based on the AP's own characteristics (the first eigenvector of the vertex) and its neighbor's influence characteristics (the second eigenvector of the edge) by learning a large number of data sets containing RF switch state labels. These characteristics may include AP signal strength, channel occupancy, distribution of neighbor APs, business traffic demand, etc.

[0172] Once the model training is completed, the trained parameter configuration model can be used to configure the frequency band parameters of the new first access point. Specifically, the trained parameter configuration model will predict the optimal RF switch state of each AP based on the first eigenvector of the input vertex and the second eigenvector of the edge, combined with the learned knowledge and experience, and determine the optimal operating frequency range of each device or link accordingly.

[0173] Among them, the trained parameter configuration model will be able to predict the optimal RF switch state of each AP in a given network environment. For example, the model can automatically recommend where to turn on or off the AP's RF to optimize network coverage and performance. For example, as the network environment changes, the model can adjust the AP's RF state in real time to adapt to changes in traffic and new user access requests. For example, when an AP fails or is under maintenance, the model can quickly adjust the RF state of surrounding APs to ensure service continuity. For example, when adding new APs or performing network upgrades, the model can provide RF configuration recommendations to ensure compatibility between new and old devices and optimization of network performance.

[0174] In one embodiment, if the parameter configuration tag is a broadband coverage tag, the trained parameter configuration model is used to perform bandwidth parameter configuration on the first access point.

[0175] Among them, the parameter configuration label can be a label for broadband coverage, a label that characterizes the coverage of the target channel or the non-coverage of the target channel, for example, the bandwidth configuration is used to intelligently allocate appropriate bandwidth resources to each wireless access point or link to ensure network transmission efficiency and fairness. In the process of bandwidth configuration, the AP's own characteristics (the first eigenvector of the vertex) and its neighbor's influence characteristics (the second eigenvector of the edge) are fully considered, so as to give a more balanced and reasonable bandwidth configuration result.

[0176] In wireless communication networks, the allocation of bandwidth resources is the key to ensuring efficient network operation. When parameter configuration tags are set as broadband coverage tags, these tags appear in the form of "covered target channel" or "uncovered target channel", which actually reflects the bandwidth occupancy of APs on different channels. By using this occupancy as a training tag for the model, the model can learn how to make the best bandwidth configuration decisions in a complex network environment.

[0177] Intelligent bandwidth allocation can dynamically adjust bandwidth resources according to the actual situation of the network, ensuring that key businesses have sufficient bandwidth support while optimizing the transmission efficiency and fairness of the overall network.

[0178] In step S105, the initial parameter configuration model learns the mapping relationship between the feature vectors of multiple first access points (including the first feature vectors of vertices and the second feature vectors of edges) and the broadband coverage labels through iterative training. These feature vectors fully reflect the AP's own characteristics (such as processing power, signal strength, business requirements, etc.) and its relationship with neighboring APs (such as interference, spectrum overlap, etc.). By making full use of this feature information, the model can more accurately understand the actual situation of the network and make more reasonable bandwidth configuration decisions accordingly.

[0179] During bandwidth configuration, the trained parameter configuration model comprehensively considers the AP's own characteristics and the influence of its neighbors. This means that when allocating bandwidth resources, the model not only considers the needs of a single AP, but also the balance and fairness of the entire network. In this way, the model can avoid excessive concentration or waste of bandwidth resources and ensure that each AP can obtain appropriate bandwidth resources to support its business operations. Ultimately, this balanced and reasonable bandwidth configuration result will help improve the overall transmission efficiency and user experience of the network.

[0180] This embodiment is applicable to wireless communication networks of various sizes, especially in scenarios that require efficient bandwidth management and optimized user experience. For example, in scenarios such as large data centers, enterprise parks, or smart cities, by applying this intelligent bandwidth configuration model, the transmission efficiency and stability of the network can be significantly improved, and the operation and maintenance costs can be reduced. In addition, with the development and application of new technologies such as 5G and IoT, the demand for intelligent bandwidth configuration will be more urgent, and this embodiment will also show greater application potential and advantages.

[0181] In one embodiment, if the parameter configuration label is a transmit power label, the trained parameter configuration model is used to perform power parameter configuration on the first access point, wherein the transmit power is between a first transmit power and a second transmit power, the first transmit power represents the minimum transmit power supported by the first access point, and the second transmit power represents the maximum transmit power supported by the first access point.

[0182] The parameter configuration tag may be a transmit power tag. For example, the power configuration is used to set the transmit power of each wireless device, balance coverage and power consumption, and reduce unnecessary electromagnetic pollution. In the power configuration process, the AP's own characteristics (the first eigenvector of the vertex) and its neighbor's influence characteristics (the second eigenvector of the edge) are fully considered, so as to provide a more balanced and reasonable power configuration result.

[0183] In wireless communication networks, transmit power is one of the key factors affecting coverage, signal quality, power consumption, and electromagnetic environment. Reasonable transmit power configuration can effectively reduce energy consumption and electromagnetic pollution while ensuring communication quality, thereby improving the overall efficiency of the network. However, due to the complex and changeable network environment, manual configuration of transmit power is often difficult to achieve the optimal effect, so intelligent algorithms are needed to achieve automatic configuration.

[0184] In the embodiment of the present application, the parameter configuration tags are labels of transmit power, which represent the optimal transmit power values ​​of each access point (AP) in different network environments. These values ​​are usually obtained through experimental testing, simulation or expert experience, aiming to balance coverage and power consumption while reducing interference to other devices. The transmit power label is between the minimum transmit power (first transmit power) and the maximum transmit power (second transmit power) supported by the AP, providing a clear optimization target for model training.

[0185] During the training process, the initial parameter configuration model receives a large number of training samples, each of which contains the feature vector of the first access point (the first feature vector of the vertex and the second feature vector of the edge) and its corresponding transmit power label. The model gradually establishes the mapping relationship between the feature vector and the transmit power by learning these samples. As the number of iterations increases, the model's ability to predict the transmit power configuration will continue to improve until it converges to a stable state.

[0186] Eigenvectors play a vital role in the model training process. The first eigenvector of the vertex describes the characteristics of the AP itself, such as hardware performance, current load, business requirements, etc., while the second eigenvector of the edge reflects the mutual influence between APs, such as channel interference, spectrum overlap, and the transmit power of neighboring APs. By comprehensively considering these two types of eigenvectors, the model can understand the network environment more comprehensively and make more reasonable transmit power configuration decisions accordingly.

[0187] During the power configuration process, the trained parameter configuration model fully considers the AP's own characteristics and the influence of its neighbors, striving to minimize power consumption and reduce electromagnetic pollution while ensuring communication quality. Through the intelligent prediction of the model, each AP can be configured with a transmit power value that meets its own needs and takes into account the influence of its neighbors, thereby achieving a balanced and reasonable configuration effect.

[0188] Among them, the trained parameter configuration model can predict the most appropriate transmit power configuration based on the characteristics of each AP and the interactions between them. For example, in a specific environment, the trained parameter configuration model may recommend that some APs use higher transmit power to support the connection needs of long-distance devices while avoiding interference caused by channel overlap with other APs. At the same time, the model may also recommend that some APs maintain lower transmit power to reduce unnecessary power consumption and electromagnetic pollution, especially in indoor environments or equipment-intensive areas.

[0189] In one embodiment, if the parameter configuration tag is a tag of channel deployment, the trained parameter configuration model is used to perform channel parameter configuration on the first access point.

[0190] Among them, the parameter configuration label can be a label for channel deployment, that is, a label that characterizes a deployed target channel or an undeployed target channel. For example, channel configuration is used to reasonably arrange the wireless channels used by various devices in a multi-channel environment, reduce co-channel interference, and improve the overall network throughput and stability. In the channel configuration process, the AP's own characteristics (the first eigenvector of the vertex) and its neighbor influence characteristics (the second eigenvector of the edge) are fully considered, thereby providing a more balanced and reasonable channel configuration result.

[0191] In wireless communication networks, channel deployment labels serve as key inputs for parameter configuration and directly guide the training process of the channel parameter configuration model. These labels clearly indicate which channels have been deployed and used by specific devices (such as APs) and which channels have not yet been deployed. By combining these labels with the feature vectors of the APs, the model can learn how to intelligently configure channel resources based on the network environment and business needs.

[0192] In step S105, the initial parameter configuration model gradually learns how to predict the optimal channel configuration scheme based on the AP's feature vectors (including the first feature vector of the vertex and the second feature vector of the edge) through iterative training. These feature vectors fully describe the AP's own properties (such as signal strength, processing power, business requirements, etc.) and its interaction with neighboring APs (such as co-channel interference, spectrum overlap, etc.). By comprehensively considering these factors, the model can formulate a balanced and reasonable channel configuration strategy to reduce co-channel interference and improve the overall performance of the network.

[0193] The trained parameter configuration model fully considers the characteristics of the AP itself and its neighbor influence characteristics during the channel configuration process. The model not only focuses on the communication needs of a single AP, but also considers the coordination and stability of the entire network. Through the intelligent channel allocation algorithm, the model can ensure that key services obtain sufficient channel resources while avoiding interference from non-key services to key services. In addition, the model can dynamically adjust the channel configuration plan according to changes in network conditions to adapt to different business needs and network environments.

[0194] Applying the trained channel configuration model to the actual wireless network environment can significantly improve the performance and stability of the network. Through intelligent channel configuration, the network can better cope with the challenges of increasing devices and complex services, reduce co-channel interference, and improve network throughput. At the same time, because the model can automatically adapt to network changes and make corresponding adjustments, the need and cost of manual intervention are reduced.

[0195] It should be noted that when configuring any of the frequency band, bandwidth, power, and channel parameters in the wireless network, the training data set used by the trained parameter configuration model is matched according to the parameters, that is, for different parameters, the corresponding training data sets are different.

[0196] The embodiment of the present application provides a training method for a parameter configuration model, including: performing an environmental scan on a first access point to be configured with parameters, obtaining the strength of a signal sent by a second access point received by the first access point; determining multiple adjacent access points of the first access point from the second access point according to the signal strength and a preset signal strength threshold; determining a feature vector of the first access point according to the scanning results of the first access point on different channels and the scanning results of adjacent access points on different channels; obtaining parameter configuration labels of multiple first access points; iteratively training an initial parameter configuration model according to the feature vectors and parameter configuration labels of multiple first access points to obtain a trained parameter configuration model, so as to configure the parameters of the first access point through the trained parameter configuration model. The present application improves the accuracy of wireless network parameter configuration by mapping the wireless network to the corresponding wireless network topology structure, so as to perform adaptive and fully automatic wireless network parameter configuration at the topology level from the network topology level.

[0197] See also Figure 2 , Figure 2 A flow chart of a parameter configuration method provided in an embodiment of the present application. Figure 2 As shown, the following steps are included:

[0198] Step S201: Performing an environmental scan on a first access point to be configured with parameters, and obtaining the strength of a signal sent by a second access point and received by the first access point.

[0199] In step S201, a comprehensive environment scan is first performed on the first access point (AP) whose parameters are to be configured. This scanning process usually includes receiving wireless signals from other access points (i.e., the second access point) in the surrounding environment and measuring the strength of these signals. The strength of the signal is an important indicator for evaluating the distance and interference between two access points. By using a dedicated scanning tool (such as a Wi-Fi scanner) or software, these signal strength data can be captured and recorded in real time. These data will serve as the basis for subsequent steps to determine the network environment of the first access point.

[0200] Step S202: determining a plurality of adjacent access points of the first access point from the second access points according to the signal strength and a preset signal strength threshold.

[0201] In step S202, according to the signal strength data collected in step S201, second access points adjacent to the first access point can be further screened. This screening process is usually based on a preset signal strength threshold. If the strength of a signal emitted by a second access point measured at the first access point exceeds the signal strength threshold, the second access point is considered to be adjacent to the first access point. In this way, multiple adjacent access points of the first access point can be accurately determined. These adjacent access points will have an important impact on the parameter configuration of the first access point.

[0202] Step S203: determining a feature vector of the first access point according to the scanning results of the first access point on different channels and the scanning results of adjacent access points on different channels.

[0203] In step S203, in order to more comprehensively describe the network environment of the first access point, it is necessary to analyze its scanning results on different channels. At the same time, it is also necessary to consider the scanning results of adjacent access points on different channels. These scanning results will provide key information about channel interference, channel occupancy, etc. By analyzing this information, a feature vector of the first access point can be constructed, and the feature vector of the first access point can include the first feature vectors of m vertices and the second feature vectors of n edges. The feature vector not only contains the attributes of the first access point itself (the first feature vector of the vertex), but also reflects the interaction relationship between it and the adjacent access points (the second feature vector of the edge). These feature vectors are important inputs for the subsequent parameter configuration model.

[0204] Step S204: performing parameter configuration processing on the feature vector of the first access point based on the trained parameter configuration model to obtain parameter configuration information corresponding to the first access point;

[0205] The trained parameter configuration model is obtained by training according to the training method of the parameter configuration model in any of the above embodiments.

[0206] In step S204, the system uses the trained parameter configuration model to perform parameter configuration processing on the feature vector of the first access point. This model is trained based on a large amount of historical data and prior knowledge, and can predict the optimal parameter configuration solution based on the input feature vector. According to different training labels (RF switch status, channel deployment, transmission power, broadband coverage, etc.), the model will output corresponding parameter configuration information, so as to be applicable to different types of parameter configuration tasks.

[0207] In an embodiment, if the trained parameter configuration model is obtained based on the tag training of the radio frequency switch state, the frequency band parameter configuration of the first access point is performed based on the trained parameter configuration model to obtain the frequency band parameter configuration information corresponding to the first access point.

[0208] RF switch state configuration: If the model is trained based on the label of the RF switch state, it will predict the RF switch state that the first access point should turn on or off on different frequency bands based on the feature vector to achieve optimal utilization of frequency band resources.

[0209] In an embodiment, if the trained parameter configuration model is obtained based on label training of channel deployment, channel parameter configuration is performed on the first access point based on the trained parameter configuration model to obtain channel parameter configuration information corresponding to the first access point.

[0210] Channel deployment configuration: If the model is trained based on channel deployment labels, it will assign the best wireless channel to the first access point to reduce co-channel interference and improve network performance. This usually involves evaluating multiple channels and selecting those with the least interference and the best performance for configuration.

[0211] In an embodiment, if the trained parameter configuration model is obtained based on the label training of the transmit power, the power parameter configuration of the first access point is performed based on the trained parameter configuration model to obtain the power parameter configuration information corresponding to the first access point.

[0212] Transmit power configuration: In the scenario of transmit power configuration, the model will predict the transmit power level that the first access point should adopt based on the feature vector. This helps to balance the relationship between coverage, signal quality and energy consumption to achieve the best network performance.

[0213] In an embodiment, if the trained parameter configuration model is obtained based on label training of broadband coverage, bandwidth parameter configuration is performed on the first access point based on the trained parameter configuration model to obtain bandwidth parameter configuration information corresponding to the first access point.

[0214] Bandwidth configuration: For models trained based on broadband coverage labels, it will configure appropriate bandwidth parameters for the first access point. This helps ensure that the network can support high-speed data transmission and multimedia applications while avoiding unnecessary bandwidth waste.

[0215] The embodiment of the present application provides a parameter configuration method, including: performing an environmental scan on a first access point to be configured with parameters, obtaining the strength of a signal sent by a second access point received by the first access point; determining multiple adjacent access points of the first access point from the second access point according to the signal strength and a preset signal strength threshold; determining a feature vector of the first access point according to the scanning results of the first access point on different channels and the scanning results of adjacent access points on different channels; performing parameter configuration processing on the feature vector of the first access point based on a trained parameter configuration model, obtaining parameter configuration information corresponding to the first access point. The present application improves the accuracy of wireless network parameter configuration by mapping the wireless network to the corresponding wireless network topology structure, so as to perform adaptive and fully automatic wireless network parameter configuration at the topology structure level from the network topology structure level.

[0216] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0217] The following is an embodiment of the device of the present application. For details not described in detail, please refer to the corresponding method embodiment described above.

[0218] Figure 3 A schematic diagram of the structure of a training device for a parameter configuration model provided in an embodiment of the present application is shown. For ease of description, only the parts related to the embodiment of the present application are shown. The training device 300 for the parameter configuration model includes:

[0219] The first scanning unit 301 is used to perform an environment scan on a first access point to be configured with parameters, and obtain the strength of a signal sent by a second access point and received by the first access point;

[0220] A first determining unit 302, configured to determine a plurality of neighboring access points of the first access point from the second access point according to the signal strength and a preset signal strength threshold;

[0221] A second determining unit 303, configured to determine a feature vector of the first access point according to a scanning result of the first access point on different channels and a scanning result of an adjacent access point on different channels;

[0222] An acquiring unit 304 is configured to acquire parameter configuration tags of multiple first access points;

[0223] The training unit 305 is configured to iteratively train the initial parameter configuration model according to the feature vectors and parameter configuration labels of the plurality of first access points to obtain a trained parameter configuration model, so as to perform parameter configuration on the first access points through the trained parameter configuration model.

[0224] In one embodiment, the training device 300 for the parameter configuration model further includes:

[0225] A mapping unit is used to map the wireless network where the first access point is located to obtain a topological structure corresponding to the wireless network; wherein the wireless network topological structure includes m vertices and n edges, m is an integer greater than or equal to 1, and n is an integer greater than or equal to 0; the vertices of the topological structure indicate the m access points in the wireless network, the edges in the topological structure indicate the connection relationship between any two access points in the topological structure, and the weight of the edge indicates the strength of the signal.

[0226] In an embodiment, the second determining unit 303 may be used to: obtain scanning results of the first access point on different channels and scanning results of adjacent access points on different channels; determine first feature vectors of m vertices according to the scanning results of the first access point on different channels and scanning results of adjacent access points on different channels; determine second feature vectors of n edges according to the scanning results of the first access point on different channels and scanning results of adjacent access points on different channels; determine a feature vector of the first access point according to the first feature vectors of the m vertices and the second feature vectors of the n edges.

[0227] In an embodiment, when the second determining unit 303 determines the first feature vectors of m vertices according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels, it can be used to: determine the first initial feature vectors of the m vertices according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels; denoise the first initial feature vectors of the m vertices to obtain the denoised first initial feature vectors; for each vertex of the m vertices, perform neighbor sampling on each vertex to obtain the neighbor nodes of each vertex; aggregate the feature vectors of the neighbor nodes of each vertex to obtain the aggregated features of each vertex; determine the first feature vector of each vertex based on the aggregated features of each vertex and the denoised first initial feature vector; summarize the first feature vectors of each vertex to obtain the first feature vectors of the m vertices.

[0228] In an embodiment, when the second determining unit 303 determines the second eigenvectors of the n edges according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels, it can be used to: determine the second initial eigenvectors of the n adjacency matrices corresponding to the n edges according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels; perform denoising on the second initial eigenvectors of the n adjacency matrices corresponding to the n edges to obtain the denoised second initial eigenvectors of the n adjacency matrices, wherein the n edges correspond to the n adjacency matrices one-to-one; for each of the n edges, obtain a denoised eigenvector matching each edge from the denoised second initial eigenvectors of the n adjacency matrices; use the denoised eigenvector matching each edge as the second eigenvector of each edge; and summarize the second eigenvectors of each edge to obtain the second eigenvectors of the n edges.

[0229] In an embodiment, the training unit 305 may be used to iteratively train the initial parameter configuration model according to the feature vectors and parameter configuration labels of the plurality of first access points, and obtain the trained parameter configuration model when the loss function satisfies a preset condition.

[0230] In one embodiment, the loss function includes a feature mapping function and a feature edge mapping function;

[0231] The training unit 305 can be used to: obtain the updated first feature vectors corresponding to the m vertices based on the first feature vectors and feature mapping functions of the m vertices; obtain the updated second feature vectors corresponding to the n edges based on the second feature vectors and feature edge mapping functions of the n edges; train the initial parameter configuration model based on the updated first feature vectors corresponding to the m vertices, the updated second feature vectors corresponding to the n edges, and the parameter configuration labels, and obtain the trained parameter configuration model when the feature mapping function and the feature edge mapping function meet the preset number of iterations.

[0232] In one embodiment, if the parameter configuration tag is a tag of a radio frequency switch state, the trained parameter configuration model is used to perform frequency band parameter configuration on the first access point.

[0233] In one embodiment, if the parameter configuration tag is a tag of channel deployment, the trained parameter configuration model is used to perform channel parameter configuration on the first access point.

[0234] In one embodiment, if the parameter configuration label is a transmit power label, the trained parameter configuration model is used to perform power parameter configuration on the first access point, wherein the transmit power is between a first transmit power and a second transmit power, the first transmit power represents the minimum transmit power supported by the first access point, and the second transmit power represents the maximum transmit power supported by the first access point.

[0235] In one embodiment, if the parameter configuration tag is a broadband coverage tag, the trained parameter configuration model is used to perform bandwidth parameter configuration on the first access point.

[0236] Figure 4 A schematic diagram of the structure of a parameter configuration device provided in an embodiment of the present application is shown. For the sake of convenience, only the parts related to the embodiment of the present application are shown. The parameter configuration device 400 includes:

[0237] The second scanning unit 401 is configured to perform an environment scan on the first access point to be configured with parameters, and obtain the strength of a signal sent by the second access point and received by the first access point;

[0238] A third determining unit 402 is configured to determine a plurality of neighboring access points of the first access point from the second access point according to the signal strength and a preset signal strength threshold;

[0239] A fourth determining unit 403, configured to determine a feature vector of the first access point according to a scanning result of the first access point on different channels and a scanning result of an adjacent access point on different channels;

[0240] A configuration unit 404 is configured to perform parameter configuration processing on the feature vector of the first access point based on the trained parameter configuration model to obtain parameter configuration information corresponding to the first access point;

[0241] The trained parameter configuration model is obtained by training according to the training method of the parameter configuration model in any of the above embodiments.

[0242] This application Figure 5 A schematic diagram of a computer device is provided. Figure 5 As shown, the computer device 500 of this embodiment includes: a processor 501, a memory 502, and a computer program 503 stored in the memory 502 and executable on the processor 501. When the processor 501 executes the computer program 503, the steps in the training method embodiment of each parameter configuration model are implemented, for example Figure 1 Alternatively, when the processor 501 executes the computer program 503, the steps in the above-mentioned parameter configuration method embodiments are implemented, for example Figure 2Alternatively, when the processor 501 executes the computer program 503, the functions of each module / unit in the training device embodiment of each parameter configuration model are implemented, for example Figure 3 Alternatively, when the processor 501 executes the computer program 503, the functions of the modules / units in the above-mentioned parameter configuration device embodiments are realized, for example Figure 4 Functionality of modules / units 401 to 404 shown.

[0243] The present application also provides a readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, it is used to implement the training method or parameter configuration method of the parameter configuration model provided in the various embodiments described above.

[0244] Among them, the readable storage medium can be a computer storage medium or a communication medium. The communication medium includes any medium that facilitates the transmission of a computer program from one place to another. The computer storage medium can be any available medium that can be accessed by a general or special-purpose computer. For example, a readable storage medium is coupled to a processor so that the processor can read information from the readable storage medium and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application-specific integrated circuit (ASIC). In addition, the ASIC can be located in a user device. Of course, the processor and the readable storage medium can also exist in a communication device as discrete components. The readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, and an optical data storage device, etc.

[0245] The present application also provides a program product, which includes an execution instruction, which is stored in a readable storage medium. At least one processor of the device can read the execution instruction from the readable storage medium, and at least one processor executes the execution instruction so that the device implements the training method or parameter configuration method of the parameter configuration model provided by the various embodiments described above.

[0246] In the embodiments of the above-mentioned devices, it should be understood that the processor may be a central processing unit, or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. The general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc. The steps of the method disclosed in the present application may be directly embodied as being executed by a hardware processor, or may be executed by a combination of hardware and software modules in the processor.

[0247] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A training method for a parameter configuration model, characterized in that: include: Performing an environmental scan on a first access point to be configured with parameters, to obtain a strength of a signal sent by a second access point and received by the first access point; Determining a plurality of neighboring access points of the first access point from the second access point according to the strength of the signal and a preset signal strength threshold; determining a feature vector of the first access point according to a scanning result of the first access point on different channels and a scanning result of the adjacent access point on different channels; Obtain parameter configuration tags of a plurality of the first access points; An initial parameter configuration model is iteratively trained according to the feature vectors and parameter configuration labels of the plurality of first access points to obtain a trained parameter configuration model, so as to perform parameter configuration on the first access points through the trained parameter configuration model.

2. The training method of the parameter configuration model according to claim 1, characterized in that: The method further comprises: Mapping the wireless network where the first access point is located to obtain a topological structure corresponding to the wireless network; Among them, the wireless networking topology structure includes m vertices and n edges, m is an integer greater than or equal to 1, and n is an integer greater than or equal to 0; the vertices of the topology structure indicate the m access points in the wireless networking, the edges in the topology structure indicate the connection relationship between any two access points in the topology structure, and the weight of the edge indicates the strength of the signal.

3. The training method of the parameter configuration model according to claim 2, characterized in that: The determining, according to the scanning result of the first access point on different channels and the scanning result of the adjacent access point on different channels, the feature vector of the first access point includes: Obtaining scanning results of the first access point on different channels and scanning results of the adjacent access points on different channels; Determine first eigenvectors of the m vertices according to scanning results of the first access point on different channels and scanning results of the adjacent access points on different channels; Determine the second eigenvectors of the n edges according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels; The eigenvector of the first access point is determined according to the first eigenvectors of the m vertices and the second eigenvectors of the n edges.

4. The training method of the parameter configuration model according to claim 3, characterized in that: The determining, according to the scanning result of the first access point on different channels and the scanning result of the adjacent access point on different channels, the first feature vectors of the m vertices comprises: Determine the first initial feature vectors of the m vertices according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels; De-noising the first initial eigenvectors of the m vertices to obtain a de-noised first initial eigenvector; For each vertex among the m vertices, perform neighbor sampling on each vertex to obtain neighbor nodes of each vertex; Aggregating feature vectors of neighboring nodes of each vertex to obtain aggregated features of each vertex; Determine a first eigenvector of each vertex based on the aggregated features of each vertex and the first initial eigenvector after denoising; The first eigenvector of each vertex is aggregated to obtain the first eigenvectors of the m vertices.

5. The training method of the parameter configuration model according to claim 3, characterized in that: The determining, according to the scanning result of the first access point on different channels and the scanning result of the adjacent access point on different channels, the second feature vectors of the n edges comprises: Determine, according to the scanning results of the first access point on different channels and the scanning results of the adjacent access points on different channels, second initial eigenvectors of the n adjacency matrices corresponding to the n edges; Performing denoising processing on the second initial eigenvectors of the n adjacency matrices corresponding to the n edges to obtain the denoised second initial eigenvectors of the n adjacency matrices, wherein the n edges correspond one-to-one to the n adjacency matrices; For each of the n edges, obtaining a denoised feature vector matching the each edge from the denoised second initial feature vectors of the n adjacency matrices; Using the denoised feature vector matched with each edge as the second feature vector of each edge; The second eigenvector of each edge is aggregated to obtain the second eigenvectors of the n edges.

6. The training method of the parameter configuration model according to claim 3, characterized in that: The iterative training of the initial parameter configuration model according to the feature vectors and parameter configuration labels of the plurality of first access points to obtain a trained parameter configuration model includes: The initial parameter configuration model is iteratively trained according to the feature vectors and parameter configuration labels of the plurality of first access points, and a trained parameter configuration model is obtained when the loss function satisfies a preset condition.

7. The training method of the parameter configuration model according to claim 6, characterized in that: The loss function includes a feature mapping function and a feature edge mapping function; The iterative training of the initial parameter configuration model according to the feature vectors and parameter configuration labels of the plurality of first access points, and obtaining the trained parameter configuration model when the loss function satisfies a preset condition, comprises: Based on the first eigenvectors of the m vertices and the eigenmapping function, obtaining updated first eigenvectors corresponding to the m vertices; Based on the second eigenvectors of the n edges and the characteristic edge mapping function, obtaining updated second eigenvectors corresponding to the n edges; Based on the updated first feature vectors corresponding to the m vertices, the updated second feature vectors corresponding to the n edges, and the parameter configuration labels, the initial parameter configuration model is trained, and the trained parameter configuration model is obtained when the feature mapping function and the feature edge mapping function meet the preset number of iterations.

8. The training method of the parameter configuration model according to claim 1, characterized in that: If the parameter configuration tag is a tag of a radio frequency switch state, the trained parameter configuration model is used to perform frequency band parameter configuration on the first access point.

9. The training method of the parameter configuration model according to claim 1, characterized in that: If the parameter configuration label is a label of channel deployment, the trained parameter configuration model is used to perform channel parameter configuration on the first access point.

10. The training method of the parameter configuration model according to claim 1, characterized in that: If the parameter configuration label is a transmit power label, the trained parameter configuration model is used to perform power parameter configuration on the first access point, wherein the transmit power is between a first transmit power and a second transmit power, the first transmit power represents a minimum transmit power supported by the first access point, and the second transmit power represents a maximum transmit power supported by the first access point.

11. The training method of the parameter configuration model according to claim 1, characterized in that: If the parameter configuration tag is a broadband coverage tag, the trained parameter configuration model is used to perform bandwidth parameter configuration on the first access point.

12. A parameter configuration method, characterized in that: include: Performing an environmental scan on a first access point to be configured with parameters, to obtain a strength of a signal sent by a second access point and received by the first access point; Determining a plurality of neighboring access points of the first access point from the second access point according to the strength of the signal and a preset signal strength threshold; determining a feature vector of the first access point according to a scanning result of the first access point on different channels and a scanning result of the adjacent access point on different channels; Performing parameter configuration processing on the feature vector of the first access point based on the trained parameter configuration model to obtain parameter configuration information corresponding to the first access point; The trained parameter configuration model is obtained by training according to the parameter configuration model training method described in any one of claims 1-11.

13. A training device for a parameter configuration model, characterized in that: include: A first scanning unit, configured to perform an environment scan on a first access point to be configured with parameters, to obtain a strength of a signal sent by a second access point and received by the first access point; A first determining unit, configured to determine a plurality of neighboring access points of the first access point from the second access point according to the strength of the signal and a preset signal strength threshold; A second determining unit, configured to determine a feature vector of the first access point according to a scanning result of the first access point on different channels and a scanning result of the adjacent access point on different channels; An acquiring unit, configured to acquire parameter configuration tags of a plurality of the first access points; A training unit is used to iteratively train an initial parameter configuration model according to the feature vectors and parameter configuration labels of the plurality of first access points to obtain a trained parameter configuration model, so as to perform parameter configuration on the first access points through the trained parameter configuration model.

14. A parameter configuration device, characterized in that: include: A second scanning unit, configured to perform an environment scan on a first access point to be configured with parameters, to obtain a strength of a signal sent by a second access point and received by the first access point; A third determining unit, configured to determine a plurality of neighboring access points of the first access point from the second access point according to the signal strength and a preset signal strength threshold; a fourth determining unit, configured to determine a feature vector of the first access point according to a scanning result of the first access point on different channels and a scanning result of the adjacent access point on different channels; a configuration unit, configured to perform parameter configuration processing on the feature vector of the first access point based on the trained parameter configuration model to obtain parameter configuration information corresponding to the first access point; The trained parameter configuration model is obtained by training according to the parameter configuration model training method described in any one of claims 1-11.

15. A computer device, characterized in that: comprising a memory, and one or more processors communicatively connected to the memory; The memory stores instructions that can be executed by the one or more processors, and the instructions are executed by the one or more processors so that the one or more processors implement the training method of the parameter configuration model as described in any one of claims 1-11 or the parameter configuration method as described in claim 12.

16. A computer-readable storage medium, characterized in that: It includes a program or an instruction, which, when executed on a computer, implements the training method of the parameter configuration model according to any one of claims 1 to 11 or the parameter configuration method according to claim 12.

Images