Intelligent WiFi management strategy method based on Internet of Things

Abstract

The invention discloses an intelligent WiFi management strategy method based on the Internet of Things, and the method comprises the steps: integrating a multi-source sensor on a client, collecting data, generating an initial strategy through employing the collected data, aggregating the uploaded data through a server, constructing a global environment state matrix based on the uploaded data, and optimizing the initial strategy. Policy-environment bidirectional construction is carried out based on the client and the server; the client side generates a personalized strategy based on a local environment, the server side generates a reference strategy based on a global environment state matrix, and a hierarchical strategy system is formed through combination of co-modeling and a confrontation mechanism; a discriminator is arranged at a server side, after a client side generates a personalized strategy, the personalized strategy is transmitted to the server side, the server side carries out cyclic confrontation based on the discriminator, through strategy-environment bidirectional construction, the strategy can adapt to environment changes more quickly, traditional one-way data flow is broken through, a bidirectional interaction channel of environment data and strategy instructions is established, and the strategy-environment interaction efficiency is improved. And the performance reduction caused by strategy lag is reduced.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) WiFi technology, and in particular to an intelligent WiFi management strategy method based on the Internet of Things. Background Technology

[0002] With the rapid development of Internet of Things (IoT) technology, the number of smart devices is exploding. From smart home devices to industrial automation sensors, various terminal devices are increasingly reliant on wireless networks. As the most widely used wireless local area network (WLAN) technology, WiFi's management efficiency and stability directly affect the overall performance and user experience of IoT systems. However, traditional WiFi management methods have revealed many limitations when facing the complex and ever-changing IoT environment, making it difficult to meet the needs of efficient, intelligent, and adaptive management.

[0003] Traditional methods rely on static configuration and manual tuning, maintaining network operation through preset parameters such as channel allocation and power control. However, in IoT scenarios, the highly dynamic nature of devices (such as AGVs and frequently moving wearable devices) leads to real-time changes in signal requirements, making static configurations unable to respond quickly and prone to signal interruptions or resource waste. Furthermore, the diverse types of IoT devices (such as mobile phones, sensors, and industrial controllers) support different WiFi standards (such as 802.11n / ac / ax), making it difficult for traditional methods to uniformly optimize compatibility, resulting in frequency band conflicts or connection failures. In addition, the complexity of the physical environment (such as obstacles and electromagnetic interference) and user behavior (such as high-density access) further exacerbates signal attenuation and interference. Traditional methods lack adaptive mechanisms, making it difficult to maintain network stability and becoming a key bottleneck restricting the expansion of IoT applications. Summary of the Invention

[0004] This application provides an IoT-based intelligent WiFi management strategy method. Through bidirectional construction (co-evolution) of policy and environment, the policy can adapt to environmental changes more quickly, breaking the traditional unidirectional data flow and establishing a bidirectional interaction channel between environmental data and policy instructions. This reduces performance degradation caused by policy lag, and personalized policies significantly reduce the need for manual parameter tuning and lower operation and maintenance costs. Through adversarial and co-evolution mechanisms, the overall intelligence level of the system is improved, and a dynamic, adaptive, and intelligently evolving WiFi management system is constructed, realizing the transformation from static policy issuance to dynamic co-evolution optimization.

[0005] This application provides a smart WiFi management strategy method based on the Internet of Things, including: S101 integrates multi-source sensors on the client side, collects data based on the multi-source sensors, generates an initial policy using the collected data, aggregates the uploaded data on the server side through a long connection, constructs a global environment state matrix based on the uploaded data, optimizes the initial policy using the global environment state matrix, and performs bidirectional construction of policy and environment based on the client and server sides. S102, the client generates a personalized strategy based on the local environment, and the server generates a baseline strategy based on the global environment state matrix. The personalized strategy and the baseline strategy are combined through co-evolution and adversarial mechanisms to form a hierarchical strategy system. S103, A discriminator is set on the server side. The discriminator is used to discriminate the strategy. After the client generates a personalized strategy, it transmits the personalized strategy to the server side. The server side performs a cyclical adversarial operation based on the discriminator.

[0006] Preferably, the bidirectional construction of the strategy-environment includes the client generating an initial strategy based on the collected environmental data, the server constructing a global environmental state matrix based on the environmental data to optimize the initial strategy, and the optimized strategy being fed back to the client to adjust the client's environmental data collection behavior.

[0007] Preferably, the co-evolution refers to the collaboration between the client and the server through two-way interaction between the policy and the environment, thereby achieving policy and environment coordination; the adversarial refers to the interactive, competitive and cooperative process between the client and the server around policy generation and evaluation.

[0008] Preferably, the discriminator includes a security discrimination module, a performance discrimination module, and a compatibility discrimination module. The security discrimination module is used to verify policy risks based on the rule engine and behavior prediction model. The performance discrimination module is used to evaluate policy latency, bandwidth, and energy consumption. The compatibility discrimination module is used to detect the matching degree between the policy and the device model.

[0009] Preferably, it further includes: S201: Collect device attribute information, calculate the comprehensive membership degree using a Gaussian function based on the device attribute information, and associate the calculated comprehensive membership degree with the server's strategy; whereby the device attribute information includes geographical location attributes, signal strength, and device type; S202, calculate the client's influence gradient function based on the comprehensive membership degree, and select strategies based on comprehensive membership degree and influence degree.

[0010] Preferably, the Gaussian function calculation formula for geographic location attributes is as follows: ,in, A Gaussian function for geographic location attributes. The mean, The standard deviation is given; the Gaussian function for signal strength properties is: ,in, The Gaussian function representing the signal strength property. As the center value, The steepness is given by the Gaussian function for the device type attribute, calculated using the following formula: ,in, The Gaussian function for the device type attribute. For type threshold, Steepness.

[0011] Preferably, the membership degree is obtained by weighted averaging of Gaussian functions calculated from geographic location attributes, signal strength, and device type attributes. The formula for calculating the membership degree is as follows: ,in, The overall membership degree represents the overall membership degree of a device to a certain policy group. For index variables, The total number of attributes considered for the device. Let i be the weight of the i-th attribute. Let be the membership degree of the device to a certain policy group on the i-th attribute.

[0012] Preferably, the influence gradient function is a double exponential decay-growth function, specifically: the influence decay formula for strategy group A is: ,in, This indicates the diminishing effect of strategy group A. To maximize the influence of the initial strategy, take =1, , These are the attenuation and growth coefficients, respectively, ranging from 0.1 to... ≤0.5, 0.1≤ ≤0.5, determined experimentally. For the equipment movement time and distance increment, The overall membership degree of the device to strategy group A; the formula for the growth of influence of strategy group B: ,in, This indicates the growth influence of strategy group B. To maximize the influence of the initial strategy, take =1, , These are the attenuation and growth coefficients, respectively, ranging from 0.1 to... ≤0.5, 0.1≤ ≤0.5 was determined experimentally. For the equipment movement time and distance increment, This represents the overall membership degree of the device to strategy group B.

[0013] Preferably, it further includes: S301, collect strategy data, construct a strategy space based on the strategy data, use algorithms to detect defect types in the strategy space, and construct a defect dynamics model based on the identified defect types; S302 designs a phase change control strategy by actively injecting defect types.

[0014] Preferably, the phase change control strategy includes: injecting strategy vortices when the system becomes rigid, with the disturbance amplitude and frequency adjusted according to the degree of rigidity; and repairing the chaotic state through strategy voids when the system becomes chaotic, with the repair intensity and inspection frequency adjusted according to the range of chaos.

[0015] One or more technical solutions provided in this application have at least the following technical effects or advantages: By constructing (co-evolving) the policy and environment in a two-way manner, the policy can adapt to environmental changes more quickly, breaking the traditional one-way data flow, establishing a two-way interactive channel between environmental data and policy instructions, reducing performance degradation caused by policy lag, personalized policies significantly reduce the need for manual parameter tuning, reduce operation and maintenance costs, improve the overall intelligence level of the system through adversarial and co-evolution mechanisms, build a dynamic adaptive and intelligent evolution WiFi management system, and realize the transformation from static policy distribution to dynamic co-evolution optimization; By using fuzzy policy boundaries and influence gradient functions, the system acquires the ability to handle uncertainty. This shift from deterministic management to uncertainty adaptation enables the system to better cope with the inherent randomness and fuzziness in the IoT environment, making policy switching smooth. Through the policy superposition mechanism, probabilistic selection is performed based on the membership degree calculated by the fuzzy policy boundary, which can more flexibly adapt to complex scenarios, solve the traditional binary policy partitioning problem, transform discrete binary decisions into continuous membership relationships, enhance the system's adaptability, flexibility and robustness in complex IoT environments, and achieve smooth transition and intelligent adaptation of policy execution. Phase change control strategies enable rapid recovery from disorder, reducing fault recovery time by an estimated 50%, significantly improving the system's resilience in the face of faults and allowing it to better adapt to complex environmental changes. By introducing topological defect-induced phase change control, the limitations of traditional WiFi management are overcome, achieving a dynamic balance between ordered and disordered states in the intelligent WiFi strategy system. This enhances system resilience, adaptability, and resource utilization, enabling the system to self-repair and upgrade. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating an intelligent WiFi management strategy method based on the Internet of Things according to the present invention. Figure 2 This is a schematic diagram illustrating the strategy selection process based on comprehensive membership and influence in this invention. Figure 3 This is a schematic diagram illustrating the design process of the phase change control strategy for this invention. Detailed Implementation

[0017] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the present invention may be more thorough and complete.

[0018] It should be noted that the terms "vertical," "horizontal," "up," "down," "left," "right," and similar expressions used in this article are for illustrative purposes only and do not represent the only possible implementation.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0020] Example 1: Figure 1 This is a flowchart illustrating an IoT-based intelligent WiFi management strategy method according to an embodiment of the present invention, including: S101 integrates multi-source sensors on the client side, collects data based on the multi-source sensors, generates an initial policy using the collected data, aggregates the uploaded data on the server side through a long connection, constructs a global environment state matrix based on the uploaded data, optimizes the initial policy using the global environment state matrix, and performs bidirectional construction of policy and environment based on the client and server sides. Specifically, different types of sensors are integrated into the existing WiFi interface module of the client system. These sensors include a WiFi signal detection module for acquiring WiFi signal information; a GPS positioning unit for accurately determining the device's geographical location; a motion accelerometer for sensing the device's motion status; and a battery status monitoring module for real-time monitoring of battery power. These multi-source sensors collect data in real-time on signal strength, network latency, packet loss rate, device movement trajectory, movement speed, battery power, and CPU load in the local environment. Signal strength reflects the strength of the WiFi signal; network latency reflects the time loss of data during network transmission; packet loss rate is the proportion of data packets lost during data transmission; and the device movement trajectory (composed of a sequence of latitude and longitude coordinates) is also collected. The system records the device's movement path in space; movement speed indicates how fast the device moves; battery level displays the remaining battery power; and CPU load reflects the CPU's workload. Accelerometer data is used to determine whether the device is moving or stationary. When the device is moving, the data acquisition frequency is increased to 3 times per second; when the device is stationary, due to the relatively slow environmental changes, the data acquisition frequency is reduced to once per minute. The acquired data undergoes preprocessing, employing a sliding window averaging algorithm to denoise signal strength, network latency, and other data, with the window size set to 5 data points. This algorithm smooths out noise in the data, making it more accurate and reliable. Simultaneously, a compression algorithm is used to compress the data, controlling the compression rate between 30% and 50%. This ensures data integrity while significantly reducing the amount of data that needs to be uploaded to the server, reducing network transmission pressure.

[0021] The rules for generating the initial policy are preset based on the collected data. For example, when the signal strength is below -70dBm, the access point (AP) will be switched. After the client obtains the data, it monitors and analyzes the data in real time. If the monitored data meets the preset rules for generating the initial policy, the initial policy generation process is triggered. For example, when the client detects that the current WiFi signal strength is below -70dBm, it will generate a draft policy for switching the access point (AP). The draft includes a list of candidate APs to be switched and the priority order of switching, among other preliminary information.

[0022] The server aggregates uploaded data through long-lived connections, also known as persistent connections. This means that once a connection is established between the client and server, it remains open for a period of time, unlike short-lived connections which close immediately after a data transfer. During this open period, the client and server can continuously interact with each other multiple times without repeatedly establishing and disconnecting connections. The server constructs a global environment state matrix based on the uploaded data. This matrix is ​​indexed by time, with device IDs as columns and environmental data metrics (such as signal strength and network latency) as rows. Data mining algorithms (cluster analysis) are used to analyze the data in the global environment state matrix to identify common patterns across devices. For example, by comparing the signal strength changes of multiple devices in the same area, regional signal interference patterns can be identified. Based on the global environment state matrix, the server uses statistical analysis methods to optimize policy templates. For example, by analyzing a large amount of environmental data, regional signal interference patterns are identified. If strong signal interference is found in a certain area, the server will adjust the AP switching priority order, prioritizing APs with less interference during AP switching in that area to improve network connection stability and quality.

[0023] Client environment data enables the server to optimize and adjust the initial policy, forming a path from the client to the server. Then, the optimized policy from the server affects the device's behavior in the client environment, forming a path from the server to the client. Based on the client-to-server path and the server-to-client path, a two-way construction of policy and environment is formed.

[0024] S102, the client generates a personalized strategy based on the local environment, and the server generates a baseline strategy based on the global environment state matrix. The personalized strategy and the baseline strategy are combined through co-evolution and adversarial mechanisms to form a hierarchical strategy system. Furthermore, an attention mechanism model (Transformer) is used to collect historical network environment data and corresponding successful strategies. This historical data and successful strategies are used as a training set, and the attention mechanism model is trained using a self-supervised learning approach. The global environment state matrix is ​​input into the trained attention mechanism model, which outputs a baseline strategy based on its internal algorithm. This baseline strategy is a general solution for problems commonly encountered in multiple devices or regions. The generated strategy is sent to the client via a grouped tree-like concatenation method, thus generating the baseline strategy. A generative network is deployed on the client, employing a model generative adversarial network structure that has undergone knowledge distillation compression. Knowledge distillation is a model compression technique that compresses the knowledge of the large Transformer model on the server to the client's generative network through soft-object transfer. The client uses the generative network to generate personalized strategies based on its local environment. This local environment refers to information closely related to the surrounding environment collected by the client device itself, such as the device's movement trajectory (obtained through a GPS positioning unit) and real-time signal quality (obtained through a WiFi signal detection module). For example, an AGV (Automated Guided Vehicle) generates a fine-tuning strategy for pre-connecting to the optimal AP based on its current location data and real-time signal quality. As the vehicle moves, it continuously acquires its own location information and the quality of the surrounding WiFi signal. The generated network analyzes and processes this data, adjusting the advance time of AP switching. If it detects that the signal quality in a certain area ahead is about to deteriorate, the vehicle will switch to an AP with a better signal in advance to minimize switching delay and ensure the stability of the vehicle's network connection, thereby guaranteeing its normal operation and task execution.

[0025] Personalized strategies and baseline strategies form a hierarchical strategy system through co-evolution and adversarial mechanisms. The baseline strategy defines the basic direction and constraints of the strategy under different environmental modes and provides a co-evolution framework. Personalized strategies are optimized through adversarial mechanisms. The client's personalized strategy continuously adjusts and improves itself in the interaction with the local environment. When generating personalized strategies, the client will refer to the frequency band selection suggestions in the server's baseline strategy. When optimizing the baseline strategy, the server will also consider the execution effect of the client's personalized strategy in the actual environment.

[0026] S103, A discriminator is set on the server side. The discriminator is used to discriminate the strategy. After the client generates a personalized strategy, it transmits the personalized strategy to the server side. The server side performs a cyclical adversarial based on the discriminator. The adversarial process is an interactive, competitive, and cooperative process between the client and server, revolving around strategy generation and evaluation. It aims to optimize strategies through continuous game theory. The discriminator includes a security discrimination module, a performance discrimination module, and a compatibility discrimination module. The security discrimination module is integrated on the server side. This module uses a rule engine to verify the generated strategies and introduces an LSTM model trained on historical strategy execution data as a behavior prediction model. Through analysis and learning from a large amount of historical data, it can predict the potential risks caused by the strategies. For the performance discrimination module, a network simulation environment is constructed using digital twin technology. This simulation environment can simulate key elements of the actual network, such as topology, number of devices, and traffic patterns. The strategy to be evaluated is input into the simulation environment, and the system performs a comprehensive pre-evaluation of the strategy's performance. Evaluation indicators include key performance indicators such as latency, bandwidth, and energy consumption, and a multi-objective weighted function is used for scoring. The compatibility discrimination module is based on a device knowledge graph, which contains detailed information such as device model, hardware configuration, and supported WiFi standards. By querying the device knowledge graph, the discrimination module can accurately detect the compatibility between the strategy and the device model.

[0027] After the client generates a personalized strategy, it transmits it to the server. The server's discriminator receives the personalized strategy generated by the client and performs a multi-dimensional evaluation, including three important dimensions: security, performance, and compatibility. A weighted score is calculated based on the evaluation results of these three dimensions. A high-scoring strategy with a score exceeding 80 is considered high-quality and meets network requirements, so it is retained and executed. A low-scoring strategy with a score below 60 indicates significant problems and is subject to cyclical adversarial testing. This cyclical adversarial testing involves the client adjusting its strategy generation parameters and regenerating the strategy based on the server's score feedback. Simultaneously, the server can also directly generate an optimized strategy and send it to the client. This iterative process of generation, evaluation, and optimization forms a dynamic game cycle, i.e., cyclical adversarial testing. The strategy execution effect feedback from the environment is used as training data for the discriminator and integrated into the cyclical adversarial testing in real time. Finally, when both strategy generation and discrimination accuracy reach over 95%, the system reduces the frequency of adversarial games and enters a stable state.

[0028] The technical solutions in the above embodiments of this application have at least the following technical effects or advantages: By constructing (co-evolving) the policy and environment in a two-way manner, the policy can adapt to environmental changes more quickly, breaking the traditional one-way data flow, establishing a two-way interactive channel between environmental data and policy instructions, reducing performance degradation caused by policy lag, personalized policies significantly reduce the need for manual parameter tuning, reduce operation and maintenance costs, and improve the overall intelligence level of the system through adversarial and co-evolution mechanisms, building a dynamic adaptive and intelligent evolution WiFi management system, and realizing the transformation from static policy issuance to dynamic co-evolution optimization.

[0029] Example 2: Building upon Example 1, this example introduces the concept of fuzzy policy boundaries to address the policy conflicts and coverage blind spots caused by traditional binary policy partitioning. By drawing on fuzzy logic theory, establishing an influence gradient model, and a policy superposition mechanism, this example enhances the system's adaptability and flexibility, making it particularly suitable for scenarios involving device movement and dynamic network changes in complex IoT environments, such as... Figure 2 As shown.

[0030] S201, Collect the attribute information of the device, select the membership function based on the attribute information of the device to calculate the comprehensive membership degree, and associate the calculated comprehensive membership degree with the strategy of the server. Furthermore, device attribute information is collected from both the client and server sides. This device attribute information includes geographical location attributes, signal strength, and device type. The geographical location attribute exhibits regional concentration and gradual change characteristics. Regional concentration reflects that devices tend to operate within a specific geographical area, while gradual change indicates that the device's geographical location changes continuously. Attributes such as signal strength and device type typically exhibit non-linear change characteristics. Using the Gaussian function (Sigmoid) can map the input value between 0 and 1, effectively reflecting the gradual membership relationship of devices in these attributes. Taking geographical location attributes as an example, the parameters of the Gaussian function are set, including the mean and standard deviation. The center of the factory area is usually a relatively concentrated area for device activity. A large number of devices will carry out production, transportation, and other operations in the area surrounding the factory center as a reference point. Setting the mean of the Gaussian function at the factory center allows the function to better reflect the geographical location characteristics of the devices under normal activity conditions. The activity range radius is set, and the standard deviation is set to 1 / 3 of the radius. To ensure that the Gaussian function can reasonably cover the activity range of devices when describing the geographical location membership, if the standard deviation is too small, the curve of the Gaussian function will be too steep, resulting in only devices very close to the mean having a high membership, ignoring the actual situation of the devices' activities within a certain range. If the standard deviation is too large, the curve will be too flat, making the difference in the membership of devices in different locations indistinct and failing to accurately reflect the different membership degrees of devices to strategy groups in different locations.

[0031] The Gaussian function is calculated based on geographic location attributes, signal strength, and device type attributes. The formula for calculating the Gaussian function based on geographic location attributes is as follows: ,in, A Gaussian function for geographic location attributes. The mean value represents the center point of the equipment's normal operating range (such as the center coordinates of the factory area). Standard deviation = R is the radius of the device's operating range; the Gaussian function for signal strength properties is: ,in, The Gaussian function representing the signal strength property. The center value represents the baseline value for signal strength. To control the steepness, the slope of the control curve is adjusted according to the signal fluctuation range; the formula for calculating the Gaussian function for the device type attribute is: ,in, The Gaussian function for the device type attribute. Type threshold, the critical value for distinguishing device types. To determine the steepness, adjust according to the type discrimination; the membership degree is obtained by weighted averaging the Gaussian functions calculated from the above geographical location attributes, signal strength, and device type attributes. The formula for calculating the membership degree is: ,in, The overall membership degree represents the overall membership degree of a device to a certain policy group, and is used to measure the degree of association between the device and the policy group. This is an index variable used to iterate through the various attributes of the device. The total number of attributes considered for the device. Let i be the weight of the i-th attribute. Let be the membership degree of the device to a certain policy group on the i-th attribute.

[0032] In the policy management module on the server side, the calculated comprehensive membership degree is associated with the policy. Each policy corresponds to a membership degree vector. The membership degree vector of a policy may contain multiple elements, which correspond to the membership degree calculated for different device attributes. Together, they reflect the overall applicability of the device under the policy. When the device information is updated, the comprehensive membership degree is recalculated based on the new device attribute information. Then, the policy applicability representation is updated, that is, the policy membership degree vector corresponding to the device in the policy management module is modified.

[0033] S202, calculate the client's influence gradient function based on the comprehensive membership degree, and select strategies based on comprehensive membership degree and influence degree. Specifically, the membership degree between a device and different strategy groups changes with the device's state (such as location and time). To accurately describe the smooth transition characteristics of the strategy's influence as it dynamically changes with the device, an influence gradient function needs to be defined, using a double exponential decay-growth function. The influence decay formula for strategy group A (when the device is far from A and close to B) is as follows: ,in, This indicates the diminishing effect of strategy group A. To maximize the influence of the initial strategy, take =1, , These are the attenuation and growth coefficients, respectively, ranging from 0.1 to... ≤0.5, 0.1≤ ≤0.5 was determined experimentally. For the equipment movement time and distance increment, The overall membership degree of the device to strategy group A; the formula for the growth of the influence of strategy group B (when the device is closer to B and farther from A): ,in, This indicates the growth influence of strategy group B. To maximize the influence of the initial strategy, take =1, , These are the attenuation and growth coefficients, respectively, ranging from 0.1 to... ≤0.5, 0.1≤ ≤0.5 was determined experimentally. For the equipment movement time and distance increment, The overall membership degree of the device to strategy group B is calculated. In the client strategy execution module, the influence gradient function is called to calculate the influence of each strategy based on the device's current location and time information, and the strategy influence is updated periodically.

[0034] The adaptability of a device to different strategies is determined by both the overall membership degree and the strategy influence. The calculated overall membership degree and strategy influence are combined proportionally to generate an overall weight. In stable scenarios (such as fixed devices), the membership degree weight is increased; in dynamic scenarios (such as mobile devices), the influence weight is enhanced. In the client's local strategy module, the probability distribution of the strategy superposition state is generated based on the overall weight. A random number generator is used to select the strategy, and random numbers are regenerated periodically (such as every minute). The strategy selection is adjusted according to the latest overall weight to adapt to changes in device status. The membership degree and strategy influence are dynamically adjusted based on the device's real-time operating data (such as energy consumption and response time), and the overall weight is recalculated.

[0035] The technical solutions described in the embodiments of this application have at least the following technical effects or advantages: By using fuzzy policy boundaries and influence gradient functions, the system has the ability to handle uncertainty. This shift from deterministic management to uncertainty adaptation allows the system to better cope with the inherent randomness and fuzziness in the IoT environment, making policy switching smooth. Through the policy superposition mechanism, probabilistic selection is performed based on the membership degree calculated by the fuzzy policy boundary, which can more flexibly adapt to complex scenarios, solve the traditional binary policy partitioning problem, transform discrete binary decisions into continuous membership relationships, enhance the system's adaptability, flexibility, and robustness in complex IoT environments, and achieve smooth transition and intelligent adaptation of policy execution.

[0036] Example 3: In Examples 1 and 2 above, topological defects (such as policy vortices or voids) from condensed matter physics exist in the policy space. These defects can trigger phase transitions in the policy system. Traditional management methods cannot effectively identify and utilize these defects, leading to potentially uncontrollable phase transitions, policy conflicts, performance fluctuations, or system stagnation. This example introduces a topological defect-induced phase transition control mechanism, treating the policy space as a topological manifold. By real-time detection and management of topological defects (such as policy vortices and voids), a controllable phase transition is actively induced, achieving a dynamic balance between ordered and disordered states in the system. Figure 3 As shown.

[0037] S301, collect strategy data, construct a strategy space based on the strategy data, use algorithms to detect defect types in the strategy space, and construct a defect dynamics model based on the identified defect types; Specifically, policy data is collected in existing smart WiFi networks. This policy data includes policy configuration information, device status data, and network load data. Real-time monitoring and collection of the operational status information of each device in the network are conducted, including but not limited to the device's online / offline status, signal strength, number of connected devices, device type (such as mobile phones, computers, smart home appliances, etc.), and current traffic usage. The network load is comprehensively collected across different time periods and regions. This collected data is stored in a database. A policy space is constructed based on the collected policy data, viewed as a topological manifold. A topological manifold is a geometric object mathematically used to describe a continuous and locally Euclidean object. The dimensions and structural characteristics of the policy space are determined based on the policy data, and different policy parameters (such as access control) are applied. The dimensions of the space are determined by parameters (such as routing parameters), and the interrelationships and variation patterns among these parameters are analyzed to determine the structural characteristics of the space, such as the existence of hierarchical structure and correlation. Principal Component Analysis (PCA) is used to project the original high-dimensional data into a low-dimensional space through linear transformation. The specific steps are as follows: First, the collected policy data is centered by subtracting the mean of the data; then, the covariance matrix of the data is calculated, which reflects the correlation between the dimensions; next, the covariance matrix is ​​decomposed into eigenvalues ​​to obtain eigenvalues ​​and corresponding eigenvectors; finally, the eigenvectors corresponding to the first few largest eigenvalues ​​are selected to form a projection matrix, and the original data is projected into the low-dimensional space formed by these eigenvectors, thereby achieving dimensionality reduction. Through dimensionality reduction, the complex policy space can be transformed into a form that is easier to analyze and process.

[0038] Defects are identified using the isomap algorithm, which constructs a low-dimensional manifold by preserving the geodesic distance between data points, thus maintaining the geometric structure of the data. The isomap algorithm is implemented on the server side to build a defect detection engine. This engine has functions such as data processing, algorithm computation, and result output. Simultaneously, based on the environmental policy mapping library and the format of fuzzy policy data, corresponding data interfaces are designed. The defect detection engine is used to detect defects in the policy space. It processes the input policy data using the isomap algorithm, mining structural features in the data, comparing the actual data distribution with the popular structure, and identifying the differences—the topological defects in the policy space. Defect types are identified based on these topological defects; common defect types include policy vortices and policy holes. A vortex indicates the presence of a vortex-like structure in the policy space, where data points exhibit rotational or cyclical characteristics, leading to chaotic or unstable policy execution. Policy vortices are identified by calculating the vorticity of policy interactions. Vorticity describes the rotational intensity of the fluid (data flow in the policy space). A vorticity threshold is set, determined through experiments and simulations. When the vorticity exceeds the threshold, a policy vortex is identified. A policy void refers to a missing or improperly covered region in the policy space. Voids are identified by calculating the Betti number of the policy vector. The Betti number is a topological invariant used to describe the hole structure in the topological space. When the Betti number is greater than 0, a policy void is identified. The environmental policy mapping library and fuzzy policy data are input into the defect detection engine. For detected defects, the defect detection engine outputs the defect type and intensity index. The intensity index ranges from 0 to 1, with a higher value indicating a stronger defect.

[0039] A defect dynamics model is constructed using partial differential equations (PDEs). The PDEs describe the evolution of defect strength over time. The formula for the dynamic equations is as follows: ,in, For defect strength, This is the inherent growth rate of defects. This is the environmental inhibition coefficient. The intensity of the impact of the phase transition event, For random noise, To reduce network load, historical data is used to train the defect dynamics model, and model parameters are adjusted to improve prediction accuracy. The model is validated using an independent dataset to evaluate the accuracy of the model's output defect evolution trajectory (represented by time series, such as the curve of defect intensity changing over time) and phase transition critical point prediction (setting a defect density threshold of 0.8; when the defect density exceeds the threshold, the prediction may trigger a disordered phase transition). Defect type, environmental data, and historical phase transition records are input into the trained defect dynamics model, and the defect dynamics model outputs the defect evolution trajectory and phase transition critical point prediction based on its internal equations and algorithms.

[0040] S302, a phase change control strategy is designed by actively injecting defect types; Furthermore, the system is determined to be in a rigid state based on strategy execution efficiency and a rigidity threshold. Strategy execution efficiency is the percentage of successfully executed strategies per unit time. When the strategy execution efficiency is below the rigidity threshold for three consecutive time windows (e.g., every 5 minutes), the system is considered rigid. Rigidity is characterized by fixed strategy paths and the inability of innovative strategies to take effect, leading to decreased system adaptability. The system is also determined to be in a chaotic state based on strategy conflict frequency and a chaos threshold. Strategy conflict frequency is the number of strategy conflicts detected per unit time. When the conflict frequency exceeds the chaos threshold for two consecutive time windows, excessive chaos is identified. Chaos is characterized by mutual interference between strategies and increased resource contention, leading to unpredictable system behavior. If the system is determined to be in a rigid state, the rigidity is broken by actively injecting strategy vortices. The perturbation amplitude, injection frequency, and injection timing of the strategy vortex are set according to the degree of system rigidity. If the system rigidity is low, the perturbation amplitude of the strategy vortex is set to a low amplitude for slight adjustments to strategy parameters, and the injection frequency is set to a low frequency to avoid frequent interference with normal processes. Injection is performed when the system load is low (e.g., at night). To minimize the impact on users, if the system is highly rigid, the perturbation amplitude is set to high to significantly modify the strategy logic, forcibly triggering reorganization. The injection frequency is set to high frequency to accelerate the resolution of rigidity. The rigidity trend is predicted by reinforcement learning agents, and injection is performed in advance. The strategy vortex introduces controllable conflicts to break the dependencies of the original strategy paths, forcing the system to re-explore the strategy space. If the system is determined to be in a chaotic state, the chaotic state is repaired by injecting strategy holes. The repair strength, check frequency, and injection timing of the strategy holes are set according to the chaotic state of the system. If the chaotic state of the system is local, the repair strength is set to local repair, only repairing the strategies involved in the current conflict. The check frequency is set to periodic checks, which is suitable for stable chaotic scenarios and pre-repairs before the system is under high load. If the chaotic state of the system is global, the repair strength is global repair, re-evaluating the compatibility of all strategies. The check frequency is real-time checks, and repair is performed immediately when a conflict is detected. The injection timing is dynamically adjusted by combining reinforcement learning reward signals. Strategy hole repair restores the synergistic relationship between strategies and rebuilds system order by eliminating the root cause of conflict.

[0041] The technical solutions in the above embodiments of this application have at least the following technical effects or advantages: the phase change control strategy can quickly recover from disorder, the fault recovery time is expected to be shortened by 50%, which greatly improves the system's recovery capability when facing faults, and can better adapt to complex environmental changes. By introducing phase change control induced by topological defects, the limitations of traditional WiFi management are broken through, and the intelligent WiFi strategy system achieves a dynamic balance between ordered and disordered states, improves system resilience, adaptability and resource utilization, and realizes the system's self-repair and upgrade capabilities.

[0042] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A smart WiFi management strategy method based on the Internet of Things, characterized in that, include: S101 integrates multi-source sensors on the client side, collects data based on the multi-source sensors, generates an initial policy using the collected data, aggregates the uploaded data on the server side through a long connection, constructs a global environment state matrix based on the uploaded data, optimizes the initial policy using the global environment state matrix, and performs bidirectional construction of policy and environment based on the client and server sides. S102, the client generates a personalized strategy based on the local environment, and the server generates a baseline strategy based on the global environment state matrix. The personalized strategy and the baseline strategy are combined through co-evolution and adversarial mechanisms to form a hierarchical strategy system. S103, A discriminator is set on the server side. The discriminator is used to discriminate the strategy. After the client generates a personalized strategy, it transmits the personalized strategy to the server side. The server side performs a cyclical adversarial operation based on the discriminator.

2. The intelligent WiFi management strategy method based on the Internet of Things as described in claim 1, characterized in that, The bidirectional construction of the strategy-environment includes the client generating an initial strategy based on the collected environmental data, the server constructing a global environmental state matrix based on the environmental data to optimize the initial strategy, and the optimized strategy being fed back to the client to adjust the client's environmental data collection behavior.

3. The intelligent WiFi management strategy method based on the Internet of Things as described in claim 2, characterized in that, The co-evolution refers to the collaboration between the client and the server through two-way interaction between the policy and the environment, thereby achieving policy and environment coordination; the adversarial refers to the interactive, competitive and cooperative process between the client and the server around policy generation and evaluation.

4. The intelligent WiFi management strategy method based on the Internet of Things as described in claim 3, characterized in that, The discriminator includes a security discrimination module, a performance discrimination module, and a compatibility discrimination module. The security discrimination module is used to verify policy risks based on the rule engine and behavior prediction model. The performance discrimination module is used to evaluate policy latency, bandwidth and energy consumption; the compatibility discrimination module is used to detect the matching degree between the policy and the device model.

5. The intelligent WiFi management strategy method based on the Internet of Things as described in claim 1, characterized in that, Also includes: S201: Collect device attribute information, calculate the comprehensive membership degree using a Gaussian function based on the device attribute information, and associate the calculated comprehensive membership degree with the server's strategy; whereby the device attribute information includes geographical location attributes, signal strength, and device type; S202, calculate the client's influence gradient function based on the comprehensive membership degree, and select strategies based on comprehensive membership degree and influence degree.

6. The intelligent WiFi management strategy method based on the Internet of Things as described in claim 5, characterized in that, The formula for calculating the Gaussian function for geographic location attributes is: ,in, A Gaussian function for geographic location attributes. The mean, The standard deviation is given; the Gaussian function for signal strength properties is: ,in, The Gaussian function representing the signal strength property. As the center value, The steepness is given by the Gaussian function for the device type attribute, calculated using the following formula: ,in, The Gaussian function for the device type attribute. For type threshold, Steepness.

7. The IoT-based intelligent WiFi management strategy method as described in claim 6, characterized in that, The membership degree is obtained by taking a weighted average of the Gaussian functions calculated from the geographic location attribute, signal strength, and device type attribute. The formula for calculating the membership degree is: ,in, The overall membership degree represents the overall membership degree of a device to a certain policy group. For index variables, The total number of attributes considered for the device. Let i be the weight of the i-th attribute. Let be the membership degree of the device to a certain policy group on the i-th attribute.

8. The intelligent WiFi management strategy method based on the Internet of Things as described in claim 7, characterized in that, The influence gradient function is a double exponential decay-growth function, specifically: the influence decay formula for strategy group A is as follows: ,in, This indicates the diminishing effect of strategy group A. To maximize the initial strategy influence, take =1, , These are the attenuation and growth coefficients, respectively, ranging from 0.1 to... ≤0.5, 0.1≤ ≤0.5, determined experimentally. For the increment of equipment movement time and distance, The overall membership degree of the device to strategy group A; the formula for the growth of influence of strategy group B: ,in, This indicates the growth influence of strategy group B. To maximize the initial strategy influence, take =1, , These are the attenuation and growth coefficients, respectively, ranging from 0.1 to... ≤0.5, 0.1≤ ≤0.5 was determined experimentally. For the increment of equipment movement time and distance, This represents the overall membership degree of the device to strategy group B.

9. The intelligent WiFi management strategy method based on the Internet of Things as described in claim 1, characterized in that, Also includes: S301, collect strategy data, construct a strategy space based on the strategy data, use algorithms to detect defect types in the strategy space, and construct a defect dynamics model based on the identified defect types; S302 designs a phase change control strategy by actively injecting defect types.

10. The intelligent WiFi management strategy method based on the Internet of Things as described in claim 9, characterized in that, The phase change control strategy includes: injecting strategy vortices when the system becomes rigid, with the disturbance amplitude and frequency adjusted according to the degree of rigidity; and repairing the chaotic state through strategy voids when the system becomes chaotic, with the repair intensity and inspection frequency adjusted according to the range of chaos.

Images