Wireless network high-speed switching automatic test optimization method based on edge computing

Abstract

The invention discloses a wireless network high-speed switching automatic test optimization method based on edge computing, and relates to the field of wireless network switching. An improved gravitational search algorithm is combined with a GIS deployment node, multi-dimensional data is collected through an intelligent heterogeneous sensor and reinforcement learning, preprocessing is carried out through a denoising auto-encoder based on a GAN, modeling is carried out through space-time diagram convolution and an LSTM mixed model, a test case is generated through reinforcement learning Monte Carlo tree search, and testing is carried out on edge nodes in a distributed mode. And an optimization strategy is formulated through multi-agent deep reinforcement learning, and strategy implementation and feedback are realized through SDN and a block chain. According to the method, multi-dimensional acquisition, intelligent modeling, automatic testing and deep reinforcement learning optimization are realized, the high-speed switching performance is improved, the time delay and failure rate are reduced, the testing efficiency is improved, resource allocation is optimized, the network security is enhanced, and stable and efficient operation of the wireless network is guaranteed in an omnibearing manner.

Description

Technical Field

[0001] The present invention relates to the technical field of wireless network handover, and particularly to an automated test optimization method for high-speed handover of wireless networks based on edge computing. Background Art

[0002] With the rapid development of technologies such as mobile Internet and Internet of Things, the application scope of wireless networks has been continuously expanding, and the requirements for network performance and stability have also been increasing day by day. In a wireless network environment, the mobility of user equipment makes network handover a common operation. For example, in scenarios such as high-speed trains and subways, user equipment needs to frequently hand over between different base stations to ensure the continuity of network connection. However, the existing wireless network handover mechanisms have many problems in high-speed mobile scenarios.

[0003] Traditional wireless network handovers mainly rely on a single index such as signal strength for judgment. When the signal strength is lower than a certain threshold, the device will trigger a handover operation. However, this method is inadequate in high-speed mobile scenarios. On the one hand, due to the high moving speed, the change of signal strength is very rapid, and a single threshold judgment is prone to problems such as frequent handovers or untimely handovers. Frequent handovers will increase the handover delay, reduce the user experience, and may even lead to connection interruptions; while untimely handovers will cause the device to be in a low-quality network environment for a long time, affecting the stability of data transmission. On the other hand, the existing handover mechanisms lack a comprehensive perception of the network environment and cannot comprehensively consider various factors such as bandwidth, packet loss rate, and latency, resulting in poor network performance after handover.

[0004] Currently, wireless network automated tests mainly adopt traditional test methods, which usually require a large amount of manual intervention and have low test efficiency. Moreover, traditional test methods are difficult to simulate complex real network environments and cannot accurately evaluate the performance of wireless networks in high-speed handover scenarios. In addition, existing test result analysis methods often can only provide some basic statistical data, are difficult to deeply explore the correlation relationships between network parameters, and cannot provide effective guidance for network optimization.

[0005] With the development of edge computing technology, it provides a new idea for the optimization of high-speed handover of wireless networks. Edge computing can place computing and data storage close to the data source, reduce data transmission latency, and improve network response speed. However, how to effectively apply edge computing technology to the automated test and optimization of high-speed handover of wireless networks is still an urgent problem to be solved. Summary of the Invention

[0006] The automated test optimization method for high-speed handover of wireless networks based on edge computing proposed by the present invention aims to solve the problems mentioned in the above existing technologies.

[0007] To achieve the above object, the present invention adopts the following technical solutions: An optimized method for automated testing of high-speed handover in a wireless network based on edge computing, including: Edge computing node deployment step: Optimize the deployment of edge computing nodes by using the improved gravitational search algorithm IGSA combined with geographical information system GIS data. Construct a comprehensive objective function considering network topology, geographical environment and expected traffic distribution, and plan the node positions with the help of GIS data; Multi-dimensional data collection step: Deploy an intelligent heterogeneous sensor network, fuse visible light communication VLC and traditional radio frequency RF sensing technologies to collect data, dynamically adjust the sampling strategy according to reinforcement learning, and adaptively sample frequency and data type emphasis through formulas; Data preprocessing step: Use the denoising autoencoder DAGAN based on the generative adversarial network GAN to purify the original data, adopt an adaptive normalization method to process the data, reduce the dimensionality of multi-source data through tensor decomposition technology, and retain the high-order feature relationships; Handover scenario modeling step: Construct a hybrid model based on the spatio-temporal graph convolutional network ST-GCN and the long short-term memory network LSTM. ST-GCN captures spatial associations, and LSTM learns the change rules of time series. Introduce an attention gating mechanism AGC to focus on key spatio-temporal features; Automated test case generation step: Generate test cases based on Monte Carlo tree search MCTS-RL of reinforcement learning, construct a reward function through performance indicators, and dynamically adjust the test priority using the fuzzy clustering algorithm; Test execution and result analysis: Deploy a distributed test execution framework on edge computing nodes, use a real-time stream data analysis engine to analyze test results, and locate problem areas through the deep association rule mining algorithm combined with self-organizing mapping SOM; Optimization strategy formulation: Use multi-agent deep reinforcement learning MA-DRL to formulate optimization strategies. Agents are responsible for optimizing different network parameters, and explore and learn the optimal strategy through collaborative competition; Strategy implementation and feedback: Adopt software-defined network SDN technology to implement optimization strategies, adjust network configurations through interfaces, establish a blockchain-based feedback mechanism, and use smart contracts to verify and analyze data.

[0008] Furthermore, it also includes: Network security monitoring step: Deploy an intrusion detection system based on federated learning and generative adversarial network FL-GAN. At each edge node, use local data to train the generator to generate normal traffic samples, and the discriminator identifies abnormal traffic. Through the federated learning mechanism, each node collaboratively trains on the premise of protecting data privacy. When an anomaly is detected, start an emergency response smart contract to isolate the risk area and notify the administrator.

[0009] Furthermore, it also includes: Edge computing resource dynamic allocation steps: Use an adaptive resource allocation algorithm to construct a resource allocation agent. Based on the real-time network load, task priority, and node resource status, dynamically adjust the computing and storage to allocate network resources through the Q-learning algorithm. The formula is , where is the Q value of taking action in state s, is the learning rate, r is the reward, γ is the discount factor, is the next state.

[0010] Furthermore, the multi-dimensional data collection step adaptively changes the sampling frequency and data type focus according to the state-action-reward mechanism. The formula is , is the probability of taking action in state s, is the state-action value function, is the temperature parameter; Introduce the quantum key distribution (QKD) data collection technology. Encrypt the collected data using the quantum encryption principle and record the data collection timestamp and hash value in combination with the distributed ledger technology.

[0011] Furthermore, the switching scenario modeling step introduces the attention gating mechanism (AGC). Focus on the key spatio-temporal features through the formula. The formula is where is the gating signal, σ is the activation function, is the weight matrix, is the hidden state at the previous moment, is the current input; Adopt the method of combining variational autoencoder (VAE) and generative adversarial network (GAN) to expand the scenario data. VAE learns the real switching scenario distribution and generates latent samples. GAN optimizes the generated samples and transfers knowledge from similar network scenario models through transfer learning technology.

[0012] Furthermore, the automated test case generation step uses the Bayesian optimization test case generation algorithm to construct a Bayesian model to evaluate the impact of different test cases on network performance. Select the optimal test case through the expected improvement (EI) criterion. The formula is , where is the current optimal solution, is the objective function value.

[0013] Furthermore, the test execution and result analysis step uses an interpretable analysis method based on deep learning to interpret the test results. Generate explanations for the prediction results of the convolutional neural network (CNN) and recurrent neural network (RNN) models to locate the key factors affecting the switching performance.

[0014] Furthermore, the optimization strategy formulation step introduces a multi-objective optimization algorithm based on game theory, constructs a game model considering different requirements, and obtains the optimal handover strategy by solving the Nash equilibrium.

[0015] Furthermore, the strategy implementation and feedback step uses digital twin technology to construct a virtual network model, evaluates the strategy effect through simulation operation, and applies the optimized strategy in the virtual network to the real network by using the interaction between the virtual network and the real network.

[0016] Furthermore, an intelligent monitoring and early warning system is established. The Internet of Things and big data analysis technologies are used to monitor the hardware of edge computing nodes, network connections, and system performance. An anomaly detection model based on deep learning is used to identify abnormal behaviors. Natural language processing technology is combined to generate easy-to-understand early warning information, and various monitoring data are associated through knowledge graph technology.

[0017] Compared with the existing technologies, the beneficial effects of the present invention are as follows: In terms of network handover performance, through multi-dimensional data collection and comprehensive analysis, various factors such as signal strength, bandwidth, packet loss rate, and latency are comprehensively considered, and a more reasonable handover strategy is formulated, avoiding the limitations of traditional methods that only judge based on signal strength. This enables devices to perform handovers more timely and accurately in high-speed mobile scenarios, reduces handover delay and handover failure rate, improves the continuity and stability of network connections, and greatly enhances the user's network usage experience.

[0018] In terms of test efficiency, an automated test case generation and distributed test execution framework are used to realize the automation of high-speed wireless network handover tests. Reducing manual intervention improves test efficiency and enables rapid and comprehensive evaluation of network performance in different scenarios. At the same time, through a real-time stream data analysis engine and a deep association rule mining algorithm, the test results can be deeply analyzed to mine the association relationships between network parameters, providing strong support for network optimization.

[0019] In terms of resource utilization, the adaptive resource allocation algorithm based on reinforcement learning can dynamically adjust the resource allocation of edge computing nodes according to the real-time network load and task priorities, improving resource utilization efficiency and reducing operating costs.

[0020] In terms of network security, an intrusion detection system based on federated learning and generative adversarial networks is deployed, which can effectively identify and prevent abnormal behaviors in the network, ensuring the security of the wireless network. At the same time, the feedback mechanism based on blockchain and the emergency response intelligent contract ensure the authenticity and immutability of data, improving the reliability and traceability of the system.

[0021] In summary, the method of the present invention can effectively improve the performance and stability of high-speed handover in wireless networks, improve the test efficiency, optimize resource utilization, and ensure network security, and has important application value and market prospects. Brief Description of the Drawings

[0022] Figure 1 It is a schematic block diagram of an automated test optimization method for high-speed handover in a wireless network based on edge computing proposed by the present invention; Figure 2 It is a schematic diagram for comparing the handover success rates of different methods; Figure 3 It is a schematic diagram for comparing the handover delays in different scenarios; Figure 4 It is a schematic diagram for comparing the test efficiencies. Detailed Embodiments

[0023] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", etc. indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention.

[0025] In addition, the terms "first" and "second" are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the said features. In the description of the present invention, "a plurality of" means two or more unless otherwise specifically defined. In addition, the terms "mounted", "connected", and "coupled" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected or indirectly connected through an intermediate medium, and it may be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances. The present invention will be further described in detail below with reference to the accompanying drawings.

[0026] Referring to Figures 1 to 4 : An optimized method for automated testing of high-speed handover in a wireless network based on edge computing, comprising: Edge computing node deployment step: When deploying edge computing nodes, first collect network topology information in the wireless network coverage area, such as base station distribution, signal transmission paths, etc., and geographical information system (GIS) data, including terrain, building distribution, etc. At the same time, predict the expected traffic distribution in different regions according to historical traffic data and service requirements. Use the improved gravitational search algorithm (IGSA) to optimize the node positions. Initialize a set of random node positions as the initial population, and each node position is regarded as an object with mass and position attributes. By calculating the gravitational force between objects, update the positions and velocities of the objects, and gradually iterate to find the optimal node deployment positions. During the calculation process, use the comprehensive objective function , where is the distance from the node to the critical service area, is the construction and operation cost of the node, is the signal strength, is the average signal strength, , , can be adjusted according to the actual situation. For example, in an area with high signal coverage requirements, the weight of can be increased. Mark the optimized node positions on the GIS map and conduct on-site deployment to ensure that the nodes can efficiently process network data.

[0027] Multi-dimensional data acquisition steps: Deploy an intelligent heterogeneous sensor network, and reasonably distribute visible light communication (VLC) sensors and traditional radio frequency (RF) sensors within the wireless network coverage area. VLC sensors can be installed on indoor lighting devices to collect network data in the indoor environment; RF sensors can be installed at positions such as base stations and utility poles to collect outdoor network signals. The sensors dynamically adjust the sampling strategy using reinforcement learning. Define the state of the sensor as the real-time fluctuations of the network, such as the change rate of signal strength and the fluctuation amplitude of bandwidth; the actions include adjusting the sampling frequency and the emphasis on data types. For example, when the signal strength changes drastically, increase the sampling frequency and focus on collecting signal strength data. The reward function is designed based on the accuracy and timeliness of the collected data, and a positive reward is given when the collected data can accurately reflect the network state and be transmitted in a timely manner. By continuously interacting with the network environment, the sensors gradually learn the optimal sampling strategy to comprehensively collect network data such as signal strength, bandwidth, packet loss rate, and latency, as well as information such as the location, speed, and direction of mobile terminals.

[0028] Data preprocessing steps: For the collected raw data, first use a denoising autoencoder based on a generative adversarial network (GAN) (DAGAN) for denoising. Train the generator to generate simulated noise data, and the discriminator to distinguish real noise from simulated noise. Through continuous adversarial training, improve the denoising effect. For example, for signal strength data affected by electromagnetic interference, after being processed by DAGAN, the noise interference can be removed and the real signal strength value can be restored. Adaptive normalization method is used for data normalization. According to the dynamic mean and standard deviation of the data, use the formula to normalize the data. This can make different types of data have the same scale, facilitating subsequent analysis and processing. Tensor decomposition technology is used to reduce the dimensionality of multi-source data. Represent the collected multi-dimensional data in tensor form, and decompose the high-dimensional tensor into low-dimensional tensors through tensor decomposition algorithms, retaining the high-order feature relationships in the data, reducing data redundancy, and improving data processing efficiency.

[0029] Switching scenario modeling steps: Construct a hybrid model based on a spatio-temporal graph convolutional network (ST-GCN) and a long short-term memory network (LSTM). Regard the base stations and mobile terminals in the network as graph nodes, and the connections between nodes represent the signal transmission relationships between them to construct a graph structure. ST-GCN performs convolutional operations on the graph structure to capture the correlations of network parameters in the spatial dimension, such as the signal interference relationships between different base stations. LSTM processes time series data to learn the changing patterns of network parameters over time, such as the fluctuations of signal strength over time. Introduce an attention gating mechanism (AGC), and calculate the gating signal through the formula where σ is the activation function, is the weight matrix, is the hidden state at the previous moment, is the current input. The gating signal is used to control the degree of attention of the model to different spatio-temporal features, focusing on the features that have a greater impact on the handover decision, and improving the model's ability to depict complex handover scenarios. A large amount of historical data is used to train the model, and the model parameters are continuously adjusted to enable the model to accurately predict the network handover situation in different scenarios.

[0030] Steps for generating automated test cases: Generate test cases based on reinforcement learning Monte Carlo tree search (MCTS-RL). Define the state of the agent as the current network environment, including network load, signal strength, mobile terminal speed, etc.; the actions include selecting different combinations of test parameters, such as handover thresholds, signal strength evaluation algorithms, etc. The reward function is designed according to the performance metrics of the test results, such as handover success rate, latency, etc., and a positive reward is given when the test results meet the expected goals. The agent explores in the simulated network environment, constructs a search tree through the Monte Carlo tree search algorithm, and selects the optimal test action. During the search process, the nodes of the search tree are continuously expanded, the value of each node is evaluated, and the node with the largest value is selected for the next exploration. Use the fuzzy clustering algorithm to classify the generated test cases, and dynamically adjust the priorities of the test cases according to different network scenarios and test objectives. For example, test cases for high-speed mobile scenarios are given higher priorities to ensure test coverage of critical scenarios.

[0031] Steps for test execution and result analysis: Deploy a distributed test execution framework on edge computing nodes, and distribute the generated test cases to different edge nodes for parallel execution. Use a real-time stream data analysis engine (such as Apache Flink) to perform real-time processing and analysis of the data during the test. During the test, network performance metrics, such as handover success rate, latency, packet loss rate, etc., are continuously collected. Use a deep association rule mining algorithm (such as an extended Eclat algorithm) to mine the association relationships between network parameters. For example, analyze the association between signal strength and packet loss rate to find the key factors affecting handover performance. Combine self-organizing mapping (SOM) neural networks to map high-dimensional test data onto a two-dimensional plane to visually present the distribution of handover performance. By observing the clustering situation in the SOM map, quickly locate the areas with poor network performance and provide a basis for network optimization.

[0032] Optimization Strategy Formulation Step: Use multi-agent deep reinforcement learning (MA-DRL) to formulate an optimization strategy. Assign different parameter optimization tasks in the network to multiple agents. For example, one agent is responsible for optimizing the handover threshold, and another agent is responsible for optimizing the signal strength evaluation algorithm, etc. Each agent explores and learns in the network environment according to its own task. Through cooperation and competition among agents, with the goal of optimizing the overall performance of the network, they continuously adjust their strategies. Use the Q-learning algorithm to update the Q-value of the agent, and the formula is , where is the learning rate, r is the reward, γ is the discount factor, and is the next state. Through continuous iterative learning, the agent gradually finds the optimal optimization strategy to improve the performance of high-speed handover in the wireless network.

[0033] Strategy Implementation and Feedback Step: Use software-defined network (SDN) technology to implement the optimization strategy. Through the southbound interface of the SDN controller, the configuration of network devices is adjusted in real time, such as modifying the handover parameters of the base station and adjusting the signal strength evaluation algorithm. At the same time, establish a feedback mechanism based on blockchain. Each edge node encrypts and uploads the network performance data after the implementation of the strategy to the blockchain. Use smart contracts to automatically verify and analyze this data to determine whether the strategy is effective. If the network performance is improved after the implementation of the strategy, then continue to apply this strategy; if the performance deteriorates, then feedback the new data to the handover scenario model and the optimization strategy module to update and optimize the model, adjust the optimization strategy, and form a closed-loop optimization process.

[0034] In the present invention, the following steps are further included: Steps for network security monitoring: Deploy an intrusion detection system based on federated learning and generative adversarial networks (FL-GAN). On each edge node, use the rich local network traffic data to train the generator. The generator adopts a deep neural network architecture and learns and extracts the features of local normal network traffic data through a combination of multi-layer convolutional neural networks and deconvolutional neural networks, and then generates realistic normal traffic samples. The discriminator is also built based on a deep neural network, with a convolutional neural network as the main structure, responsible for screening the input traffic samples and accurately identifying abnormal traffic. With the help of the federated learning mechanism, each edge node realizes collaborative training on the premise of strictly protecting local data privacy. Specifically, each node only uploads the updated parameters of the model instead of the original data, aggregates the parameters on the central server through secure aggregation technology, and then distributes the updated model to each node for continuous iterative optimization, thus significantly improving the recognition ability of the intrusion detection model for diverse attacks (such as DDoS attacks, SQL injection attacks, etc.). Once the discriminator successfully detects abnormal traffic, the system immediately triggers an emergency response smart contract based on blockchain technology. This smart contract has pre-set detailed rules and execution logic, ensures the reliability of instruction execution through the immutable feature of the blockchain, automatically isolates the risk area, and timely notifies the network administrator in a reliable communication manner to achieve rapid response and effective disposal of network security threats.

[0035] In the present invention, the following steps are further included: Steps for dynamic allocation of edge computing resources: Adopt an adaptive resource allocation algorithm based on reinforcement learning to construct a resource allocation agent. This agent collects network load information based on real-time perception of the network environment, including but not limited to the traffic rate, the number of data packets of each edge node, etc., to measure the current busyness of the network; at the same time, obtain task priorities, for example, tasks with high real-time requirements (such as high-definition video stream transmission, industrial control instruction transmission, etc.) need to ensure resource supply first; it also monitors the node resource status, such as the CPU usage rate of the computing node, the remaining memory, and the storage space occupancy rate of the storage node, etc. The agent dynamically adjusts the allocation of computing, storage, and network resources through the Q-learning algorithm (the formula is , where is the Q value of taking action in state s, is the learning rate, r is the reward, γ is the discount factor, is the next state) to ensure efficient resource utilization and stable switching performance.

[0036] In the present invention, the multi-dimensional data acquisition step introduces a data acquisition technology that guarantees data transmission security based on quantum key distribution (QKD). Quantum key distribution is based on the basic principles of quantum mechanics and uses the quantum states of photons to generate and distribute encryption keys. During the data acquisition process, the sending end randomly prepares a single-photon sequence in different quantum states (such as polarization states) to encode key information. The receiving end then measures the received photons using the corresponding measurement basis. Due to the non-clonability of quantum states and the property that measurement causes quantum state collapse, any eavesdropping behavior by a third party will be detected by the sending end and the receiving end. Once it is determined that the key distribution process is secure, the generated quantum key is used to encrypt the acquired data, ensuring that even if the data is intercepted during transmission to the edge node, it cannot be stolen or tampered with. To further enhance data credibility, distributed ledger technology is combined. Each sensor participates as a ledger node. When the data acquisition is completed, the sensor records an accurate data acquisition timestamp, which is calibrated based on a time synchronization protocol (such as the Network Time Protocol NTP) to ensure the accuracy of time. In addition, the sensor also calculates the hash value of the acquired data. The hash function uses a secure cryptographic hash algorithm such as SHA-256 to map the data to a hash value of a fixed length. These timestamps and hash values are recorded in the distributed ledger. The data in the ledger is interconnected through the chain structure and cryptographic technology of the blockchain and has the property of being tamper-proof, enabling the source and integrity of the data to be reliably verified, thus comprehensively guaranteeing the security and credibility of multi-dimensional data acquisition.

[0037] In the present invention, in the switching scenario modeling step, the variational autoencoder (VAE) and the generative adversarial network (GAN) are combined to augment the scenario data. As a generative model, the variational autoencoder (VAE) maps the real switching scenario data to a low-dimensional latent space through an encoder to learn its distribution characteristics. The encoder network is usually composed of multiple layers of neural networks, which extract features from the input switching scenario data (such as signal strength, node position, network topology, etc.) to obtain the mean and variance parameters of the latent space. By using the reparameterization trick to sample from the latent distribution, and then the decoder restores the sampled latent vector to a switching scenario data sample, so as to learn the distribution of the real switching scenario and generate latent samples. The generative adversarial network (GAN) further optimizes the samples generated by the VAE. The GAN consists of a generator and a discriminator. The generator receives the latent samples generated by the VAE and processes them to generate data closer to the real scenario. The discriminator is responsible for judging whether the input data comes from the real scenario or the samples generated by the generator. In the continuous adversarial training process, the generator tries to generate more realistic data to deceive the discriminator, and the discriminator continuously improves its discrimination ability, thereby enhancing the diversity and authenticity of the generated samples. In addition, through transfer learning techniques, knowledge is transferred from similar network scenario models. For example, in network scenarios with similar topological structures or communication modes, the feature representations, parameters, etc. learned by the trained model are transferred to the current switching scenario model. This reduces the amount of data and computational resources required for model training, accelerates the training process of the current model, and at the same time improves the generalization ability of the model, enabling it to better adapt to complex and changing wireless network switching scenarios.

[0038] In the present invention, in the automated test case generation step, a test case generation algorithm based on Bayesian optimization is adopted. First, a Bayesian model is constructed. This model is based on probability theory and uses prior knowledge and continuously acquired new data to update the understanding of the relationship between network performance and test cases. In this model, different test cases are regarded as input variables, which cover various factors affecting high-speed switching of wireless networks, such as changes in network topology, node movement speed, signal interference intensity, etc. And network performance indicators, such as handover success rate, delay time, packet loss rate, etc. are used as the objective function values. Based on the constructed Bayesian model, the optimal test cases are selected by maximizing the expected improvement (EI) criterion. The formula is the core expression of this criterion. Among them, represents the currently determined optimal solution, which is the combination of test case parameters that can make the network performance reach the relatively best state found in the previous test and analysis process; is the objective function value corresponding to any test case , reflecting the performance of the network under this test case. E represents taking the expectation, means only considering when the difference between the two at that time, which measures from the current test case to the optimal solution the expected benefit that can be brought. By continuously applying this criterion to evaluate and screen a large number of potential test cases, while ensuring sufficient test coverage for various scenarios and influencing factors of high-speed handover in wireless networks, the number of test cases can be significantly reduced. A large number of redundant tests are avoided, thus greatly improving the test efficiency, more accurately locating network performance problems and optimizing them.

[0039] In the present invention, the test execution and result analysis steps adopt an interpretable analysis method based on deep learning (such as the LIME algorithm) to interpret the test results. Taking the LIME algorithm (Local Interpretable Model-agnostic Explanations algorithm) as an example, it can provide an intuitive and easy-to-understand explanation for the prediction results of deep learning models. In the high-speed handover test of wireless networks, models such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs) are often used for performance prediction. CNNs are good at processing data with a grid structure and perform well in analyzing network features related to spatial distribution (such as the distribution of signal strength at different geographical locations); RNNs, due to their good processing ability for time series data, are suitable for analyzing the change of network performance over time (such as the fluctuation of handover delay at consecutive time points). The LIME algorithm works based on these models. It perturbs the local data near the model prediction results to generate a series of new data samples. Based on these new samples, a simple and interpretable surrogate model (such as a linear regression model) is constructed. The coefficients and feature importances of this surrogate model can clearly show which input features (such as network topology, node movement speed, signal interference strength, etc.) have a key impact on the model's prediction results. In this way, testers can deeply understand the decision-making basis of the model. Instead of just facing a prediction result given by the deep learning model, they can clearly know why the model makes such a judgment. Furthermore, the key factors affecting the high-speed handover performance of wireless networks can be quickly located, such as whether the signal interference in a certain area causes an increase in the handover failure rate, or whether the node movement is too fast resulting in the handover delay exceeding the standard. This greatly improves the efficiency of problem troubleshooting and solution, and helps to more targeted optimize the performance of high-speed handover in wireless networks.

[0040] In the present invention, in the optimization strategy formulation step, a multi-objective optimization algorithm based on game theory is introduced to comprehensively consider the requirements of different stakeholders in the network. Users pursue a high-quality network experience, paying attention to low latency, high bandwidth, and high stability during handover, expecting seamless network connection handover during movement to ensure the smooth operation of various services (such as high-definition video streaming, real-time games, etc.); operators focus on the efficient utilization of network resources, control of operating costs, and maintenance of service quality, striving to achieve profitability while meeting user needs; service providers hope that the services they provide are not affected during network handover, ensuring the reliability and availability of services. Based on these different requirements, a game model is constructed. In this model, each stakeholder is regarded as a game participant, and they each have different strategy spaces. For example, users can choose different device configurations and network access modes; operators can adjust base station layouts and resource allocation strategies; service providers can optimize service deployment architectures and data transmission protocols, etc. The strategy choices of each participant will affect each other and generate corresponding benefits or costs. The optimal handover strategy is determined by solving the Nash equilibrium. The Nash equilibrium refers to a stable state in a game where, after all participants have chosen their respective strategies, no single party can improve its own benefit by unilaterally changing its strategy. In the wireless network handover scenario, through complex mathematical calculations and iterative solution processes, a strategy combination that balances the interests of all parties is found. At this time, network performance is guaranteed, the user experience is improved, and at the same time, operating costs can be effectively controlled, achieving a win-win situation for all parties and maximizing the overall network benefit.

[0041] In the present invention, the strategy implementation and feedback step uses digital twin technology to construct a virtual network model. The digital twin technology is used to construct a highly realistic virtual network model, which comprehensively maps the topological structure of the real wireless network, accurately covering the positions and connection relationships of base stations, access points, terminal devices, etc.; details the physical layer characteristics of the network, including signal propagation, attenuation, interference, etc.; and also carefully models the protocol layer and application layer of the network, such as the operation mechanisms of various network protocols and the traffic characteristics of different service applications. After the virtual network model is constructed, the optimization strategy is rehearsed in it. During the rehearsal process, by simulating different network operation scenarios, such as high-traffic periods, user fast-moving scenarios, etc., the strategy effect is comprehensively evaluated. With the help of advanced simulation tools and algorithms, accurate calculations and real-time monitoring of network performance indicators (such as handover success rate, latency, throughput, etc.) are carried out. During this process, potential problems can be discovered in advance, such as local network congestion that may occur after the strategy is implemented, and handover failures caused by excessive load on certain devices. Using the real-time interaction mechanism between the virtual network and the real network, efficient migration of the strategy is achieved. By continuously adjusting and optimizing the strategy in the virtual network, when the effectiveness of the strategy is determined, with the help of technologies such as the Internet of Things and big data, the optimized strategy parameters in the virtual network are quickly and accurately transmitted to each control unit and device in the real network. At the same time, the real network feeds back real-time operation data to the virtual network to further calibrate and optimize the virtual model, forming a continuously iterative and two-way optimized closed-loop system, significantly improving the accuracy and efficiency of strategy implementation, and ensuring the stable and efficient operation of high-speed handover of the wireless network.

[0042] In the present invention, an intelligent monitoring and early warning system is established. With the help of Internet of Things technology, various sensors are deployed at edge computing nodes to collect the operating parameters of hardware devices (such as CPU, GPU, memory modules, etc.) in real time, including temperature, utilization rate, voltage, etc.; at the same time, the network connection status is monitored, such as signal strength, packet loss rate, network latency, etc.; and system performance indicators, such as task processing speed, resource allocation efficiency, etc. The collected data is aggregated to a big data analysis platform through a high-speed communication link. In terms of anomaly detection, an autoencoder model based on deep learning is adopted. The autoencoder consists of an encoder and a decoder. The encoder compresses the high-dimensional monitoring data into a low-dimensional feature representation, and the decoder then attempts to restore it. Under normal operating conditions, the reconstruction error is small. Once an anomaly occurs in the network or device, the data features change, and the reconstruction error will increase significantly, thereby identifying abnormal behaviors. Combining natural language processing technology, the system can convert the detected anomaly information into understandable early warning information. For example, using a text generation algorithm, complex technical indicator anomalies are converted into intuitive expressions such as "The CPU temperature of the edge node is too high, which may affect the network switching performance. Please check the cooling system in time", which is convenient for operation and maintenance personnel to understand. In addition, through knowledge graph technology, correlation analysis is performed on various monitoring data. The knowledge graph stores and represents the semantic relationships between data in a graph structure. For example, it associates hardware failures with information such as network performance degradation and system error logs. Based on this, the system can intelligently diagnose the root cause of the problem and provide targeted handling suggestions, such as "It is detected that the connection between a certain base station and the edge node is interrupted, which may be a fiber optic cable failure. It is recommended to check the corresponding line". Through this series of technical means, the stable operation of the wireless network is guaranteed in all aspects, potential problems are discovered and solved in a timely manner, and the reliability of high-speed switching of the wireless network in the edge computing environment is ensured.

[0043] The above is only a preferred specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. An automated test optimization method for high-speed handover of wireless networks based on edge computing, characterized in that, It includes the following steps: Edge computing node deployment step: Use the improved gravitational search algorithm IGSA combined with geographic information system GIS data to optimize the deployment of edge computing nodes. Consider the network topology, geographical environment, and expected traffic distribution to construct a comprehensive objective function, and use GIS data to plan the node locations; Multi-dimensional data collection step: Deploy an intelligent heterogeneous sensor network, fuse visible light communication VLC and traditional radio frequency RF sensing technologies to collect data, dynamically adjust the sampling strategy according to reinforcement learning, and adaptively sample frequencies and data type emphases through formulas; Data preprocessing step: Use the denoising autoencoder DAGAN based on the generative adversarial network GAN to purify the original data, adopt an adaptive normalization method to process the data, and reduce the dimensions of multi-source data through tensor decomposition technology while retaining high-order feature relationships; Switching scenario modeling step: Construct a hybrid model based on the spatio-temporal graph convolutional network ST-GCN and the long short-term memory network LSTM. ST-GCN captures spatial associations, and LSTM learns the changing rules of time series. Introduce an attention gating mechanism AGC to focus on key spatio-temporal features; Automated test case generation step: Generate test cases based on Monte Carlo tree search MCTS-RL with reinforcement learning, construct a reward function through performance metrics, and dynamically adjust the test priorities using a fuzzy clustering algorithm; Test execution and result analysis: Deploy a distributed test execution framework on edge computing nodes, use a real-time stream data analysis engine to analyze test results, and locate problem areas through a deep association rule mining algorithm combined with self-organizing mapping SOM; Optimization strategy formulation: Use multi-agent deep reinforcement learning MA-DRL to formulate optimization strategies. Agents are responsible for optimizing different network parameters, and explore and learn the optimal strategy through collaborative competition; Strategy implementation and feedback: Use software-defined network SDN technology to implement optimization strategies, adjust network configurations through interfaces, establish a blockchain-based feedback mechanism, and use smart contracts to verify and analyze data.

2. The automated test optimization method for high-speed handover of a wireless network based on edge computing according to claim 1, wherein It also includes: Network security monitoring step: Deploy an intrusion detection system based on federated learning and generative adversarial network FL-GAN. On each edge node, use local data to train a generator to generate normal traffic samples, and a discriminator to identify abnormal traffic. Through the federated learning mechanism, each node collaboratively trains while protecting data privacy. When an anomaly is detected, start an emergency response smart contract to isolate the risk area and notify the administrator.

3. The automated test optimization method for high-speed handover of a wireless network based on edge computing according to claim 1, wherein It also includes: Steps for dynamic allocation of edge computing resources: An intelligent agent for resource allocation is constructed using an adaptive resource allocation algorithm. Based on the real-time network load, task priority, and node resource status, the computing and storage allocation of network resources is dynamically adjusted through the Q-learning algorithm. The formula is , where is the Q value of taking action in state s, is the learning rate, r is the reward, γ is the discount factor, is the next state.

4. The automated test optimization method for high-speed handover of a wireless network based on edge computing according to claim 1, wherein The multi-dimensional data acquisition step adaptively changes the sampling frequency and the emphasis on data types according to the state-action-reward mechanism, and the formula is , is the probability of taking action in state s, is the state-action value function, is the temperature parameter; Introduce data collection technology based on quantum key distribution QKD, encrypt the collected data using the principle of quantum encryption, and combine distributed ledger technology to record the data collection timestamps and hash values.

5. The automated test optimization method for high-speed handover of a wireless network based on edge computing according to claim 1, wherein The switching scene modeling step introduces the attention gating mechanism AGC, which focuses on key spatio-temporal features through a formula. The formula is where is the gating signal, σ is the activation function, is the weight matrix, is the hidden state at the previous moment, is the current input; Adopt a method that combines variational autoencoder VAE and generative adversarial network GAN to expand scenario data. VAE learns the real switching scenario distribution and generates latent samples, and GAN optimizes the generated samples. Transfer knowledge from similar network scenario models through transfer learning technology.

6. The automated test optimization method for high-speed handover of a wireless network based on edge computing according to claim 1, wherein The steps of automated test case generation use a test case generation algorithm based on Bayesian optimization to construct a Bayesian model to evaluate the impact of different test cases on network performance, and select the optimal test case through the Expected Improvement (EI) criterion maximization. The formula is , where is the current optimal solution, is the objective function value.

7. The automated test optimization method for high-speed handover of a wireless network based on edge computing according to claim 1, characterized in that The test execution and result analysis step uses a deep learning-based interpretability analysis method to interpret test results, generates explanations for the prediction results of convolutional neural network CNN and recurrent neural network RNN models, and locates the key factors affecting switching performance.

8. The automated test optimization method for high-speed handover of a wireless network based on edge computing according to claim 1, wherein The optimization strategy formulation step introduces a multi-objective optimization algorithm based on game theory, constructs a game model considering different requirements, and obtains the optimal switching strategy by solving the Nash equilibrium.

9. The automated test optimization method for high-speed handover of a wireless network based on edge computing according to claim 1, wherein The strategy implementation and feedback step uses digital twin technology to construct a virtual network model, evaluates the strategy effect through simulation operation, uses the interaction between the virtual network and the real network, and applies the optimized strategy in the virtual network to the real network.

10. The automated test optimization method for high-speed handover of a wireless network based on edge computing according to claim 1, wherein An intelligent monitoring and early warning system is established. The Internet of Things and big data analysis technologies are used to monitor the hardware of edge computing nodes, network connections, and system performance. An anomaly detection model based on deep learning is used to identify abnormal behaviors. Natural language processing technology is combined to generate easy-to-understand early warning information, and various types of monitoring data are associated through knowledge graph technology.

Images