An intelligent switching system based on the frequency of communication between an adapter and a device

Abstract

This invention relates to the field of intelligent communication equipment technology, specifically to an intelligent switching system based on the communication frequency between an adapter and a device. The system includes: a communication status acquisition module for acquiring real-time communication parameters, instantaneous power consumption, and link attenuation characteristics between the adapter and the device; a status feature analysis module for fusing data and constructing a dynamic graph of communication quality and device requirements using a self-supervised spatiotemporal frequency elasticity coefficient model, quantifying the nonlinear correlation between frequency adaptation and response delay; a frequency decision module for generating a distributed switching strategy through federated transfer reinforcement learning, combined with knowledge graph temporal causal chain reasoning, to output target frequency parameters; and a dynamic control feedback module for executing switching through a three-objective dynamic equilibrium algorithm, real-time monitoring of communication jitter and data fidelity, and automatically generating a frequency optimization report with device aging compensation coefficients. This solves the problems of static feature analysis, inefficient decision-making mechanisms, and lack of control feedback in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of intelligent communication equipment technology, specifically to an intelligent switching system based on the communication frequency between an adapter and a device. Background Technology

[0002] In traditional adapter and device communication systems, frequency management methods, such as fixed frequency presets or manual configuration based on experience, are typically used to ensure the stability of basic communication connections. During the factory or initial deployment phase, adapters and devices are pre-set with a set of standard communication frequencies based on device type and communication protocol, or technicians manually adjust these frequencies to suit the current environment using debugging tools. This approach plays a fundamental role in static or simple dynamic scenarios: on the one hand, fixed frequencies reduce the complexity of communication protocols and decrease the computational load during device initialization; on the other hand, manual configuration allows for optimized frequency selection for specific environments, ensuring the reliability of basic data interactions such as command issuance and status reporting, meeting the basic requirement of "being able to communicate" between devices. This is a crucial technical means supporting device interconnection in fields such as consumer electronics and the Industrial Internet of Things (IIoT).

[0003] While adapter-device communication frequency switching technology is widely used in smart device interaction and energy efficiency optimization, existing technologies still have significant limitations. Current parameter acquisition is largely limited to basic signal strength and transmission rate, lacking real-time synchronous acquisition of instantaneous power consumption and link attenuation characteristics. This makes it impossible to construct a complete communication status profile, resulting in insufficient data support and difficulty in comprehensively assessing the correlation between communication quality, device energy consumption, and link stability. Feature analysis is static: it relies heavily on fixed thresholds or static models, lacking self-supervised spatiotemporal frequency elasticity coefficient modeling, failing to quantify the nonlinear relationship between frequency adaptation and response delay, and exhibiting weak model generalization ability in dynamic scenarios. The decision-making mechanism is inefficient: it is primarily centralized, lacking distributed collaborative capabilities through federated transfer learning, and does not incorporate knowledge graph temporal causal chain reasoning. In distributed heterogeneous scenarios, decision latency is high, and the adaptability of target frequency parameters is poor. Regulation feedback is lacking: there is no real-time monitoring of communication jitter and data fidelity after switching, and there is a lack of device aging compensation mechanisms, making it difficult to form a decision-execution-feedback-optimization closed loop, resulting in insufficient multi-objective optimization and full lifecycle performance guarantees. These shortcomings limit the energy efficiency improvement, stability, and long-term service performance of communication systems, requiring innovative technical solutions to overcome these bottlenecks. Summary of the Invention

[0004] This application provides an intelligent switching system based on the communication frequency between the adapter and the device to solve the problems of static feature analysis, inefficient decision-making mechanism, and lack of control feedback in the prior art.

[0005] The first aspect of this application provides an intelligent switching system based on the communication frequency between an adapter and a device, comprising: a communication status acquisition module, a status feature analysis module, a frequency decision module, and a dynamic adjustment feedback module; wherein, the communication status acquisition module is used to acquire real-time communication parameters, instantaneous power consumption, and link attenuation characteristics between the adapter and the device; the status feature analysis module is used to fuse data, construct a communication quality-device demand dynamic graph through a self-supervised spatiotemporal frequency elasticity coefficient model, and quantify the nonlinear correlation between frequency adaptation and response delay; the frequency decision module, based on the nonlinear correlation, generates a distributed switching strategy through federated transfer reinforcement learning, and outputs target frequency parameters by combining knowledge graph temporal causal chain reasoning; the dynamic adjustment feedback module, based on the distributed switching strategy, executes switching through a three-objective dynamic equilibrium algorithm, monitors communication jitter and data fidelity in real time, and automatically generates a frequency optimization report with device aging compensation coefficients.

[0006] Preferably, the communication status acquisition module includes a real-time communication parameter acquisition unit, an instantaneous power consumption monitoring unit, and a link attenuation characteristic evaluation unit. The real-time communication parameter acquisition unit is used to acquire real-time communication parameters between the adapter and the device using protocol analysis tools, and to record key indicators such as transmission rate, latency, and bit error rate. The instantaneous power consumption monitoring unit is used to synchronously acquire instantaneous power consumption data between the adapter and the device using a current sensor and a voltage measurement device, and to analyze the power consumption characteristics under different operating conditions. The link attenuation characteristic evaluation unit is used to evaluate the attenuation characteristics of the link between the adapter and the device using signal strength detection and quality analysis algorithms, identify signal interference sources, and quantify the link quality change trend.

[0007] Preferably, the state feature analysis module includes a data fusion processing unit, a dynamic graph construction unit, and a nonlinear correlation quantization unit. The data fusion processing unit extracts and fuses features from the original data using a self-supervised learning algorithm, eliminating noise interference and dimensional differences between data. The dynamic graph construction unit constructs a communication quality-equipment requirement dynamic graph based on the fused data, performing multi-scale spatiotemporal correlation between time-series features and spatial distribution features. The nonlinear correlation quantization unit analyzes the changes in the elasticity coefficient in the dynamic graph to quantify the nonlinear correlation strength and dynamic evolution law between frequency adaptation and response delay.

[0008] Preferably, the frequency decision module includes a federated transfer reinforcement learning unit and a temporal causal chain reasoning unit. The federated transfer reinforcement learning unit performs distributed policy training and optimization using a federated transfer reinforcement learning algorithm to generate a distributed switching policy. The temporal causal chain reasoning unit performs temporal causal analysis using a knowledge graph temporal causal chain reasoning algorithm and, in conjunction with the distributed switching policy, outputs target frequency parameters.

[0009] Preferably, the formula for the federated transfer reinforcement learning algorithm is:

[0010]

[0011] in, The value when the decision-making strategy reaches its optimal state; The environmental conditions in which the decision is made; This refers to the specific operation selected in the current state. The expected state transition probability is based on the environment. For immediate rewards; To represent the discount factor, the value range is 0 ≤ ... ≦1; To select the action that will bring the greatest value from all possible next actions; The next state; Next action.

[0012] Preferably, the dynamic control feedback module includes an algorithm execution unit and a report generation unit. The algorithm execution unit uses a three-objective dynamic equilibrium algorithm to dynamically weigh and decide on key indicators such as latency, bandwidth, and reliability during the switching process. The report generation unit collects and stores real-time status data by monitoring communication jitter amplitude and data fidelity indicators in real time, and combines the equipment aging compensation coefficient to perform correlation analysis between the monitoring data and the switching results, automatically generating an optimization report containing frequency adjustment suggestions.

[0013] The second aspect of this application provides an intelligent switching method based on the communication frequency between an adapter and a device, comprising: acquiring real-time communication parameters between the adapter and the device; fusing the real-time communication parameters between the adapter and the device, constructing a communication quality-device demand dynamic graph through a self-supervised spatiotemporal frequency elasticity coefficient model, and quantifying the nonlinear correlation strength and dynamic evolution law between frequency adaptation and response delay by analyzing the changes in elasticity coefficients in the dynamic graph; based on the nonlinear correlation strength and dynamic evolution law, performing distributed policy training and optimization through federated transfer reinforcement learning to generate a distributed switching strategy, and outputting target frequency parameters by combining knowledge graph temporal causal chain reasoning; and based on the distributed switching strategy, performing switching through a three-objective dynamic equilibrium algorithm, monitoring communication jitter and data fidelity in real time, collecting and storing real-time status data, and automatically generating a frequency optimization report with device aging compensation coefficients.

[0014] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement a method for intelligent switching of communication frequency between an adapter and a device as described in the above embodiments.

[0015] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a method for intelligent switching of communication frequencies between an adapter and a device as described in the above embodiments.

[0016] A fifth aspect of this application provides a computer program product, including a computer program or instructions, for implementing a method for intelligent switching of communication frequencies between an adapter and a device as described in the above embodiments.

[0017] Therefore, this application has the following beneficial effects:

[0018] This application's embodiments acquire real-time communication parameters, instantaneous power consumption, and link attenuation characteristics of the adapter and device through a communication status acquisition module, effectively capturing dynamic changes in communication status and basic link quality data, providing comprehensive perception support for subsequent analysis. Combined with a self-supervised spatiotemporal frequency elasticity coefficient model from the status feature analysis module, it fuses multi-source data to construct a dynamic graph of communication quality and device requirements, quantifying the nonlinear correlation between frequency adaptation and response delay, and improving the depth and breadth of communication status feature analysis. Relying on a federated transfer reinforcement learning algorithm in the frequency decision module, it generates a distributed switching strategy, combines knowledge graph temporal causal chain reasoning, and outputs target frequency parameters, enhancing the adaptability and accuracy of frequency decisions in different scenarios. The dynamic control feedback module executes the switching strategy through a three-objective dynamic equilibrium algorithm, monitors communication jitter and data fidelity in real time, automatically generates a frequency optimization report with device aging compensation coefficients, continuously optimizes communication performance, and dynamically compensates for device aging, improving the long-term stability and reliability of the system. Thus, it solves the problems of static feature analysis, inefficient decision-making mechanisms, and lack of control feedback in existing technologies.

[0019] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0020] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0021] Figure 1 This is a schematic diagram of a smart switching system based on the communication frequency between an adapter and a device, according to an embodiment of this application.

[0022] Figure 2 This is a schematic diagram of a communication status acquisition module according to an embodiment of this application;

[0023] Figure 3This is a schematic diagram of a state feature analysis module provided according to an embodiment of this application;

[0024] Figure 4 This is a schematic diagram of a frequency decision module provided according to an embodiment of this application;

[0025] Figure 5 This is a schematic diagram of a dynamic control feedback module provided according to an embodiment of this application;

[0026] Figure 6 This is a schematic diagram of an intelligent switching system provided according to an embodiment of this application;

[0027] Figure 7 This is a schematic diagram of an intelligent switching system based on the communication frequency between an adapter and a device, according to an embodiment of this application.

[0028] Figure 8 This is a flowchart illustrating an intelligent switching method based on the communication frequency between an adapter and a device, according to an embodiment of this application.

[0029] Figure 9 This is a schematic diagram of an intelligent switching method based on the communication frequency between an adapter and a device according to an embodiment of this application;

[0030] Figure 10 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of this application. Detailed Implementation

[0031] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0032] The following describes an intelligent switching system based on the communication frequency between an adapter and a device, according to an embodiment of this application, with reference to the accompanying drawings. Addressing the high detection error rate mentioned in the background section, this application provides an intelligent switching system based on the communication frequency between an adapter and a device. In this system, a communication status acquisition module acquires real-time communication parameters, instantaneous power consumption, and link attenuation characteristics of the adapter and device, effectively capturing dynamic changes in communication status and basic link quality data, providing comprehensive perception support for subsequent analysis. Combined with a self-supervised spatiotemporal frequency elasticity coefficient model from the status feature analysis module, multi-source data is fused to construct a dynamic graph of communication quality and device requirements, quantifying the nonlinear correlation between frequency adaptation and response delay, and improving the depth and breadth of communication status feature analysis. A distributed switching strategy is generated using a federated transfer reinforcement learning algorithm in the frequency decision module, combined with knowledge graph temporal causal chain reasoning, to output target frequency parameters, enhancing the adaptability and accuracy of frequency decisions in different scenarios. A dynamic adjustment feedback module executes the switching strategy through a three-objective dynamic equilibrium algorithm, monitors communication jitter and data fidelity in real time, automatically generates a frequency optimization report with device aging compensation coefficients, continuously optimizes communication performance, and dynamically compensates for device aging, improving the long-term stability and reliability of the system. This solves the problems of static feature analysis, inefficient decision-making mechanisms, and lack of regulatory feedback in existing technologies.

[0033] Figure 1 This is a schematic diagram of the structure of an intelligent switching system based on the communication frequency between the adapter and the device, provided in an embodiment of this application.

[0034] This application provides an intelligent switching system based on the communication frequency between an adapter and a device. The system 10 includes:

[0035] The system includes a communication status acquisition module 100, a status feature analysis module 200, a frequency decision module 300, and a dynamic control feedback module 400.

[0036] The communication status acquisition module 100 is used to acquire real-time communication parameters, instantaneous power consumption, and link attenuation characteristics between the adapter and the device; the status feature analysis module 200 is used to fuse data, construct a dynamic graph of communication quality and device demand through a self-supervised spatiotemporal frequency elasticity coefficient model, and quantify the nonlinear correlation between frequency adaptation and response delay; the frequency decision module 300 generates a distributed switching strategy based on nonlinear correlation through federated transfer reinforcement learning, and outputs the target frequency parameters by combining knowledge graph temporal causal chain reasoning; the dynamic adjustment feedback module 400 executes switching based on the distributed switching strategy through a three-objective dynamic equilibrium algorithm, monitors communication jitter and data fidelity in real time, and automatically generates a frequency optimization report with device aging compensation coefficient.

[0037] It is understood that in this embodiment, the communication status acquisition module acquires real-time communication parameters, instantaneous power consumption, and link attenuation characteristics of the adapter and device, effectively capturing dynamic changes in communication status and basic data on link quality, providing comprehensive perception support for subsequent analysis. Combined with the self-supervised spatiotemporal frequency elasticity coefficient model of the status feature analysis module, multi-source data is fused to construct a dynamic graph of communication quality and device requirements, quantifying the nonlinear correlation between frequency adaptation and response delay, and improving the depth and breadth of communication status feature analysis. Relying on the federated transfer reinforcement learning algorithm of the frequency decision module, a distributed switching strategy is generated, combined with knowledge graph temporal causal chain reasoning, to output target frequency parameters, enhancing the adaptability and accuracy of frequency decisions in different scenarios. The dynamic control feedback module executes the switching strategy through a three-objective dynamic equilibrium algorithm, monitors communication jitter and data fidelity in real time, automatically generates a frequency optimization report with device aging compensation coefficients, continuously optimizes communication performance, and dynamically compensates for device aging, improving the long-term stability and reliability of the system. Thus, it solves the problems of static feature analysis, inefficient decision-making mechanisms, and lack of control feedback in existing technologies.

[0038] In this embodiment of the application, the communication status acquisition module 100 includes: Figure 2 As shown, there are a real-time communication parameter acquisition unit, an instantaneous power consumption monitoring unit, and a link attenuation characteristic evaluation unit.

[0039] The real-time communication parameter acquisition unit is used to obtain real-time communication parameters between the adapter and the device through protocol analysis tools, and record key indicators such as transmission rate, latency and bit error rate; the instantaneous power consumption monitoring unit is used to synchronously collect instantaneous power consumption data between the adapter and the device using current sensors and voltage measurement devices, and analyze the power consumption characteristics under different operating conditions; the link attenuation characteristic evaluation unit is used to evaluate the attenuation characteristics of the link between the adapter and the device through signal strength detection and quality analysis algorithms, identify signal interference sources and quantify the link quality change trend.

[0040] It should be noted that the signal strength detection and quality analysis algorithm is as follows:

[0041]

[0042] in, For path; For loss; Distance The function; Distance from the transmitter Path loss at the location; For reference distance The function; For reference distance Path loss at the location; A dimensionless parameter that reflects the degree of environmental impact on path loss; This refers to logarithmic operations with base 10. This is the ratio of the current distance to the reference distance. This is the shadowing fading term of the log-normal distribution.

[0043] It is understood that, by integrating a real-time communication parameter acquisition unit, an instantaneous power consumption monitoring unit, and a link attenuation characteristic evaluation unit, this embodiment of the application can comprehensively acquire key real-time communication indicators between the adapter and the device, instantaneous power consumption data and power consumption characteristics under different operating states, link attenuation characteristics and signal interference source information, and link quality change trends. It captures the dynamic changes of the adapter and device during communication, power consumption, and link connection processes, providing multi-dimensional data support for subsequent adapter and device communication performance optimization, power consumption control, and link stability improvement. The real-time communication parameter acquisition unit can accurately acquire real-time communication parameters between the adapter and the device through protocol analysis tools, clearly recording transmission rate, latency, and... The bit error rate (BER) is a key indicator that ensures high accuracy in communication performance evaluation and provides a core basis for judging the operating status of the communication link. The instantaneous power consumption monitoring unit can use current sensors and voltage measurement devices to simultaneously collect instantaneous power consumption data of the adapter and device, deeply analyze the power consumption characteristics under different operating conditions, effectively adapt to the diverse operating scenarios of the device, and ensure accurate capture and analysis of power consumption changes. The link attenuation characteristic evaluation unit can accurately evaluate the attenuation characteristics of the link between the adapter and the device through signal strength detection and quality analysis algorithms, accurately identify signal interference sources and quantify the link quality change trend, ensure effective perception of the link status in complex signal environments, and guarantee the reliability of link quality evaluation.

[0044] For example, in a smart factory automotive manufacturing scenario, a company built a monitoring system for industrial robot controllers (adapters) and articulated motors (equipment): A SerialTek PCIe 6.0 protocol analyzer was used to collect real-time communication parameters between the two, stably capturing a 128Gbps transmission rate, ≤500ns latency, and <1e-12 bit error rate, while simultaneously recording protocol anomalies; a J&DIPCS-100A current sensor and a high-precision voltage measuring device were used to monitor instantaneous power consumption in three states: high-speed operation (peak 500W), precise positioning (average 300W), and standby (<50W), triggering an early warning if the power consumption deviated 15% from the model value; a SmartMesh network based on the IEEE 802.15.4eTSCH protocol was used to collect RSSI (-65~-95dBm) and SNR (15~30dB) in real time to assess link attenuation, locate the 2.45GHz interference source of the welding equipment through spectrum analysis, and then optimize the link using 10ms channel frequency hopping and dynamic power control (1~20mW).

[0045] In this embodiment of the application, the state feature analysis module 200 includes: Figure 3 As shown, there are a data fusion processing unit, a dynamic graph construction unit, and a nonlinear correlation quantization unit.

[0046] The data fusion processing unit extracts and fuses features from the original data using a self-supervised learning algorithm, eliminating noise interference and dimensional differences between data. The dynamic map construction unit constructs a communication quality-equipment requirement dynamic map based on the fused data, and correlates time series features with spatial distribution features at multiple scales in a spatiotemporal manner. The nonlinear correlation quantification unit analyzes the changes in elasticity coefficients in the dynamic map to quantify the nonlinear correlation strength and dynamic evolution law between frequency adaptation and response delay.

[0047] It should be noted that self-supervised learning algorithms:

[0048]

[0049] in, The name of the loss function; This is the scaling factor for the average loss; The upper and lower bounds for summation are determined by iterating through all elements in the batch. One sample; For the currently processed number One sample; It is the natural logarithm; Convert the value within the parentheses to exponential form; It is a similarity measurement function; For the first The query feature vector of each original sample; For the first Positive sample feature vectors of samples; For temperature parameters; For the first One candidate sample; To forcibly exclude cases where a sample is compared to itself; For the first The negative sample feature vector of each candidate sample.

[0050] By analyzing the changes in elasticity coefficients in the dynamic spectrum, the nonlinear correlation strength and dynamic evolution law between frequency adaptation and response delay are quantified. The multi-source fusion data in the dynamic spectrum are cleaned, decomposed into multiple scales, and feature extracted to obtain frequency adaptation variables and response delay variables. Then, drawing on the concept of elasticity in economics, a nonlinear elasticity coefficient is defined. The dynamic evolution law in the time dimension, spatial dimension, and frequency-delay dimension is analyzed.

[0051] It is understood that the embodiments of this application employ a self-supervised learning algorithm through a data fusion processing unit to perform feature extraction and fusion processing on the original data, effectively eliminating noise interference and dimensional differences between data to ensure data reliability. The dynamic graph construction unit, based on the fused data, constructs a communication quality-equipment requirement dynamic graph, performing multi-scale spatiotemporal correlation between time series features and spatial distribution features to provide spatiotemporal support for collaborative analysis of communication and equipment. The nonlinear correlation quantification unit analyzes the changes in elasticity coefficients in the dynamic graph, quantifies and displays the nonlinear correlation strength and dynamic evolution law between frequency adaptation and response delay, improves the accuracy of data feature extraction, the multi-scale spatiotemporal correlation, and the dynamic nature of correlation quantification, and provides a reliable analytical basis for communication quality optimization and equipment requirement adaptation research through structured dynamic graph output, reducing the probability of noise residue, dimensional imbalance, and correlation quantification deviation.

[0052] For example, in a certain intelligent traffic management system, the data fusion processing unit integrates data from cameras, radar, and floating car GPS. It automatically extracts core features such as vehicle speed, queue length, and turning intention through a self-supervised learning algorithm. Simultaneously, it eliminates dimensional biases caused by differences in sampling frequencies between different sensors (e.g., 10Hz for cameras and 20Hz for radar), and filters false alarm signals in rainy or foggy weather through a noise reduction network. The dynamic map construction unit, based on the fused real-time data, constructs a dynamic map of "traffic flow-signal timing-equipment status": in the time dimension, it captures the periodic fluctuations of westbound traffic flow during the morning rush hour; in the spatial dimension, it correlates the lane-level traffic flow distribution of five adjacent intersections within a 3-kilometer radius, forming a multi-scale spatiotemporal correlation network. The nonlinear correlation quantification unit analyzes the elasticity coefficient changes of "green light extension frequency" and "average vehicle delay" in the map. It finds that when the green light duration exceeds 55 seconds, for every 5-second increase in green light time, vehicle delay actually increases by 12%, exhibiting a significant nonlinear "threshold effect." It also dynamically visualizes the evolution trend of response delay under different phase combinations, providing a decision-making basis for optimizing signal timing schemes.

[0053] In this embodiment of the application, the frequency decision module 300 includes: Figure 4 As shown, there are federated transfer reinforcement learning units and temporal causal chain reasoning units.

[0054] The federated transfer reinforcement learning unit trains and optimizes distributed policies using the federated transfer reinforcement learning algorithm to generate distributed switching policies; the temporal causal chain reasoning unit performs temporal causal analysis using the knowledge graph temporal causal chain reasoning algorithm, and outputs target frequency parameters in conjunction with the distributed switching policy.

[0055] It should be noted that the knowledge graph temporal causal chain reasoning algorithm is as follows:

[0056]

[0057] in, To quantify the overall causal reliability of a time-series causal chain; It is a sequential causal chain; For regularization terms; The length of the temporal causal chain; This is the first triplet; As the main body; For relationship; As an object; For timestamps; It is a time decay function; To represent the first The triplet and the first The time interval of each triplet.

[0058] It is understood that the embodiments of this application employ a federated transfer reinforcement learning unit to perform distributed training and optimization of the policy using a federated transfer reinforcement learning algorithm, effectively carrying out distributed iterative upgrades and precise refinement of the policy to generate a reliable distributed switching policy; the temporal causal chain reasoning unit relies on the knowledge graph temporal causal chain reasoning algorithm to perform temporal causal analysis, and combines the distributed switching policy to synergistically integrate temporal causal logic and policy information to provide causal basis for determining frequency parameters, outputting accurate target frequency parameters; through distributed policy training optimization and temporal causal collaborative analysis, the system improves the distributed efficiency of policy training, the temporal accuracy of causal analysis, and the output accuracy of target frequency parameters, and provides reliable parameter support for relevant scenarios through the structured output of target frequency parameters, reducing policy optimization deviation and the probability of frequency parameter misjudgment.

[0059] For example, in the final assembly workshop of a smart car manufacturing plant, multiple adapters are connected to PLC controllers, LiDAR sensors, and robotic arm actuators, respectively. The communication frequency needs to be dynamically adjusted to balance real-time performance and network load. At this point, the frequency decision module begins operation: the federated transfer reinforcement learning unit first collects historical interaction data from equipment on each production line (welding, painting, and final assembly) over the past three months—including communication latency, packet loss rate, and equipment energy consumption at different frequencies (100Hz-300Hz). Considering the significant differences in production cycles across production lines and the sensitivity of the data, the federated learning framework, without transmitting the original data, utilizes transfer learning to transfer the experience of the welding production line, where "200Hz performs best during peak hours with high load," to similar scenarios on the current final assembly production line (such as periods of concentrated material transportation). Combined with the equipment models in the final assembly workshop (such as communication protocol upgrades for new LiDAR models) and the network topology (addition of 5G edge gateways), a distributed switching strategy is trained and generated, initially planning that "PLCs and sensors use 180Hz, and robotic arms and controllers use 250Hz." At the same time, the temporal causal chain inference unit synchronously connects to the workshop IoT platform to obtain the real-time status of the equipment (such as the lidar temperature rising from 45℃ to 58℃, the robotic arm motor load jumping from 65% to 82%, and the current bandwidth utilization of the 5G gateway reaching 75%). Through a pre-built knowledge graph (nodes cover "equipment temperature", "load intensity", "network bandwidth", and "communication protocol version", and edges are defined as causal relationships such as "temperature increases by 5℃ → communication error rate +2%" and "load exceeds 80% → data packet size increases by 15%), the temporal causal chain analysis reveals that "the continuous increase in robotic arm load + 5G gateway bandwidth approaching saturation" will lead to "communication latency possibly exceeding the 50ms threshold within the next 8 minutes". Although the 250Hz frequency of the robotic arm in the original federated migration strategy can handle the normal load, there is a risk of exceeding the latency limit by 20% under the current causal chain. Ultimately, the time-series causal chain inference unit, combined with a distributed strategy, dynamically adjusts the target frequency parameters of the robotic arm and controller to 280Hz (increased by 30Hz to compensate for bandwidth pressure), while maintaining 180Hz for the PLC and sensors (because their causal chain displays the current temperature and load stability), ensuring the real-time performance and reliability of communication between equipment in the final assembly workshop.

[0060] In this embodiment of the application, the formula for the federated transfer reinforcement learning algorithm is as follows:

[0061]

[0062] in, The value when the decision-making strategy reaches its optimal state; The environmental conditions in which the decision is made; This refers to the specific operation selected in the current state. The expected state transition probability is based on the environment. For immediate rewards; To represent the discount factor, the value range is 0 ≤ ... ≦1; To select the action that will bring the greatest value from all possible next actions; The next state; Next action.

[0063] It is understood that the embodiments of this application protect the data privacy of each distributed node and avoid the risks of centralization by leveraging the characteristics of federated learning where the data remains stationary while the model moves. They also rely on transfer learning to reuse the knowledge of mature nodes to solve the problem of sparse or heterogeneous data on some nodes, thereby improving training efficiency. Furthermore, they combine reinforcement learning mechanisms for dynamic policy optimization and ensure the policy's generalization ability across multiple node differences through federated aggregation and transfer adaptation. Ultimately, they efficiently complete the collaborative training and optimization of distributed policies, generating a distributed switching policy that is "globally adaptable, locally usable, and dynamically adjustable." This policy not only provides high-quality dynamic input for subsequent temporal causal chain inference units but also, through its combination with temporal causal analysis, lays a crucial foundation for accurately outputting target frequency parameters.

[0064] For example, in a multi-adapter communication scenario for smart home devices, a household deploys three types of devices: a smart speaker (requiring medium-to-high bandwidth real-time interaction), door and window sensors (low-power long-term connection), and a security camera (intermittent high-definition video transmission). These devices connect to the home gateway via Wi-Fi, ZigBee, and Bluetooth adapters, respectively. Traditional single-device training is prone to poor policy generalization due to insufficient samples, while centralized training poses a risk of device data privacy leakage. In this case, the federated transfer reinforcement learning unit optimizes communication frequency switching through the following process: First, using the gateway as the federated server and each device adapter as a client, it only encrypts and uploads the gradient parameters of the local reinforcement learning model (such as the value function update of Q-learning) without uploading the original communication data (such as signal strength and latency logs). Second, the server aggregates the parameters of each client to form a global basic policy (such as "when the camera is activated, the Wi-Fi adapter should prioritize switching to the 2.4GHz band to avoid Bluetooth interference"), and uses transfer learning to apply the "frequency avoidance rules when multiple devices compete for channels" accumulated in previous office scenarios. "The system is then adapted to the current home environment to correct any deviations in the initial strategy. Finally, each device trains locally based on the global strategy. For example, when a smart speaker makes a video call at night, it combines the current Wi-Fi load (packet loss rate monitored locally) and its own battery level (power consumption needs to be reduced when it is below 30%). It then attempts to switch to the 5GHz band or temporarily reduce the sampling frequency using an ε-greedy strategy. If the communication latency does not exceed the limit and the power saving is significant after verification, the reward value for this action (e.g., +5 points for a 10ms reduction in latency) is fed back to the local model. This ultimately forms a distributed communication frequency switching strategy that is adapted to the characteristics of home devices, ensures privacy and security, and can be dynamically optimized."

[0065] In this embodiment of the application, the dynamic control feedback module 400 includes, as follows: Figure 5 As shown, the algorithm execution unit and the report generation unit are shown.

[0066] The algorithm execution unit uses a three-objective dynamic equilibrium algorithm to dynamically weigh and make decisions on key indicators such as latency, bandwidth, and reliability during the handover process. The report generation unit collects and stores real-time status data by monitoring communication jitter amplitude and data fidelity indicators in real time, and combines the equipment aging compensation coefficient to perform correlation analysis on the monitoring data and handover results, and automatically generates an optimization report containing frequency adjustment suggestions.

[0067] It should be noted that, by combining the equipment aging compensation coefficient, the correlation analysis between monitoring data and switching results is to quantify the performance degradation caused by long-term operation of the equipment. When analyzing real-time monitoring data and historical switching results, this coefficient is introduced into the analysis model as a key correction variable to identify whether the performance degradation is due to changes in the external communication environment or the aging of the equipment itself. Based on this, frequency adjustment suggestions that are more in line with the actual health status of the equipment are proposed, and intelligent optimization based on long-term performance is carried out.

[0068] It is understood that the embodiments of this application employ a three-objective dynamic equilibrium algorithm through the algorithm execution unit to dynamically weigh and decide on key indicators such as latency, bandwidth, and reliability during the handover process, effectively achieving dynamic optimization of key indicators to ensure the scientific nature of handover decisions. The report generation unit relies on real-time monitoring of communication jitter amplitude and data fidelity indicators to collect and store real-time status data, and combines the equipment aging compensation coefficient to conduct correlation analysis between monitoring data and handover results, automatically generating an optimization report containing frequency adjustment suggestions, providing data support and directional guidance for parameter iteration during the handover process. Through the synergy of dynamic weighing of key indicators and correlation analysis of monitoring data, the dynamic adaptability of handover decisions, the accuracy of monitoring analysis, and the pertinence of optimization suggestions are improved. The structured optimization report output provides reliable data basis for continuous improvement of the handover process, reducing the probability of indicator imbalance and decision deviation.

[0069] For example, such as Figure 6 As shown, in a smart home scenario, when a user initiates a 4K high-definition video call, the system's dynamic control feedback module begins operation: the algorithm execution unit, through a three-objective dynamic equalization algorithm, monitors in real time that the latency of the current Wi-Fi channel suddenly increases from 30ms to 80ms, with bandwidth utilization reaching 92%. Simultaneously, the reliability of the BLE sensor network drops to 85% due to co-channel interference. At this point, the algorithm automatically triggers a frequency switching decision, switching the video stream to the 5GHz band to reduce latency to 25ms, and allocating independent sub-channels to the BLE sensors, restoring reliability to 98%. Simultaneously, by dynamically adjusting the bandwidth quotas of each device, it ensures that the overall network bandwidth utilization remains at the optimized threshold of 75%. The report generation unit, using an oscilloscope, monitors the communication jitter amplitude in real time and discovers that the periodic jitter of a certain ZigBee temperature and humidity sensor increases from 5ps to 18ps. Combined with the aging compensation coefficient (0.85) of the device, which has been used for 3 years, analysis shows that the crystal oscillator frequency offset causes the data fidelity to decrease to 91%. The system automatically generates an optimization report, recommending switching the sensor's communication frequency from 2.4GHz to 868MHz and performing a firmware upgrade during off-peak hours at night to compensate for aging errors. In this process, the algorithm execution unit and the report generation unit form a closed loop, improving the overall system's communication stability by 22%, reducing device power consumption by 15%, and advancing the warning time for potential fault risks by 72 hours.

[0070] This application proposes an intelligent switching system based on the communication frequency between an adapter and a device. A communication status acquisition module acquires real-time communication parameters, instantaneous power consumption, and link attenuation characteristics of the adapter and device, effectively capturing dynamic changes in communication status and basic link quality data, providing comprehensive perceptual support for subsequent analysis. In conjunction with a state feature analysis module, a self-supervised spatiotemporal frequency elasticity coefficient model fuses multi-source data to construct a dynamic graph of communication quality and device requirements, quantifying the nonlinear correlation between frequency adaptation and response delay, and improving the depth and breadth of communication status feature analysis. A federated transfer reinforcement learning algorithm in the frequency decision module generates a distributed switching strategy, combined with knowledge graph temporal causal chain reasoning, outputting target frequency parameters to enhance the adaptability and accuracy of frequency decisions in different scenarios. A dynamic control feedback module executes the switching strategy through a three-objective dynamic equilibrium algorithm, monitoring communication jitter and data fidelity in real time, automatically generating a frequency optimization report with device aging compensation coefficients, continuously optimizing communication performance and dynamically compensating for device aging, improving the long-term stability and reliability of the system. This solves the problems of static feature analysis, inefficient decision-making mechanisms, and lack of control feedback in existing technologies.

[0071] The following will illustrate an intelligent switching system based on the communication frequency between the adapter and the device through a specific embodiment, such as... Figure 7 As shown, it includes:

[0072] In a large-scale intelligent manufacturing park for new energy vehicle power batteries, an intelligent switching system based on the communication frequency between the adapter and the equipment is being deeply integrated into the entire production chain from electrode manufacturing to cell packaging. Through the closed-loop collaboration of the communication status acquisition module, status feature analysis module, frequency decision module and dynamic control feedback module, it has solved the communication frequency adaptation problem of various types of equipment in the park (including core production equipment such as electrode rolling mills, winding machines, liquid injection machines, and formation cabinets, logistics equipment such as AGV handling robots and stacker cranes in automated warehouses, and environmental monitoring equipment such as temperature and humidity sensors and pressure sensors) in a complex industrial environment. Its operation process shows a high degree of accuracy and dynamic adaptability.

[0073] The communication status acquisition module serves as the system's data entry point. Through miniature radio frequency detectors and edge computing nodes deployed at the interfaces of various devices, it acquires multi-dimensional communication parameters between different devices and their corresponding adapters in real time at a high sampling frequency of 10ms / time. For example, when the electrode mill is performing 18μm thick electrode rolling, the communication frequency band between the adapter and the device dynamically switches between the 2.4GHz and 5GHz dual-band. The module synchronously acquires data showing that the actual transmission rate fluctuates between 280Mbps and 320Mbps, and the instantaneous power consumption increases from 1200W under no-load to 1800W under full-load as the mill load changes. Simultaneously, a signal attenuation analyzer captures the 12dB-18dB link attenuation caused by the metal frame within the workshop to the 2.4GHz signal, as well as the attenuation of multiple coils... The module attenuates 5dB-8dB of interference generated during concurrent operation of the AGV handling robot. For the AGV handling robot, the module not only tracks its real-time position coordinates (e.g., X=35.2m, Y=18.7m, Z=0.5m) and moving speed (1.2m / s), but also records the difference in link attenuation between densely packed shelving areas (severe signal obstruction) and open aisles (smooth signal). It even includes the slight impact of ambient temperature and humidity (e.g., average daily temperature of 32℃ and humidity of 68% in summer workshops) on the communication signal (attenuation change of 0.3dB / ℃) in the collection scope. After all data is initially filtered by the edge node (to remove outliers caused by instantaneous sensor failures), it is synchronized to the local cache server and cloud database in encrypted format, providing complete and reliable raw data support for subsequent analysis.

[0074] The state feature analysis module undertakes the core functions of data fusion and correlation mining. First, it standardizes the collected multi-source heterogeneous data (communication parameters as time-series data, power consumption as numerical data, and link attenuation as feature data) using data cleaning algorithms, mapping the parameter indicators of different devices to a normalized range of 0-1. Then, it introduces a self-supervised spatiotemporal frequency elasticity coefficient model. This model uses the operating data of 120 core devices in the park over the past six months (covering different batches of battery cell production, different equipment load peaks, and different seasonal environmental changes) as training samples to automatically learn the spatiotemporal correlation between communication frequency and equipment operating status, thereby constructing a dynamic graph of "communication quality-equipment demand"—for example, in the winding process. In the process, when the equipment needs to achieve an electrode alignment accuracy of ±0.5mm (requirement level marked as level 9), the graph clearly shows that the current communication latency of 22ms in the 2.4GHz band can no longer meet the requirements. The model calculates that the spatiotemporal frequency elasticity coefficient at this time is 0.65 (the closer the coefficient is to 1, the better the frequency adaptability), quantifying the nonlinear correlation that "a 10% increase in frequency can reduce latency by 15%". For the liquid injection machine that has been used for 2 years, the module also found that due to the aging of the radio frequency module, the link attenuation in the 2.4GHz band is 7dB higher than that of new equipment. Therefore, the equipment aging factor is included as a key feature in the dynamic graph, so that the graph can not only reflect the real-time communication status, but also predict the frequency adaptability trend after the long-term operation of the equipment.

[0075] Based on the aforementioned dynamic graph, the frequency decision module generates a distributed switching strategy using a federated transfer reinforcement learning algorithm. The federated nature of this algorithm is reflected in the fact that five production workshops, two quality inspection centers, and one logistics dispatch center within the park are used as independent training nodes. Each node only uploads model parameters, not raw production data (effectively protecting the core privacy of battery production processes). Global model optimization is achieved through federated aggregation. Transfer learning transfers the "high-precision communication strategy" validated in the quality inspection center to the liquid injection process, which has stringent data transmission requirements, significantly shortening the strategy iteration cycle. In specific decision-making, the system also introduces knowledge graph temporal causal chain reasoning, mining causal relationships in historical data (e.g., "When the formation cabinet is in a 40℃ environment, and the communication delay in the 5GHz band is >20ms, the cell formation pass rate decreases by 1.2%"; "When the AGV uses the 2.4GHz band in a densely populated shelf area, the communication delay..."). The failure rate is 3 times that of 5GHz. A time-series causal chain of "ambient temperature → link attenuation → communication delay → production quality" is constructed. When the temperature in the area where a certain forming cabinet is located rises to 38℃, the inference module automatically predicts that if the current 5GHz band is maintained, the delay will rise to 21ms, thereby triggering a frequency adjustment decision and finally outputting differentiated target frequency parameters: the target frequency of the electrode mill is set to 5.2GHz when fully loaded (to ensure that the transmission rate is stable above 300Mbps) and switches to 2.4GHz when idle (to reduce standby power consumption). The AGV uses the 5.5GHz enhanced band in the densely packed shelf area (to compensate for 12dB of link attenuation) and switches to 2.4GHz in the open channel (to balance power consumption and efficiency). The aging forming cabinet that has been used for more than 3 years has an additional frequency compensation coefficient of 0.15 on the basis of 5GHz (to offset the signal loss caused by hardware aging).

[0076] The dynamic control feedback module, acting as the system's execution and optimization terminal, performs frequency switching operations using a three-objective dynamic balancing algorithm (Objective 1: Communication stability, requiring communication jitter <10ms and data fidelity >99.97%; Objective 2: Power consumption optimization, reducing power consumption by 10%-15% compared to the static frequency scheme; Objective 3: Delaying equipment aging, reducing hardware RF module losses by 15% through frequency compensation). During the switching process, the module first establishes a communication link between the adapter and the device in advance using a pre-synchronization mechanism (switching time is controlled within 500ms to avoid data transmission interruption). Then, it monitors key indicators in real time after the switch—for example, after the winding machine switches from 2.4GHz to 5.3GHz, the communication jitter is monitored to decrease from 18ms to 7ms within one minute, and the transmission error rate of the electrode alignment parameters decreases from 0.08%. The instantaneous power consumption decreased from 1750W to 1620W, reaching 0.02%, and all three indicators met the three-objective balance requirements. For AGV fleets, the module also calculates differentiated aging compensation coefficients for equipment with different usage durations (e.g., the compensation coefficient for an AGV used for 1 year is 0.05, and for an AGV used for 3 years it is 0.12). Based on daily communication jitter data (e.g., an AGV jitters for more than 12ms in a specific area for 3 consecutive days), it automatically generates a frequency optimization report with equipment aging compensation coefficients. The report not only includes the current frequency parameters and optimization suggestions for each device (e.g., for the electrode mill A, which has accumulated 18,000 hours of operation, it is recommended to check the RF module interface monthly and adjust the frequency compensation coefficient to 0.18), but also includes equipment maintenance cycle warnings (e.g., the RF module of the stacker crane B is expected to reach the aging threshold in 6 months, and spare parts need to be purchased in advance). After six months of operation in the park, the intelligent switching system has significantly improved efficiency: the number of communication interruptions for core production equipment has decreased from 12 times per month to 3 times; the average power consumption of the equipment has decreased from 1500W to 1320W (saving approximately 120,000 yuan in electricity costs annually); the data transmission latency between power battery processes has decreased from an average of 22ms to 9ms; the first-pass yield of battery cells has increased from 96.8% to 99.1% (reducing losses from defective products by approximately 300,000 yuan per month); and the average lifespan of the equipment's radio frequency modules has increased from 3 years to 3.8 years, significantly reducing maintenance costs. In extreme scenarios, such as summer... The high temperatures during the summer season caused an 8dB increase in overall signal attenuation in the workshop. The system was able to complete frequency compensation adjustments for all core equipment within 10 seconds, ensuring the communication stability of critical equipment such as the electrode rolling mill and winding machine. During holidays when the park was operating at low load, the system would automatically switch non-critical equipment such as environmental monitoring sensors and idle AGVs to the 2.4GHz low-frequency band and reduce the sampling frequency to 50ms / time, further optimizing energy consumption. This fully demonstrated the adaptability and practicality of the system in complex industrial scenarios, providing a replicable solution for communication optimization in the field of intelligent manufacturing of new energy vehicles.

[0077] In summary, the embodiments of this application, through a targeted communication status acquisition, status feature analysis, and dynamic adjustment feedback mechanism, can stably address the communication frequency adaptation challenges of various types of equipment in complex industrial environments. Combined with high-frequency acquisition and edge computing, it achieves real-time data processing and transmission. Relying on a self-supervised spatiotemporal frequency elasticity coefficient model, federated transfer reinforcement learning algorithms, and knowledge graph temporal causal chain reasoning, it achieves electrode alignment accuracy of ±0.5mm and a fast switching response of ≤500ms. Combined with multi-dimensional feature fusion and differentiated compensation strategies, it can flexibly adapt to different equipment types, load states, and environmental conditions, precisely controlling communication frequency and transmission rate. Simultaneously, it constructs a data chain, effectively improving the stability and efficiency of the communication system, reducing equipment energy consumption and maintenance costs, and providing strong support for subsequent production optimization and decision support.

[0078] Next, referring to the accompanying drawings, a method for intelligent switching based on the communication frequency between the adapter and the device, according to an embodiment of this application, is described.

[0079] like Figure 8 As shown, this intelligent switching method based on the communication frequency between the adapter and the device includes the following steps:

[0080] In step S101, the real-time communication parameters between the adapter and the device are obtained.

[0081] Understandably, this application obtains real-time communication parameters between the adapter and the device as the basic input for subsequent intelligent frequency switching decisions. Through the collaborative acquisition of multi-dimensional communication indicators and real-time status data, it can comprehensively capture the communication status characteristics under different operating conditions. This means that it can retain key communication strength and latency data under complex scenarios such as high device load and signal interference, and accurately obtain dynamic connection status information when the adapter power fluctuates or the device moves rapidly. This provides high-quality real-time data support for subsequent communication quality assessment, frequency adaptation analysis, and switching strategy generation, thereby improving the adaptability to complex operating conditions and the accuracy of frequency switching.

[0082] In step S102, the real-time communication parameters of the adapter and the device are integrated, and a communication quality-device demand dynamic spectrum is constructed through a self-supervised spatiotemporal frequency elasticity coefficient model. By analyzing the changes in elasticity coefficients in the dynamic spectrum, the nonlinear correlation strength and dynamic evolution law between frequency adaptation and response delay are quantified and displayed.

[0083] Among them, the self-supervised spatiotemporal frequency elasticity coefficient model is a model that uses self-supervised learning to process data with temporal dynamics and spatial distribution characteristics, combines frequency domain analysis to capture multi-dimensional correlation patterns, and outputs elasticity coefficients that quantify the sensitivity of the influence between variables.

[0084] It is understood that the embodiments of this application, by employing a self-supervised spatiotemporal frequency elasticity coefficient model, can address situations where the real-time communication parameters between the adapter and the device are dynamically changing, and the correlation between frequency adaptation and response delay is complex and nonlinear. By fusing real-time communication parameters to construct a dynamic spectrum of communication quality and device requirements, and analyzing the changes in elasticity coefficients in the dynamic spectrum, the nonlinear correlation strength and dynamic evolution law between frequency adaptation and response delay are quantified and displayed. This effectively avoids the problem that traditional models are unable to capture the dynamic nature of communication parameters in the spatiotemporal dimension and cannot accurately characterize the nonlinear correlation and dynamic evolution of the two. This provides a clearer basis for correlation characteristics and dynamic law reference for subsequent optimization of frequency adaptation strategies and control of response delay, improving the accuracy of matching communication quality with device requirements and the effectiveness of dynamic adjustment.

[0085] It should be noted that the Communication Quality-Equipment Demand Dynamic Map is a dynamic correlation analysis tool that focuses on communication network scenarios and establishes a real-time mapping relationship between communication quality performance and the equipment requirements that support that quality through visualization and data-driven methods. This allows for closed-loop management of quality-based demand and demand-based quality assurance.

[0086] The elasticity coefficient is a quantitative indicator that measures the sensitivity of one variable to changes in another variable, and is usually expressed as the ratio of their relative rates of change.

[0087] Quantized display frequency adaptation is a technology that automatically adjusts the screen refresh rate using digital algorithms to match the frame rate of the currently displayed content, thereby achieving smooth viewing and saving power.

[0088] Nonlinear correlation strength is a statistical indicator that quantifies the closeness of the nonlinear complex relationship between two or more variables to describe their degree of dependence on each other for co-change.

[0089] The law of dynamic evolution is a scientific concept that describes the inherent essential connections and inevitable trends that follow when various elements interact and influence each other during the continuous change of things over time.

[0090] Self-supervised spatiotemporal frequency elasticity coefficient model:

[0091]

[0092] in, The elastic modulus is the frequency coefficient. Rate the quality of communication; For real-time communication frequency; The change is a percentage. To correct the impact of environmental attenuation on frequency elasticity; For all variables to be functions of time; This is a nonlinear correlation adjustment coefficient; When time changes When, function The resulting minute changes; To indicate time The previous extremely small change.

[0093] For example, in a 5G private network scenario in a smart factory, 5G adapters deployed on production equipment collect and upload communication parameters with the base station in real time (including radio frequency f, end-to-end delay τ, signal-to-noise ratio SNR, and QoS requirements of equipment service types). The self-supervised spatiotemporal frequency elasticity coefficient model is constructed based on a temporal convolutional network (TCN) and a graph neural network (GNN): First, local temporal features of frequency-delay are extracted through a time-dimensional sliding window (10-second granularity) (such as the jump pattern of f in the 2.4GHz / 5GHz band and the hysteresis response of τ). At the same time, the frequency distribution and delay mean of different production line areas are aggregated using a spatial-dimensional graph structure (a topology with base stations as nodes and equipment locations as edges) to construct a spatiotemporal joint feature matrix. The model is trained by a self-supervised task (predicting the change τ Δτ when f is adjusted by Δf in the next 5 seconds) to learn the dynamic elasticity coefficient. Where t is the timestamp and p is the spatial location (corresponding to different production lines). In actual operation, the dynamic spectrum output by the model shows that during the peak production period at 8:00 AM (t=8:00-9:00), the α value of the high-frequency band (5GHz) is significantly higher than that of the low-frequency band (2.4GHz). Furthermore, as equipment QoS requirements increase (such as real-time quality inspection video transmission), α jumps from 1.2ms / MHz to 3.5ms / MHz (non-linear enhancement), reflecting that even small adjustments to the high-frequency band can cause drastic delay fluctuations. During the midday off-peak period (t=12:00-13:00), the α value of each frequency band drops below 0.5ms / MHz. As equipment switches to low-rate data reporting mode, α increases with f in a U-shaped curve (non-linear inflection point), first decreasing and then increasing, reflecting the dynamic modulation of the frequency-delay coupling relationship by changes in equipment demand. This spectrum intuitively reveals the highly sensitive characteristics of frequency adaptation to delay in "high-frequency band - high demand" scenarios, providing a quantitative basis for the intelligent scheduling of production line communication parameters.

[0094] In step S103, based on the nonlinear correlation strength and dynamic evolution law, distributed policy training and optimization are carried out through federated transfer reinforcement learning to generate distributed switching policies. Combined with knowledge graph temporal causal chain reasoning, target frequency parameters are output.

[0095] Among them, federated transfer reinforcement learning is a technology that combines the privacy protection features of federated learning, the cross-domain knowledge transfer capabilities of transfer learning, and the interactive optimization mechanism of reinforcement learning. Under the premise that the data of multiple participants is not shared, it uses source domain knowledge to improve the efficiency of policy learning in data-scarce or complex target domains.

[0096] It is understood that the embodiments of this application utilize federated transfer reinforcement learning technology, based on the characteristics of nonlinear association strength and dynamic evolution, to efficiently advance distributed policy training and optimization and generate distributed switching policies. Even in distributed scenarios, it can stably achieve collaborative iteration and accurate adaptation of policies. Its privacy-preserving characteristics based on federated learning and the knowledge reuse capabilities of transfer learning ensure the security and efficiency of distributed policy training, providing reliable policy support for outputting target frequency parameters through knowledge graph temporal causal chain inference. It also supports knowledge transfer and policy collaboration between different distributed nodes. This overcomes the limitations of traditional centralized reinforcement learning in terms of data privacy leakage risks and distributed scenario adaptability, improves the optimization efficiency and generalization ability of distributed switching policies, ensures the accuracy and timeliness of target frequency parameter output, and provides timely and reliable policy and parameter basis for the dynamic control and precise parameter configuration of related systems.

[0097] It should be noted that distributed policy training refers to a technical method in which a policy model is trained through collaborative parallel computing in a distributed system composed of multiple computing devices or nodes, in order to improve training efficiency and support the training of large-scale data or complex models.

[0098] A knowledge graph temporal causal chain is a continuous chain of connections in a knowledge graph that depicts the dynamic transmission of causal relationships between entities and events in chronological order.

[0099] For example, an industrial park contains three independent operating entities: a photovoltaic microgrid (A), a wind power microgrid (B), and a gas turbine microgrid (C). Due to data privacy policy restrictions, each microgrid cannot directly share operational data, but their frequency stability requirements need to be coordinated through distributed switching strategies (such as real-time switching of master-slave control / droop control). Based on a federated transfer reinforcement learning framework, a lightweight policy network is first deployed locally on each microgrid. Initial policies are trained offline using their respective historical operational data (such as photovoltaic output fluctuations, control actions during load surges, and frequency deviations). The focus is on capturing the nonlinear correlation strength between "new energy output - load demand - energy storage SOC" and the dynamic evolution patterns at the intraday / weekly level (e.g., during midday when photovoltaic power generation is high, grid A tends to be the master, while wind power grid B dominates at night). Then, the policy gradients of each microgrid are aggregated through a federated server, and global policy parameters are adjusted through transfer learning to address the cold start problem of "newly connected microgrid D (lithium battery energy storage) lacking historical data." Simultaneously, the action space of the distributed switching strategy is optimized by combining a temporal causal chain constructed from a knowledge graph (e.g., setting grid A to automatically switch to slave control when SOC is below 30%, triggering grid B's master control mode). Finally, through multiple rounds of federated interaction and causal reasoning, a distributed switching strategy is generated that balances the operating costs, frequency stability (target frequency deviation ≤ 0.1Hz), and switching losses of each microgrid, achieving cross-entity coordinated frequency regulation.

[0100] In step S104, based on the distributed handover strategy, the handover is performed through a three-objective dynamic balancing algorithm, communication jitter and data fidelity are monitored in real time, real-time status data is collected and stored, and a frequency optimization report with equipment aging compensation coefficient is automatically generated.

[0101] Among them, the three-objective dynamic equilibrium algorithm is a type of multi-objective optimization algorithm that performs collaborative optimization on three potentially conflicting objective functions under dynamically changing environments or constraints in order to find a trade-off optimal solution that balances the needs of the three.

[0102] Three-objective dynamic equilibrium algorithm:

[0103]

[0104] in, For equipment At any moment Overall utility value Weighting coefficients for communication jitter targets; Normalized value of communication jitter; equipment At any moment ; Weighting coefficients for data fidelity targets; This is a normalized value for data fidelity. Weighting coefficients for the switching cost objectives; The normalized value for the switching cost; This is the equipment aging compensation coefficient.

[0105] It is understood that the embodiments of this application, by employing a three-objective dynamic equalization algorithm based on a distributed handover strategy, efficiently execute system handover operations, monitor communication jitter and data fidelity in real time, and can still stably achieve high-fidelity data transmission and low-jitter communication in dynamic network environments; its multi-objective optimization capability can ensure the balance and adaptability of handover decisions, provide consistency and reliability for the collection and storage of real-time status data, and support the automatic generation of frequency optimization reports for equipment aging compensation coefficients; it makes up for the limitations of traditional single-index handover strategies, improves the overall performance and robustness of the system, ensures that frequency optimization can accurately adapt to changes in equipment aging, and provides a timely and reliable data foundation for communication quality assessment and network optimization.

[0106] It should be noted that the equipment aging compensation coefficient is a specific coefficient used to correct relevant parameters and results of equipment in order to offset the performance degradation or accuracy reduction caused by long-term use and aging, so as to ensure its use effect.

[0107] For example, in an industrial IoT sensor network scenario in a smart factory, the three-objective dynamic balancing algorithm performs real-time optimization with the core objectives of "minimizing communication latency, maximizing data fidelity, and balancing equipment load": The algorithm module deployed on the edge gateway collects the overall network status every 50ms - including the communication jitter value of each sensor node (e.g., the standard deviation of the transmission delay of node A in the last 10 transmissions is 12ms, and that of node B is 8ms), data fidelity index (e.g., the data packet error rate of node C rises to 0.3% due to circuit aging, while that of node D remains at 0.1%), and equipment load (the CPU utilization of node E reaches 85%, while that of node F is only 40%). At the same time, it calls the equipment aging model (based on the exponential decay function of running time, e.g., node C has been running for 3 years, with an aging compensation coefficient k=1.2, and node D has been running for 1 year, with k=1.0). The algorithm employs an improved multi-objective particle swarm optimization (MOPSO) approach to perform Pareto front analysis on the three objectives, dynamically calculating the comprehensive cost function for each node (cost = 0.4 × latency × (1+k) + 0.3 × bit error rate × (1+k) + 0.3 × load). Ultimately, it selects to switch 30% of the high-frequency sampled data stream from node C to node F (low load and minimal aging), and reduces the transmission frequency of node C from 10Hz to 8Hz (to compensate for signal attenuation caused by aging). Simultaneously, it increases the transmission priority of node B to offset the jitter fluctuations of node A. After implementing this strategy, the average latency across the entire network decreased by 18%, the data packet loss rate decreased to below 0.15%, and the standard deviation of device load narrowed from 22% to 9%. The final frequency optimization report clearly indicates the current aging compensation coefficient for each node (e.g., node C with k=1.2 requires special attention and is recommended to be replaced within 3 months), achieving synergistic optimization of communication performance, data quality, and device lifespan.

[0108] This application proposes an intelligent switching method based on the communication frequency between an adapter and a device. The method acquires real-time communication parameters, instantaneous power consumption, and link attenuation characteristics of the adapter and device through a communication status acquisition module, effectively capturing dynamic changes in communication status and basic link quality data, providing comprehensive perceptual support for subsequent analysis. Combined with a self-supervised spatiotemporal frequency elasticity coefficient model from a status feature analysis module, it fuses multi-source data to construct a dynamic graph of communication quality and device requirements, quantifying the nonlinear correlation between frequency adaptation and response delay, and improving the depth and breadth of communication status feature analysis. A distributed switching strategy is generated using a federated transfer reinforcement learning algorithm in the frequency decision module, combined with knowledge graph temporal causal chain reasoning, to output target frequency parameters, enhancing the adaptability and accuracy of frequency decisions in different scenarios. A dynamic control feedback module executes the switching strategy through a three-objective dynamic equilibrium algorithm, monitors communication jitter and data fidelity in real time, automatically generates a frequency optimization report with device aging compensation coefficients, continuously optimizes communication performance, and dynamically compensates for device aging, improving the long-term stability and reliability of the system. This solves the problems of static feature analysis, inefficient decision-making mechanisms, and lack of control feedback in existing technologies.

[0109] The following will illustrate a method for intelligent switching based on the communication frequency between the adapter and the device through a specific embodiment, such as... Figure 9 As shown, it includes:

[0110] In a 5G+Industrial Internet scenario at a smart factory for new energy vehicle power batteries, the need for intelligent switching of communication frequencies between the adapter (an edge computing gateway integrating a 5G industrial module) and the laser welding equipment and vision inspection PLC controller on the battery assembly line is particularly urgent. The laser welding equipment needs to transmit welding trajectory instructions with millisecond-level low latency (peak rate 1.2Gbps), while the vision inspection PLC needs to upload 2-megapixel weld seam images at high frequency (100Hz) (bandwidth requirement 800Mbps). Both of these require dynamic adaptation of the adapter's communication frequency (supporting 2.4GHz / 5GHz / 6GHz three-band). At the same time, the dense metal equipment in the workshop makes the 5GHz band susceptible to interference (signal attenuation reaches -12dB in typical scenarios), and the high temperature and humidity environment accelerates the aging of the adapter's radio frequency devices (after 3 months of continuous operation, the transmit power attenuates by about 5%). Traditional fixed frequency or simple polling switching strategies can no longer meet the dual requirements of production cycle time (90 seconds / battery pack) and yield rate (≥99.2%).

[0111] Based on this, the factory deployed an intelligent switching system integrating the aforementioned patented technologies: First, the edge computing gateway collects communication parameters between the adapter and the device in real time through embedded sensors—including the 5G module's RSSI (Received Signal Strength, accuracy ±0.5dBm), EVM (Error Vector Amplitude, reflecting modulation quality, threshold ≤10%), uplink latency (measured under PTP clock synchronization, unit μs), packet loss rate (statistical analysis of 1000 consecutive packets), and current fluctuations of the laser welding equipment (reflecting load status, range 0-10A, accuracy ±0.1A), and temperature sensor data from the vision PLC (-40℃~85℃, accuracy ±0.5℃); simultaneously, it obtains production plan data (current batch battery pack type, remaining process time) from the MES system via the OPCUA protocol. These parameters are aggregated to the factory-level edge server at 100ms intervals, forming a multi-dimensional real-time communication parameter stream containing timestamps (accurate to μs), spatial location (based on UWB positioning, error ≤0.3m), and device ID.

[0112] To quantify the nonlinear correlation between frequency adaptation and response delay, the system invokes a self-supervised spatiotemporal frequency elasticity coefficient model for processing: On one hand, based on the time dimension, the model extracts the time-series features of parameters through an LSTM network (e.g., the RSSI fluctuation period of the 5GHz band in the past 10 minutes is 8 seconds, consistent with the frequency of the AGV vehicle), identifying periodic interference sources; on the other hand, based on the spatial dimension, a graph neural network (GNN) is used to model the positional relationship of different devices (the laser welding equipment is located in area B of the workshop, the vision PLC is located in area C, with 3 metal conveyor frames in between), calculating the spatial attenuation coefficient (the 5GHz RSSI in area B is 7dB lower than that in area C); finally, the model outputs "frequency-delay". The "latency" elasticity coefficient matrix—for example, when the adapter operates at 2.4GHz, the average delay of the laser welding equipment is 12ms (elasticity coefficient α=0.3, that is, for every 1GHz increase in frequency, the delay theoretically decreases by 0.3ms), but due to the influence of metal interference, the actual delay fluctuation standard deviation reaches 1.8ms; while the theoretical delay of the 5GHz band can be reduced to 8ms (α=0.5), but the actual delay increases sharply to 25ms (α=-0.2) due to interference. This contradiction is accurately captured by the dynamic changes of the elasticity coefficient (adjusted with time, space, and equipment load), forming a dynamic spectrum that includes timestamps, equipment pairs (adapter-laser welding / adapter-vision PLC), frequency, and elasticity coefficient.

[0113] Based on the evolution of dynamic graphs (e.g., when the temperature and humidity in the workshop are stable at night, interference in the 5GHz band decreases, and α increases from -0.2 to 0.4; during the day shift from 10:00 to 11:00, when the AGV passes through area B at high frequency, 5GHz α drops sharply to -0.5), the system initiates federated migration reinforcement learning to train policies: five similar factories across the country are treated as federated members (each factory retains local data privacy), and parameter updates are coordinated through a central server—Factory A (high temperature and high humidity environment) contributes the experience of "6GHz band EVM deteriorates to 15% when temperature > 70℃", Factory B (multi-metal obstruction) contributes the priority rule of "5GHz band switching to 2.4GHz when AGV passes", and Factory C (newly put into production, relatively new equipment) contributes the "latency advantage of direct fiber optic connection replacing 5G when adapter operation is <1000 hours". The transfer learning module maps these experiences to the feature space of the current factory (e.g., the current factory adapter has been running for 1800 hours and there is device aging). The optimized strategy priority is as follows: ① Due to the high image transmission bandwidth requirements of the vision PLC, the 5GHz band is allocated first (but EVM ≤ 12%); ② The laser welding equipment uses 5GHz (theoretical delay 8ms) during the non-passing period of AGV (e.g., 10:30-11:30), otherwise it switches to 2.4GHz (delay 12ms but high stability); ③ When the adapter temperature is > 60℃ (aging compensation coefficient β = 1.2, i.e., actual transmission power = nominal value × β), the available power of the 6GHz band is insufficient, and it is forced to switch to 2.4GHz.

[0114] To verify the causal rationality of the strategy, the system invokes the knowledge graph temporal causal chain reasoning: constructing a causal chain containing "adapter aging level (β) - RF power (P) - spatial attenuation (L) - received signal-to-noise ratio (SNR) - EVM - delay (D)", where "β↑→P↓→SNR↓ when L remains unchanged→EVM↑→D↑" is the key causal path. For example, the reasoning found that when β=1.3 (corresponding to 2000 hours of operation), even if there is no interference in the 5GHz band (L=-5dB), the SNR will still drop from 25dB to 18dB, resulting in EVM=14% (exceeding the 12% threshold of the visual PLC). Therefore, the 6GHz to 5GHz switchover plan needs to be triggered 300 hours in advance (when β=1.25). Based on the elasticity coefficient of the dynamic spectrum (α=0.4 at this time for 5GHz), the final target frequency parameters are corrected as follows: the vision PLC uses 5GHz from 8:00 to 18:00 (avoiding AGV peak hours), and switches to 6GHz from 18:00 to 8:00 the next day (less interference); the laser welding equipment is forced to use 2.4GHz during the AGV passage period (10:00-10:10, 15:00-15:10), and dynamically selects 5GHz or 6GHz according to the β value at other times.

[0115] During the strategy execution phase, the three-objective dynamic equilibrium algorithm uses "delay minimization (weight 0.4), throughput maximization (weight 0.3), and energy consumption minimization (weight 0.3)" as objective functions to adjust the frequency in real time. For example, when the vision PLC experiences a sudden surge in throughput (single frame image increases from 2MB to 5MB), the algorithm prioritizes ensuring its throughput (5GHz bandwidth of 800Mbps is sufficient), increasing the receiving latency from 8ms to 10ms. When the laser welding equipment enters the critical path (remaining process time < 30 seconds), the algorithm sacrifices some throughput (switching to 2.4GHz sacrifices 200Mbps) to stabilize the latency within 12ms to ensure welding accuracy. At the same time, the algorithm monitors the adapter temperature continuously rising (currently 55℃, β=1.15) and predicts that β will reach 1.2 in 30 minutes, so it adjusts the usage period of 6GHz from "after 18:00" to "after 17:30" in advance to avoid production line downtime caused by temporary switching.

[0116] The system synchronously monitors communication jitter (defined as 3 times the standard deviation of latency within 1 second, threshold ≤ 50μs) and data fidelity (bit error rate of laser welding commands ≤ 1e-6, SSIM of visual images ≥ 0.98) in real time via the OPCUA interface. When the jitter exceeds the threshold (e.g., 70μs jitter in 5GHz caused by an AGV passing by), a switch is immediately triggered and an alarm is sent to the MES system. Simultaneously, the collected status data (including parameters, strategy decisions, and switch results) is stored in the factory-level time-series database (InfluxDB) for subsequent model iterations. At the end of each month, the system automatically generates a frequency optimization report with equipment aging compensation coefficients: statistics on the number of frequency switches for each equipment pair this month (87 adapter-vision PLC switches, of which 23% were triggered by aging), average latency (decreased from 15ms to 11ms), and production anomalies caused by improper frequency adaptation (4 fewer than the previous month). Combined with the monthly average growth rate of β (0.05 / 100 hours), the system predicts that the availability of the 6GHz band needs to be focused on next month (recommended to replace aging adapters).

[0117] Through the application of this system, the factory's 5G+industrial internet communication efficiency has increased by 28%, the laser welding yield has increased from 98.7% to 99.5%, the visual inspection latency has decreased from 22ms to 12ms, and the failure rate of the adapter due to frequent switching has decreased by 40%, fully verifying the effectiveness of the intelligent switching method based on the communication frequency between the adapter and the equipment in actual industrial scenarios.

[0118] In summary, the embodiments of this application, through a real-world scenario of a smart factory for new energy vehicle power batteries, not only verified the technical effectiveness of the intelligent switching method in complex industrial environments—precisely resolving communication adaptation challenges caused by heterogeneous demands of multiple devices, environmental interference, and equipment aging—but also significantly improved production efficiency (laser welding yield increased from 98.7% to 99.5%) and equipment reliability (adapter failure rate decreased by 40%) through dynamic optimization strategies. More importantly, its end-to-end technical framework and privacy-protecting collaborative mechanisms such as federated transfer learning provide a replicable and scalable intelligent management paradigm for high-frequency communication scenarios involving multiple devices, such as 5G+Industrial Internet, driving the upgrade of industrial communication from experience-driven to data-intelligent driven.

[0119] Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0120] The memory 1001, the processor 1002, and the computer program stored on the memory 1001 and capable of running on the processor 1002.

[0121] When the processor 1002 executes the program, it implements the intelligent switching method based on the communication frequency between the adapter and the device provided in the above embodiment.

[0122] Furthermore, electronic devices also include:

[0123] Communication interface 1003 is used for communication between memory 1001 and processor 1002.

[0124] The memory 1001 is used to store computer programs that can run on the processor 1002.

[0125] The memory 1001 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0126] If the memory 1001, processor 1002, and communication interface 1003 are implemented independently, then the communication interface 1003, memory 1001, and processor 1002 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0127] Optionally, in a specific implementation, if the memory 1001, processor 1002, and communication interface 1003 are integrated on a single chip, then the memory 1001, processor 1002, and communication interface 1003 can communicate with each other through an internal interface.

[0128] The processor 1002 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0129] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described intelligent switching method based on the communication frequency between an adapter and a device.

[0130] Furthermore, this application also provides a computer program product, including a computer program or instructions, which, when executed, implement the above-described intelligent switching method based on the communication frequency between the adapter and the device.

[0131] In the description of this specification, the references to "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0132] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0133] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or more executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0134] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0135] Those skilled in the art will understand that all or part of the steps of the methods described in the above embodiments can be implemented by a program instructing related hardware, and the program can be stored in a computer-readable storage medium. When executed, the program includes one or a combination of the steps of the method embodiments.

[0136] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. An intelligent switching system based on the communication frequency between an adapter and a device, characterized in that, include: The module includes a communication status acquisition module, a status feature analysis module, a frequency decision module, and a dynamic control feedback module; among which, The communication status acquisition module is used to acquire real-time communication parameters, instantaneous power consumption, and link attenuation characteristics between the adapter and the device. The state feature analysis module is used to fuse data and construct a dynamic graph of communication quality-equipment requirements through a self-supervised spatiotemporal frequency elasticity coefficient model, quantifying the nonlinear correlation between frequency adaptation and response delay. Based on the nonlinear association, the frequency decision module generates a distributed switching strategy through federated transfer reinforcement learning, and outputs the target frequency parameter by combining knowledge graph temporal causal chain reasoning. The dynamic control feedback module, based on the distributed switching strategy, performs switching through a three-objective dynamic equilibrium algorithm, monitors communication jitter and data fidelity in real time, and automatically generates a frequency optimization report with equipment aging compensation coefficient.

2. The intelligent switching system based on the communication frequency between the adapter and the device according to claim 1, characterized in that, The communication status acquisition module includes a real-time communication parameter acquisition unit, an instantaneous power consumption monitoring unit, and a link attenuation characteristic evaluation unit. The real-time communication parameter acquisition unit uses protocol analysis tools to acquire real-time communication parameters between the adapter and the device, recording key indicators such as transmission rate, latency, and bit error rate. The instantaneous power consumption monitoring unit uses current sensors and voltage measurement devices to synchronously acquire instantaneous power consumption data between the adapter and the device, analyzing power consumption characteristics under different operating conditions. The link attenuation characteristic evaluation unit uses signal strength detection and quality analysis algorithms to evaluate the attenuation characteristics of the link between the adapter and the device, identifying signal interference sources and quantifying the link quality change trend.

3. The intelligent switching system based on the communication frequency between the adapter and the device according to claim 1, characterized in that, The state feature analysis module includes a data fusion processing unit, a dynamic graph construction unit, and a nonlinear correlation quantization unit. The data fusion processing unit extracts and fuses features from the original data using a self-supervised learning algorithm, eliminating noise interference and dimensional differences between data. The dynamic graph construction unit constructs a communication quality-equipment requirement dynamic graph based on the fused data, performing multi-scale spatiotemporal correlation between time-series features and spatial distribution features. The nonlinear correlation quantization unit analyzes the changes in elasticity coefficients in the dynamic graph to quantify the strength and dynamic evolution of the nonlinear correlation between frequency adaptation and response delay.

4. The intelligent switching system based on the communication frequency between the adapter and the device according to claim 1, characterized in that, The frequency decision module includes a federated transfer reinforcement learning unit and a temporal causal chain reasoning unit. The federated transfer reinforcement learning unit performs distributed policy training and optimization through a federated transfer reinforcement learning algorithm to generate a distributed switching policy. The temporal causal chain reasoning unit performs temporal causal analysis through a knowledge graph temporal causal chain reasoning algorithm and, in conjunction with the distributed switching policy, outputs the target frequency parameters.

5. The intelligent switching system based on the communication frequency between the adapter and the device according to claim 1, characterized in that, The formula for the federated transfer reinforcement learning algorithm is as follows: in, The value when the decision-making strategy reaches its optimal state; The environmental conditions in which the decision is made; This refers to the specific operation selected in the current state. The expected state transition probability is based on the environment. For immediate rewards; To represent the discount factor, the value range is 0 ≤ ... ≦1; To select the action that will bring the greatest value from all possible next actions; The next state; Next action.

6. The intelligent switching system based on the communication frequency between the adapter and the device according to claim 1, characterized in that, The dynamic control feedback module includes an algorithm execution unit and a report generation unit. The algorithm execution unit uses a three-objective dynamic equilibrium algorithm to dynamically weigh and make decisions on key indicators such as latency, bandwidth, and reliability during the handover process. The report generation unit collects and stores real-time status data by monitoring communication jitter amplitude and data fidelity indicators in real time, and combines the equipment aging compensation coefficient to perform correlation analysis on the monitoring data and handover results, automatically generating an optimization report containing frequency adjustment suggestions.

7. A method for intelligent switching based on the communication frequency between an adapter and a device, applicable to any one of claims 1-6, characterized in that, include: Obtain real-time communication parameters between the adapter and the device; By integrating the real-time communication parameters of the adapter and the device, a dynamic spectrum of communication quality and device demand is constructed through a self-supervised spatiotemporal frequency elasticity coefficient model. By analyzing the changes in elasticity coefficients in the dynamic spectrum, the nonlinear correlation strength and dynamic evolution law between frequency adaptation and response delay are quantified and displayed. Based on the nonlinear correlation strength and dynamic evolution law, distributed policy training and optimization are carried out through federated transfer reinforcement learning to generate distributed switching policies. Combined with knowledge graph temporal causal chain reasoning, target frequency parameters are output. Based on the distributed handover strategy, the handover is performed through a three-objective dynamic balancing algorithm, communication jitter and data fidelity are monitored in real time, real-time status data is collected and stored, and frequency optimization reports with equipment aging compensation coefficients are automatically generated.

8. An electronic device, characterized in that, include: The device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the intelligent switching method based on the communication frequency between the adapter and the device as described in claim 7.

9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed, they implement the intelligent switching method based on the communication frequency between the adapter and the device as described in claim 7.

10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed, they implement the intelligent switching method based on the communication frequency between the adapter and the device as described in claim 7.

Images