A CPS-based multi-communication link adaptive edge decision method and device

Abstract

This application discloses a CPS-based multi-communication link adaptive edge decision-making method and device, belonging to the fields of industrial automation control and power information technology. The method includes: acquiring link status parameters of multiple communication links accessing the edge device at preset time intervals; calculating the comprehensive score and average link utilization of each communication link based on the link status parameters using a preset weighted scoring method to obtain the performance evaluation results of each communication link; based on the performance evaluation results, using the edge device to execute an adaptive decision-making algorithm to select a determined set of target links and a backup link configuration as the decision result; based on the decision result, using the edge device to transmit business data; employing a CPS architecture to divide the system into multiple layers—cloud-edge-device—overcoming the limitations of existing centralized or distributed single architectures; and achieving a control closed loop through cloud-edge collaboration, which not only reduces on-site response latency but also alleviates the communication and computing load on the cloud.

Description

Technical Field

[0001] This application belongs to the fields of industrial automation control and power information technology, and specifically relates to a multi-communication link adaptive edge decision method and device based on CPS. Background Technology

[0002] With the development of the Industrial Internet of Things (IIoT) and new energy fields, traditional power automation control systems are gradually integrating emerging information technologies and evolving towards a Cyber-Physical System (CPS) architecture. Currently, monitoring and control of power plants and distribution networks mostly adopt a centralized architecture. However, in remote and harsh environments, this single communication method suffers from insufficient stability: for example, sandstorms can affect wireless signal quality, saline-alkali environments may corrode wired interfaces, and extreme temperatures also pose challenges to equipment communication performance. Existing remote terminal units (RTUs) or edge devices often only have a single network interface (such as 4G or WiFi only). When the link signal is poor or interrupted, the system cannot switch to other links in time, leading to data transmission delays or even interruptions, affecting the reliability of control.

[0003] Furthermore, the integration of cloud computing and edge computing has become a trend in improving the performance of industrial control systems. While traditional centralized cloud computing offers powerful storage and computing capabilities, network latency and bandwidth limitations prevent it from meeting the stringent real-time and reliability requirements of field control. Simultaneously, traditional centralized control is sensitive to single points of failure; a communication or central node failure can render the entire system inoperable, indicating insufficient reliability and robustness. Currently, many power IoT systems lack a robust cloud-edge collaboration mechanism, making it difficult to deploy cloud-trained intelligent algorithms to the field in a timely manner. The edge and cloud operate independently, failing to fully leverage their collaborative advantages. Summary of the Invention

[0004] To address the aforementioned problems, this application provides a CPS-based multi-communication link adaptive edge decision-making method, the improvement of which includes: The link status parameters of various communication links of the access edge device are obtained at preset time intervals. Based on the link status parameters, a preset weighted scoring method is used to calculate the comprehensive score and average link utilization of each communication link, and the performance evaluation results of each communication link are obtained. Based on the performance evaluation results, an adaptive decision-making algorithm is executed by the edge device to select the determined target link set and backup link configuration as the decision result; Based on the decision results, edge devices are used to transmit business data; The adaptive decision-making algorithm is deployed from the cloud to edge devices.

[0005] Optionally, the step of using an adaptive decision-making algorithm executed by the edge device based on the performance evaluation results to select the determined target link set and backup link configuration as the decision result includes: Based on the performance evaluation results, and based on the preset target number of links N and the preset threshold, the edge devices select the top N links with the comprehensive score as the determined target link set, and select them as backup links according to the attributes of the communication links themselves. When the average link utilization exceeds the preset threshold, additional links are activated. When the overall score of the backup link exceeds that of the target link and the duration is not less than the preset hysteresis time, the backup link will be switched to the new target link, and the target link with the declining overall score will be downgraded to the backup link. The preset threshold and hysteresis time are remotely issued by the cloud platform and take effect in real time.

[0006] Optionally, the step of using edge devices to transmit business data based on the decision results includes: The data type of the business data is identified based on the business data level identifier or port number; the business data is divided into control instruction data and monitoring data. Based on the decision results, for control command data, edge devices send control command data completely through a single link and synchronize redundant copies through a backup link. For monitoring data, edge devices use a multi-path transmission mechanism to send monitoring data in fragments in parallel. When a failure is detected in the target link being used, the link status parameters are obtained and a backup link is selected as the target link. When the original target link recovers from the fault and its overall score is higher than that of the currently used target link, and the preset hysteresis time is continuously maintained, the original target link will be restored to the target link or the link status parameters will be obtained to select a new target link.

[0007] Optionally, the step of using edge devices to transmit business data based on the decision results further includes: If all target links are unavailable, the system enters offline caching mode to cache the service data to be transmitted; when any communication link becomes available again, the system exits offline caching mode and retransmits the cached service data in batches. In the offline caching mode, at least the most recent control commands and monitoring data for a preset duration are cached. When the cache is full, the earliest data is discarded using a first-in-first-out strategy.

[0008] Optionally, the link state parameters include one or more of the following: Signal strength, bandwidth limit, current bandwidth usage, round-trip time, packet loss rate, or stability index.

[0009] Optionally, the calculation formula for the preset weighted scoring method is:

[0010] in, The overall score for the communication link is represented by RSSI (Signal Strength Index), L (Round-Trip Time), PL (Package Loss Rate), and SI (Stability Index). The weighting coefficients of the RSSI (Signal Strength Index) are used to represent the signal strength index. This represents the weighting coefficient of the round-trip delay L index. This represents the weighting coefficient of the packet loss rate (PL) metric. This represents the weighting coefficient of the Stability Index (SI). This represents the normalization / mapping function for RSSI. This represents a penalized normalization / mapping function for L. This represents the penalized normalization / mapping function for PL.

[0011] Based on the same inventive concept, this application also provides a CPS-based multi-communication link adaptive edge decision-making device, the improvement of which includes: The acquisition unit is used to acquire link status parameters of various communication links of the access edge device at preset time intervals; The performance analysis unit is used to calculate the comprehensive score and average link utilization of each communication link based on the link status parameters using a preset weighted scoring method, and obtain the performance evaluation results of each communication link. The adaptive decision unit is used to select the determined target link set and backup link configuration as the decision result by using an adaptive decision algorithm executed by the edge device based on the performance evaluation results. The execution unit is used to transmit business data using edge devices based on the decision results; The adaptive decision-making algorithm is deployed from the cloud to edge devices.

[0012] Optionally, the step of using an adaptive decision-making algorithm executed by the edge device based on the performance evaluation results to select the determined target link set and backup link configuration as the decision result includes: Based on the performance evaluation results, and based on the preset target number of links N and the preset threshold, the edge devices select the top N links with the comprehensive score as the determined target link set, and select them as backup links according to the attributes of the communication links themselves. When the average link utilization exceeds the preset threshold, additional links are activated. When the overall score of the backup link exceeds that of the target link and the duration is not less than the preset hysteresis time, the backup link will be switched to the new target link, and the target link with the declining overall score will be downgraded to the backup link. The preset threshold and hysteresis time are remotely issued by the cloud platform and take effect in real time.

[0013] Optionally, the step of using edge devices to transmit business data based on the decision results includes: The data type of the business data is identified based on the business data level identifier or port number; the business data is divided into control instruction data and monitoring data. Based on the decision results, for control command data, edge devices send control command data completely through a single link and synchronize redundant copies through a backup link. For monitoring data, edge devices use a multi-path transmission mechanism to send monitoring data in fragments in parallel. When a failure is detected in the target link being used, the link status parameters are obtained and a backup link is selected as the target link. When the original target link recovers from the fault and its overall score is higher than that of the currently used target link, and the preset hysteresis time is continuously maintained, the original target link will be restored to the target link or the link status parameters will be obtained to select a new target link.

[0014] Optionally, the step of using edge devices to transmit business data based on the decision results further includes: If all target links are unavailable, the system enters offline caching mode to cache the service data to be transmitted; when any communication link becomes available again, the system exits offline caching mode and retransmits the cached service data in batches. In the offline caching mode, at least the most recent control commands and monitoring data for a preset duration are cached. When the cache is full, the earliest data is discarded using a first-in-first-out strategy.

[0015] Optionally, the link state parameters include one or more of the following: Signal strength, bandwidth limit, current bandwidth usage, round-trip time, packet loss rate, or stability index.

[0016] Optionally, the calculation formula for the preset weighted scoring method is:

[0017] in, The overall score for the communication link is represented by RSSI (Signal Strength Index), L (Round-Trip Time), PL (Package Loss Rate), and SI (Stability Index). The weighting coefficients of the RSSI (Signal Strength Index) are used to represent the signal strength index. This represents the weighting coefficient of the round-trip delay L index. This represents the weighting coefficient of the packet loss rate (PL) metric. This represents the weighting coefficient of the Stability Index (SI). This represents the normalization / mapping function for RSSI. This represents a penalized normalization / mapping function for L. This represents the penalized normalization / mapping function for PL.

[0018] This application provides a CPS-based multi-communication link adaptive edge decision-making method and apparatus, comprising: acquiring link status parameters of multiple communication links accessing the edge device at preset time intervals; calculating the comprehensive score and average link utilization of each communication link based on the link status parameters using a preset weighted scoring formula to obtain the performance evaluation results of each communication link; based on the performance evaluation results, using the edge device to execute an adaptive decision-making algorithm to select a determined set of target links and a backup link configuration as the decision result; and using the edge device to transmit service data based on the decision result; wherein the adaptive decision-making algorithm is deployed from the cloud to the edge device; The system adopts a CPS architecture, dividing it into multiple layers: cloud, edge, and endpoint. A collaborative mechanism enables both centralized management and distributed autonomy. This layered collaborative intelligent system leverages the global optimization capabilities of the cloud while ensuring real-time local control, overcoming the limitations of existing centralized or distributed single architectures. The terminal and edge gateway of this application can automatically select the optimal communication method based on link quality, maintaining uninterrupted data transmission even in harsh environments and significantly improving network reliability. Redundancy backup is achieved through technologies such as MPTCP link aggregation and fast switching, avoiding single points of failure and ensuring that control commands and monitoring data are delivered stably and with low latency. This application achieves a closed-loop control system through a collaborative approach of cloud-based algorithm training and edge-based decision-making. By bringing complex cloud-based algorithm models down to the edge for real-time execution, not only is on-site response latency reduced, but the communication and computational load on the cloud is also lessened.

[0019] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 A flowchart of a CPS-based multi-communication link adaptive edge decision-making method provided in this application is shown. Figure 2The figure shows the implementation steps of a preferred embodiment of the CPS-based multi-communication link adaptive edge decision method provided in this application; Figure 3 This paper presents a structural diagram of a CPS-based multi-communication link adaptive edge decision-making device provided in this application. Figure 4 This paper illustrates an implementation structural diagram of a preferred embodiment of a CPS-based multi-communication link adaptive edge decision device provided in this application. Detailed Implementation

[0022] In terms of software architecture, traditional industrial control systems employ a relatively closed software deployment model, making upgrades and maintenance difficult. Field controllers often run as firmware or proprietary programs, lacking modern software management mechanisms. Updating control logic or algorithms frequently requires system downtime, posing a high risk. Containerization technology is a mature application in the IT field, enabling rapid deployment, isolated operation, and fault self-healing by encapsulating applications within containers. However, in the current industrial control field, container technology is less widely used. Software updates for field edge devices often require manual intervention, hindering smooth transitions or canary releases, thus limiting system flexibility and continuous evolution capabilities.

[0023] Security is also a critical issue for power CPS systems. Power systems are critical national infrastructure, and cyberattacks targeting power monitoring systems have drawn significant attention in recent years. Traditional industrial control network protocols (such as Modbus and DNP3) lack encryption and authentication mechanisms, making them vulnerable to tampering and eavesdropping attacks. Although international standards such as the IEC 62351 series of power system communication security standards have been developed to add security extensions like TLS encryption and role authentication to power protocols such as IEC 61850, many existing power field devices have not fully implemented these specifications, posing security risks. Furthermore, industrial field networks typically require partitioning and isolating production control networks from office networks, and authenticating and auditing access devices to prevent unauthorized access and potential intrusions. However, deploying robust access control and intrusion detection / prevention mechanisms in traditional systems is difficult, and edge layer security capabilities need improvement.

[0024] Finally, the reliability and scalability of current power edge systems also need improvement. On the one hand, the lack of link redundancy mechanisms in network communication makes it difficult to guarantee the continuity of data transmission; on the other hand, many systems have closed architectures and support limited industrial protocols, making it difficult for users to customize control logic according to their own needs, hindering the rapid deployment of new services. For example, wind power equipment from different manufacturers may use their own protocols, requiring significant development and adaptation work during integration. The lack of a unified programmable platform makes it difficult for field engineers to quickly define or modify business logic to respond to production demands.

[0025] The existing technology has the following shortcomings: (1) It lacks a hierarchical collaborative control architecture that integrates new-generation information technology, making it difficult to simultaneously consider global optimization and local real-time control; (2) It lacks an intelligent switching mechanism for multiple communication links, resulting in insufficient communication reliability of the system in harsh environments; (3) It lacks software means for flexible deployment and maintenance of control applications at the edge, making system upgrades and maintenance inconvenient; (4) It lacks sufficient collaboration between the cloud and the edge, making it difficult to promptly sink cloud intelligence to local execution to form a rapid closed loop; (5) Its network security protection capabilities need to be strengthened, failing to meet the requirements of the latest power safety standards; (6) Its communication and control lack highly reliable mechanisms, lacking self-recovery means when encountering link fluctuations or device disconnections; (7) Its system scalability is limited, making it difficult to support heterogeneous device access and user-defined logic. In order to solve the above problems, it is necessary to provide a new technical solution.

[0026] This application addresses the shortcomings of existing power CPS systems in terms of intelligent architecture, communication reliability, software deployment, cloud-edge collaboration, security protection, and scalability. It proposes a solution integrating multiple emerging technologies. Specifically, the technical challenges are: how to construct a layered, collaborative intelligent control system capable of automatically optimizing data transmission using multiple communication links, maintaining reliable communication even in harsh environments; simultaneously supporting flexible deployment and self-healing control applications at the edge, and forming a closed-loop control system in collaboration with the cloud; ensuring the system complies with power network security standards, and providing high reliability and scalable operational support.

[0027] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, 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, 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.

[0028] Example 1 This application provides a CPS-based adaptive edge decision-making method for multiple communication links, such as... Figure 1 ,include: The link status parameters of various communication links of the access edge device are obtained at preset time intervals. Based on the link status parameters, a preset weighted scoring method is used to calculate the comprehensive score and average link utilization of each communication link, and the performance evaluation results of each communication link are obtained. Based on the performance evaluation results, an adaptive decision-making algorithm is executed by the edge device to select the determined target link set and backup link configuration as the decision result; Based on the decision results, edge devices are used to transmit business data; The adaptive decision-making algorithm is deployed from the cloud to edge devices.

[0029] The adaptive decision-making algorithm is deployed from the cloud to edge devices in the following ways, including but not limited to: packaging the decision-making algorithm into a container image and pushing it to the edge device through a container orchestration system; updating decision strategy parameters through configuration files; and synchronizing the weights of the machine learning model trained in the cloud through a model parameter update mechanism. The deployment process employs hot update or canary release mechanisms to ensure uninterrupted service on edge devices. Optionally, the step of using an adaptive decision-making algorithm executed by the edge device based on the performance evaluation results to select the determined target link set and backup link configuration as the decision result includes: Based on the performance evaluation results, and based on the preset target number of links N and the preset threshold, the edge devices select the top N links with the comprehensive score as the determined target link set, and select them as backup links according to the attributes of the communication links themselves. When the average link utilization exceeds the preset threshold, additional links are activated. When the overall score of the backup link exceeds that of the target link and the duration is not less than the preset hysteresis time, the backup link will be switched to the new target link, and the target link with the declining overall score will be downgraded to the backup link. To avoid link switching jitter, the hysteresis timer is not started when the difference between the overall score of the backup link and the target link is less than the preset difference threshold ΔQ; the switch is only performed when the score difference exceeds ΔQ and the hysteresis time continues.

[0030] The preset threshold and hysteresis time are remotely issued by the cloud platform and take effect in real time.

[0031] Optionally, the step of using edge devices to transmit business data based on the decision results includes: The data type of the business data is identified based on the business data level identifier or port number; the business data is divided into control instruction data and monitoring data. Based on the decision results, for control command data, edge devices send control command data completely through a single link and synchronize redundant copies through a backup link. For monitoring data, edge devices use a multi-path transmission mechanism to send monitoring data in fragments in parallel. When a failure is detected in the target link being used, the link status parameters are obtained and a backup link is selected as the target link. When the original target link recovers from the fault and its overall score is higher than that of the currently used target link, and the preset hysteresis time is continuously maintained, the original target link will be restored to the target link or the link status parameters will be obtained to select a new target link.

[0032] Optionally, the step of using edge devices to transmit business data based on the decision results further includes: If all target links are unavailable, the system enters offline caching mode to cache the service data to be transmitted; when any communication link becomes available again, the system exits offline caching mode and retransmits the cached service data in batches. In the offline caching mode, at least the most recent control commands and monitoring data for a preset duration are cached. When the cache is full, the earliest data is discarded using a first-in-first-out strategy.

[0033] Optionally, the link state parameters include one or more of the following: Signal strength, bandwidth limit, current bandwidth usage, round-trip time, packet loss rate, or stability index.

[0034] Optionally, the calculation formula for the preset weighted scoring method is:

[0035] in, The overall score for the communication link is represented by RSSI (Signal Strength Index), L (Round-Trip Time), PL (Package Loss Rate), and SI (Stability Index). The weighting coefficients of the RSSI (Signal Strength Index) are used to represent the signal strength index. This represents the weighting coefficient of the round-trip delay L index. This represents the weighting coefficient of the packet loss rate (PL) metric. This represents the weighting coefficient of the Stability Index (SI). This represents the normalization / mapping function for RSSI. This represents a penalized normalization / mapping function for L. This represents the penalized normalization / mapping function for PL, where each weight coefficient satisfies... + + + =1; the output value range of each normalization function is [0, 100], ensuring the comprehensive score. The value range is [0, 100]; when the result of any normalization function is less than 0, the value is 0.

[0036] Optionally, the normalized mapping function may take the following form: Signal strength normalization function (RSSI) employs a linear mapping: (RSSI)=(RSSI-RSSImin) / (RSSImax-RSSImin)×100; Where RSSImin is the lower limit threshold of signal strength (e.g., -100dBm) and RSSImax is the upper limit threshold of signal strength (e.g., -30dBm). Round-trip delay penalty normalization function (L) Uses reverse mapping: F2(L)=max(0,100-k1×L) where k1 is the delay penalty coefficient and L is the round-trip delay (in milliseconds). The greater the delay, the lower the score. Packet loss rate penalty normalization function (PL) uses reverse mapping: (PL) = max(0, 100 - k2 × PL), where k2 is the packet loss rate penalty coefficient, and PL is the packet loss rate (unit: percentage). The higher the packet loss rate, the lower the score. The above function parameters can be adjusted according to the actual application scenario, and this application is not limited to the specific form mentioned above.

[0037] Optional, weight w1 w4 dynamically adjusts based on service level and historical link performance using a Bayesian self-learning algorithm.

[0038] Optional, weight The weights are dynamically adjusted based on service priority and historical link performance using a Bayesian self-learning algorithm. Specifically, this includes: classifying data transmission into multiple service levels based on service type, such as real-time control (high priority), video surveillance (medium priority), and status reporting (low priority); employing Bayesian inference to update the posterior probability distribution of each weight coefficient based on the joint distribution of historical transmission success rate and link status parameters; increasing the values ​​of latency weight w2 and packet loss rate weight w3 for high-priority services; and increasing the value of bandwidth-related weight w1 for low-priority services. The weight update cycle can be set to once per hour or once per day to balance algorithm convergence speed and system stability.

[0039] Optionally, the offline caching mode caches at least the control commands and monitoring data from the most recent 30 minutes. When the cache is full, a first-in-first-out (FIFO) strategy is used to discard the oldest data.

[0040] Example 2 Based on the same inventive concept, this application also provides a preferred embodiment of a CPS-based multi-communication link adaptive edge decision-making method, such as... Figure 2 ,include: The edge device simultaneously establishes connections with at least two different types of communication links, including any two of cellular mobile network links, wireless local area network links, and low power wide area network links, and collects the signal strength RSSI, round-trip time L, packet loss rate PL, and stability index SI for each communication link in real time.

[0041] 1. Communication Link Status Acquisition Steps: Edge devices connect to at least two different types of communication links (e.g., cellular mobile networks, Wi-Fi wireless LANs, fiber optic Ethernet, or low-power wide area network links such as LPWAN) and collect the status parameters of each link in real time. Typical link status parameters include signal strength, bandwidth limit, current bandwidth usage, latency, packet loss rate, and link availability information. Acquiring these parameters provides basic data for subsequent evaluation, ensuring that decisions are based on evidence. The stability index SI is used to characterize the performance fluctuation of the communication link within a certain time window. Its calculation method is: SI = 100 - σ(Q) / μ(Q) × 100, where σ(Q) represents the standard deviation of the comprehensive score over the most recent N sampling periods, and μ(Q) represents the mean of the comprehensive score over the most recent N sampling periods; a higher stability index indicates a more stable link. Optionally, N ranges from 10 to 60, corresponding to an observation window of 10 seconds to 1 minute.

[0042] Calculate the overall score for each communication link. .

[0043] 2. Link Performance Evaluation Steps: Based on the acquired link status parameters, edge nodes perform quantitative analysis of the performance of each communication link. For example, a preset link quality calculation algorithm can be used to comprehensively calculate the link quality score or estimated link transmission rate by considering factors such as bandwidth and latency of each link. Furthermore, based on link transmission rate and historical traffic data, the average link utilization rate of each link can be calculated to represent the degree of utilization of link resources. Through this evaluation step, the system can understand the relative advantages and disadvantages and load conditions of each communication link.

[0044] When the backup link The connection continuously exceeds the primary link and the duration is not less than the preset hysteresis time. When necessary, the backup link will be switched to the new primary link, and / or the link with the declining score will be downgraded to a backup link.

[0045] 3. Adaptive Decision-Making Steps: Based on the performance metrics of each link obtained from the evaluation, the edge device executes an adaptive decision-making algorithm to select the optimal communication link or link combination for data transmission. The decision-making process may include the following strategies: First, all links are sorted from high to low according to their quality score or utilization rate; then, based on a preset target number of links or a quality threshold, a target link set for the current transmission task is determined. For example, under a pre-set strategy of using a maximum of N links simultaneously, the top N links in terms of quality are selected as the effective link set; or when the quality of a major link drops below a threshold, a backup link is introduced to maintain communication quality. This adaptive decision-making fully considers the dynamic changes in link status and can flexibly switch or use multiple links in parallel under different scenarios to ensure the real-time communication needs of the CPS system.

[0046] The Multipath Transmission Control Protocol (MPTCP) is used to send service data in parallel on the target link set and reassemble it at the receiving end.

[0047] When any path failure is recovered and its Higher than the current primary link and continues After a certain period of time, the system will automatically switch back or select a new set of target links; if all links are unavailable, it will enter offline caching mode and retransmit the cached data in batches after the links are restored.

[0048] 4. Data Transmission Execution Steps: Based on the target link set determined in the previous step, the edge device transmits service data through the selected links. In a single-link scenario, data packets are directed to the currently optimal link interface for transmission; in a multi-link parallel usage scenario, a multi-path transmission mechanism can be used to fragment and send data in parallel. For example, a Multipath Transmission Control Protocol (MPTCP) can be used to establish multiple sub-connections at the transport layer, enabling a data stream to be transmitted simultaneously on multiple network paths. MPTCP allows simultaneous data transmission using multiple paths, providing higher total bandwidth, better load balancing, and higher reliability. By executing this step, the system implements the results of adaptive decisions, completing the reliable transmission of data from the edge to the target. If a significant change in link status is detected during operation (e.g., current link interruption or performance degradation), this method can return to step 1 to obtain the updated link status and iteratively execute the above steps, thereby achieving continuous adaptive optimization.

[0049] Working Principle and Function: The above steps form a closed-loop control process. Step 1 provides environmental awareness, ensuring the accuracy of decision-making; Step 2 quantifies complex multi-link situations through algorithmic analysis, laying the foundation for comparison and selection; Step 3 is the core decision-making stage, realizing intelligent scheduling of communication links at the edge; Step 4 completes the specific data transmission task, ensuring data is transmitted through the optimal path. The entire method relies on edge computing, pushing computational decision-making down to the data source closer to the physical devices, thereby reducing latency caused by cloud involvement and improving the real-time performance and robustness of the CPS system.

[0050] The method described in this application can be widely applied to IoT and industrial control scenarios that require multi-network access. For example, in smart manufacturing, the edge controller can simultaneously connect to industrial Ethernet, 5G mobile networks, and LPWAN (Low-Power Wide-Area Network) sensor networks. By adaptively selecting communication links through this method, efficient transmission of production data and reliable communication for remote equipment control can be achieved.

[0051] For example, an edge gateway device in a smart factory uses the CPS-based multi-communication link adaptive edge decision-making method of this application for multi-link adaptive communication decisions. As an edge computing node in a CPS environment, this gateway connects to sensors and controllers (physical devices) within the factory and possesses various network interfaces, including industrial Ethernet, Wi-Fi, cellular mobile network modules (e.g., 4G / 5G modules), and LPWAN modules (e.g., transceivers supporting narrowband IoT NB-IoT). During normal operation, the gateway needs to upload data collected from the field to a cloud monitoring platform and receive control commands from the cloud. Since different network links have varying bandwidths, coverage, and reliability, the method of this application ensures that the gateway selects the optimal communication link based on actual conditions, improving data transmission efficiency and reducing latency.

[0052] Step 1: Communication Link Status Acquisition. After the edge gateway starts, it initializes all available network interface modules and establishes a link monitoring task to periodically acquire link status parameters. For example, for Wi-Fi links, it calls the wireless network driver to obtain the current signal strength (RSSI) and link rate limit; for 4G / 5G cellular links, it queries the current signal quality (such as RSRP, SINR) and available bandwidth through the API provided by the operator; for LPWAN links such as NB-IoT, it reads the received signal strength indication and the latency of the most recent communication. In this embodiment, it is assumed that the edge device collects the status of each link once per second and stores the parameter values ​​in shared memory for subsequent modules to read. For example, the parameters collected at a certain moment are as follows: Wi-Fi link signal strength is -50dBm, bandwidth limit is 100Mbps, latency is 5ms; 4G link signal quality is good (e.g., RSRP -90dBm), estimated bandwidth is 20Mbps, latency is 20ms; NB-IoT link signal strength is medium, bandwidth is only tens of kbps, latency is relatively high, about 100ms. Through this step, the gateway fully perceives the real-time status of each connected heterogeneous network link.

[0053] Step 2: Link Performance Evaluation. The gateway's built-in link evaluation module reads the aforementioned link status parameters and calculates and analyzes the communication performance of each link. In this embodiment, the evaluation module uses different combinations of indicators for different types of links: for high-speed links (Wi-Fi, 4G), bandwidth and latency are mainly considered, while for low-speed links (NB-IoT), reliable coverage and power consumption are emphasized. The evaluation module first estimates the actual achievable transmission rate of each link based on the most recent historical data. For example, under current signal conditions, the Wi-Fi link is estimated to stably provide 80Mbps throughput, the 4G link can provide 15Mbps throughput, and the NB-IoT link is limited to only about 40kbps due to the rate cap. Next, the average link utilization rate of each link is calculated, defined as the ratio of actual occupied bandwidth to available bandwidth. For example, if the Wi-Fi link actually occupies 40Mbps of bandwidth within a certain time window, its utilization rate is 50%; if the 4G link occupies 10Mbps, its utilization rate is about 66%; and if the NB-IoT link occupies 0.02Mbps, its utilization rate is about 50%. Simultaneously, the evaluation module integrates information such as latency and packet loss rate to generate a link quality score Q for ranking (assuming a maximum score of 100, Q=90 for Wi-Fi links, Q=75 for 4G links, and Q=60 for NB-IoT links). Through this step, the edge gateway obtains a quantitative list of link performance indicators, including the bandwidth capacity, utilization percentage, and overall quality score for each link. This data provides an objective basis for subsequent selection decisions. It is worth noting that the evaluation algorithm can employ different methods as needed, such as using a weighted formula to synthesize multiple indicators into a score, or using a machine learning model to predict link stability. This embodiment uses a preset formula to calculate the score to ensure fast and controllable calculation.

[0054] Step 3: Adaptive Decision Making. After understanding the performance of each link, the edge gateway enters the adaptive decision-making stage. The decision-making module operates according to the following strategy: First, the links are sorted from high to low according to their quality score Q: Wi-Fi is the highest (90), followed by 4G (75), and finally NB-IoT (60). The system pre-sets the maximum number of links that can be used simultaneously to be 2, and requires that the average utilization rate of the selected links is not less than 50%. Based on this strategy, the decision-making module prioritizes selecting the Wi-Fi link with the highest score as the main transmission channel. At the same time, it checks the score and utilization rate of the second-best link, 4G. If the threshold is met, 4G is included in the transmission set, forming a target link set containing two links: {Wi-Fi, 4G}. Due to its low bandwidth and lowest score, NB-IoT is not used as the main transmission path when a high-speed link is available. However, its low-power wide-area coverage characteristics make it suitable as a redundant backup link: the decision-making module sets NB-IoT to send a small number of critical control commands or temporarily take over basic data communication when the first two links fail. The decision results therefore include the main link set and the backup link configuration. It should be noted that the decision-making process in this application is dynamic: the gateway continuously and periodically re-evaluates the links. If a Wi-Fi link interruption is detected, the decision module will remove it from the set in real time and activate the backup NB-IoT link to ensure uninterrupted data transmission. Alternatively, when the NB-IoT link quality significantly improves and there are a large number of small data packets suitable for low-speed transmission, the decision module may also choose to switch some non-real-time data to NB-IoT transmission to reduce the load on the high-speed link. This adaptive characteristic ensures that the CPS system maintains the optimal communication strategy in changing environments.

[0055] Step 4: Data Transmission Execution. Based on the decision obtained in Step 3, the edge gateway sends out the actual data stream according to the specified link scheme. In this embodiment, since the target link set includes both Wi-Fi and 4G links, the gateway enables the MPTCP multipath transmission mechanism: this mechanism simultaneously establishes a primary connection for accessing the Internet via Wi-Fi and a backup connection via 4G cellular at the transport layer. When the actual data packets are sent, the MPTCP protocol stack divides them into two sub-streams according to the load, one part going through Wi-Fi and the other through 4G, thereby aggregating the bandwidth of the two links. Specifically, for video surveillance data from the factory site (high bandwidth requirements but not extremely sensitive to latency), MPTCP automatically fragments it for transmission on both Wi-Fi and 4G links to achieve a high throughput close to "the sum of the two bandwidths"; while for real-time control commands (small data volume but requiring reliable low latency), the system mainly sends them through the Wi-Fi link, and uses MPTCP to keep the backup sub-stream on the 4G link ready at any time. Once Wi-Fi latency increases or packets are lost, MPTCP automatically transfers these critical small data packets to the 4G path for transmission, ensuring the timely delivery of control commands. For control command data with high real-time requirements, a single-link complete transmission is used, with redundant copies synchronized via a backup link. For monitoring data with high bandwidth requirements but high latency tolerance, multi-path fragmentation and parallel transmission are employed. Data types are identified by service level identifiers or port numbers. Furthermore, for ultra-low-speed data such as device status heartbeats reported via the NB-IoT link, the gateway directly uses the NB-IoT module for transmission without going through MPTCP (because it uses an independent IoT protocol stack). This data volume is extremely small, and the ultra-long-range coverage of NB-IoT ensures that the system can maintain basic status reporting regardless of Wi-Fi / 4G fluctuations. Through this transmission execution scheme, each of the three links leverages its strengths: the high-speed link bears the main load and serves as a backup, while the low-speed link provides a safety net and special applications. This verifies the effectiveness of the proposed method in complex CPS applications.

[0056] Cyclic Execution and Optimization: Steps 1 to 4 above are repeated at preset time intervals, allowing the system to continuously adapt to environmental changes. For example, link quality is reassessed and decisions are updated every 5 seconds. After a period of operation, the edge gateway can statistically analyze the long-term utilization of each link and adjust policy parameters, such as increasing the target link utilization threshold to more actively integrate link resources, or adjusting the load distribution weight of the multi-path algorithm to optimize performance. Furthermore, this method also supports collaboration with a higher-level management platform: the cloud can issue policy instructions (e.g., specifying to minimize the use of a link when its cost is high), and the edge device incorporates this as one of the constraints in its decision-making, thereby achieving edge-cloud collaborative network resource optimization.

[0057] This application has the following significant advantages: Layered and Collaborative Intelligent Architecture: This application integrates traditional automation control with emerging information technology, adopting a CPS architecture to divide the system into multiple layers: cloud, edge, and terminal. A collaborative mechanism enables both centralized management and distributed autonomy. This layered and collaborative intelligent system leverages the global optimization capabilities of the cloud while ensuring local real-time control requirements, overcoming the limitations of existing centralized or distributed single architectures.

[0058] High reliability through multi-link adaptive communication: This application supports multiple communication links such as 4G / 5G / WiFi / LPWAN, and the terminal and edge gateway can automatically select the optimal communication method based on link quality. Compared with existing single communication solutions, it can maintain uninterrupted data transmission in harsh environments, significantly improving network reliability. Redundancy backup is achieved through MPTCP link aggregation and fast switching technologies to avoid single points of failure and ensure that control commands and monitoring data can be delivered stably and with low latency.

[0059] Edge autonomous decision-making and containerized deployment: Utilizing container technologies (such as K3s) to deploy control applications on edge gateways enables rapid loading and updating of local logic. Edge containers possess self-healing capabilities, automatically restarting in the event of a failure to ensure continuous operation of the control system; the canary release mechanism ensures a smooth and safe upgrade process, reducing downtime risks. This IT-driven software architecture endows edge nodes with autonomous decision-making capabilities and high flexibility, significantly improving the maintenance efficiency and reliability of industrial field systems.

[0060] Cloud-edge collaborative closed-loop control: This application achieves a control closed loop through a collaborative approach of cloud-based algorithm training and edge-based decision-making. Complex cloud-based algorithm models are deployed to the edge for real-time operation, reducing both on-site response latency and cloud communication and computational load. The system can continuously iterate and optimize control strategies based on on-site feedback, enabling new energy power stations to operate efficiently and safely under various conditions, demonstrating adaptive optimization capabilities unmatched by existing technologies.

[0061] Comprehensive security protection capabilities: This application strictly complies with power safety standards such as IEC 62351, and incorporates end-to-end data encryption and authentication mechanisms to ensure secure and reliable communication links. It also provides multiple security defenses, including access control, network isolation, access auditing, and intrusion detection / prevention, enhancing the system's ability to resist network attacks and unauthorized access, and meeting the stringent security requirements of the power industry for control systems.

[0062] High availability and fault self-recovery: Through multi-network redundancy, QoS buffering, and intelligent switching strategies, this application achieves a high-availability design at the network layer, capable of adapting to complex and ever-changing communication environments. The edge node disconnection self-recovery mechanism ensures that the device quickly re-enters the network and synchronizes data after a failure, reducing downtime. Overall, the system possesses extremely high reliability and robustness, guaranteeing uninterrupted operation of power monitoring and control services.

[0063] Excellent scalability and compatibility: This application provides open programmable interfaces and industrial protocol adaptations, allowing users to flexibly customize business logic as needed and easily integrate third-party devices and applications. Compared to traditional closed systems, this system boasts strong compatibility and convenient expansion, adapting to future technology upgrades and functional expansions, demonstrating excellent versatility and long-term application value.

[0064] Example 3 Based on the same inventive concept, this application also provides a CPS-based multi-communication link adaptive edge decision-making device, such as... Figure 3 ,include: The acquisition unit is used to acquire link status parameters of various communication links of the access edge device at preset time intervals; The performance analysis unit is used to calculate the comprehensive score and average link utilization of each communication link based on the link status parameters using a preset weighted scoring method, and obtain the performance evaluation results of each communication link. The adaptive decision unit is used to select the determined target link set and backup link configuration as the decision result by using an adaptive decision algorithm executed by the edge device based on the performance evaluation results. The execution unit is used to transmit business data using edge devices based on the decision results; The adaptive decision-making algorithm is deployed from the cloud to edge devices.

[0065] Optionally, the step of using an adaptive decision-making algorithm executed by the edge device based on the performance evaluation results to select the determined target link set and backup link configuration as the decision result includes: Based on the performance evaluation results, and based on the preset target number of links N and the preset threshold, the edge devices select the top N links with the comprehensive score as the determined target link set, and select them as backup links according to the attributes of the communication links themselves. When the average link utilization exceeds the preset threshold, additional links are activated. When the overall score of the backup link exceeds that of the target link and the duration is not less than the preset hysteresis time, the backup link will be switched to the new target link, and the target link with the declining overall score will be downgraded to the backup link. The preset threshold and hysteresis time are remotely issued by the cloud platform and take effect in real time.

[0066] Optionally, the step of using edge devices to transmit business data based on the decision results includes: The data type of the business data is identified based on the business data level identifier or port number; the business data is divided into control instruction data and monitoring data. Based on the decision results, for control command data, edge devices send control command data completely through a single link and synchronize redundant copies through a backup link. For monitoring data, edge devices use a multi-path transmission mechanism to send monitoring data in fragments in parallel. When a failure is detected in the target link being used, the link status parameters are obtained and a backup link is selected as the target link. When the original target link recovers from the fault and its overall score is higher than that of the currently used target link, and the preset hysteresis time is continuously maintained, the original target link will be restored to the target link or the link status parameters will be obtained to select a new target link.

[0067] Optionally, the step of using edge devices to transmit business data based on the decision results further includes: If all target links are unavailable, the system enters offline caching mode to cache the service data to be transmitted; when any communication link becomes available again, the system exits offline caching mode and retransmits the cached service data in batches. In the offline caching mode, at least the most recent control commands and monitoring data for a preset duration are cached. When the cache is full, the earliest data is discarded using a first-in-first-out strategy.

[0068] Optionally, the link state parameters include one or more of the following: Signal strength, bandwidth limit, current bandwidth usage, round-trip time, packet loss rate, or stability index.

[0069] Optionally, the calculation formula for the preset weighted scoring method is:

[0070] in, The overall score for the communication link is represented by RSSI (Signal Strength Index), L (Round-Trip Time), PL (Package Loss Rate), and SI (Stability Index). The weighting coefficients of the RSSI (Signal Strength Index) are used to represent the signal strength index. This represents the weighting coefficient of the round-trip delay L index. This represents the weighting coefficient of the packet loss rate (PL) metric. This represents the weighting coefficient of the Stability Index (SI). This represents the normalization / mapping function for RSSI. This represents a penalized normalization / mapping function for L. This represents the penalized normalization / mapping function for PL.

[0071] Example 4 Based on the same inventive concept, this application also provides a preferred embodiment of a CPS-based multi-communication link adaptive edge decision-making device, such as... Figure 4 ,include: The structure and functions of each component module of the system are as follows: Multi-link communication interface module: A collection of hardware interfaces for connecting multiple heterogeneous communication networks, including but not limited to cellular mobile communication modules (such as 4G / 5G modules), Wi-Fi wireless transceiver modules, Ethernet ports, and LPWAN modules (such as LoRa or NB-IoT transceivers). This module provides the system with physical layer and data link layer access capabilities, enabling edge devices to simultaneously access different networks in the CPS environment. Each interface module can provide real-time link status information output interfaces, such as signal strength indicators and telecom operator service quality indicators.

[0072] Link Status Acquisition Module: This functional unit acquires link status parameters from the aforementioned communication interface modules. It periodically calls the drivers or management APIs of each interface to collect data representing link performance, such as signal quality, bandwidth capacity, latency, and bit error rate, and preprocesses some of the raw data (e.g., noise filtering and normalization). The acquisition module then summarizes the acquired multi-link status parameters and provides them to the system memory or message bus for access by subsequent processing modules. Through this module, the operational status of various communication links can be monitored synchronously, providing timely and accurate input for system decision-making.

[0073] Link Performance Analysis Module: This module calculates and analyzes collected link status parameters to obtain standardized link performance metrics. It includes an algorithm library capable of performing operations such as bandwidth estimation, link utilization calculation, and comprehensive quality scoring. For example, the analysis module can calculate the average link utilization based on the ratio of the transmission throughput of each link to its maximum bandwidth over the past few seconds, and calculate the link quality score (Q-value) according to a predefined formula or model. The output of this module is a set of comparable link performance metrics corresponding to each communication link available in the system at the current moment. The performance analysis module can be implemented as a software thread or firmware logic circuit in an edge device; its operating frequency and algorithm complexity can be configured as needed to fully evaluate link status while ensuring real-time performance.

[0074] The adaptive decision-making module is the core control unit of the system, responsible for making optimal link selection decisions based on link performance indicators. This module implements an adaptive policy algorithm. Its inputs are the link performance data provided by the analysis module and preset decision policy parameters (such as the upper limit of the number of target links, performance thresholds, etc.). The output is the decision result, including the determined set of target communication links and the usage allocation of each target link (e.g., primary / backup). Internally, the decision-making module can use a rule engine or an embedded artificial intelligence model to execute the decision logic and can interact with remote management commands to obtain policy updates as needed. This module runs on the processor of the edge device and has certain real-time requirements, ensuring rapid policy updates after changes in link status. Through the adaptive decision-making module, the system achieves autonomous optimization of communication resource allocation locally, thereby enabling CPS edge nodes to have intelligent network scheduling capabilities. The decision policy is stored in the form of an editable configuration file, which can be remotely modified by the cloud platform during operation and take effect immediately.

[0075] The Communication Control and Execution Module controls the actual data transmission process based on the results output by the Adaptive Decision Module. This module interacts directly with the Multi-Link Communication Interface Module, responsible for sending application data according to the links and methods specified in the decision. Its functions include: configuring the network protocol stack (e.g., enabling or disabling a socket connection on a link), triggering the establishment of multi-path transmission protocol sessions such as MPTCP, managing the mapping and load balancing of data streams across different links, and monitoring link anomalies during transmission. Once the decision module determines the target link set, the communication control module activates the corresponding interface and executes transmission according to the set's contents. For example, when the target set contains two links, this module will initiate a data splitting mechanism, distributing traffic proportionally across the two links; if the target set has only a single link, all data will be transmitted through that link, while other links will be closed or left idle to save resources. When the communication control module detects a failure in a currently used link, it can immediately notify the decision module to re-route and temporarily switch the data stream to other available links to ensure uninterrupted communication. In terms of hardware implementation, this module is typically implemented by the network processor of the edge device or the network stack of the operating system kernel, or it can be accomplished through a dedicated communication scheduling chip. The communication control and execution module enables multi-link parallel transmission and substream reassembly by enabling the MPTCP function in the operating system kernel.

[0076] The aforementioned modules work in tandem: the communication interface module provides basic connectivity, the status acquisition module and performance analysis module jointly realize link status perception and evaluation, the adaptive decision-making module makes intelligent choices based on this, and the communication execution module is responsible for the final data transmission action. The entire system revolves around a closed-loop design of "perception-analysis-decision-execution," aligning with the real-time perception and control architecture requirements of the CPS system. In applications, this system can be integrated into devices such as edge computing gateways, vehicle terminals, and drone controllers, enabling them to have adaptive communication capabilities in multi-network environments. Through the close cooperation of each module, the system can achieve dynamic optimal utilization of multiple communication links, meeting the requirements of communication reliability, low latency, and flexibility in CPS scenarios.

[0077] The specific steps include: System Hardware Deployment: The system in this embodiment is integrated into an industrial edge gateway. This gateway is equipped with multiple network interface hardware components: including a gigabit Ethernet port, a Wi-Fi wireless network card (supporting the 802.11ac protocol), a 4G LTE cellular communication module, and an LPWAN communication module supporting LoRaWAN. The gateway's processor is an embedded quad-core ARM processor running a real-time operating system, with sufficient built-in memory for data buffering and algorithm execution. These hardware components together constitute a multi-link communication interface module, laying the foundation for the system's connection to external networks. Specifically, the Ethernet port connects to the factory's wired network, the Wi-Fi network card connects to the local wireless network (covering mobile devices), the 4G module accesses the public mobile communication network to connect to a remote cloud server, and the LoRaWAN module is used to collect data from remotely distributed battery-powered sensors. This interface configuration enables the gateway to simultaneously achieve high-speed, local, remote, and wide-area low-power communication capabilities.

[0078] Module functionality implementation: The link status acquisition module is implemented as follows: A dedicated link monitoring daemon runs within the gateway's operating system, acting as the link status acquisition module. This daemon periodically reads the status parameters of each network interface through the network interface statistics and driver interface provided by the kernel. For example, it reads the Ethernet port link speed and error frame count every 2 seconds; calls the Wi-Fi card driver API to obtain the current RSSI and connection rate; queries signal strength (such as RSSI value) and base station cell congestion information through the AT commands of the 4G module; and reads the signal-to-noise ratio of the most recent downlink through the serial port interface of the LoRaWAN module. The collected data is then aggregated and written to shared memory or a message queue by the daemon. Through this implementation, the link status acquisition module can continuously and stably provide real-time updated link status parameters to meet the needs of subsequent processing.

[0079] The link performance analysis module is implemented as follows: A separate analysis thread or service within the gateway's software system processes and analyzes data from the status acquisition module. This module first performs necessary unit conversions and normalization on the data from different links. For example, it converts Wi-Fi's RSSI to a quality percentage and maps 4G's RSRP to a signal level. Then, it weights and summarizes the various indicators according to their importance. Next, the analysis module calculates the performance indicators for each link. In this example, the performance indicators include an estimated available bandwidth and a link quality score Q. For Ethernet ports, since they connect to internal wired networks, the bandwidth is typically fixed and the packet loss rate is extremely low, so the highest score Q≈100 can be directly assigned to its gigabit speed and zero packet loss. For Wi-Fi links, the available bandwidth is estimated based on RSSI and channel noise (e.g., RSSI -55dBm corresponds to approximately 50Mbps of available bandwidth) and combined with the current latency of 10ms to calculate a score Q≈85. For 4G links, the available bandwidth is estimated at 10Mbps based on RSRP -95dBm, with a latency of approximately 50ms, resulting in a score Q≈70. LoRa links have extremely low bandwidth (tens of kbps) and high latency (on the order of 1 second), with a score Q≈40, but their unique value lies in ultra-long-distance coverage and low power consumption. This analysis module packages these results into a performance index data structure for the decision-making module to read. The analysis module is implemented in C to improve efficiency and provides an external query interface, allowing the decision-making module to obtain the latest analysis results whenever needed. This link performance analysis module effectively transforms the complex and diverse link states into unified quantitative indicators, providing an objective basis for intelligent decision-making.

[0080] Adaptive Decision Module Implementation: The gateway's software control layer includes a core decision engine (which can be viewed as a decision module). In this implementation, the decision engine is implemented as a rule-based inference system according to actual needs, and its rule policies can be defined and dynamically adjusted by configuration files. The initial rules are set as follows: prioritize using the link with the highest score; if the score of the second-highest-scoring link is not lower than 80% and the current utilization rate of the main link exceeds 70%, then enable the second-highest-scoring link to share the load; maintain a low-speed link as a redundancy backup. The decision module checks link performance metrics every second. In the example data above, Ethernet, with a score of 100, is found to be the highest and most usable, so it is selected as the primary link. Wi-Fi, with a score of 85, meets the requirement of at least 80% utilization, and when Ethernet utilization is high, it is added for shared transmission (assuming high video traffic within the factory and Ethernet bandwidth nearing saturation, Wi-Fi is used to offload some video data). 4G, with a score of 70, is below the 80% threshold and is not currently used as the active link, but can quickly take over when the primary link becomes unavailable; therefore, 4G is marked as a backup link. LoRa, with a score far too low, is only suitable for sending low-speed monitoring data and is therefore continuously used as the device heartbeat reporting channel, not participating in high-speed data transmission. This decision result is sent to the communication control execution module. It should be noted that the decision module supports more complex strategies, such as using fuzzy logic or machine learning models to handle multi-dimensional inputs; however, to meet the clarity requirements of the patent application, an interpretable rule-based approach is used here. Through the implementation of the adaptive decision module, the system autonomously completes the intelligent selection and switching of multiple communication links locally, providing flexible and reliable communication guarantees for the CPS system.

[0081] The communication control execution module implements the following: In the gateway's network subsystem, communication control is accomplished through the operating system's multipath transmission support and the application-layer routing program. First, the gateway's operating system kernel enables MPTCP (Linux kernel 5.x supports MPTCP). This means that as long as the application creates an MPTCP-enabled socket connection, the underlying layer can automatically use multiple links to transmit data packets in parallel based on kernel path management. Based on this, the communication control module implements the "primary Ethernet + secondary Wi-Fi + backup 4G" scheme from the previous decision module: the application-layer transmission process establishes an MPTCP connection with the cloud server, initially using the Ethernet interface for transmission; then, according to the decision instruction, it calls the system's setsockopt interface to add the Wi-Fi interface as the second path for this MPTCP session. In this way, data flows are sent simultaneously via Ethernet and Wi-Fi, achieving load balancing. Simultaneously, the communication control module keeps the 4G interface in standby mode through the system network management interface: it does not join the main MPTCP session but maintains a TCP handshake channel with the cloud server. Once an Ethernet or Wi-Fi outage is detected, the 4G interface is immediately added to the MPTCP session to take over the traffic. For LoRa heartbeat data, the communication control module uses a separate thread to periodically report via the LoRaWAN gateway protocol, without affecting the main data stream. Throughout the process, the communication control module is also responsible for monitoring the actual traffic and errors of each interface, and promptly reporting any anomalies to the decision-making module (e.g., if the Wi-Fi packet error rate is high, it suggests that the decision-making module reduce its allocation). Through the above implementation, the module successfully coordinates the transmission of multiple links: main service data is dynamically balanced across high-speed links via MPTCP, critical small data is redundantly guaranteed through a multi-path mechanism, and ultra-long-distance monitoring data is reliably delivered via LPWAN. This multi-link collaborative transmission mechanism fully demonstrates the advantages of the system in maintaining communication stability and efficiency in complex network environments.

[0082] Module Collaboration and System Advantages: In this embodiment, the modules are decoupled in software architecture but work closely together through message queues, giving the system excellent scalability and stability. For example, if new communication links (such as satellite links) need to be supported in the future, only the corresponding interface modules need to be added and their indicators incorporated into the decision rules, without changing the overall architecture. System operation results show that when the wired network is good, the system mainly uses Ethernet transmission to ensure the lowest latency; when the wired network is disrupted or congested, Wi-Fi and 4G immediately intervene to offload traffic and maintain performance; and even if all high-speed links are briefly interrupted, the LoRaWAN link can still maintain the basic data link without interruption, enabling the CPS device to operate without downtime. Through edge-autonomous link selection, the system avoids the latency and single point of failure that may have been caused by the previous unified scheduling by the cloud, and can respond more agilely to changes in the field network conditions.

[0083] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A multi-communication link adaptive edge decision-making method based on CPS, characterized in that, include: The link status parameters of various communication links of the access edge device are obtained at preset time intervals. Based on the link status parameters, a preset weighted scoring method is used to calculate the comprehensive score and average link utilization of each communication link, and the performance evaluation results of each communication link are obtained. Based on the performance evaluation results, an adaptive decision-making algorithm is executed by the edge device to select the determined target link set and backup link configuration as the decision result; Based on the decision results, edge devices are used to transmit business data; The adaptive decision-making algorithm is deployed from the cloud to edge devices.

2. The method as described in claim 1, characterized in that, The step of using an adaptive decision-making algorithm executed by the edge device based on the performance evaluation results to select a determined set of target links and a backup link configuration as the decision result includes: Based on the performance evaluation results, and based on the preset target number of links N and the preset threshold, the edge devices select the top N links with the comprehensive score as the determined target link set, and select them as backup links according to the attributes of the communication links themselves. When the average link utilization exceeds the preset threshold, additional links are activated. When the overall score of the backup link exceeds that of the target link and the duration is not less than the preset hysteresis time, the backup link will be switched to the new target link, and the target link with the declining overall score will be downgraded to the backup link. The preset threshold and hysteresis time are remotely issued by the cloud platform and take effect in real time.

3. The method as described in claim 2, characterized in that, The step of transmitting business data using edge devices based on the decision results includes: The data type of the business data is identified based on the business data level identifier or port number; the business data is divided into control instruction data and monitoring data. Based on the decision results, for control command data, edge devices send control command data completely through a single link and synchronize redundant copies through a backup link. For monitoring data, edge devices use a multi-path transmission mechanism to send monitoring data in fragments in parallel. When a failure is detected in the target link being used, the link status parameters are obtained to select a backup link as the target link. When the original target link recovers from the fault and its overall score is higher than that of the currently used target link, and the preset hysteresis time is continuously maintained, the original target link will be restored to the target link or the link status parameters will be obtained to select a new target link.

4. A method as described in claim 3, characterized in that, The method of transmitting business data using edge devices based on decision results also includes: If all target links are unavailable, the system enters offline caching mode to cache the service data to be transmitted; when any communication link becomes available again, the system exits offline caching mode and retransmits the cached service data in batches. In the offline caching mode, at least the most recent control commands and monitoring data for a preset duration are cached. When the cache is full, the earliest data is discarded using a first-in-first-out strategy.

5. A method as described in claim 3, characterized in that, The link state parameters include one or more of the following: Signal strength, bandwidth limit, current bandwidth usage, round-trip time, packet loss rate, or stability index.

6. A method as described in claim 5, characterized in that, The calculation formula for the preset weighted scoring method is: in, The overall score for the communication link is represented by RSSI (Signal Strength Index), L (Round-Trip Time), PL (Package Loss Rate), and SI (Stability Index). The weighting coefficients of the RSSI (Signal Strength Index) are used to represent the signal strength index. This represents the weighting coefficient of the round-trip delay L index. This represents the weighting coefficient of the packet loss rate (PL) metric. This represents the weighting coefficient of the Stability Index (SI). This represents the normalization / mapping function for RSSI. This represents a penalized normalization / mapping function for L. This represents the penalized normalization / mapping function for PL.

7. A multi-communication link adaptive edge decision-making device based on CPS, characterized in that, include: The acquisition unit is used to acquire link status parameters of various communication links of the access edge device at preset time intervals; The performance analysis unit is used to calculate the comprehensive score and average link utilization of each communication link based on the link status parameters using a preset weighted scoring method, and obtain the performance evaluation results of each communication link. The adaptive decision unit is used to select the determined target link set and backup link configuration as the decision result by using the edge device to execute an adaptive decision algorithm based on the performance evaluation results. The execution unit is used to transmit business data using edge devices based on the decision results; The adaptive decision-making algorithm is deployed from the cloud to edge devices.

8. The apparatus as described in claim 7, characterized in that, The step of using an adaptive decision-making algorithm executed by the edge device based on the performance evaluation results to select a determined set of target links and a backup link configuration as the decision result includes: Based on the performance evaluation results, and based on the preset target number of links N and the preset threshold, the edge devices select the top N links with the comprehensive score as the determined target link set, and select them as backup links according to the attributes of the communication links themselves. When the average link utilization exceeds the preset threshold, additional links are activated. When the overall score of the backup link exceeds that of the target link and the duration is not less than the preset hysteresis time, the backup link will be switched to the new target link, and the target link with the declining overall score will be downgraded to the backup link. The preset threshold and hysteresis time are remotely issued by the cloud platform and take effect in real time.

9. The apparatus as described in claim 8, characterized in that, The step of transmitting business data using edge devices based on the decision results includes: The data type of the business data is identified based on the business data level identifier or port number; the business data is divided into control instruction data and monitoring data. Based on the decision results, for control command data, edge devices send control command data completely through a single link and synchronize redundant copies through a backup link. For monitoring data, edge devices use a multi-path transmission mechanism to send monitoring data in fragments in parallel. When a failure is detected in the target link being used, the link status parameters are obtained to select a backup link as the target link. When the original target link recovers from the fault and its overall score is higher than that of the currently used target link, and the preset hysteresis time is continuously maintained, the original target link will be restored to the target link or the link status parameters will be obtained to select a new target link.

10. The apparatus as described in claim 9, characterized in that, The method of transmitting business data using edge devices based on decision results also includes: If all target links are unavailable, the system enters offline caching mode to cache the service data to be transmitted; when any communication link becomes available again, the system exits offline caching mode and retransmits the cached service data in batches. In the offline caching mode, at least the most recent control commands and monitoring data for a preset duration are cached. When the cache is full, the earliest data is discarded using a first-in-first-out strategy.

11. The apparatus as claimed in claim 10, characterized in that, The link state parameters include one or more of the following: Signal strength, bandwidth limit, current bandwidth usage, round-trip time, packet loss rate, or stability index.

12. The apparatus as claimed in claim 11, characterized in that, The calculation formula for the preset weighted scoring method is: in, The overall score for the communication link is represented by RSSI (Signal Strength Index), L (Round-Trip Time), PL (Package Loss Rate), and SI (Stability Index). The weighting coefficients of the RSSI (Signal Strength Index) are used to represent the signal strength index. This represents the weighting coefficient of the round-trip delay L index. This represents the weighting coefficient of the packet loss rate (PL) metric. This represents the weighting coefficient of the Stability Index (SI). This represents the normalization / mapping function for RSSI. This represents a penalized normalization / mapping function for L. This represents the penalized normalization / mapping function for PL.

Images