Soft bus ad hoc network method and device based on intelligent prediction and storage medium

Abstract

The invention provides a soft bus ad hoc network method and device based on intelligent prediction and a storage medium, and the method comprises the steps: a group leader node collects historical record data of a group member node, carries out the preprocessing of the historical record data, obtains a standard data set, and sends the standard data set to a cloud; the cloud terminal trains a time sequence prediction model based on the standard data set and a preset business process and issues the time sequence prediction model to the group leader node; the group leader node responds to the target task, calls a time sequence prediction model to generate corresponding prediction link information, and converts the prediction link information into a connection instruction containing a link priority; and the group leader node schedules the group member node to pre-establish the soft bus connection according to the link priority in the connection instruction before the actual data interaction demand of the target task occurs. A prediction model is trained in combination with a standard data set and a business process, prediction is carried out, pre-scheduling is carried out before an actual data interaction demand occurs, and connection is established, so that waiting during connection of a business flow and a data flow is reduced.

Description

Technical Field

[0001] This invention belongs to the field of automated production technology, and in particular relates to a method, device and storage medium for a soft bus self-organizing network based on intelligent prediction. Background Technology

[0002] In industrial automated production systems, a large number of devices need to work collaboratively. To achieve control command transmission and data exchange between devices, they are typically connected via various communication methods such as soft bus, Ethernet, and Wi-Fi. However, because many low-power devices cannot be online for extended periods, an on-demand connection approach is used to balance production efficiency and cost. This approach has the following drawbacks:

[0003] Currently, soft bus networking requires manual triggering, has a low degree of automation, and is difficult to meet the needs of large-scale networking scenarios. Furthermore, the networking relationship cannot adapt quickly to changes in the production process, resulting in low automation and production efficiency. Summary of the Invention

[0004] In view of this, the present invention provides a method, system, device and storage medium for self-organizing soft bus based on intelligent prediction, in order to solve the problems of low automation and inability to adapt to changes in production processes in the prior art when manually triggering soft bus networking.

[0005] A first aspect of the present invention provides a soft bus self-organizing network method based on intelligent prediction, applied to a production system, the production system including a cloud, at least one group leader node, and multiple group member nodes, each group leader node being connected to the cloud and at least one group member node respectively; the method includes:

[0006] The group leader node collects historical data from the group member nodes, preprocesses the historical data to obtain a standard dataset, and sends it to the cloud. The historical data includes historical running data, real-time status data, and network performance data.

[0007] The cloud platform trains a time-series prediction model based on the standard dataset and preset business processes, and then distributes the trained time-series prediction model to the group leader node; the time-series prediction model is used to output predicted link information according to the task.

[0008] The group leader node responds to the target task by calling the time series prediction model to generate corresponding predicted link information and converting the predicted link information into connection instructions containing link priorities.

[0009] Before the actual data interaction requirement of the target task occurs, the group leader node schedules the group member nodes to pre-establish a soft bus connection according to the link priority in the connection instruction.

[0010] A second aspect of the present invention provides a production system, including a cloud, at least one group leader node and a plurality of group member nodes, wherein each group leader node is connected to the cloud and at least one group member node respectively;

[0011] The group leader node is used to collect historical data from the group member nodes, preprocess the historical data to obtain a standard dataset, and send it to the cloud. The historical data includes historical running data, real-time status data, and network performance data.

[0012] The cloud is used to train a time-series prediction model based on the standard dataset and preset business processes, and to distribute the trained time-series prediction model to the group leader node; the time-series prediction model is used to output predicted link information according to the task.

[0013] The group leader node is also used to respond to the target task by calling the time series prediction model to generate corresponding prediction link information and converting the prediction link information into a connection instruction containing link priority.

[0014] The group leader node is also used to schedule the group member nodes to pre-establish a soft bus connection based on the link priority in the connection instruction before the actual data interaction requirement of the target task occurs.

[0015] A third aspect of the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the intelligent prediction-based soft bus self-organizing network method as described in the first aspect above.

[0016] A fourth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the intelligent prediction-based soft bus self-organizing network method as described in the first aspect above.

[0017] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0018] By integrating historical patterns, real-time status, and network quality data, and training a predictive model based on business processes, the link prediction results are both accurate and reasonable. By pre-scheduling and establishing connections based on the priority in the predicted instructions before actual data interaction needs occur, communication establishment delays during task execution are completely eliminated, reducing waiting time when business flows and data flows are connected. Simultaneously, the priority-based scheduling strategy ensures that critical control links are established first when resources are limited, thereby improving overall system response efficiency and resource utilization while guaranteeing the determinism and stability of core production processes, and increasing automation and production efficiency. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of a soft bus self-organizing network method based on intelligent prediction provided in an embodiment of the present invention;

[0021] Figure 2 This is a flowchart of a soft bus self-organizing network method based on intelligent prediction provided in an embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of data interaction between different entities provided in an embodiment of the present invention;

[0023] Figure 4 This is a schematic diagram of a production system provided in an embodiment of the present invention;

[0024] Figure 5 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the present invention. However, those skilled in the art will recognize that the present application may be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted to avoid unnecessary detail that could obscure the description of the present application.

[0026] First, it is necessary to clarify the definition of the soft bus in this invention:

[0027] Softbus is a core concept in industrial automation and distributed systems. It's a software-level communication architecture that acts like a virtual "data highway," enabling different devices, applications, or services with different functions to easily, efficiently, and in a standardized manner discover, connect, and exchange data, much like hardware components plugged into a physical motherboard. The connections made by a softbus include software processes, services, applications, or heterogeneous devices (through software proxies).

[0028] The technical solution of the present invention will be illustrated below through specific embodiments.

[0029] Reference Figure 1 The diagram illustrates a method for self-organizing a soft bus based on intelligent prediction, provided by an embodiment of the present invention. This method can be executed by a production system, which can be implemented in hardware and / or software and can be configured in an electronic device.

[0030] The production system comprises a cloud platform, at least one team leader node, and multiple team member nodes. Each team leader node is connected to both the cloud platform and at least one team member node. The cloud platform is responsible for core computing and model training, the team leader node is responsible for local coordination, data aggregation, and instruction scheduling, and the team member nodes are responsible for specific production tasks (terminal devices). Specifically, the cloud platform, team leader nodes, and team member nodes can be connected via a local area network.

[0031] like Figure 1 As shown, the soft bus self-organizing network method based on intelligent prediction includes:

[0032] S101. The group leader node collects historical data from the group member nodes, preprocesses the historical data to obtain a standard dataset, and sends it to the cloud.

[0033] Specifically, within a production workshop, at least one leader node (such as a server) and multiple member nodes (such as industrial equipment like robotic arms, AGVs, and welding robots) are deployed. The leader node periodically collects data from all the member nodes it is connected to and stores it as historical records.

[0034] Historical records include historical operational data, real-time status data, and network performance data.

[0035] Historical operational data refers to the interaction logs generated by member nodes when performing various production tasks in the past. This includes historical connection information and active period information. Historical connection information reflects the network topology between devices, while active period information reflects the connection status between member nodes at different time windows. Optionally, the interaction logs may include: which other nodes (reflecting connection relationships) established connections with at what time (reflecting active periods), the duration of the connections, and the type of data transmitted. This data reflects the collaboration patterns and time regularities between devices.

[0036] Real-time status data includes instantaneous performance metrics of member nodes at the time of collection, such as CPU utilization, memory usage, and current task queue length. This data reflects the current health and available capacity of the devices. In addition, real-time status data also includes environmental data, which reflects the impact of the environment on device (member node) interactions. Real-time data helps determine whether a device is capable of handling networking tasks. If a device is operating under high load, joining the network may cause it to lag, leading to service interruptions. Introducing this data helps avoid the impact of the environment on the network; when a device is under high load, a backup device can be used for networking to ensure link stability.

[0037] Network performance data refers to real-time quality indicators of network links, including round-trip latency between member nodes, bandwidth utilization, and packet loss rate. These indicators reflect the transmission capacity of the link. Different services have different network performance requirements; introducing this data ensures that the link, once established, meets the network needs of the corresponding service, reducing packet loss and network congestion, and minimizing performance waste.

[0038] The team leader node locally aggregates and preprocesses the aforementioned raw data, including: cleaning (removing obvious outliers), imputation (filling missing values ​​with the mean or previous values), normalization (scaling data of different dimensions to a uniform range), and feature extraction (e.g., extracting abstract features such as collaboration density and active periods from historical connection logs). After preprocessing, a standardized dataset with a uniform format and reliable quality is generated and uploaded in batches to the cloud server.

[0039] S102. The cloud-based system trains a time-series prediction model based on a standard dataset and preset business processes, and then distributes the trained time-series prediction model to the team leader node.

[0040] The time-series prediction model is used to predict link information based on task output.

[0041] After receiving standardized datasets from various production areas, the cloud server combines these datasets with the pre-defined business processes of each area (e.g., a fixed sequence of processes such as stamping, welding, painting, and inspection, and the dependencies of each process on equipment) to train the model. The training objective of the model is to learn the mapping from task features to the optimal connection scheme. After training, the cloud performs lightweight processing on the model to adapt to the computing capabilities of edge devices, and then distributes the model file to the corresponding team leader node.

[0042] S103. The group leader node responds to the target task by calling the time series prediction model to generate the corresponding predicted link information and converts the predicted link information into a connection instruction containing link priority.

[0043] When the production management system issues a new target task (such as "start the welding operation of batch A") to the team leader node, the team leader node immediately initiates the forecasting process:

[0044] Combining real-time status data and network performance data collected at the current moment, the locally deployed time-series prediction model is invoked. The model integrates historical patterns, current system status, and business processes to output predicted link information for the target task. The leader node converts this predicted link into executable connection instructions.

[0045] Specifically, converting the predicted link information into a connection instruction containing link priority includes:

[0046] The group leader node generates a structured connection instruction for each link. The fields in the connection instruction include at least: link identifier, initiating node identifier, target node identifier, communication protocol, authentication method, planned start time, predicted connection duration, link priority, and the identifier of the preceding link it depends on. All generated connection instructions are sorted according to link priority, and instructions with the same link priority are sorted according to planned start time.

[0047] Each instruction is a structured data object, containing at least:

[0048] Link identifier: also known as link ID, is a unique identifier.

[0049] Initiator / Target Node Identifier: Specifies which two member nodes need to be connected.

[0050] Communication protocols and authentication methods: such as MQTT over TLS.

[0051] Planned start time and predicted connection duration: Specifies the connection window.

[0052] Link priority: For example, mark the link between the welding robot and the central controller as P0.

[0053] Pre-link identifier: Indicates that this link can only be started after another link has been established.

[0054] The link priorities are represented by P0-P2, where P0 represents critical links such as the controller-actuator link, P1 represents supporting links such as auxiliary data acquisition and equipment status monitoring links, and P2 represents non-critical links such as log upload links.

[0055] Finally, all connection commands are sorted from highest to lowest link priority, and those with the same priority are sorted from earliest to latest scheduled start time, forming a queue of commands to be executed.

[0056] S104. Before the actual data interaction requirement of the target task occurs, the group leader node schedules the group member nodes to establish a soft bus connection in advance according to the link priority in the connection instruction.

[0057] When the leader node receives a connection command, it does not wait for the production task to reach the stage requiring data transmission, but immediately begins scheduling to eliminate connection establishment delays: it schedules relevant member nodes according to the link priority in the connection command, pre-establishing soft bus connections. When the production task reaches the stage requiring data interaction, the communication link is already established, reducing waiting time.

[0058] The intelligent prediction-based soft bus self-organizing network method of this invention integrates three types of data: historical patterns, real-time status, and network quality. By training a prediction model in conjunction with business processes, the method ensures that link prediction results are both accurate and reasonable. By pre-scheduling and establishing connections based on the priority in the prediction instructions before actual data interaction needs occur, the communication establishment delay during task execution is completely eliminated, reducing waiting time when business flows and data flows are connected. Simultaneously, the priority-based scheduling strategy ensures that critical control links are established first when resources are limited, thereby improving the overall system response efficiency and resource utilization while guaranteeing the determinism and stability of core production processes.

[0059] In an optional embodiment, the standard dataset includes comprehensive features from multiple tasks, and the training process of the time-series prediction model includes:

[0060] For each task's comprehensive features in the standard dataset, corresponding standard link information is determined. This standard link information includes a standard list of current task links and a list of subsequent associated task links. The task's comprehensive features and a preset business process are input into the time-series prediction model to output predicted link information that conforms to the preset business process. This predicted link information includes a predicted list of current task links and a list of subsequent associated task links. The model loss value is determined based on the difference between the predicted link information and the standard link information. If the model loss value is greater than a preset loss threshold, the parameters of the time-series prediction model are updated based on the model loss value, and training continues iteratively. If the model loss value is less than or equal to the preset loss threshold, the time-series prediction model training is considered complete.

[0061] Task comprehensive features are extracted from historical data and are used to describe a complete feature vector of a historical production task. It integrates historical collaboration patterns, real-time device status snapshots, and network status snapshots prior to the task's execution.

[0062] Standard link information is extracted from interaction logs (historical operational data) and represents connection schemes that were actually successfully established and verified as valid in that historical task. It includes the actual connection node pairs (topology), the priority determined based on business impact, and the actual duration.

[0063] Regarding the source of standard link information, the system has recorded the complete execution process through the monitoring component when historical tasks were actually executed, specifically including the following logs:

[0064] Network connection logs record which specific device IPs / IDs established socket connections or data streams during task execution, as well as the start and end timestamps of the connections (from which the actual connection duration can be calculated).

[0065] Controller command log: Records which devices the host computer or scheduling system sent control commands to, thus identifying the core execution device of the task.

[0066] Business system logs: record when a production task was started and ended.

[0067] For example, for Task ID-123: Consumables Production, by correlating and mining these logs, it can be automatically reconstructed that during the execution of the consumables production task, a connection was established between Device A (robotic arm) and Device B (vision sensor) for 15 seconds, and a connection was established between Device A and Device C (PLC controller) for 30 seconds. These are the device list and connection duration in the standard link information, and the standard link information corresponding to each comprehensive task characteristic can be obtained based on historical operation data.

[0068] During model training, the model is repeatedly fed with the combined features of these tasks and attempts to output its own predictions (predicted link information). The model's goal is to make its predictions infinitely close to the corresponding standard link information. Specifically, after each prediction, the system quantifies the difference between the prediction result and the standard answer, which is the model loss value. It comprehensively measures the accuracy of the predicted topology, the correctness of the priority ranking, and the error of the duration prediction. When the model's average prediction loss on a large number of samples decreases below a preset threshold, it indicates that its predictions are accurate and stable enough. At this point, training is complete, and the model can be put into practical use.

[0069] The model's input includes not only data features but also pre-defined business processes, which represent business logic such as process dependencies and equipment functional constraints. This is equivalent to providing the model with a production rule manual when making predictions, forcing its predictions to conform to business logic such as process dependencies and equipment functional constraints. This ensures that the output prediction chain information is engineering-feasible and logically sound, rather than predictions based solely on data correlation.

[0070] In existing technologies, when business processes change, network relationships cannot adapt quickly. However, in this embodiment, the learning of the time series prediction model relies on both data-driven (finding patterns from history) and knowledge-driven (adhering to business process rules), making its predictions both flexible and reliable.

[0071] Furthermore, when training the model in this embodiment, the standard link information includes a standard current task link list and a subsequent associated task link list, and the predicted link information includes a predicted current task link list and a subsequent associated task link list; the subsequent associated task link list is a list of associated task links related to the links in the current task link list.

[0072] The current task link list contains all the link information (topology, priority, duration) that task A actually establishes during its execution.

[0073] The list of subsequent related task links is a summary of all links established by one or more subsequent tasks (such as task B and task C) that actually occur immediately after task A is completed within a preset time window (e.g., the next 30 minutes).

[0074] For example, suppose the historical logs show:

[0075] 09:00: The task "Visual Inspection of Welds (Task_Inspection)" begins.

[0076] 09:00-09:05: This task actually established the link between the camera and the image processing server.

[0077] 09:06: The system automatically started the subsequent task "Generate Detection Report (Task_Report)".

[0078] 09:06-09:10: Task_ reports that the link between the image processing server and the database has actually been established.

[0079] Constructed training samples:

[0080] Input (simulating the state at 09:00): Characteristics of the "Weld Visual Inspection" task + system state at 09:00 + business process rules (stipulating that "a report must be generated after inspection").

[0081] Output labels (real history):

[0082] Current task chain list: Camera - Image Processing Server;

[0083] List of subsequent related tasks: Image processing server - Database.

[0084] Through training with a massive number of such samples, the model will internalize two patterns:

[0085] In-task patterns: which devices are needed and how to connect for a specific type of task, i.e., learning to predict the current task link list.

[0086] Inter-task patterns: After completing a certain type of task, based on business processes and time patterns, what type of task is likely to be launched next, and thus what links need to be prepared in advance, i.e., learning and predicting the list of subsequent related task links.

[0087] In an optional embodiment, the standard link information includes standard link topology, standard link priority, and standard link connection duration; the predicted link information includes predicted link topology, predicted link priority, and predicted link connection duration. The model loss value is determined based on the difference between the predicted link information and the standard link information, including:

[0088] Calculate the topology overlap rate based on the standard link topology and the predicted link topology; calculate the priority ranking error based on the standard link priority and the predicted link priority; calculate the connection duration prediction error based on the standard link connection duration and the predicted link connection duration; calculate the model loss value based on the topology overlap rate, priority ranking error, and connection duration prediction error.

[0089] It's important to note that a task linking list is an ordered sequence that lists the steps or nodes required to complete a task. For example, a process might be [Task A - Task B - Task C]. The task linking list emphasizes the order and path. Link topology, on the other hand, refers to the complete graph structure of the entire task network, consisting of all nodes (tasks) and edges (connections). It includes not only a single path but also all possible paths, branches, merging points, and all connections between nodes. Link topology emphasizes the overall structure and connectivity.

[0090] The topology overlap rate measures whether the model accurately predicts the set of device pairs that need to establish connections. This is the most basic accuracy metric for link prediction, ensuring that no critical connections are missed (high recall) and that unnecessary connections are not incorrectly predicted (high precision).

[0091] Precision = TP / (TP + FP);

[0092] Recall rate = TP / (TP + FN);

[0093] TP indicates a true positive (predicted connection, and the connection actually occurs); TN indicates a true negative (predicted no connection, and the connection also does not occur); FP indicates a false positive (predicted connection, but the connection does not occur); FN indicates a false negative (predicted no connection, but the connection actually occurs).

[0094] Connection duration error = (predicted duration - actual duration) / actual duration.

[0095] Priority ranking error measures whether the model accurately understands the order of business importance among different links. This ensures that, during resource scheduling, the critical links (P0) predicted by the model are consistent with the actual business urgency, which is crucial for ensuring production certainty.

[0096] Connection duration prediction error measures whether the model accurately estimates the time resources required for each link. Accurate duration prediction is fundamental for efficient resource reservation and intelligent lifecycle management, preventing premature resource release or timeout.

[0097] In each training iteration, the model's predicted information is quantitatively compared with the standard answer across the three dimensions mentioned above. The calculated topological overlap rate (generally, the higher the better), priority error, and duration error (generally, the lower the better) are combined using a weighted formula to obtain a total model loss value. This total loss value serves as the sole feedback signal, guiding optimization algorithms (such as gradient descent) to adjust model parameters. To reduce the total loss, the model must simultaneously improve its performance across the three dimensions. The weights of each indicator reflect the business focus and can be set according to actual needs; this embodiment does not impose any restrictions on this.

[0098] This embodiment incorporates priority and duration—two core business attributes—into the optimization objectives, allowing the model output to be further optimized beyond connectivity, better meeting the needs of production scheduling and resource management. It also avoids the model falling into overfitting based on a single metric. For example, optimizing only the topology overlap rate might lead the model to blindly predict a large number of long-term connections, wasting resources; thus, the embodiment helps the model find a balance between accuracy, importance, and economy.

[0099] In an optional embodiment, scheduling member nodes to pre-establish soft bus connections based on link priority in the connection command includes:

[0100] The leader node sends resource requests to the relevant member nodes sequentially according to the link priority in the connection command; the member nodes verify whether their local resources meet the requirements for establishing a connection based on the received resource requests, and report the request results back to the leader node; if the resource requests of all member nodes in a link are successfully verified, the leader node initiates the soft bus connection establishment process between the relevant member nodes; if the verification of any resource request fails, the leader node re-initiates the resource request to the backup node and triggers the soft bus connection establishment process.

[0101] The backup node can be specified in the connection command or the predicted link information.

[0102] The long node processes instructions sequentially according to priority and sends resource requests as probes to the initiating node and the target node (both ends of a link) specified in each instruction. This ensures that resource requests for high-priority links are processed first, while also checking the status of nodes at both ends of the link, resulting in higher efficiency.

[0103] Each member node independently verifies its local resources (CPU, memory, ports, specific hardware, etc.) and returns a "yes / no" boolean result to the group leader node. The group leader node does not need to maintain the precise resource status of all nodes in real time, reducing the complexity of the central node.

[0104] The leader node issues the final link establishment command only when both ends of a link report success, at which point the nodes begin the handshake, authentication, and other establishment processes. This ensures that the connection is established with sufficient resources. If any node fails, a switchover is immediately triggered. The leader node does not retry or wait; instead, it directly sends the same resource requests to the backup nodes based on the pre-planned list of backup nodes.

[0105] In an optional embodiment, it further includes:

[0106] The leader node determines the core link based on the link priority; after the core link is established, it initiates a business startup request to trigger business interactions between member nodes.

[0107] Typically, business rules pre-define the highest priority (e.g., P0 level) links as core links. These links are directly related to the safety of production tasks, core control flow, or necessary real-time data flow (e.g., controller-actuator, safety PLC-emergency stop device). The team leader node filters all links that need to be established based on the link priority field defined in the connection instructions, and continuously monitors the establishment status of all links marked as core links. Only when all these core links are reported as successfully established does the team leader node initiate a business startup request to the upper-level production management system or business module. Upon approval of the request or as a direct trigger signal, the team leader node immediately sends a command to relevant team member nodes to begin business interaction. Team member nodes then begin transmitting production data or control commands through the established high-quality links, ensuring that business operations do not start before critical communication paths are ready, thus avoiding production interruptions and data loss due to network problems.

[0108] In an optional embodiment, the connection instruction includes the predicted connection duration of the link, and the method further includes:

[0109] After a business interaction is triggered, the team member node calculates the remaining connection duration based on the link's elapsed runtime and the predicted connection duration; it then determines whether the remaining connection duration is less than a preset duration threshold; if so, it acquires execution data related to the business task at a preset period and uploads it to the team leader node.

[0110] In this embodiment, after the business interaction begins, a duration-based monitoring loop is initiated. The predicted connection duration is the total time estimated by the model based on task characteristics, reserved to ensure business completion. The remaining duration is the difference between the predicted connection duration and the actual running time. The system presets an early warning threshold (e.g., 10% of the predicted connection duration or a fixed value). When the remaining connection duration is less than the preset threshold, it indicates that the connection is about to enter a critical phase. Once the critical phase is entered, the system increases the monitoring intensity from basic heartbeat to enhanced monitoring mode. Team member nodes begin collecting and reporting execution data related to the business task at shorter intervals (preset intervals).

[0111] The execution data reported near the end of the connection period (such as task completion percentage, processing speed, and buffer status) serves as the direct basis for the team leader node to determine whether the task can be completed normally before the connection times out. Furthermore, during most of the connection's stable period, the system may only maintain a low-cost heartbeat keep-alive. This on-demand monitoring design, which only initiates detailed data reporting in the final critical window, ensures the acquisition of necessary information while minimizing system communication and processing overhead.

[0112] In an optional embodiment, the connection instruction includes the predicted connection duration of the link, and the method further includes:

[0113] After a business interaction is triggered, the member node determines whether the corresponding business task has been completed within the predicted connection duration. If so, it disconnects and releases resources. If not, it sends a delay request to the leader node. The leader node decides whether to approve the delay request based on system resource information. If the delay request is approved, it updates the connection duration of the link where the member node is located. If the delay request is not approved, it controls the switching of the current connection to a path composed of backup nodes.

[0114] During business interactions, member nodes assess whether their assigned subtasks can be completed within the predicted connection window. Based on real-time monitored task execution data (such as processing volume and transmission progress), they delegate decision-making authority to the node most familiar with the task details, enabling rapid initial assessment. Decision-making authority then returns to the team leader node, which possesses overall coordination capabilities. Based on the request and resource availability, two recovery paths are initiated: If the team leader node determines sufficient system resources, it approves the delay request and updates the connection duration of the member node's link upon approval; for example, the updated connection duration is half of the original connection duration. If the team leader node determines system resources are strained (e.g., resources need to be reserved for higher-priority tasks), it rejects the delay request. For example, the team leader node might reject the current task's delay request to ensure sufficient resources for an upcoming higher-priority task, forcing it to switch to a backup path, thereby maximizing overall system efficiency.

[0115] To clearly describe the self-organizing network process of the soft bus based on intelligent prediction, we will now combine... Figure 2 , Figure 3 Let me explain. Figure 2 This is a flowchart of a soft bus self-organizing network method based on intelligent prediction. Figure 3 This is a diagram illustrating data interaction between different entities. It should be noted that... Figure 2 , Figure 3 In this context, the group leader is the group leader node, and the group members are the group member nodes. The number of group leader nodes and group member nodes is merely an example and is not intended to limit the invention.

[0116] Specifically, the intelligent prediction-based soft bus self-organizing network method includes the following process:

[0117] I. Data Collection.

[0118] 1. Collect historical records, specifically including historical operational data, real-time status data, and network performance data. Specifically, such as... Figure 3 As shown, the data acquisition command can be issued by an operator, who can be either an automated system or a staff member.

[0119] 2. Data Aggregation and Preprocessing: The above data is uploaded to the group leader node for aggregation, preliminary preprocessing, removal of outliers, supplementation of default values, normalization, extraction of key features, and formation of a standard dataset. After that, it is uploaded to the cloud in batches to avoid devices directly transmitting data to the cloud and reduce the pressure on the cloud.

[0120] II. Training of the time series prediction model.

[0121] 1. Model Input and Output: The uploaded standard dataset and business process are taken as input and fed into the prediction model. The output is the predicted link information, including the list of devices to be connected and the link topology, link priority, and link connection duration.

[0122] 2. Model Evaluation and Tuning: Compare the model's predicted information with manually labeled standard link information, quantify the differences in various metrics, including topology overlap rate (including precision and recall), link priority ranking differences, and connection duration error, to evaluate the model. Adjust the model until it meets the following requirements: prediction precision and recall greater than 85%, connection duration prediction error less than 20%, and priority ranking error (calculated using the Kendall coefficient formula) greater than 0.6.

[0123] 3. Model Update: Pull incremental datasets on a T+1 day cycle or according to the data increment ratio, and update the model with the new data to ensure that the model adapts to business changes;

[0124] 4. Model distribution: After the trained model has been lightweighted, it is distributed to each group leader node in batches.

[0125] III. Task Assignment.

[0126] 1. Generate a link list: When the business module issues a task request, the time series prediction model predicts and outputs two types of link lists. The first is the current task link list, which is the link between all devices that need to be established for this task. The second is the subsequent related task link list, which predicts the links required for tasks that need to be executed in the short term based on the business process (such as the need to connect process B after process A) and historical task information (such as process C being active at a certain time). This allows for advance planning of link resources.

[0127] 2. Convert prediction results into instructions: parse the link prediction results and convert them into specific operation commands.

[0128] 3. Instruction sorting: All operations are sorted in descending order of priority score (priority_score), and those with the same priority are sorted in ascending order of start time (start_time) to ensure that the core link is established first.

[0129] 4. Instruction Issuance: The sorted instructions are issued sequentially to the corresponding nodes.

[0130] IV. Business Execution.

[0131] 1. The team leader issues connection commands in batches according to link priority, ensuring that critical links are connected first. The connection process is as follows:

[0132] Resource Request: The group leader sends a resource request to the group member nodes. The group member nodes check whether they meet the resource request. If they do, the request is successful; otherwise, the request fails. If the group leader receives a request failure message, a backup node is selected as the alternative path. If there is no backup node or the backup node fails to request resources, the next backup group member node is queried. If all nodes fail to request resources, the model is triggered to readjust the prediction results.

[0133] Establishing a connection: After a resource application is successful, team members establish a connection through the soft bus and synchronously report the connection status of the team leader node. If the team leader node receives a connection failure message, it selects a backup node as an alternative path. If the backup node fails to connect, it continues to query the next backup node. If all nodes fail to connect, the model is triggered to readjust the prediction results.

[0134] 2. If the current link has dependent links, the group leader node will monitor the status of the preceding links in real time, and automatically trigger the connection of subsequent links after the preceding links are stable.

[0135] 3. After the link is established, each node starts monitoring the connection duration countdown. When the connection duration is less than the threshold, it periodically obtains execution data and uploads it to the group leader node. If the task is completed, it actively performs a disconnection operation to release network, CPU, memory and other resources, and reports the execution data to the group leader. If the connection duration exceeds the threshold and the task is not completed, the group member requests a new connection duration from the group leader. The duration is half of the previous connection duration. The group leader determines whether the current resources are sufficient. If they are sufficient, the request is approved. If not, the group leader switches to the backup path.

[0136] 4. Once all core links are established, the team leader initiates a business startup request, triggering interaction between the business devices.

[0137] 5. The team leader monitors all nodes in real time and quickly switches to a backup link when a link failure is detected.

[0138] V. Feedback Optimization.

[0139] 1. Data Feedback: During business execution, each node regularly reports data to the team leader node: link connection success rate, network performance data (latency, packet loss rate), and abnormal event records (interruption reasons, number of reconnections, etc.). The team leader node then summarizes and uploads the data to the cloud.

[0140] 2. Model optimization: After receiving feedback data, the cloud uses the new data to adjust the prediction model and improve prediction accuracy.

[0141] It should be noted that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0142] Reference Figure 4 The diagram illustrates a production system provided by an embodiment of the present invention, which may specifically include the following modules:

[0143] It includes a cloud 401, at least one group leader node 402 and multiple group member nodes 403, each of the group leader nodes 402 being connected to the cloud 401 and at least one of the group member nodes 403 respectively. Figure 4 As an example of a production system, the group leader node, cloud, and group member nodes described below will no longer be labeled.

[0144] The group leader node is used to collect historical data from the group member nodes, preprocess the historical data to obtain a standard dataset, and send it to the cloud. The historical data includes historical running data, real-time status data, and network performance data.

[0145] The cloud is used to train a time-series prediction model based on the standard dataset and preset business processes, and to distribute the trained time-series prediction model to the group leader node; the time-series prediction model is used to output predicted link information according to the task.

[0146] The group leader node is also used to respond to the target task by calling the time series prediction model to generate corresponding prediction link information and converting the prediction link information into a connection instruction containing link priority.

[0147] The group leader node is also used to schedule the group member nodes to pre-establish a soft bus connection based on the link priority in the connection instruction before the actual data interaction requirement of the target task occurs.

[0148] Optionally, the standard dataset includes comprehensive features from multiple tasks, and the cloud is used for:

[0149] For each task's comprehensive features in the standard dataset, corresponding standard link information is determined. The standard link information includes a standard current task link list and a list of subsequent associated task links.

[0150] The task's comprehensive features and the preset business process are input into the time-series prediction model to output predicted link information that conforms to the preset business process. The predicted link information includes a list of predicted current task links and a list of subsequent associated task links.

[0151] The model loss value is determined based on the difference between the predicted link information and the standard link information;

[0152] If the model loss value is greater than the preset loss threshold, the parameters of the time series prediction model are updated according to the model loss value, and training continues iteratively.

[0153] If the model loss value is less than or equal to the preset loss threshold, then the time series prediction model training is considered complete.

[0154] Optionally, the standard link information includes standard link topology, standard link priority, and standard link connection duration, and the predicted link information includes predicted link topology, predicted link priority, and predicted link connection duration.

[0155] When determining the model loss value based on the difference between the predicted link information and the standard link information, the cloud platform is used for:

[0156] Calculate the topology overlap rate based on the standard link topology and the predicted link topology;

[0157] Calculate the priority sorting error based on the standard link priority and the predicted link priority;

[0158] Calculate the connection duration prediction error based on the standard link connection duration and the predicted link connection duration;

[0159] The model loss value is calculated based on the topology overlap rate, the priority sorting error, and the connection duration prediction error.

[0160] Optionally, the group leader node is used to generate structured connection instructions for each link. The fields in the connection instructions include at least: link identifier, initiating node identifier, target node identifier, communication protocol, authentication method, planned start time, predicted connection duration, link priority, and the identifier of the preceding link on which it depends. All the generated connection instructions are sorted according to the link priority, and instructions with the same link priority are sorted according to the planned start time.

[0161] Optionally, the group leader node is configured to process each connection instruction sequentially according to the link priority in the connection instruction, and send resource requests to the group member nodes specified by the connection instruction;

[0162] The member node is used to verify whether the local resources meet the requirements for establishing a connection based on the received resource request, and to report the application result to the group leader node.

[0163] If the resource requirements of all member nodes in a link are successfully verified, the leader node initiates the process of establishing a soft bus connection between the relevant member nodes.

[0164] If the resource requirement verification fails, the group leader node will re-initiate the resource requirement to the backup node and trigger the soft bus connection establishment process.

[0165] Optionally, the group leader node is used to determine the core link based on the link priority; after the core link is established, it initiates a service startup request to trigger service interaction between the group member nodes.

[0166] Optionally, the connection instruction includes the predicted connection duration of the link;

[0167] After a business interaction is triggered, the member node is used to calculate the remaining connection duration based on the link's running time and the predicted connection duration; determine whether the remaining connection duration is less than a preset duration threshold; if so, obtain execution data related to the business task at a preset period and upload it to the leader node.

[0168] Optionally, the connection instruction includes the predicted connection duration of the link.

[0169] After a business interaction is triggered, the member node is used to determine whether the corresponding business task has been completed within the predicted connection duration; if yes, the connection is disconnected and resources are released; if no, a delay request is sent to the leader node.

[0170] The group leader node is used to decide whether to approve the delay request based on system resource information. When the delay request is approved, it updates the connection duration of the link where the group member node is located. When the delay request is not approved, it controls the current connection to be switched to a path composed of backup nodes.

[0171] The present invention provides a production system that can implement the steps in the aforementioned embodiments of the intelligent prediction-based soft bus self-organizing network method.

[0172] It should be noted that the module division in the various production systems provided in the above embodiments is illustrative and only represents one logical functional division. In actual implementation, other division methods may also be used. Furthermore, the functional modules in the various embodiments of this invention can be integrated into a single processor, exist as separate physical entities, or be integrated into a single module. The integrated modules described above can be implemented in hardware or as software functional modules.

[0173] If the integrated module is implemented as a software functional module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the technical solution of the embodiments of the present invention can be embodied in the form of a computer program product, which is stored in a computer storage medium and includes several instructions to cause an electronic device or processor to execute all or part of the steps of the methods in the various embodiments of the present invention. The aforementioned computer storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0174] Furthermore, the production system provided in the above embodiments and the embodiment of the method for self-organizing soft bus based on intelligent prediction belong to the same concept. For details of its specific implementation process, please refer to the method embodiment, which will not be repeated here.

[0175] Reference Figure 5 The diagram illustrates an electronic device according to an embodiment of the present invention. Figure 5 As shown, the electronic device in this embodiment of the invention includes: a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps in the above-described embodiment of the intelligent prediction-based soft bus self-organizing network method. Alternatively, when the processor executes the computer program, it implements the functions of each module in the above-described production system embodiment.

[0176] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete this application. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which can be used to describe the execution process of the computer program in the electronic device.

[0177] The electronic device may be a desktop computer, a cloud server, or other computing device. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that... Figure 5 This is merely one example of an electronic device and does not constitute a limitation on the electronic device. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0178] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0179] The memory can be an internal storage unit of the electronic device, such as a hard drive or RAM. Alternatively, it can be an external storage device, such as a plug-in hard drive, Smart Media Card (SMC), Secure Digital (SD) card, Flash Card, etc. Furthermore, the memory can include both internal and external storage units. The memory is used to store the computer program and other programs and data required by the electronic device. The memory can also be used to temporarily store data that has been output or will be output.

[0180] This invention also discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the intelligent prediction-based soft bus self-organizing network method as described in the foregoing embodiments.

[0181] This invention also discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements the intelligent prediction-based soft bus self-organizing network method as described in the foregoing embodiments.

[0182] This invention also discloses a computer program product that, when run on a computer, causes the computer to execute the intelligent prediction-based soft bus self-organizing network method described in the foregoing embodiments.

[0183] The embodiments described above are only used to illustrate the technical solutions of this application, and are not intended to limit it. 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, and should all be included within the protection scope of this application.

Claims

1. A soft bus ad hoc network method based on intelligent prediction, characterized in that, The method is applied to a production system, the production system comprising a cloud, at least one group leader node and a plurality of group member nodes, each of the group leader nodes being connected with the cloud and at least one of the group member nodes respectively; the method comprising: The group leader node collects historical record data of the group member nodes, pre-processes the historical record data to obtain a standard data set, and sends the standard data set to the cloud; the historical record data comprises historical running data, real-time state data and network performance data; The cloud trains a time series prediction model based on the standard data set and a preset business process, and distributes the trained time series prediction model to the group leader nodes; the time series prediction model is used to output prediction link information according to a task; The group leader node generates corresponding prediction link information by calling the time series prediction model in response to a target task, and converts the prediction link information into connection instructions containing link priority; The group leader node establishes a soft bus connection in advance according to the link priority in the connection instructions before the actual data interaction demand of the target task occurs.

2. The method of claim 1, wherein, The standard data set comprises a plurality of task comprehensive features, and the cloud trains a time series prediction model based on the standard data set and a preset business process, comprising: The cloud determines corresponding standard link information for each task comprehensive feature in the standard data set, the standard link information comprising a standard current task link list and a subsequent associated task link list; The task comprehensive features and the preset business process are input into the time series prediction model to output prediction link information conforming to the preset business process, the prediction link information comprising a predicted current task link list and a subsequent associated task link list; A model loss value is determined according to the difference between the prediction link information and the standard link information; If the model loss value is greater than a preset loss threshold, the parameters of the time series prediction model are updated according to the model loss value, and iterative training is continued; If the model loss value is less than or equal to the preset loss threshold, it is determined that the time series prediction model is trained.

3. The method of claim 2, wherein, The standard link information comprises standard link topology, standard link priority and standard link connection duration, and the prediction link information comprises prediction link topology, prediction link priority and prediction link connection duration; The model loss value is determined according to the difference between the prediction link information and the standard link information, comprising: A topology structure coincidence rate is calculated according to the standard link topology and the prediction link topology; A priority sorting error is calculated according to the standard link priority and the prediction link priority; A connection duration prediction error is calculated according to the standard link connection duration and the prediction link connection duration; The model loss value is calculated according to the topology structure coincidence rate, the priority sorting error and the connection duration prediction error.

4. The method of claim 1, wherein, The prediction link information is converted into connection instructions containing link priority, comprising: The group leader node generates a structured connection instruction for each link, and fields in the connection instruction at least include: link identification, initiator node identification, target node identification, communication protocol, authentication method, planned start time, predicted connection duration, link priority, and dependent pre-link identification; All generated connection instructions are sorted according to link priority, and instructions with the same link priority are sorted according to planned start time.

5. The method of claim 1, wherein, The group leader node generates a structured connection instruction for each link, and fields in the connection instruction at least include: link identification, initiator node identification, target node identification, communication protocol, authentication method, planned start time, predicted connection duration, link priority, and dependent pre-link identification; The group leader node processes each connection instruction in turn according to the link priority in the connection instruction, and sends resource requirements to the group member nodes specified in the connection instruction; The group member nodes verify whether local resources meet the requirements of establishing a connection based on the received resource requirements, and feed back application results to the group leader node; If the resource requirements of the group member nodes in a link are all verified successfully, the group leader node initiates a soft bus connection establishment process between the related group member nodes; If there is a resource requirement verification failure, the group leader node reinitiates the resource requirement to a backup node and triggers a soft bus connection establishment process.

6. The method according to any one of claims 1 to 5, characterized in that, Further comprising: The group leader node determines a core link according to the link priority; After the core link is established, a business start application is initiated to trigger business interaction between the group member nodes.

7. The method of claim 6, wherein, The connection instruction includes the predicted connection duration of the link, and the method further comprises: After the business interaction is triggered, the group member nodes calculate the remaining connection duration according to the running duration of the link and the predicted connection duration; Determine whether the remaining connection duration is less than a preset duration threshold; If yes, acquire and upload execution data related to the business task to the group leader node at a preset period.

8. The method of claim 6, wherein, The connection instruction includes the predicted connection duration of the link, and the method further comprises: After the business interaction is triggered, the group member nodes determine whether the corresponding business task has been completed within the predicted connection duration; if yes, disconnect the connection and release the resources; if no, send a time extension application to the group leader node; The group leader node decides whether to approve the time extension application according to system resource information; when the time extension application is approved, the group leader node updates the connection duration of the link where the group member node is located; when the time extension application is not approved, the group leader node controls the current connection to be switched to a path composed of backup nodes.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the intelligent prediction-based soft bus ad hoc network method of any one of claims 1-8.

10. A computer-readable storage medium storing a computer program, the computer program comprising instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 9. The computer program is executed by the processor to implement the intelligent prediction-based soft bus ad hoc network method of any one of claims 1-8.

Images