An industrial production resource intelligent scheduling system and method based on 5G communication

Abstract

This invention relates to the field of knowledge construction and discloses an intelligent scheduling system and method for industrial production resources based on 5G communication. The system includes: a metadata mapping module that converts resource coordinates and rates transmitted back from the network into entity nodes and matrices in a knowledge network; a dynamic edge-building module that calculates the interference degree based on the ratio of the sum of rates to the relative distance, constructs constraint edges when the threshold is exceeded, and reduces the weights through a linear decay mechanism of confidence when the network is discontinuous; and a resource scheduling module that generates anti-interference scheduling instructions based on the network containing constraint edges. This invention converts spatial changes and rate fluctuations in situ into topological changes in the knowledge network, transforms the spatial attributes of physical production factors into knowledge graph representations with business-related logic, thereby reducing global numerical optimization to local topological conflict detection, eliminating the search space for invalid solutions, and mitigating topological oscillations caused by packet loss noise from a management decision-making perspective, thus improving the intelligence level of manufacturing resource organization and scheduling.

Description

Technical Field

[0001] This invention belongs to the field of knowledge construction technology, and in particular relates to an intelligent scheduling system and method for industrial production resources based on 5G communication. Background Technology

[0002] Current system architectures typically employ a pre-defined rule engine paired with a static database. This architecture uses discrete physical device operating state parameters acquired from underlying sensor networks as numerical input, which are then fed into an operations research-based scheduling algorithm to solve for a multi-dimensional parameter matrix under rigid boundary constraints. This provides discrete, time-series instruction flows to each processing unit, logistics resource, and assembly station along the entire production line. In the actual operational flow of this architecture, the processor compares the received time-series physical state parameters with prior process path parameters, constructing constraint boundaries within the logical storage space that represent the mutual exclusion, cooperation, and sequential order of resources. Through the matrix encapsulation of these fixed relationships, complex industrial processes are made more efficient. The physical conditions on site are precipitated into a static logical network processing mechanism that the rule engine can directly identify and compare. With the deployment of high-speed, high-concurrency communication networks in the digital workshop, the frequency of state changes of physical production resources has increased significantly. This leads to high-frequency transient disorder and congestion during the flow of high-dimensional time-varying conditions. The static rule framework that the traditional architecture implicitly relies on cannot extract these transient interference trends caused by the dense intersection of physical entities within a short time window. This manifests as the rate of change of physical entity space exceeding the efficiency of logical network topology reconstruction. As a result, the system still outputs scheduling instructions that conform to static rules but are spatiotemporally misaligned, directly causing a chain reaction of technical problems that lead to the collapse of the entire production sequence.

[0003] The problem lies in the traditional system's lack of real-time knowledge construction capabilities for dynamically changing physical elements. It cannot effectively aggregate discrete physical parameters into manufacturing knowledge nodes with business semantics, leading to production scheduling decisions detached from actual on-site business logic. Faced with complex and ever-changing industrial environments, existing architectures not only struggle to overcome limitations in the underlying hardware physical entity form and sensing node layout, but also suffer from deficiencies in top-level software control methods. For example, Chinese invention patent application CN121526448A discloses a knowledge graph-based auxiliary method and system for mine production scheduling. This method extracts multi-source scheduling logs and text data through a large language model, constructs a structured dataset containing entity attributes and temporal relationships, and uses comparative learning for three-dimensional processing. Tuple optimization is used to assist scheduling decisions. In-depth analysis reveals that the implicit premise for the establishment of such software control methods is that the underlying data source has long-term temporal stability and allows offline non-real-time semantic reasoning. However, in the 5G high-frequency transient disorder congestion scenario focused on in this application, the spatial coordinates and movement speed of physical resources change abruptly in milliseconds. The aforementioned scheme relies on offline text semantic parsing and large language model feature calculation operation mechanism. There is a fundamental mismatch between the core presupposition and actual boundary conditions and the actual working conditions for the direct in-situ mapping of transient disturbances of high-frequency physical quantities. The complicated semantic conversion steps inevitably introduce unacceptable time delays, making it impossible to directly convert interference potential energy into the underlying topological constraints of blocking physical collisions within the transient time window.

[0004] Therefore, the technical problem to be solved by this invention is how to convert the physical characteristics of production resources collected at high frequency into a logical graph topology network with spatiotemporal attributes, and how to smooth out transient operating condition disturbances through the dynamic growth and confidence decay of local constraint edges, so as to realize the rapid graph reasoning and reconstruction of scheduling constraint boundaries of large-scale entity networks. Summary of the Invention

[0005] This invention aims to solve the problems of lagging spatiotemporal graph topology construction and excessive constraints in reasoning of large-scale entity networks under highly dynamic operating conditions.

[0006] In this technical solution, an intelligent scheduling system for industrial production resources based on 5G communication is provided. The system includes: The time-varying multi-source metadata mapping module is used to receive time-varying multi-dimensional state representation parameters of industrial production resources transmitted back via the 5G network. The time-varying multi-dimensional state representation parameters include the spatial coordinates and movement speed of the industrial production resources. The module converts the industrial production resources into entity knowledge nodes in the knowledge structure network and constructs the corresponding spatial state vector matrix from the spatial coordinates and movement speed. The topology knowledge dynamic edge construction module is used to periodically extract the spatial state vector matrix corresponding to each of the two entity knowledge nodes, and calculate the spatial interference degree based on the ratio of the sum of the two movement speeds to the relative distance between the two entity knowledge nodes. The topology knowledge dynamic edge construction module is also used to construct an associated constraint edge between the two entity knowledge nodes when the spatial interference degree exceeds the set interference judgment threshold, and to start a confidence maintenance mechanism through historical time series. When the 5G network is intermittent for a short time, the confidence of the associated constraint edge is linearly decayed. When the confidence decays to zero and no new time-varying multidimensional state representation parameters are received, the associated constraint edge is removed from the knowledge structure network.

[0007] The knowledge-driven resource scheduling module is used to generate interference-free scheduling instructions for industrial production resources based on a knowledge structure network containing associated constraint edges.

[0008] Preferably, the system also includes a graph folding unit; the graph folding unit is used to identify a set of entity knowledge nodes that exhibit strong correlation and whose interference potential energy is lower than the calibrated interference threshold within a defined time window, merge the set of entity knowledge nodes into aggregated knowledge supernodes and superimpose attribute vectors in the knowledge structure network, and directly transform the topological correlation level of the knowledge structure network through topological reconstruction, thereby reducing the computational load of the knowledge structure network when performing continuous reasoning in long-process industrial sites.

[0009] Preferably, when the confidence of the associated constraint edge is linearly decayed, the topological knowledge dynamic edge construction module follows the following inheritance refinement logic: within a defined control period, based on the duration of the time-varying multidimensional state representation parameters that have not been received, the weight value of the associated constraint edge is gradually reduced in an arithmetic progression.

[0010] Preferably, the topological knowledge dynamic edge construction module also includes the following functions: periodically aggregating the spatial interference degree of all associated constraint edges in the knowledge structure network to generate a time-varying interference feature vector, which is used to characterize the transient interference trend of global production resources.

[0011] Preferably, the system also includes a safety takeover module; the safety takeover module is used to continuously monitor the variance of the time-varying interference feature vector and record the calibration interval of the time-varying interference feature vector under normal operating conditions; when the variance exceeds three times the standard deviation of the normal distribution and it is determined that the underlying sensor array has a physical offset, the control topology knowledge dynamic edge construction module stabilizes the edge construction constraints within the range of prior safe topology constraints, instead of directly calculating the abnormal time series data; the safety takeover module is also used to backward switch and lock to the prior knowledge subgraph solidified by the previous steady-state time series slice, and output the maintenance instruction to the knowledge-driven resource scheduling module as the only scheduling benchmark to maintain the scheduling state of industrial production resources until the variance falls back to the calibration interval.

[0012] Preferably, when identifying a strongly correlated state, the graph folding unit follows the following inheritance refinement logic: it calculates the historical collaborative interaction frequency between any two entity knowledge nodes in the entity knowledge node set, and determines that the entity knowledge node set is in a strongly correlated state when the historical collaborative interaction frequency is greater than a preset association frequency threshold and the interference potential energy is lower than the calibrated interference threshold.

[0013] Preferably, when the graph folding unit superimposes attribute vectors onto the aggregated knowledge supernode, it follows the following inheritance refinement logic: it extracts the spatial state vector matrix of all entity knowledge nodes in the entity knowledge node set in the dimensionality-reduced topological space constructed by the graph folding unit, generates centroid state vectors through mean clustering, and writes the centroid state vectors as attribute vectors into the aggregated knowledge supernode.

[0014] Preferably, when the knowledge-driven resource scheduling module generates anti-interference scheduling instructions, it follows the following inheritance refinement logic: it extracts the topological constraint boundaries corresponding to each associated constraint edge in the knowledge structure network, converts the topological constraint boundaries into the boundary safe reachable intervals and safe timing sequences of the logic control layer, and generates anti-interference scheduling instructions to control the movement of each industrial production resource.

[0015] Preferably, the system also includes an alarm module, which is used to output a warning signal representing the timing conflict state when the topology knowledge dynamic edge construction module removes the associated constraint edge and the knowledge-driven resource scheduling module determines that the anti-interference scheduling instructions of each industrial production resource have a logical conflict.

[0016] A method for intelligent scheduling of industrial production resources based on 5G communication includes: Step S1: Receive the time-varying multidimensional state representation parameters of industrial production resources transmitted back via the 5G network. The time-varying multidimensional state representation parameters include the spatial coordinates and movement speed of the industrial production resources. Convert the industrial production resources into entity knowledge nodes in the knowledge structure network, and construct the corresponding spatial state vector matrix from the spatial coordinates and movement speed. Step S2: Periodically extract the spatial state vector matrix corresponding to each of the two entity knowledge nodes, and calculate the spatial interference degree based on the ratio of the sum of the two movement speeds to the relative distance between the two entity knowledge nodes; the topology knowledge dynamic edge construction module is also used to construct an associated constraint edge between the two entity knowledge nodes when the spatial interference degree exceeds the set interference judgment threshold, and to start a confidence maintenance mechanism through historical time series, so that the confidence degree of the associated constraint edge is linearly decayed when the 5G network is intermittent for a short time, and the associated constraint edge is removed from the knowledge structure network when the confidence degree decays to zero and no new time-varying multidimensional state representation parameters are received; Step S3: Based on the knowledge structure network containing associated constraint edges, generate interference-free scheduling instructions for industrial production resources.

[0017] Compared with existing technologies, the present invention, an intelligent scheduling system and method for industrial production resources based on 5G communication, has the following advantages: 1. In the intelligent scheduling of industrial production resources using 5G communication, the three-dimensional spatial coordinates and transient power requirements of each resource entity are extracted and mapped to the attributes of knowledge nodes in the graph structure. The transient interference potential energy between any two knowledge nodes is calculated based on the power change gradient, Euclidean relative distance, and prior process path logical dependency coefficient. When the transient interference potential energy is greater than a preset mutual exclusion threshold, mutually exclusive constraint edges representing resource competition are directly generated. This transforms the multi-dimensional parameter matrix operation that relies on static rule table matching and global numerical iteration in traditional scheduling into closed-loop subgraph topology detection for mutually exclusive constraint edges in the dynamic industrial knowledge graph. The spatiotemporal interference constraint relationship of the entire industrial site grows into a dynamic topology in real time as data flows in. The calculation of the scheduling sequence degenerates into the extraction of safe paths in the graph, thereby significantly reducing the search space for invalid solutions and improving the computational efficiency of millisecond-level rescheduling under highly dynamic conditions.

[0018] 2. To address the transient signal loss caused by strong electromagnetic interference and high-frequency sensor noise, when the transient interference potential energy falls below the mutual exclusion threshold, the system adds a lifecycle attribute containing semantic confidence to the generated mutual exclusion constraint edges. According to the preset decay period, the weight of the associated constraint edges is gradually reduced, so that the graph does not immediately remove the associated edges due to short-term signal discontinuity. Instead, the confidence is maintained through historical time series until the confidence decays linearly to zero and no new data is received to activate it. Only then are the topological edges removed from the logical graph. The time axis causal consistency verification mechanism constructed in this way effectively smooths out the frequent flipping of the edge construction logic caused by the disturbance of the underlying physical quantity, and avoids the chain collapse of the entire production time sequence.

[0019] 3. By identifying sets of fine-grained knowledge nodes that exhibit strong correlation and low interference potential within a specific time window through graph folding units, these sets are merged into aggregated knowledge supernodes and attribute vectors are superimposed in the logical topology. This directly compresses the graph reasoning depth of large-scale production lines through spatial dimensionality reduction, effectively reducing the computing load on the processor when processing large-scale entity networks. It also avoids the risk of storage medium write deadlock caused by frequent graph reconstruction under high-concurrency conditions, enabling the knowledge representation process under high-dimensional time-varying conditions to have long-term stable operation capabilities in large-scale long-process industrial sites. Attached Figure Description

[0020] Figure 1 This is the control flow diagram of the intelligent scheduling system for industrial production resources using 5G communication according to the present invention. Figure 2 This is a hardware deployment diagram of the intelligent scheduling system for industrial production resources using 5G communication according to the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0022] A smart scheduling system for industrial production resources based on 5G communication, the system includes: The time-varying multi-source metadata mapping module is used to receive time-varying multi-dimensional state representation parameters of industrial production resources transmitted back via the 5G network. The time-varying multi-dimensional state representation parameters include the spatial coordinates and movement speed of the industrial production resources. The module converts the industrial production resources into entity knowledge nodes in the knowledge structure network and constructs the corresponding spatial state vector matrix from the spatial coordinates and movement speed. The topology knowledge dynamic edge construction module is used to periodically extract the spatial state vector matrix corresponding to each of the two entity knowledge nodes, and calculate the spatial interference degree based on the ratio of the sum of the two movement speeds to the relative distance between the two entity knowledge nodes. The topology knowledge dynamic edge construction module is also used to construct an associated constraint edge between the two entity knowledge nodes when the spatial interference degree exceeds the set interference judgment threshold, and to start a confidence maintenance mechanism through historical time series. When the 5G network is intermittent for a short time, the confidence of the associated constraint edge is linearly decayed. When the confidence decays to zero and no new time-varying multidimensional state representation parameters are received, the associated constraint edge is removed from the knowledge structure network.

[0023] The knowledge-driven resource scheduling module is used to generate interference-free scheduling instructions for industrial production resources based on a knowledge structure network containing associated constraint edges.

[0024] Preferably, the system also includes a graph folding unit; the graph folding unit is used to identify a set of entity knowledge nodes that exhibit strong correlation and whose interference potential energy is lower than the calibrated interference threshold within a defined time window, merge the set of entity knowledge nodes into aggregated knowledge supernodes and superimpose attribute vectors in the knowledge structure network, and directly transform the topological correlation level of the knowledge structure network through topological reconstruction, thereby reducing the computational load of the knowledge structure network when performing continuous reasoning in long-process industrial sites.

[0025] Preferably, when the confidence of the associated constraint edge is linearly decayed, the topological knowledge dynamic edge construction module follows the following inheritance refinement logic: within a defined control period, based on the duration of the time-varying multidimensional state representation parameters that have not been received, the weight value of the associated constraint edge is gradually reduced in an arithmetic progression.

[0026] Preferably, the topological knowledge dynamic edge construction module also includes the following functions: periodically aggregating the spatial interference degree of all associated constraint edges in the knowledge structure network to generate a time-varying interference feature vector, which is used to characterize the transient interference trend of global production resources.

[0027] Preferably, the system also includes a safety takeover module; the safety takeover module is used to continuously monitor the variance of the time-varying interference feature vector and record the calibration interval of the time-varying interference feature vector under normal operating conditions; when the variance exceeds three times the standard deviation of the normal distribution and it is determined that the underlying sensor array has a physical offset, the control topology knowledge dynamic edge construction module stabilizes the edge construction constraints within the range of prior safe topology constraints, instead of directly calculating the abnormal time series data; the safety takeover module is also used to backward switch and lock to the prior knowledge subgraph solidified by the previous steady-state time series slice, and output the maintenance instruction to the knowledge-driven resource scheduling module as the only scheduling benchmark to maintain the scheduling state of industrial production resources until the variance falls back to the calibration interval.

[0028] Preferably, when identifying a strongly correlated state, the graph folding unit follows the following inheritance refinement logic: it calculates the historical collaborative interaction frequency between any two entity knowledge nodes in the entity knowledge node set, and determines that the entity knowledge node set is in a strongly correlated state when the historical collaborative interaction frequency is greater than a preset association frequency threshold and the interference potential energy is lower than the calibrated interference threshold.

[0029] Preferably, when the graph folding unit superimposes attribute vectors onto the aggregated knowledge supernode, it follows the following inheritance refinement logic: it extracts the spatial state vector matrix of all entity knowledge nodes in the entity knowledge node set in the dimensionality-reduced topological space constructed by the graph folding unit, generates centroid state vectors through mean clustering, and writes the centroid state vectors as attribute vectors into the aggregated knowledge supernode.

[0030] Preferably, when the knowledge-driven resource scheduling module generates anti-interference scheduling instructions, it follows the following inheritance refinement logic: it extracts the topological constraint boundaries corresponding to each associated constraint edge in the knowledge structure network, converts the topological constraint boundaries into the boundary safe reachable intervals and safe timing sequences of the logic control layer, and generates anti-interference scheduling instructions to control the movement of each industrial production resource.

[0031] Preferably, the system also includes an alarm module, which is used to output a warning signal representing the timing conflict state when the topology knowledge dynamic edge construction module removes the associated constraint edge and the knowledge-driven resource scheduling module determines that the anti-interference scheduling instructions of each industrial production resource have a logical conflict.

[0032] A method for intelligent scheduling of industrial production resources based on 5G communication includes: Step S1: Receive the time-varying multidimensional state representation parameters of industrial production resources transmitted back via the 5G network. The time-varying multidimensional state representation parameters include the spatial coordinates and movement speed of the industrial production resources. Convert the industrial production resources into entity knowledge nodes in the knowledge structure network, and construct the corresponding spatial state vector matrix from the spatial coordinates and movement speed. Step S2: Periodically extract the spatial state vector matrix corresponding to each of the two entity knowledge nodes, and calculate the spatial interference degree based on the ratio of the sum of the two movement speeds to the relative distance between the two entity knowledge nodes; the topology knowledge dynamic edge construction module is also used to construct an associated constraint edge between the two entity knowledge nodes when the spatial interference degree exceeds the set interference judgment threshold, and to start a confidence maintenance mechanism through historical time series, so that the confidence degree of the associated constraint edge is linearly decayed when the 5G network is intermittent for a short time, and the associated constraint edge is removed from the knowledge structure network when the confidence degree decays to zero and no new time-varying multidimensional state representation parameters are received; Step S3: Based on the knowledge structure network containing associated constraint edges, generate interference-free scheduling instructions for industrial production resources.

[0033] Example 1: When a system faces concurrent multi-source data in a discrete manufacturing environment, multiple heavy-duty mobile carts and multi-joint robotic arms converge densely in a shared material handling area. Due to the transient changes in the physical coordinates and movement speeds of each mobile entity, the traditional system architecture, which uses a preset rule table to match a static database, lacks the ability to construct real-time knowledge of time-varying physical elements. This leads to a semantic gap between the static constraint paths of the system's internal rule world and the actual working conditions of the physical environment. Because the system cannot extract the transient interference trends generated by the convergence of physical entities within a preset time window, it outputs control commands that result in spatiotemporal misalignment, causing a chain reaction of production sequence collapses and interruptions in topology reasoning of the physical network. To solve the above technical problems, this application discloses an intelligent scheduling system for industrial production resources based on 5G communication. The time-varying multi-source metadata mapping module receives, in real-time, time-varying multi-dimensional state representation parameters of industrial production resources transmitted back via the 5G network. The state representation parameters include the spatial coordinates and movement speed of industrial production resources. This module converts industrial production resources into entity knowledge nodes in a knowledge structure network, and constructs a corresponding spatial state vector matrix based on the spatial coordinates and movement speed, thereby establishing a digital topological foundation for the underlying physical state changes within the logical space. Specifically, when constructing this matrix, the time-varying multi-source metadata mapping module uses the independent number of the entity knowledge node as the addressing index, extracts the parsed entity's three-dimensional spatial coordinate position quantities along the X, Y, and Z axes as the first row vector of the matrix, and simultaneously extracts the entity's movement speed scalar along the corresponding three spatial axes in the current sampling period as the second row vector. After vertical feature alignment, these are combined to generate a fixed two-row, three-column two-dimensional relationship matrix. The topological knowledge dynamic edge construction module periodically extracts the spatial state vector matrices corresponding to two entity knowledge nodes, and calculates the spatial interference degree based on the ratio of the sum of the two movement speeds to the relative distance between the two entity knowledge nodes. The specific mathematical formula is as follows: ,in, For spatial interference, Let be the scalar of the movement rate of the first entity knowledge node. Let be the scalar value representing the movement rate of the second entity knowledge node. This represents the Euclidean distance between two entity knowledge nodes.

[0034] After obtaining the spatial interferometry, the topology knowledge dynamic edge construction module compares the spatial interferometry with a preset interference judgment threshold. When the spatial interferometry exceeds the set interference judgment threshold, the module constructs an association constraint edge between two entity knowledge nodes and initiates a confidence maintenance mechanism through historical time series. When the 5G network experiences short-term intermittent transmissions and no new time-varying multidimensional state representation parameters are received, the module initiates a confidence linear decay mechanism following an arithmetic progression. Within a defined control period, the weight values ​​of the association constraint edges are reduced at a fixed ratio to avoid frequent topology flips caused by transient signal disturbances. This continues until the confidence decays to zero and no new parameters are received, at which point the association constraint edge is removed from the knowledge structure network. Based on this, the graph folding unit identifies sets of entity knowledge nodes exhibiting strong association and interference potential energy below the calibrated interference threshold within a defined time window. By calculating the historical collaborative interaction frequency between any two entity knowledge nodes in this set, if the historical collaborative interaction frequency is greater than a preset association frequency threshold, the set is determined to be in a strong association state. This unit then places the nodes in the knowledge structure network... The set is merged into an aggregated knowledge supernode and attribute vectors are superimposed. The spatial state vector matrix of all entity knowledge nodes in the set is extracted through mean clustering and centroid state vector is generated. The centroid state vector is written into the aggregated knowledge supernode as an attribute vector, which reduces the computational load of the knowledge structure network during continuous reasoning in long-process industrial sites. According to the convex hull boundary closure theorem of computational geometry, the exclusive boundary of the space of multiple rigid bodies depends on the outermost vertices forming a closed envelope surface. After the graph folding unit is written into the centroid state vector, the coordinate matrix of the vertices of the physical shape of all industrial production resources in the entity knowledge node set is extracted. The Graham scan algorithm is executed to calculate the polar angle sorting of the coordinate matrix. The feature points of the polar angle periphery are connected in sequence to generate the minimum convex polygon space boundary of the set. The coordinate set of each vertex of the minimum convex polygon space boundary is added to the aggregated knowledge supernode as a spatial exclusive attribute. When the topology knowledge dynamic edge construction module processes the edge construction logic with other nodes, the shortest Euclidean distance from the external entity knowledge node to the minimum convex polygon space boundary is used instead of the straight distance to a single centroid node for input spatial interference calculation process.

[0035] When a sudden hardware disturbance at the production site causes a physical shift in the underlying sensor array, the safety takeover module continuously monitors the time-varying interference feature vector generated by the convergence of the spatial interference degrees of all associated constraint edges and calculates the variance of the time-varying interference feature vector. When it undergoes an abnormal jump and exceeds three times the standard deviation of the normal distribution, the safety takeover module controls the topology knowledge dynamic edge construction module to stabilize the edge construction constraints within the prior safe topology constraint range, stops the direct dynamic calculation based on the current abnormal time series data, and reverses and locks to the prior knowledge subgraph solidified by the previous steady-state time series slice. The knowledge-driven resource scheduling module, based on the knowledge structure network containing associated constraint edges or prior knowledge subgraphs, extracts the topology constraint boundaries corresponding to each associated constraint edge through a depth-first traversal algorithm and converts them into the boundary safe reachable intervals and safe time series of the logic control layer. Based on this, it generates anti-interference scheduling instructions to control the movement of various industrial production resources. At the same time, when the topology knowledge dynamic edge construction module removes associated constraint edges and the anti-interference scheduling instructions generate logical conflicts, the alarm module outputs a warning signal representing the time series conflict state.

[0036] By transforming the spatial changes and rate fluctuations of the physical site into topological changes of the knowledge structure network in situ at the network layer, the system transforms the discrete spatial attributes of manufacturing elements into knowledge graph representations with business-related logic. This reduces the numerical optimization calculation of the global multidimensional parameter matrix to local topological conflict detection, mitigates topological oscillations caused by communication packet loss noise from the management decision-making level, and establishes a causal closed loop for the transmission of time-varying constraints across the entire link using the system's inherent multi-module collaborative mechanism. This enables the organization and scheduling of industrial production resources to maintain a continuous steady state of operation in an environment of high-frequency and drastic fluctuations in discrete physical conditions.

[0037] Example 2: When the system faces strong electromagnetic interference in a discrete manufacturing environment, in the experimental platform constructed by the semi-physical industrial network simulation system, multiple data sources transmit the three-dimensional spatial coordinates and movement speed of a heavy-duty mobile cart and a multi-joint robotic arm through a spatial sensing unit with a 100Hz sampling frequency. The experimental platform is used to simulate a manufacturing environment containing electromagnetic interference and channel fading. The control cycle of the time-varying multi-source metadata mapping module is determined by a trade-off rule between the real-time performance of data acquisition and the data processing load of the system. Specifically, when the data receiving bandwidth of the time-varying multi-source metadata mapping module is 10MHz... Within the 20MHz range, the time-varying characteristic changes non-monotonically with the data concurrency, and is determined to be 10ms under this operating condition. To verify the time-varying constraint transfer characteristics of the system, the experiment uses the running trajectory stream collected by the physical experimental platform as the raw input data, and superimposes Gaussian white noise with a signal-to-noise ratio of 20dB in the communication link to simulate power frequency interference harmonics. The experimental group uses a time-varying multi-source metadata mapping module to receive spatial coordinates and movement speed, transforming the spatial transition of physical resources into a spatial state vector matrix. In order to compensate for the time lag loss introduced by the 5G transmission delay, a quantization compensation operator is injected when calculating the spatial interferometry. Quantization compensation operator The network latency change rate is adjusted in real time based on the current control cycle. This is used to correct the deviation in entity knowledge node state caused by network jitter at the component level. When performing this component-level correction, the time-varying multi-source metadata mapping module continuously extracts the difference between the round-trip latency of network data packets in the previous cycle and the latency in the current cycle to obtain the latency change rate. When the latency change rate shows a positive divergence trend, a weight constant greater than 1 is called as an operator through table lookup. The assignment amplifies the estimated margin of interference. When the rate of change converges negatively, a weight constant less than 1 is assigned, thereby stabilizing the disordered network communication state into a definite kinematic correction factor. Based on the principle of rigid body relative kinematic velocity vector projection, the interference trend is determined by the projection component of the velocity vector on the line connecting the centers of mass of the two entities. The time-varying multi-source metadata mapping module collects the angle between the velocity vector direction of the first entity knowledge node and the line connecting the centers of mass. And the angle between the velocity vector direction of the second entity knowledge node and the line connecting the centroid. To calculate the equivalent interference rate The calculation formula is: ,in, Equivalent interference rate; The dimensionless values ​​of the quantization compensation operator range from 0.90 to 1.10. and These are the scalars representing the movement rates of the first and second entity knowledge nodes, respectively. and These represent the velocity vector angles, and the equivalent interference rate is obtained from the topological knowledge dynamic edge construction module. Then directly replace the spatial interferometry formula scalar and numerator calculate.

[0038] To establish the rationality of the numerical range, the experiment used the three-point support method to test the interference judgment threshold and the 5Hz benchmark as the correlation frequency threshold to verify the boundary limits. The experimental group operated at the lower limit of the correlation frequency threshold (1Hz), the normal median (5Hz), and the upper limit (10Hz), respectively. Under the condition that the correlation frequency threshold was set to 1Hz, due to the excessively low exclusion baseline, the graph folding unit triggered node aggregation actions at high frequencies, causing data overload in the attribute vectors within the aggregated knowledge supernodes. The inference load of the global architecture exhibited irregular fluctuations after 12.6 minutes of continuous operation. When the correlation... When the frequency threshold is set to 10Hz, the high sorting priority prevents the set of strongly correlated entity knowledge nodes from meeting the aggregation conditions, resulting in a 3.42-fold increase in the inference load of the global architecture. However, when the correlation frequency threshold is within a 5Hz working window, the graph folding unit extracts the centroid state vector through mean-based clustering, achieving smooth convergence of the computational load while ensuring the topological structure remains intact. Simultaneously, in the transparent verification of spatial interference calculation, the topological knowledge dynamic edge-building module calculates the spatial interference according to the formula, the specific mathematical formula of which is: ,in, Spatial interferometry is a dimensionless parameter. Let be the scalar of the movement rate of the first entity knowledge node. Let be the scalar value representing the movement rate of the second entity knowledge node. To mitigate the dimensional conflicts arising from different feature parameters in numerical calculations, the processor, when calculating the spatial interferometry, incorporates a standardized time base coefficient with the unit of seconds (s) and a constant value of 1. This coefficient is used to calculate the frequency dimension of the equation. The relative proximity rate quotient is multiplied by the time reference coefficient to legally convert it into a dimensionless pure numerical scalar for subsequent comparison logic; when the movement rate scalar of the first entity knowledge node is detected to be 1.22 m / s, the movement rate scalar of the second entity knowledge node is 0.78 m / s, and the relative distance between the two nodes decreases to 0.41 m, the spatial interference degree... The calculated value suddenly changed from 1.46 under normal conditions to 4.88. This data trajectory slice confirms the inevitability of the causal response of multi-feature combination in the dynamic growth of local constrained edges.

[0039] To further demonstrate the inventiveness of the proposed solution, an experiment was conducted with a control group containing a conventional static rule engine system and a control group without the linear decay mechanism of confidence. A gradient control system for 5G communication packet loss rate was also constructed, encompassing low, medium, and high levels of disruption. Under low-intensity interference (1% packet loss), both the experimental group and the control group without the linear decay mechanism maintained the stability of associated constraint edges using historical time series analysis, resulting in a topology misclassification rate of approximately 2.3% for both groups. However, when the packet loss rate increased to a medium-intensity gradient of 5%, the lack of weights hindered the development of a more robust and efficient system. The dynamic mitigation mechanism removed the linear confidence decay mechanism. In the control group, the associated constraint edges experienced high-frequency flickering and topology flipping, causing the topology state misjudgment rate to surge to 28.6%. In contrast, the experimental group, with its linear confidence decay mechanism that reduced weights by a fixed 10% increment within a 10ms control period, stabilized the topology state misjudgment rate within a control range of 4.1%. Under high-intensity conditions with a packet loss rate of 15%, the control group without the linear confidence decay mechanism experienced a global topology inference collapse due to sudden network interruptions, and the system failed to output a safe value due to node mismatch. The system provides fully reachable instructions, while the experimental group disconnects failed edge constructions when the confidence level decays to 0. Furthermore, the safety takeover module triggers a logic judgment instruction when the variance of the time-varying interference feature vector exceeds three times the standard deviation. This controls the edge construction constraints to remain stable within the prior safe topological constraints and then reverses to the prior knowledge subgraph. This reduces the conflict rate of global scheduling instructions under high-intensity network interference conditions by 64.3% compared to the control group, exhibiting a nonlinear performance mutation effect caused by confidence decay, graph folding, and multi-unit collaboration in safety takeover. Based on the measured data flow analysis of the aforementioned multi-dimensional comparison system, this experiment directly confirms that the proposed method has definite physical feasibility in non-ideal high-dynamic industrial electromagnetic environments. The system transforms discrete physical quantities into nodes and edge construction constraint logic within the graph space, converting the global multi-dimensional parameter matrix optimization solution into local knowledge network conflict detection. This mitigates topological oscillations caused by network channel packet loss noise, ensuring that the anti-interference scheduling instructions generated based on the knowledge structure network have definite reachable boundaries at the physical execution layer. This achieves a stable engineering closed loop for multi-module collaborative transmission in intelligent scheduling scenarios of long-process, high-concurrency industrial resources.

[0040] Example 3: When the system faces the situation of deep interweaving of three-dimensional spatial trajectories between a heavy-duty mobile trolley and a multi-joint robotic arm in a high-concurrency, high-volume material turnover environment, the moving components in the physical environment experience dynamic offsets due to transient acceleration changes and mechanical dead zone limitations, resulting in spatiotemporal motion interference conflicts within local narrow channels. If the system cannot convert the abstract topological constraint boundaries in the knowledge structure network into motion physical quantity constraints that can be responded to by the electrical drive mechanism, or if the prior safety topological constraint range stored in the system becomes outdated due to lack of timely updates, it will cause spatial overlap between the scheduling control commands output by the logic control layer and the actual motion envelope of the physical entities, thereby triggering physical-mechanical collision faults or control loop closures in the multi-machine system. To determine the interference threshold in the dynamic edge-building module based on topological knowledge during prolonged suspension, an intelligent scheduling system for industrial production resources based on 5G communication acquires calibration input parameters. These parameters include the geometric envelope radius of the heavy-duty mobile vehicle (1.5m) collected by a laser rangefinder and the maximum aspect ratio radius of the multi-joint robotic arm (2.2m) extracted by an angle encoder. The simulation environment is hosted by a processor with a clock speed of 3.2GHz and 16GB of memory. The dynamic edge-building module calculates the interference threshold based on the geometric constraints formed by the mechanical braking displacement and control response delay of the heavy-duty mobile vehicle and the multi-joint robotic arm. The calculation formula is as follows: ,in, The threshold for interference determination. This represents the maximum moving speed of the heavy-duty mobile trolley. The maximum end effector speed of the multi-joint robotic arm. For the system's control cycle, This is the minimum safe distance allowed when the two components decelerate. Let be the geometric envelope radius of the heavy-duty mobile vehicle. Given the maximum aspect ratio radius of the end effector of the multi-joint robotic arm, when the maximum moving speed is set to 2.0 m / s, the maximum end effector speed is set to 1.5 m / s, the control cycle is set to 0.01 s, and the minimum safe distance is set to 5.2 m, the topology knowledge dynamic edge construction module calculates the interference judgment threshold using a formula. The value is 0.0233, thus converting the topological determination criterion of the graph into a physical constraint based on the mechanical motion limit.

[0041] The safety takeover module reconstructs the prior safety topology constraint range and prior knowledge subgraph. The system maintains a sliding time storage matrix with a capacity of 500 time-series slices in the storage area to record the transition matrix of associated constraint edges under steady-state conditions. When updating the sliding time storage matrix, the system reduces the statistical weight of historical topology data exceeding 60 seconds, increasing the feature contribution value of steady-state slices within the current control cycle. During system operation, the process judgment quantization module collects the arrival time series of each resource through photoelectric position sensors and calculates the instantaneous range of the time-varying interference feature vector. When the variance of the time-varying interference feature vector within five consecutive sampling points is greater than three times the standard deviation determined by historical normal operating conditions, and the number of associated constraint edges in the current knowledge structure network is less than 20m... When the growth slope within s is greater than 1.5, the system determines that the physical position of the field sensor array has shifted. The variance in the above determination logic is greater than 3 times the standard deviation, which is a feature representation of the monitoring trend inside the processor. Its actual execution process is as follows: the processor retrieves the time-varying interference feature variance value obtained by discrete calculation and performs square root dimensionality reduction processing, converts it into real-time standard deviation, and then compares it with the threshold of 3 times the standard deviation of the historical normal distribution with the same numerical magnitude, so as to ensure the rigor and consistency of the determination trigger operation in statistical derivation; the safety takeover module disconnects the data cascade from the current computing link and reverses the graph network control state to the prior knowledge subgraph solidified in the previous period in the sliding time storage matrix, eliminating the lag when the topology data undergoes time evolution.

[0042] The knowledge-driven resource scheduling module reads the global adjacency matrix in the knowledge structure network at the current moment and initializes a depth-first traversal stack with the entity knowledge node as the root node. This module searches the graph branches sequentially along each associated constraint edge. When the topological connection between any two entity knowledge nodes satisfies the interference tree structure, it calculates the intersection of the two sets of spatial state vector matrices, extracting the center coordinates of the spatially overlapping region and the overlapping time period on the time axis. Based on Minkowski and continuous collision detection theories, the equivalent motion ray of the interference judgment between two moving rigid bodies is geometrically intersected with the boundary of the extended obstacle. The knowledge-driven resource scheduling module extracts the two-dimensional projection contours of the heavy-duty mobile vehicle and the rotating base of the multi-joint robotic arm on the horizontal plane and calculates their minimum circumcircle radius. and Using the current coordinates of the heavy-duty mobile vehicle as a fixed origin, generate the radius. A static circular extended envelope boundary is used to extract the relative displacement of the multi-joint robotic arm during the overlapping period and convert it into a ray vector. The intersection point of the ray vector and the quadratic equation of the static circular extended envelope boundary is calculated, and the coordinate value corresponding to the first real root is output as the center coordinate of the spatial overlapping area. The knowledge-driven resource scheduling module converts the center coordinate of the spatial overlapping area into the boundary safe reachable interval of the electrical control layer. A mechanical safety margin of 1.2 is multiplied outward in the three-dimensional axial directions of the center coordinate to establish the prohibited physical coordinate boundary of the mobile vehicle in the shared material receiving area. The overlapping period is converted into a prohibited time window at the time control layer. A 50ms communication protection delay is added at the beginning and end of the overlapping period to calculate the control time limit of the joint servo motor driver of the multi-joint robotic arm. The coordinate boundary and the forbidden time window are input as constraint variables into the path planner. Through the cascade deduction of discrete topology, the path planner outputs anti-interference scheduling instructions to control various industrial production resources. The path planner sends the anti-interference scheduling instructions to the joint servo motor drivers and hub motor controllers of the motion components, controlling multiple heavy-duty mobile trolleys and multi-joint robotic arms to generate physical displacements according to the boundary safe reach range under the condition of high-concurrency material interweaving. This smooths the topology fluctuations caused by network channel packet loss noise. In industrial production environments where network packet loss or sensor offset occurs, the path planner controls the movement of the moving parts of various industrial production resources to flow within the boundary safe reach range and the forbidden time window, keeping the overall production timing conflict rate below 0.1%, so that the organization and scheduling state of industrial production resources converges to a continuous steady state.

[0043] Example 4: When the system faces the initial calibration conditions of deploying new industrial production resources on-site or adding new heavy-duty mobile vehicles to the network, the topological differences in the spatial distribution of 5G base stations in different workshops introduce initial position offset residuals caused by multipath effects in the time-varying multi-source metadata mapping module, resulting in a deviation between the initial coordinates of entity knowledge nodes and the physical space axis. Before putting industrial production resources into resource scheduling, the system initiates a pre-calibration state, using passive calibration benchmarks arranged at standard coordinate points in the shared material connection area to transmit test pulses. The on-site 5G antenna array collects a continuous 1000-cycle signal transmission delay sequence, and calculates the inherent channel fading factor in the current spatial environment using the least squares method. .

[0044] The time-varying multi-source metadata mapping module will map the inherent channel fading factor. The underlying environment parameter table is written to the underlying memory. When the time-varying multidimensional state characterization parameters are received from the 5G network, the inherent channel fading factor is called. The coordinate measurement noise caused by multipath reflection is corrected, and the initial spatial position residual is controlled within 5mm, so that the generated spatial state vector matrix completes the alignment of the coordinate axes. Each heavy-duty mobile car and multi-joint robotic arm continues to flow according to the corrected anti-interference scheduling instructions. The time-varying constraint relationship of the whole link generates causal transmission in the unified knowledge structure network. The physical organization and scheduling state of the multi-machine system converges to the continuous steady state under high dynamic discrete working conditions.

[0045] Example 5: When the system faces the zero-point drift of the spatial position sensor caused by the deterioration of the on-site working environment, the three-dimensional spatial coordinates fed back by the position sensor during the flow of the heavy-duty mobile trolley in the shared material transfer area will generate a continuously increasing zero-point factor deviation. This zero-point factor deviation causes the spatial state vector matrix obtained by the downstream topology knowledge dynamic edge construction module to contain incorrect position feature sources, resulting in polarity reversal calculation results when calculating the spatial interference degree, outputting non-realistic associated constraint edges, causing the path planner to generate anti-interference scheduling instructions with logical conflicts, leading to interference collisions of mechanical moving parts when the heavy-duty mobile trolley passes through narrow channels or frequent shutdowns of the servo system due to misjudgment of boundaries.

[0046] To correct coordinate deviation, the system activates the adaptive topology compensation module to correct zero-point factor deviation. The initial input parameters of the adaptive topology compensation module consist of a 100Hz high-frequency synchronous positioning pulse sequence returned by adjacent normally operating reference anchor point units. At this time, the processor's processing throughput boundary is within the preset specifications. The time-varying multi-source metadata mapping module continuously monitors the attribute vector transition rate inside the aggregated knowledge supernode through the graph folding unit. The adaptive topology compensation module periodically extracts entity knowledge nodes containing zero-point factor deviation, calculates the relative displacement difference sequence between the entity knowledge node and the reference anchor point unit within 50 consecutive control cycles, and calculates the coordinate correction amount along the current coordinate axis through a time-series difference algorithm. The formula for calculating the coordinate correction amount is as follows: ,in, For coordinate correction, This represents the positional drift rate of the entity knowledge node within the sampling window. Given a control cycle set for the system, with the current position drift rate set to 0.15 m / s and a control cycle set to 0.01 s, the output coordinate correction is calculated using a formula. The calibration value is 0.0015m. This value is embedded in the spatial state vector matrix as an attribute pre-feature, thereby blocking the second-order physical dissipation loss caused by sensor aging at the component layer, restoring the corrected three-dimensional spatial coordinates to the reference value range, and controlling the heavy-duty mobile trolley to move within the safe time sequence boundary.

[0047] The embodiments of this application have been described above with reference to the accompanying drawings. Unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other. This application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit of this application and the scope of protection of this invention, and all of these forms are within the protection scope of this application.

Claims

1. An intelligent scheduling system for industrial production resources based on 5G communication, characterized in that, The system includes: The time-varying multi-source metadata mapping module is used to receive time-varying multi-dimensional state representation parameters of industrial production resources transmitted back via the 5G network. The time-varying multi-dimensional state representation parameters include the spatial coordinates and movement speed of the industrial production resources. The module converts the industrial production resources into entity knowledge nodes in the knowledge structure network and constructs the corresponding spatial state vector matrix from the spatial coordinates and movement speed. The topology knowledge dynamic edge construction module is used to periodically extract the spatial state vector matrix corresponding to each of the two entity knowledge nodes, and calculate the spatial interference degree based on the ratio of the sum of the two movement speeds to the relative distance between the two entity knowledge nodes. The topology knowledge dynamic edge construction module is also used to construct an associated constraint edge between the two entity knowledge nodes when the spatial interference degree exceeds the set interference judgment threshold, and to start a confidence maintenance mechanism through historical time series. When the 5G network is intermittent for a short time, the confidence of the associated constraint edge is linearly decayed. When the confidence decays to zero and no new time-varying multidimensional state representation parameters are received, the associated constraint edge is removed from the knowledge structure network. The knowledge-driven resource scheduling module is used to generate interference-free scheduling instructions for industrial production resources based on a knowledge structure network containing associated constraint edges.

2. The intelligent scheduling system for industrial production resources based on 5G communication according to claim 1, characterized in that, The system also includes a graph folding unit; the graph folding unit is used to identify a set of entity knowledge nodes that are strongly correlated and whose interference potential energy is lower than the calibrated interference threshold within a defined time window. In the knowledge structure network, the set of entity knowledge nodes is merged into aggregated knowledge supernodes and attribute vectors are superimposed. The topological association level of the knowledge structure network is directly transformed through topological reconstruction, thereby reducing the computational load of the knowledge structure network when performing continuous reasoning in long-process industrial sites.

3. The intelligent scheduling system for industrial production resources based on 5G communication according to claim 1, characterized in that, When the confidence of associated constraint edges decreases linearly, the topology knowledge dynamic edge construction module follows the following inheritance refinement logic: within a defined control period, based on the duration of the time-varying multidimensional state representation parameters that have not been received, the weight values ​​of associated constraint edges are reduced step by step in an arithmetic progression.

4. The intelligent scheduling system for industrial production resources based on 5G communication according to claim 1, characterized in that, The topology knowledge dynamic edge construction module also includes the following functions: periodically aggregating the spatial interference degree of all associated constraint edges in the knowledge structure network to generate time-varying interference feature vectors, which are used to characterize the transient interference trend of global production resources.

5. The intelligent scheduling system for industrial production resources based on 5G communication according to claim 4, characterized in that, The system also includes a safety takeover module; the safety takeover module is used to continuously monitor the variance of the time-varying interference feature vector and record the calibration interval of the time-varying interference feature vector under normal operating conditions; when the variance exceeds three times the standard deviation of the normal distribution and it is determined that the underlying sensor array has a physical offset, the control topology knowledge dynamic edge construction module stabilizes the edge construction constraints within the range of prior safe topology constraints, instead of directly calculating the abnormal time series data; the safety takeover module is also used to backward switch and lock to the prior knowledge subgraph solidified by the previous steady-state time series slice, and output maintenance instructions to the knowledge-driven resource scheduling module as the only scheduling benchmark to maintain the scheduling status of industrial production resources until the variance falls back to the calibration interval.

6. The intelligent scheduling system for industrial production resources based on 5G communication according to claim 2, characterized in that, When identifying a strongly correlated state, the graph folding unit follows the following inheritance refinement logic: it calculates the historical collaborative interaction frequency between any two entity knowledge nodes in the entity knowledge node set, and determines that the entity knowledge node set is in a strongly correlated state when the historical collaborative interaction frequency is greater than the preset association frequency threshold and the interference potential energy is lower than the calibrated interference threshold.

7. The intelligent scheduling system for industrial production resources based on 5G communication according to claim 2, characterized in that, When the graph folding unit superimposes attribute vectors onto the aggregated knowledge supernode, it follows the following inheritance refinement logic: it extracts the spatial state vector matrix of all entity knowledge nodes in the entity knowledge node set in the dimensionality-reduced topological space constructed by the graph folding unit, generates the centroid state vector through mean clustering, and writes the centroid state vector as an attribute vector into the aggregated knowledge supernode.

8. The intelligent scheduling system for industrial production resources based on 5G communication according to claim 1, characterized in that, When the knowledge-driven resource scheduling module generates anti-interference scheduling instructions, it follows the following inheritance refinement logic: it extracts the topological constraint boundaries corresponding to each associated constraint edge in the knowledge structure network, converts the topological constraint boundaries into the boundary safe reachable intervals and safe timing sequences of the logic control layer, and generates anti-interference scheduling instructions to control the movement of each industrial production resource.

9. The intelligent scheduling system for industrial production resources based on 5G communication according to claim 1, characterized in that, The system also includes an alarm module, which is used to output a warning signal representing the timing conflict state when the topology knowledge dynamic edge construction module removes associated constraint edges and the knowledge-driven resource scheduling module determines that the anti-interference scheduling instructions of each industrial production resource have a logical conflict.

10. A method for intelligent scheduling of industrial production resources based on 5G communication, used to implement the intelligent scheduling system for industrial production resources based on 5G communication as described in claim 1, characterized in that, include: Step S1: Receive the time-varying multidimensional state representation parameters of industrial production resources transmitted back via the 5G network. The time-varying multidimensional state representation parameters include the spatial coordinates and movement speed of the industrial production resources. Convert the industrial production resources into entity knowledge nodes in the knowledge structure network, and construct the corresponding spatial state vector matrix from the spatial coordinates and movement speed. Step S2: Periodically extract the spatial state vector matrix corresponding to each of the two entity knowledge nodes, and calculate the spatial interference degree based on the ratio of the sum of the two movement speeds to the relative distance between the two entity knowledge nodes; the topology knowledge dynamic edge construction module is also used to construct an associated constraint edge between the two entity knowledge nodes when the spatial interference degree exceeds the set interference judgment threshold, and to start a confidence maintenance mechanism through historical time series, so that the confidence degree of the associated constraint edge is linearly decayed when the 5G network is intermittent for a short time, and the associated constraint edge is removed from the knowledge structure network when the confidence degree decays to zero and no new time-varying multidimensional state representation parameters are received; Step S3: Based on the knowledge structure network containing associated constraint edges, generate interference-free scheduling instructions for industrial production resources.

Images