A virtual-real collaborative communication scheduling method suitable for a digital twin workshop

Abstract

The application provides a virtual-real collaborative communication scheduling method suitable for a digital twin workshop; comprising: (1) a virtual-real collaborative communication scheduling method for a digital twin workshop low-speed subnetwork; (2) a virtual-real collaborative communication scheduling method for a digital twin workshop high-speed subnetwork. The application has advantages in reducing data transmission delay and improving network efficiency through DAG model optimization selection of multi-cycle paths. In addition, the application can ensure the real-time and reliability of transmission in a complex network environment, so that the virtual model can timely reflect the changes of the physical world, and ensure the efficient operation of the digital twin workshop. Compared with the traditional technology, the application has obvious improvement in network congestion, data transmission real-time and system resource utilization, and improves the efficiency and reliability of virtual-real collaboration in the digital twin system.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of industrial wireless networks; in particular, it relates to a virtual-real collaborative communication scheduling method suitable for digital twin workshops. BACKGROUND

[0002] With the promotion of new generation information technology, many countries have launched advanced development plans aimed at realizing intelligent manufacturing, which relies on the interoperability, integration and fusion between the physical world and the virtual world in the manufacturing process. Intelligent manufacturing requires real-time data collection, analysis and application in the production process, which relies on highly accurate and reliable technical support. In this context, digital twin technology emerged, creating an accurate virtual copy of physical entities, enabling various aspects of the production process to be simulated and optimized in a virtual environment. Data is the driving force behind digital twins, and real-time data transmission is crucial to ensuring the effectiveness of digital twin systems.

[0003] Industrial wireless networks (IWN) are an important part of digital twin systems, responsible for collecting data from the physical world and transmitting it to the virtual copy. In the field of IWN, there are several key technical standards, such as ISA100.11a, WirelessHART and WIA-PA, which are essential for achieving efficient operation of the network. As application scenarios continue to expand and deepen, IWN faces more stringent real-time transmission requirements, requiring faster response to sensor data updates to ensure that the digital twin system can timely reflect the current state of physical entities. In the industrial scenario of digital twins, sensors need to monitor the status of workshop elements and transmit data in real time to the central control system. For example, in an aircraft assembly workshop, by creating a virtual digital model of the aircraft, the assembly process can be simulated in a virtual environment, allowing potential problems to be identified and resolved before actual production. By monitoring the assembly line in real time based on multi-source, multi-dimensional data and fine-grained analysis, engineers can adjust strategies and methods in a timely manner to ensure the quality of aircraft assembly.

[0004] With the development of microelectronics and wireless communication technology, sensors and actuators have been optimized in terms of cost, size and intelligence, and are widely used in industrial automation, smart cities and environmental monitoring. Standardization and interoperability of devices have improved seamless integration of devices and systems, leading to a surge in the number of nodes that generate large amounts of data that need to be transmitted in real time and periodically to the central control system for analysis and decision-making. However, the surge in data can cause some nodes in the data link to be overloaded, increasing the risk of network congestion and data delay, and posing challenges to the stability and reliability of the network. At the same time, network scheduling is also more complex, requiring flexible handling of multi-cycle, multi-length data packets and providing effective routing for each node to ensure stable data transmission.

[0005] Digital twin technology plays a revolutionary role in intelligent manufacturing by collecting data from sensors in real time, accurately simulating and optimizing production processes in a virtual environment. Wireless technology, as the "nervous system" of digital twin, connects sensors throughout the intelligent factory. As more and more sensors are integrated into digital twin systems, wireless sensor networks must meet diverse needs, especially in terms of low latency, to maintain real-time synchronization between the digital twin and the physical entity. To meet the needs of real-time data transmission, many technologies have been proposed and applied. For example, OFDMA is considered an efficient multi-user wireless communication technology that can improve spectrum utilization by allocating subcarriers to different users in the form of multiple resource units (RUs). TDMA is widely used in TT communication scenarios, such as time-triggered Ethernet and time-sensitive networks, which require high reliability and determinism.

[0006] However, as industrial scenarios become increasingly complex, IWN faces increasing challenges, particularly in the virtual-real collaboration of digital twin workshops. Both low-speed subnets and high-speed subnets have high demands for network real-time performance and reliability. Traditional TDMA and OFDMA scheduling methods, although they can meet the needs of real-time communication in specific scenarios, have limitations in scalability, load balancing, and scheduling complexity, which limit their application in large-scale industrial workshops.

[0007] Traditional load balancing methods in low-speed subnets often fail to fully consider multi-source communication flows in the network, causing some nodes to carry excessive traffic and become bottlenecks in network performance. In high-speed subnets, the random access method using OFDMA can improve the flexibility of multi-user access, but in high-load situations, it is prone to conflicts, leading to increased communication delays and transmission uncertainty, making it difficult to meet the strict requirements of digital twin workshops for latency. To address these issues, existing technologies have attempted to improve scheduling algorithms, such as precise scheduling methods based on SMT theory, to improve system performance. However, these methods often have high computational complexity and cannot be effectively applied in large-scale, multi-node industrial environments. SUMMARY

[0008] The purpose of the present application is to provide a virtual-real collaborative communication scheduling method suitable for digital twin workshops.

[0009] The present application is implemented by the following technical solutions:

[0010] The present invention relates to a virtual-real collaborative communication scheduling method applicable to digital twin workshops, including: (1) a virtual-real collaborative communication scheduling method for low-speed subnets in digital twin workshops; and (2) a virtual-real collaborative communication scheduling method for high-speed subnets in digital twin workshops.

[0011] Preferably, the specific steps of the virtual-real collaborative communication scheduling method for low-speed subnets in digital twin workshops are as follows:

[0012] Step 1, Multi-source single-sink network model

[0013] This invention focuses primarily on uplink data from low-speed terminal system nodes to the gateway. Employing this unidirectional communication model simplifies the analysis process while preserving the generality of key communication challenges, as uplink traffic is typically much higher than downlink traffic due to the continuous transmission of sensor data. Furthermore, digital twin systems exhibit significant asymmetry, meaning that the data flow from the physical entity to the digital replica is continuous, while the control commands from the digital replica to the physical entity are relatively few. Therefore, focusing on unidirectional links effectively captures the main traffic characteristics in these networks. This invention uses a DAG model to describe the multi-hop topology of the IWN, chosen because it effectively represents the unidirectional communication and hierarchical relationships between nodes in the network. In this model, terminal system nodes are abstracted as vertices, and the reachability and communication direction between nodes are represented by directed edges.

[0014] Step 2, Optimize the algorithm

[0015] First, a loss function is defined to guide the search and optimization process to minimize the error;

[0016] The algorithm's evaluation function measures the load balancing rate of each layer, and its design takes into account the difference between the actual load and the theoretical average load. Specifically, for each layer in the network, the difference between the actual load of a node and the theoretical average load of that layer is derived, and the square root of the sum of the squares of these differences is taken to obtain a comprehensive load balancing metric. This comprehensive load balancing metric will guide the algorithm to find the optimal path to achieve load balancing.

[0017] Taking a multi-hop topology as an example, starting from the gateway node, it is divided into layers one through four from top to bottom; let V layer τ represents the set of all nodes at a certain level. LCM Let LayerNum be the least common multiple of the transmission periods of all communication streams; let LayerNum be the total number of layers, |V layer | indicates the number of all nodes in the hierarchy, node v k The actual load is v kLoad can be calculated after the path selection scheme is determined; since the first and last layers do not forward and their loads are fixed, these two layers are not considered in the calculation; based on the idea of ​​hierarchical load balancing, the difference between the actual load of a single node and the theoretical average load of the hierarchy is used as the load balancing metric, and the loss function is in the form of root mean square error, expressed as:

[0018]

[0019] Among them, V layer |V represents the set of all nodes at a certain level. layer | represents the total number of nodes in the hierarchy, τ LCM Let ω be the least common multiple of the transmission periods of all communication flows, ω be the link transmission delay of the communication flows, LayerNum be the total number of layers, and node v be the least common multiple of the transmission periods of all communication flows. k The actual load is v k .Load can be calculated after the path selection scheme is determined;

[0020] Heuristic algorithms are used to optimize the path so that the actual load of the nodes is close to the average load. Specifically, the aurora optimization algorithm is adopted. This algorithm is inspired by the aurora produced by the convergence of high-energy particles of solar wind at the Earth's poles. By analyzing the motion of high-energy particles and studying the basic principles of physics, the motion of particles is simulated.

[0021] The routing problem addressed in this invention is defined as a multi-source single-sink problem. Its core objective is to find the optimal path scheme within the IWN (Integrated Winding Network) to achieve efficient data aggregation. Specifically: First, for each source node, the algorithm finds all shortest paths with the same hop count; the set of these paths forms the basis of the solution space for network path selection. After constructing a path set for each source node, these sets are merged to construct the solution space for the entire network's path selection. The design principles of PLO (Progressive Loop) are utilized to construct a PLO optimization mechanism oriented towards hierarchical load balancing to achieve network load balancing. The construction of the hierarchical load balancing PLO optimization mechanism includes:

[0022] Mapping from individual to path selection scheme: In PLO, each individual represents a potential solution; in HLB-PLO, each individual is mapped to a path combination, that is, each high-energy particle represents a potential communication path selection scheme in the network.

[0023] Dimension-to-communication flow mapping: Each dimension in PLO can be mapped to the path selection of each communication flow in HLB-PLO; in this way, the optimization objective of each dimension is associated with the path selection of the communication flow.

[0024] Fitness function mapping: The fitness function in PLO needs to be redefined to adapt to the scenario-specific objectives; In HLB-PLO, the fitness function is set as the HLB loss function that measures the load balancing performance of the path selection scheme.

[0025] Optimization of the mapping of objectives: The objective of PLO is to find the individual with the best fitness, while the objective of HLB-PLO is to find the path selection scheme with the minimum HLB loss function value.

[0026] The optimization algorithm operates within the solution space, which involves not only the algorithm's search efficiency but also its ability to accurately evaluate the performance of each path selection scheme. After overall optimization, the path selection scheme with the maximum fitness (i.e., the lowest HLB loss) is finally selected.

[0027] Preferably, the specific steps of the virtual-real collaborative communication scheduling method for high-speed subnets in digital twin workshops are as follows:

[0028] Step 1, OFDMA Network Model

[0029] To achieve real-time and deterministic transmission, this invention employs a star topology, in which a group of terminal nodes are wirelessly connected to a base station; the communication system uses TDMA to achieve deterministic media access, where time is divided into time slots (e.g., TTI);

[0030] This system achieves simultaneous streaming transmission by allocating adjacent RUs to each terminal node, with each RU containing an equal number of data subcarriers; consider a system containing a set of TT streams, denoted as F, and assume that the transmission of each TT stream requires only one TTI; such as Figure 3 As shown, TT flow f i ∈F by terminal device ED i With period P i It is generated periodically, with an initial generation time of G. i Then, for this flow, draw from the set {1,2,...,N}. RU In the frequency domain, adjacent RUs are allocated. To maximize the utilization of frequency domain resources, the number of RUs allocated (denoted as RL) is... i ) is set to support f i Minimum required value;

[0031] Step 2, TTI-RU constraint model

[0032] This invention constructs a scheduling constraint model based on TTI-RU using integer programming, called TTI-RU-IP; specifically, it is an OFDMA-based time-triggered network with an O i and RU i Constraints, where O i Represents communication flow f iThe initial transmission time of ∈F, RU i Indicates assignment to f i The smallest index among all RUs ∈ F. Since time-triggered streams are generated and sent periodically, this invention only focuses on the scheduling strategy within its initial period.

[0033] The scheduling constraint model includes:

[0034] Constraint 1: Delay Constraint: Each stream should be transmitted immediately after generation, and its corresponding time window should not exceed its period, expressed by the following formal expression:

[0035]

[0036] Constraint 2: RU Continuity Constraint: Each flow should be assigned an RU from the set, represented by the following constraint:

[0037]

[0038] Constraint 3: TTI-RU Collision-Free Constraint: Once any two flows are assigned to the same RU, the TTIs they occupy during transmission must not overlap. This is expressed by the following constraint:

[0039]

[0040] Where G(P) i ,P j ) represents P i and P j The greatest common divisor.

[0041] This invention proposes a hierarchical load balancing routing algorithm for virtual-physical collaboration of low-speed data in digital twin workshops. It primarily optimizes multi-cycle communication flows by establishing a Directed Acyclic Graph (DAG) model. The Polar Lights Optimization (PLO) algorithm is employed to achieve global load balancing, avoiding node overload. This method ensures data transmission stability in low-speed subnets and improves the system's real-time response capability. For virtual-physical collaboration of high-speed data in digital twin workshops, this invention proposes a TT scheduling strategy based on OFDMA. This strategy optimizes resource allocation by establishing a constraint model based on Transmission Time Interval (TTI) and RU (Return on Request), ensuring low-latency and collision-free transmission.

[0042] The present invention has the following advantages:

[0043] (1) This invention improves the performance of virtual-real collaborative communication in low-speed subnets through load balancing and path optimization. PLO achieves hierarchical load balancing, ensuring balanced load distribution among nodes in the network and avoiding network bottlenecks in traditional methods, thereby improving the overall stability of the system.

[0044] (2) This invention optimizes the selection of multi-cycle paths using a DAG model, offering advantages in reducing data transmission latency and improving network efficiency. Furthermore, this invention ensures real-time and reliable transmission in complex network environments, enabling the virtual model to reflect changes in the physical world promptly and ensuring the efficient operation of the digital twin workshop. Compared to traditional technologies, this invention demonstrates significant improvements in network congestion, data transmission real-time performance, and system resource utilization, enhancing the efficiency and reliability of virtual-physical collaboration in the digital twin system.

[0045] (3) This invention proposes a hierarchical load balancing routing algorithm for virtual-physical collaborative transmission in low-speed subnets, mainly by optimizing multi-cycle communication flows through the establishment of a DAG model. PLO is used to achieve global load balancing, avoiding node overload. For virtual-physical collaborative transmission in high-speed subnets, this invention proposes a TT scheduling strategy based on OFDMA. This strategy optimizes resource allocation by establishing a constraint model based on TTI and RU, ensuring low latency and collision-free transmission. Attached Figure Description

[0046] Figure 1 This is a schematic diagram of the mapping principle involved in the method of this invention;

[0047] Figure 2 This is a schematic diagram of the method of the present invention;

[0048] Figure 3 This is a flowchart of the PLO sub-process in this invention;

[0049] Figure 4 This is a diagram illustrating the topology and scheduling strategy;

[0050] Figure 5 This is a typical scene diagram of a digital twin workshop. Detailed Implementation

[0051] The present invention will now be described in detail with reference to specific embodiments. It should be noted that the following embodiments are merely further illustrations of the present invention, but the scope of protection of the present invention is not limited to the following embodiments.

[0052] Example 1

[0053] This embodiment relates to a digital twin system in a manufacturing workshop, where a low-speed subnet is responsible for monitoring and reporting the real-time status (such as temperature, humidity, vibration, etc.) of each device. In this example, sensor nodes are densely distributed in the workshop, and the generated sensor data needs to be uploaded to the central control system for real-time updates of the digital twin model. Traditional scheduling methods can lead to network congestion and transmission delays when node load increases, affecting real-time performance.

[0054] This invention constructs a Directed Acyclic Graph (DAG) model and applies HLB-PLO to achieve global load balancing in a low-speed subnet, preventing certain nodes from becoming network bottlenecks due to traffic overload. The algorithm optimizes node communication paths to balance data traffic from multiple sensors and distributes the load reasonably across the entire network. (See [link to relevant documentation]). Figure 2 As shown, the specific steps are as follows:

[0055] Step 1, Multi-source single-sink network model

[0056] This invention focuses primarily on uplink data from low-speed terminal system nodes to the gateway. Employing this unidirectional communication model simplifies the analysis process while preserving the generality of key communication challenges, as uplink traffic is typically much higher than downlink traffic due to the continuous transmission of sensor data. Furthermore, digital twin systems exhibit significant asymmetry, meaning that the data flow from the physical entity to the digital replica is continuous, while the control commands from the digital replica to the physical entity are relatively few. Therefore, focusing on unidirectional links effectively captures the key traffic characteristics in these networks. This invention uses a DAG model to describe the multi-hop topology of IWNs, see [link to DAG model]. Figure 1 As shown; this model was chosen because it can effectively represent unidirectional communication and hierarchical relationships between nodes in a network. In this model, terminal system nodes are abstracted as vertices, and the reachability and communication direction between nodes are represented by directed edges;

[0057] Step 2, Optimize the algorithm

[0058] First, a loss function is defined to guide the search and optimization process to minimize the error;

[0059] The algorithm's evaluation function measures the load balancing rate of each layer, and its design takes into account the difference between the actual load and the theoretical average load. Specifically, for each layer in the network, the difference between the actual load of a node and the theoretical average load of that layer is derived, and the square root of the sum of the squares of these differences is taken to obtain a comprehensive load balancing metric. This comprehensive load balancing metric will guide the algorithm to find the optimal path to achieve load balancing.

[0060] by Figure 1 For example, starting from the gateway node, the layers are divided into four levels from top to bottom; let V... layerτ represents the set of all nodes at a certain level. LCM Let LayerNum be the least common multiple of the transmission periods of all communication streams; let LayerNum be the total number of layers, |V layer | indicates the number of all nodes in the hierarchy, node v k The actual load is v k Load can be calculated after the path selection scheme is determined; since the first and last layers do not forward and their loads are fixed, these two layers are not considered in the calculation; based on the idea of ​​hierarchical load balancing, the difference between the actual load of a single node and the theoretical average load of the hierarchy is used as the load balancing metric, and the loss function is in the form of root mean square error, expressed as:

[0061]

[0062] Among them, V layer |V represents the set of all nodes at a certain level. layer | represents the total number of nodes in the hierarchy, τ LCM Let ω be the least common multiple of the transmission periods of all communication flows, ω be the link transmission delay of the communication flows, LayerNum be the total number of layers, and node v be the least common multiple of the transmission periods of all communication flows. k The actual load is v k .Load can be calculated after the path selection scheme is determined;

[0063] Heuristic algorithms are used to optimize the path so that the actual load of the nodes is close to the average load. Specifically, the aurora optimization algorithm is adopted. This algorithm is inspired by the aurora produced by the convergence of high-energy particles of solar wind at the Earth's poles. By analyzing the motion of high-energy particles and studying the basic principles of physics, the motion of particles is simulated.

[0064] The routing problem addressed in this invention is defined as a multi-source single-sink problem. Its core objective is to find the optimal path scheme within the IWN (Integrated Winding Network) to achieve efficient data aggregation. Specifically: First, for each source node, the algorithm finds all shortest paths with the same hop count; the set of these paths forms the basis of the solution space for network path selection. After constructing a path set for each source node, these sets are merged to construct the solution space for the entire network's path selection. A PLO optimization mechanism oriented towards hierarchical load balancing is constructed using the design principles of PLO (Progressive Loop). See [link to relevant documentation]. Figure 3 As shown, this is to achieve network load balancing; among which, the PLO optimization mechanism for hierarchical load balancing includes:

[0065] Mapping from individual to path selection scheme: In PLO, each individual represents a potential solution; in HLB-PLO, each individual is mapped to a path combination, that is, each high-energy particle represents a potential communication path selection scheme in the network.

[0066] Dimension-to-communication flow mapping: Each dimension in PLO can be mapped to the path selection of each communication flow in HLB-PLO; in this way, the optimization objective of each dimension is associated with the path selection of the communication flow.

[0067] Fitness function mapping: The fitness function in PLO needs to be redefined to adapt to the scenario-specific objectives; In HLB-PLO, the fitness function is set as the HLB loss function that measures the load balancing performance of the path selection scheme.

[0068] Optimization of the mapping of objectives: The objective of PLO is to find the individual with the best fitness, while the objective of HLB-PLO is to find the path selection scheme with the minimum HLB loss function value.

[0069] The optimization algorithm operates within the solution space, which involves not only the algorithm's search efficiency but also its ability to accurately evaluate the performance of each path selection scheme. After overall optimization, the path selection scheme with the maximum fitness (i.e., the lowest HLB loss) is finally selected.

[0070] Example 2

[0071] This embodiment relates to a high-speed subnet within the same aerospace manufacturing workshop, primarily responsible for transmitting complex data streams (such as quality inspection and equipment operation commands). These data streams require low latency and high bandwidth. In this network environment, traditional OFDMA random access methods struggle to avoid data collisions and unstable transmission latency under high load scenarios, impacting the real-time transmission of critical data.

[0072] This embodiment applies the OFDMA-based TT scheduling strategy in a high-speed subnet. Through TTI and RU constraint optimization, this scheme achieves reasonable resource allocation and ensures conflict-free, low-latency transmission. The specific steps are as follows:

[0073] Step 1, OFDMA Network Model

[0074] To achieve real-time and deterministic transmission, this invention employs a star topology, in which a group of terminal nodes are wirelessly connected to a base station; the communication system uses TDMA to achieve deterministic media access, where time is divided into time slots (e.g., TTI);

[0075] This system achieves simultaneous streaming transmission by allocating adjacent RUs to each terminal node, with each RU containing an equal number of data subcarriers; consider a system containing a set of TT streams, denoted as F, and assume that the transmission of each TT stream requires only one TTI; such as Figure 4 As shown, TT flow f i ∈F by terminal device ED i With period P iIt is generated periodically, with an initial generation time of G. i Then, for this flow, draw from the set {1,2,...,N}. RU In the frequency domain, adjacent RUs are allocated. To maximize the utilization of frequency domain resources, the number of RUs allocated (denoted as RL) is... i ) is set to support f i Minimum required value;

[0076] Step 2, TTI-RU constraint model

[0077] This invention constructs a scheduling constraint model based on TTI-RU using integer programming, called TTI-RU-IP; specifically, it is an OFDMA-based time-triggered network with an O i and RU i Constraints, where O i Represents communication flow f i The initial transmission time of ∈F, RU i Indicates assignment to f i The smallest index among all RUs ∈ F. Since time-triggered streams are generated and sent periodically, this invention only focuses on the scheduling strategy within its initial period.

[0078] The scheduling constraint model includes:

[0079] Constraint 1: Delay Constraint: Each stream should be transmitted immediately after generation, and its corresponding time window should not exceed its period, expressed by the following formal expression:

[0080]

[0081] Constraint 2: RU Continuity Constraint: Each flow should be assigned an RU from the set, represented by the following constraint:

[0082]

[0083] Constraint 3: TTI-RU Collision-Free Constraint: Once any two flows are assigned to the same RU, the TTIs they occupy during transmission must not overlap. This is expressed by the following constraint:

[0084]

[0085] Where G(P) i ,P j ) represents P i and P j The greatest common divisor.

[0086] Experimental results show that this invention improves the performance of virtual-physical collaborative communication within a high-speed subnet through an OFDMA-based time-triggered scheduling strategy. Through joint optimization of TTI and RU, this invention effectively reduces end-to-end data transmission latency and increases scheduling scale. Experiments verify that the scheduling method under joint optimization of TTI and RU exhibits lower latency and higher scheduling efficiency compared to traditional schemes. This invention maintains good real-time performance and high throughput, ensuring the efficiency and reliability of virtual-physical collaborative communication, solving the problems of insufficient real-time responsiveness and network stability in existing technologies, and improving the overall efficiency of virtual-physical collaboration within a digital twin workshop.

[0087] In a digital twin workshop, production elements such as people, machines, and materials are interconnected via a network, enabling real-time transmission of equipment monitoring, workshop management, and production status data. Figure 5 As shown. For low-speed transmission, this invention balances the load of low-speed subnets through routing resource scheduling. For high-speed transmission, this invention uses TTI-RU joint optimization scheduling to schedule time-frequency resources and improve the transmission performance of high-speed subnets.

[0088] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various modifications or variations within the scope of the claims, which do not affect the essence of the present invention.

Claims

1. A virtual-physical collaborative communication scheduling method suitable for digital twin workshops, characterized in that, include: (1) A virtual-real collaborative communication scheduling method for low-speed subnets in digital twin workshops; (2) A virtual-real collaborative communication scheduling method for high-speed subnets in digital twin workshops; The specific steps of the virtual-real collaborative communication scheduling method for low-speed subnets in digital twin workshops are as follows: Step 1, Multi-source single-sink network model The multi-hop topology of IWN is described using a DAG model. In this model, terminal system nodes are abstracted as vertices, and the reachability and communication direction between nodes are represented by directed edges. Here, IWN stands for Industrial Wireless Network, and DAG stands for Directed Acyclic Graph. Step 2, Optimize the algorithm First, a loss function is defined to guide the search and optimization process to minimize the error; For each layer of the network, the difference between the actual load of a node and its theoretical average load is derived, and the square root of the sum of squares of these differences is taken to obtain a comprehensive load balancing metric. This comprehensive load balancing metric will guide the algorithm to find the optimal path to achieve load balancing. Specifically: The difference between the actual load of a single node and the theoretical average load at each level is used as the load balancing metric. The loss function is in the form of root mean square error, expressed as: Among them, V layer |V represents the set of all nodes at a certain level. layer | represents the total number of nodes in the hierarchy, τ LCM Let ω be the least common multiple of the transmission periods of all communication flows, ω be the link transmission delay of the communication flows, LayerNum be the total number of layers, and node v be the least common multiple of the transmission periods of all communication flows. k The actual load is v k .Load can be calculated after the path selection scheme is determined; τ j Let j be the sending period of communication stream j; Heuristic algorithms are used to optimize the path so that the actual load of the nodes is close to the average load. Specifically, the Aurora Optimization Algorithm is used to simulate particle motion by analyzing the motion of high-energy particles and studying the basic principles of physics. The optimal path scheme is found in IWN to achieve efficient data aggregation. Specifically, for each source node, the algorithm finds all shortest paths with the same hop count, and the set of paths forms the basis of the solution space for network path selection. After constructing the path set for each source node, these sets are merged to construct the solution space for the path selection scheme of the entire network. The design concept of PLO is used to construct a PLO optimization mechanism for hierarchical load balancing to achieve network load balancing. The construction of the PLO optimization mechanism for hierarchical load balancing includes: where PLO is the abbreviation for Aurora Optimization Algorithm. Mapping from individual to path selection solution: In PLO, each individual represents a potential solution; Dimension-to-communication flow mapping: Each dimension in PLO can be mapped to the path selection of each communication flow in HLB-PLO; Fitness function mapping: The fitness function in PLO needs to be redefined to adapt to the contextualized objectives; Optimization of the mapping of objectives: The objective of PLO is to find the individual with the best fitness, while the objective of HLB-PLO is to find the path selection scheme with the minimum HLB loss function value.

2. The virtual-physical collaborative communication scheduling method applicable to digital twin workshops as described in claim 1, characterized in that, The specific steps of the virtual-real collaborative communication scheduling method for high-speed subnets in digital twin workshops are as follows: Step 1, OFDMA Network Model A star topology is adopted, in which a group of terminal nodes are connected to the base station wirelessly; the communication system uses TDMA, and time is divided into time slots, which are transmission time intervals (TTI). The system contains a set of TT streams, denoted as F. Each TT stream requires only one time slot for transmission. TT stream f i ∈F by terminal device ED i With period P i It is generated periodically, with an initial generation time of G. i The system achieves simultaneous streaming transmission by allocating adjacent RUs to each terminal node, with each RU containing an equal number of data subcarriers; from the set {1,2,...,N} RU In the}, adjacent RUs are allocated, and the number of allocated RUs is set to support f. i The minimum required value is denoted as RL. i ; Step 2, TTI-RU constraint model A scheduling constraint model, called TTI-RU-IP, was constructed based on TTI-RU using integer programming. This model is based on the OFDMA time-triggered network. i and RU i Constraints, where O i Represents communication flow f i The initial transmission time of ∈F, RU i Indicates assignment to f i The smallest index among all RUs ∈ F; The scheduling constraint model includes: Constraint 1: Delay Constraint: Each stream should be transmitted immediately after generation, and its corresponding time window should not exceed its period, expressed by the following formal expression: Constraint 2: RU Continuity Constraint: Each flow should be assigned an RU from the set, represented by the following constraint: Constraint 3: TTI-RU Collision-Free Constraint: Once any two flows are assigned to the same RU, the TTIs they occupy during transmission must not overlap. This is expressed by the following constraint: Where G(P) i ,P j ) represents P i and P j The greatest common divisor.

Images