5G-TSN network traffic scheduling methods, systems, equipment, and media

Abstract

This application discloses a 5G-TSN network traffic scheduling method, system, device, and medium, applied in the field of communication technology. The method includes: determining the effective priority of a service flow to be scheduled based on service flow parameters, network topology, and network status data using a two-layer priority mapping mechanism; generating decision state data for the service flow to be scheduled based on service flow parameters, effective priorities, network topology, and network status data; inputting the decision state data into a scheduling decision model to obtain the target transmission path of the service flow to be scheduled; performing hop-by-hop time slot allocation based on the target transmission path and effective priority to obtain hop-by-hop time slot allocation results; and generating the actual scheduling configuration based on the target transmission path and hop-by-hop time slot allocation results. By unifying priority mapping, path decision-making, and time slot allocation into the same scheduling closed loop, synergistic optimization of cross-domain priority mapping and intelligent scheduling is achieved.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a 5G-TSN network traffic scheduling method, system, device and medium. Background Technology

[0002] With the development of the Industrial Internet, industrial field operations are placing higher demands on the latency limits, jitter control, reliability, and real-time performance of communication networks. Fifth-generation (5G) networks offer wide-area coverage, flexible wireless access, and ultra-reliable low-latency communication capabilities. Time-Sensitive Networking (TSN) provides deterministic transmission guarantees on the Ethernet side through mechanisms such as unified clock synchronization, gating scheduling, and priority queuing. The integration of these two technologies has become an important technical route for deterministic transmission in industry.

[0003] In existing technologies, 5G-TSN networks typically achieve end-to-end transmission control through cross-domain Quality of Service (QoS) mapping, path selection, and TSN time slot scheduling. However, this approach has several technical shortcomings in practical applications. Static cross-domain QoS mapping: The conversion from 5G QoS Identifier (5QI) to TSN Priority Code Point (PCP) is completed using pre-configured rules or static mapping tables. The mapping result only reflects the static attributes of the service category and cannot reflect dynamic factors such as the timeliness of service information, the urgency of deadline constraints, and the status of network path resource occupancy. This can easily lead to inconsistencies between priority expression and actual service scheduling requirements, making it difficult to ensure the timely provision of critical services. It is difficult to balance scheduling efficiency and quality: Existing path allocation schemes usually suffer from problems such as the solution complexity increasing significantly with the increase of network size, making it difficult to meet online scheduling requirements, and limited adaptability to complex scenarios, which can easily lead to problems such as decreased scheduling success rate and increased local congestion. Priority mapping and scheduling decision-making are disconnected: Existing time slot scheduling schemes usually treat priority mapping as an independent preceding step. The mapping result is passed into the scheduling module as a fixed input, which cannot be synchronously corrected with the service status and network status. Furthermore, there is a lack of unified modeling for issues such as high-priority service protection and priority inversion suppression, resulting in insufficient stability and generalization ability of policy performance. The scheduling strategy is disconnected from the actual execution of the TSN: the existing time slot scheduling scheme only focuses on the optimization of path selection and does not fully take into account the time slot allocation constraints of the TSN. As a result, the scheduling results cannot be directly converted into executable gating configurations on the TSN side, resulting in low engineering feasibility.

[0004] In summary, existing 5G-TSN network traffic scheduling technologies cannot achieve a stable balance between deterministic latency guarantees, resource utilization, and priority consistency, making it difficult to meet the dynamic and complex scheduling needs of industrial scenarios. Summary of the Invention

[0005] This application provides a 5G-TSN network traffic scheduling method, system, device, and medium to address the problem that existing 5G-TSN network traffic scheduling technologies struggle to achieve a stable balance between deterministic latency guarantees, resource utilization, and priority consistency. The technical solution provided by this application is as follows: On the one hand, this application provides a 5G-TSN network traffic scheduling method applied to the control plane, including: The system acquires the service flow parameters of the service flow to be scheduled, as well as the network topology and network status data of the 5G-TSN network. The service flow parameters include the 5G quality of service identifier and end-to-end transmission attributes. The network topology includes each network element node and the communication links between each network element node. Based on service flow parameters, network topology map, and network status data, a two-layer priority mapping mechanism is adopted to determine the effective priority of the service flow to be scheduled. The two-layer priority mapping mechanism includes: mapping the 5G service quality identifier to the basic priority based on the preset mapping relationship, and correcting the basic priority to the effective priority based on end-to-end transmission attributes, network topology map, and network status data. Based on service flow parameters and effective priorities, as well as network topology and network status data, decision status data for the service flows to be scheduled is generated. The decision state data is input into a pre-trained scheduling decision model to obtain the target transmission path of the service flow to be scheduled; the scheduling decision model adopts a reinforcement learning model based on graph neural network and near-end policy optimization network. Based on the target transmission path and effective priority, hop-by-hop time slot allocation is performed in the TSN domain based on multiple condition constraints for the service flow to be scheduled, and the hop-by-hop time slot allocation result is obtained. Based on the target transmission path and hop-by-hop time slot allocation results, the actual scheduling configuration of the service flow to be scheduled is generated and distributed to each network element node on the target transmission path to complete the scheduling of the service flow to be scheduled.

[0006] Optionally, based on service flow parameters, network topology diagram, and network status data, a two-layer priority mapping mechanism is adopted to determine the effective priority of the service flow to be scheduled, including: Based on the preset mapping relationship between 5G service quality identifiers and TSN priority code points, the 5G service quality identifiers are mapped to the basic priority represented by TSN priority code points. Based on end-to-end transmission attributes, network topology, and network status data, the freshness of service information, urgency of deadline constraints, and tension of transmission paths of the service flows to be scheduled are determined. The effective priority is obtained by modifying the basic priority based on the freshness of business information, the urgency of deadline constraints, and the tension of transmission paths.

[0007] Optionally, end-to-end transmission attributes include source node, destination node, service flow period, end-to-end cutoff delay, and single-frame message length; Based on end-to-end transmission attributes, network topology, and network status data, the freshness of service information, urgency of deadline constraints, and tension of transmission paths for the service flows to be scheduled are determined, including: The arrival time of the most recent successful update of the receiving end of the business flow to be scheduled is determined based on the business flow cycle, and the freshness of the business information of the business flow to be scheduled is determined based on the arrival time of the most recent successful update of the receiving end. The remaining available time of the scheduled service flow is calculated based on the end-to-end cutoff delay. The lower bound of the shortest reachable end-to-end delay of the scheduled service flow is estimated based on the network topology. The urgency of the cutoff constraint of the scheduled service flow is determined based on the remaining available time and the lower bound of the shortest reachable end-to-end delay. A set of candidate transmission paths for the service flow to be scheduled is generated based on the source node, the destination node, and the network topology. The transmission path congestion of each candidate transmission path in the set is determined based on the network status data. The transmission path tension of the service flow to be scheduled is determined based on the transmission path congestion of each candidate transmission path.

[0008] Optionally, the basic priority is modified based on the freshness of business information, the urgency of deadline constraints, and the tension of transmission paths to obtain an effective priority, including: The priority adjustment amount is obtained by normalizing and weighting the freshness of business information, the urgency of deadline constraints, and the tension of transmission paths. The effective priority is determined based on the priority adjustment amount and the basic priority.

[0009] Optionally, based on service flow parameters and effective priorities, as well as network topology and network status data, decision status data for the service flow to be scheduled is generated, including: Based on the business flow parameters and effective priority, generate the business flow feature vector of the business flow to be scheduled; Based on the network topology map and network status data, a network status map of the 5G-TSN network is generated. Based on the service flow feature vector and candidate transmission path set of the service flow to be scheduled, as well as the network state diagram of the 5G-TSN network, decision state data of the service flow to be scheduled is generated; among which, the candidate transmission path set is generated based on the network topology diagram and the source node and destination node in the end-to-end transmission attributes.

[0010] Optionally, the decision state data includes service flow feature vectors, a set of candidate transmission paths, and a network state graph; The decision state data is input into a pre-trained scheduling decision model to obtain the target transmission path of the service flow to be scheduled, including: By using a graph neural network, the state graph is encoded to obtain a node embedding matrix. Based on the node embedding matrix, the path representation vector of each candidate transmission path in the candidate transmission path set is extracted. The path representation vector of each candidate transmission path is concatenated with the service flow feature vector to form the decision input data of each candidate transmission path. By optimizing the network through near-end strategies, the candidate path selection probability of each candidate transmission path is calculated based on the decision input data of each candidate transmission path, and the candidate transmission path with the highest candidate path selection probability is determined as the target transmission path of the service flow to be scheduled.

[0011] Optionally, based on the target transmission path and effective priority, hop-by-hop time slot allocation in the TSN domain based on multiple constraints is performed for the service flow to be scheduled, resulting in hop-by-hop time slot allocation results, including: Based on the effective priority, determine the priority queue that the service flow to be scheduled enters on each TSN node included in the target transmission path; Based on the priority queue that the service flow to be scheduled enters at each TSN node, the time slot search space of each hop communication link contained in the target transmission path is determined. Based on multiple constraints, in the time slot search space of each hop communication link included in the target transmission path, the earliest available transmission time slot is searched for each hop communication link to obtain the hop-by-hop time slot allocation result. Among them, the multiple constraints include link time slot mutual exclusion constraints, cross-hop timing constraints, and end-to-end cutoff delay constraints. The link time slot mutual exclusion constraint means that the same transmission time slot of the same priority queue on the same communication link is allocated to only one service flow. The cross-hop timing constraint means that the transmission time of the next hop communication link is not earlier than the sum of the transmission completion time and frame processing delay of the previous hop communication link. The end-to-end cutoff delay constraint means that the actual end-to-end delay of the service flow is less than or equal to the end-to-end cutoff delay.

[0012] On the other hand, this application provides a 5G-TSN network traffic scheduling system applied to the control plane, including: The data acquisition unit is used to acquire the service flow parameters of the service flow to be scheduled, as well as the network topology map and network status data of the 5G-TSN network; among which, the service flow parameters include the 5G service quality identifier and end-to-end transmission attributes; the network topology map includes each network element node and each communication link between each network element node; The priority mapping unit is used to determine the effective priority of the service flow to be scheduled based on service flow parameters, network topology map and network status data, using a two-layer priority mapping mechanism. The two-layer priority mapping mechanism includes: mapping the 5G service quality identifier to the basic priority based on a preset mapping relationship, and correcting the basic priority to the effective priority based on end-to-end transmission attributes, network topology map and network status data. The preprocessing unit is used to generate decision status data for the service flow to be scheduled based on service flow parameters, effective priorities, network topology, and network status data. The scheduling decision unit is used to input decision state data into a pre-trained scheduling decision model to obtain the target transmission path of the service flow to be scheduled; wherein, the scheduling decision model adopts a reinforcement learning model based on graph neural network and near-end policy optimization network; The time slot allocation unit is used to perform hop-by-hop time slot allocation in the TSN domain based on multi-condition constraint verification for the scheduled service flow based on the target transmission path and effective priority, and obtain the hop-by-hop time slot allocation result. The scheduling configuration unit is used to generate the actual scheduling configuration of the service flow to be scheduled based on the target transmission path and the hop-by-hop time slot allocation result, and to distribute it to each network element node on the target transmission path to complete the scheduling of the service flow to be scheduled.

[0013] On the other hand, this application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described 5G-TSN network traffic scheduling method.

[0014] On the other hand, this application also provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the above-described 5G-TSN network traffic scheduling method.

[0015] The beneficial effects of this application are as follows: This application, by comprehensively considering real-time service status, network topology, link resource occupancy, and TSN scheduling constraints, enables more reasonable path and time slot resource allocation in complex topology and dynamic load scenarios. This is beneficial for improving the guarantee capability of high-priority services, reducing the probability of priority inversion, and increasing the overall scheduling success rate and the stability of the scheduling strategy. Furthermore, by introducing a hop-by-hop time slot allocation and constraint verification mechanism, it ensures that the scheduling results are consistent with the actual executable configuration on the TSN side, thereby improving engineering feasibility and practical deployment value.

[0016] Other features and advantages of this application will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram outlining the general flow of the 5G-TSN network traffic scheduling method in the embodiments of this application; Figure 2 This is a schematic diagram of the two-layer priority mapping mechanism in the embodiments of this application; Figure 3 This is a schematic diagram of the scheduling decision model architecture in the embodiments of this application; Figure 4 This is a schematic diagram of the 5G-TSN network traffic scheduling framework in the embodiments of this application; Figure 5 This is a schematic diagram of the composition structure of the 5G-TSN network traffic scheduling system in the embodiments of this application; Figure 6 This is a schematic diagram of the hardware structure of the electronic device in the embodiments of this application. Detailed Implementation

[0018] To make the objectives, technical solutions, and beneficial effects of this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0019] Currently, while 5G-TSN network traffic scheduling solutions can achieve service connectivity between the 5G side and the TSN side through cross-domain collaborative transmission mechanisms and meet the needs of deterministic transmission in industrial scenarios to a certain extent, there are still significant technical shortcomings in actual scheduling processes: First, most existing cross-domain QoS mapping schemes use pre-configured rules or static mapping tables to convert 5QI to PCP. The mapping results primarily reflect the static attributes of the service category, failing to capture dynamic factors such as the timeliness of information, the urgency of deadlines, and the status of network path resource occupancy at the current scheduling moment. For industrial control and status monitoring services, the actual urgency of the same service type may vary significantly across different time windows. When network load fluctuates or local link congestion worsens, continuing to use fixed priorities for TSN queue scheduling can easily lead to inconsistencies between priority expression and the actual scheduling needs of the service. This results in critical services not being guaranteed timely during resource contention, impacting end-to-end latency and deterministic transmission performance.

[0020] Secondly, existing TSN-side path allocation and time slot scheduling methods face the challenge of balancing computational efficiency and scheduling quality in converged network scenarios. While optimization methods based on integer linear programming can achieve good scheduling results under relatively complete constraints, their complexity increases significantly with network size, the number of traffic flows, and the expansion of the path candidate set, making it difficult to meet online or near real-time scheduling requirements. Heuristic methods based on shortest paths and greedy rules, while simple to implement and computationally fast, typically rely on fixed rules for decision-making, limiting their comprehensive adaptability to link resource contention, service priority guarantees, and multi-objective scheduling requirements. In high-load or dynamic topology scenarios, they are prone to decreased overall scheduling success rate, increased local congestion, and insufficient guarantees for high-priority services. In other words, existing methods often only address some objectives under complex constraints, struggling to achieve a stable balance between deterministic latency guarantees, resource utilization, and priority consistency.

[0021] Meanwhile, while existing intelligent scheduling technologies based on graph neural networks and deep reinforcement learning have improved the ability to model topology and network states and can obtain online decision-making strategies through training, they still generally suffer from the problem of disconnect between cross-domain priority mapping and scheduling decision-making processes in 5G-TSN converged scenarios. Existing solutions typically treat priority mapping as a separate preliminary step, and then use the mapping result as a fixed input to the graph model and reinforcement learning module for path selection or time slot allocation decisions. While this approach gives the scheduling strategy a certain degree of intelligence, the priority information input remains relatively coarse-grained, making it difficult to synchronously correct for changes in service and network states. Furthermore, some solutions still favor a particular type of path optimization objective in their reinforcement learning algorithm selection and state representation methods, lacking unified modeling for issues such as high-priority service guarantees, priority inversion suppression, and cross-domain semantic consistency. This results in insufficient policy performance stability and generalization ability when network topology changes, traffic distribution changes, or resource conflicts intensify.

[0022] Therefore, to address the problems of static cross-domain priority mapping, insufficient adaptability of scheduling strategies to complex topologies, and the disconnect between priority mapping and scheduling processes in existing technologies, it is necessary to propose a new 5G-TSN network traffic scheduling method. This application introduces dynamic priority correction based on cross-domain QoS mapping and combines the topology state representation capability of graph neural networks and the online decision-making capability of reinforcement learning to achieve collaborative optimization of cross-domain priority mapping and path time slot scheduling, thereby improving the shortcomings of existing technologies in high-priority service guarantee, priority consistency, and scheduling performance in complex scenarios.

[0023] After introducing the application scenarios and design concepts of this application, the technical solutions provided by this application will be described in detail below.

[0024] This application provides a 5G-TSN network traffic scheduling method, applied to the control plane of a 5G-TSN network. (See attached document.) Figure 1 As shown in the embodiments of this application, the general flow of the 5G-TSN network traffic scheduling method is as follows: Step 101: Upon receiving a service flow to be scheduled, obtain the service flow parameters of the service flow to be scheduled, as well as the network topology and network status data of the 5G-TSN network; wherein, the service flow parameters include the 5G quality of service identifier and end-to-end transmission attributes; the network topology includes each network element node and each communication link between each network element node; the network status data includes network element node status data, network element node resource occupancy data, and communication link status data.

[0025] In this embodiment of the application, the service flow parameters of the service flow to be scheduled can be modeled as a six-tuple. ;in, and These are the source and destination nodes of the service flow to be scheduled, respectively. The service flow period for the service flow to be scheduled represents the service transmission frequency; The end-to-end cutoff delay for the pending service flow. The length of a single frame of the service flow to be scheduled. The 5G Quality of Service (5QI) identifier, defined by 3GPP, reflects the standardized QoS requirements of the scheduled service flow in terms of latency, reliability, packet loss rate, and other dimensions. Single frame message length. This is used to characterize the resource requirements of the scheduled service flow on the communication link, to determine the transmission duration or the required number of time slots for the scheduled service flow on each hop of the communication link included in the target transmission path, and is input into the scheduling decision model as part of the service flow feature vector. After determining the earliest available transmission time slot on each hop of the communication link, the transmission duration of the scheduled service flow on the corresponding communication link is calculated based on the single frame message length and the communication link bandwidth to determine the number of continuously occupied time slots.

[0026] In this embodiment, the network topology graph can be constructed as a directed graph. Specifically, the 5G-TSN network can be constructed as a directed graph based on each network element node in the 5G-TSN network and the communication links between each network element node. This is a network topology diagram for a 5G-TSN network; among which... For network element nodes, This is a set of communication links. The set of network element nodes includes 5G nodes, User Plane Function (UPF) nodes, TSN switch nodes, etc. Different types of network element nodes are distinguished by type labels in the network topology diagram, which are used to subsequently introduce structural information such as whether it is a radio access point or a TSN switch into the node status characteristics. (Communication Link Set) In this network, one part consists of 5G wireless links, primarily carrying uplink and downlink services on the access side, while the other part is a wired Ethernet link, corresponding to the TSN backplane network; each communication link... Associate a set of basic link attributes, including link bandwidth. Propagation delay Current average queuing delay Link type identifier 5G wireless links also consider additional channel quality metrics to characterize the uncertainty of wireless segments in path cost and reward.

[0027] In this embodiment, network status data may include network element node status data, network element node resource occupancy data, and communication link status data. Among them, network element node status data includes the queue occupancy rate, average queuing delay, node type, and current load information of each network element node; network element node resource occupancy data includes the resource occupancy ratio of each network element node for different priority service flows within the current scheduling window; and communication link status data includes the link bandwidth, propagation delay, current utilization rate, and available time slot ratio of each communication link.

[0028] Step 102: Based on service flow parameters, network topology diagram and network status data, a two-layer priority mapping mechanism is used to determine the effective priority of the service flow to be scheduled.

[0029] In this embodiment of the application, by adopting a two-layer priority mapping mechanism, the mapping between 5G-side QoS semantics and TSN-side priority semantics can be realized, while ensuring that the effective priority reflects the true urgency of the service flow to be scheduled at the current scheduling moment. For example... Figure 2 As shown, the two-layer priority mapping mechanism includes two layers of priority mapping mechanisms: The first layer is the static mapping layer: based on the fundamental mapping function (i.e., the preset mapping relationship) between the 5G Quality of Service Identifier (5QI) and the TSN Priority Code Point (PCP), the 5G Quality of Service Identifier is mapped to the basic priority represented by the TSN Priority Code Point; wherein, the fundamental mapping function is defined as follows: In the basic mapping function, For the 5QI set, This represents the PCP value space for service flows in the TSN; for service flows to be scheduled... Define its 5QI as The fundamental priority obtained by performing a fundamental mapping based on the fundamental mapping function can be denoted as: .

[0030] The second layer is the dynamic correction layer: Based on end-to-end transmission attributes, network topology, and network status data, after determining the freshness of service information, urgency of deadline constraints, and tension of transmission paths for the service flows to be scheduled, the basic priority is corrected based on these factors to obtain the effective priority. Specifically, this includes: 1) Business Information Freshness: Business information freshness characterizes the staleness of business information in a scheduled business flow. It can be determined based on the arrival time of the most recent successful update from the receiver of the scheduled business flow, and then the freshness of the business information in the scheduled business flow can be determined based on this arrival time. Specifically, Age of Information (AoI) can be introduced as a measure of business information freshness. (This applies to the business flow to be scheduled.) At the current decision-making moment AoI is defined as ;in, For the current decision-making moment; This is the arrival time of the most recent successful update from the receiver of the service flow to be scheduled, calculated based on the service flow cycle and the update records of the receiver. The larger the value, the older the information, and the more likely it is to prioritize ensuring the timely delivery of this business flow.

[0031] 2) Deadline Constraint Urgency: The deadline constraint urgency characterizes the scheduling urgency of a service flow under end-to-end deadline delay constraints. It calculates the remaining available time of the service flow based on the end-to-end deadline delay, estimates the lower bound of the shortest reachable end-to-end delay based on the network topology, and then determines the deadline constraint urgency of the service flow based on the remaining available time and the lower bound of the shortest reachable end-to-end delay. Specifically, it can be based on the service flow... Generation time and end-to-end cutoff delay Determine the service flow to be scheduled absolute deadline Based on the service flow to be scheduled absolute deadline and the current decision-making moment Determine the service flow to be scheduled Remaining available time Estimating the service flow to be scheduled based on network topology graph The shortest reachable end-to-end delay lower bound The shortest reachable lower bound for end-to-end latency The network topology is obtained by performing a shortest path search. The edge weight of each communication link is set as the sum of the propagation delay, average queuing delay, and fixed processing delay of that link. The minimum time delay path from the source node to the destination node is calculated based on Dijkstra's algorithm, and its path cost is used as the lower bound of the shortest reachable end-to-end delay. Define normalized deadline constraint urgency based on the earliest deadline priority strategy. as follows: (1) Among them, when near hour, Approaching 1 indicates extremely small delay margin and urgent scheduling; when Much larger hour, A value close to 0 indicates that there is still a large scheduling margin.

[0032] 3) Transmission Path Strain: Transmission path strain reflects the degree of strain on path resources. Specifically, it can be based on the traffic flow to be scheduled. source node Heshu Node And the network topology diagram of the 5G-TSN network, generating service flows to be scheduled. Candidate transmission path set Each candidate transmission path in the candidate transmission path set is a path originating from... arrive Acyclic reachable paths are identified, and then the congestion level of each candidate transmission path in the candidate transmission path set is determined based on network state data. And the transmission path congestion level based on each candidate transmission path. Determine the transmission path stress of the service flows to be scheduled For example, the transmission path congestion level for each candidate transmission path. The average value is taken as the transmission path tension of the service flow to be scheduled. Transmission path congestion The calculation formula is as follows: (2) in, Transmission path congestion is defined as the average time slot resource occupancy level of each hop communication link contained in the candidate transmission path. For the first The number of hops in each candidate transmission path; For the first Jump communication link; For the current decision-making moment; Indicates the first The time slot resource occupancy level of the hop communication link within the current gating period is calculated as follows: (3) in, The number of time slots within a gating cycle. For the first Jump communication link At this moment of decision-making The number of time slots already occupied.

[0033] 4) Effective Priority: Prioritization of the freshness of business information. Deadline constraint urgency and transmission path tension Normalization and weighting are performed to obtain the priority adjustment amount. Then, adjust the amount based on priority. and basic priority Determine effective priority Among them, priority adjustment amount The formula is as follows: (4) Considering the freshness of business information The unit of measurement is time, which can be used to measure the freshness of business information. Normalization is performed: (5) in, To determine the freshness of business information after normalization; The supercycle length is the least common multiple of the cycles of the business flows to be scheduled within the current scheduling window; To ensure the freshness of business information.

[0034] To ensure the reproducibility of the mapping mechanism, a deterministic implementation of equation (4) with linear weighting and truncation can be adopted. Specifically, the priority adjustment amount... It can be represented as follows: (6) in, These are the weighting coefficients; This is to prevent situations where priorities cross levels.

[0035] Priority-based adjustment amount and basic priority Determine effective priority : (7) in, For effective priority, Based on priority, Adjust the priority amount.

[0036] The above priority adjustment strategies can enable business flows with excessively fresh information or excessively urgent deadline constraints to receive positive priority adjustments. At the same time, they can take into account network conditions to avoid excessive priority fluctuations, so that the corrected effective priority can reflect both the business category and the actual urgency of the business under the current network conditions.

[0037] Step 103: Based on the service flow parameters and effective priorities, as well as the network topology and network status data, generate decision status data for the service flow to be scheduled; wherein, the decision status data includes the service flow feature vector, the candidate transmission path set, and the network status diagram.

[0038] In this embodiment of the application, at the current decision-making moment Based on the network topology map and network status data, a network status map of the 5G-TSN network is generated. ,in, For network element nodes, For a set of communication links, This represents the network element node status characteristics extracted from the network element node status data, including the queue occupancy rate, average queuing delay, node type, and current load information of each network element node. This represents the communication link status features extracted from the communication link status data, including the link bandwidth, propagation delay, current utilization, and available time slot percentage for each communication link. This represents the priority load statistics matrix of network element nodes based on resource usage data. It consists of priority load statistics vectors for each network element node and is used to characterize the resource usage distribution of different priority services for each network element node within the current scheduling window. For any network element node... Its priority load statistics vector is represented as ,in, Represents network element nodes The highest priority is The resource usage ratio of the business within the current window. Simultaneously, the business flow parameters and effective priorities of the business flows to be scheduled are constructed into a business flow feature vector. and the business flow feature vector Network state diagram and candidate transmission path set Decision status data that makes up the business flow to be scheduled ,in, The feature encoding result of the service flow to be scheduled is the service flow feature vector; Network state diagram; This constitutes a set of candidate transmission paths. The resulting decision state data includes both the topology and resource contention relationships of the 5G-TSN network, as well as priority and timeliness information directly related to the current service.

[0039] Step 104: Input the decision state data into the pre-trained scheduling decision model to obtain the target transmission path of the service flow to be scheduled; wherein, the scheduling decision model adopts a reinforcement learning model based on graph neural network and near-end policy optimization network.

[0040] In this embodiment, the scheduling decision model includes a graph neural network and a near-end policy optimization network. When generating the target transmission path for the service flow to be scheduled using the scheduling decision model, firstly, the graph neural network encodes the network state graph to obtain a node embedding matrix. Based on the node embedding matrix, the path representation vector of each candidate transmission path in the candidate transmission path set is extracted. The path representation vector of each candidate transmission path is then concatenated with the service flow feature vector to form the decision input data for each candidate transmission path. Next, the near-end policy optimization network calculates the candidate path selection probability of each candidate transmission path based on the decision input data, and determines the candidate transmission path with the highest selection probability as the target transmission path for the service flow to be scheduled. For details, see [link to relevant documentation]. Figure 3 As shown, in the scheduling decision model: In the state encoding phase, the graph neural network processes the network state graph. Multi-layer graph convolutional propagation is performed to obtain a node embedding matrix containing topological structure information and resource state information of neighboring nodes, including: First, the state characteristics of network element nodes are... and network element node priority load statistics matrix Concatenate them into the initial node embedding matrix:

[0041] in, The initial node embedding matrix, This indicates the state characteristics of network element nodes. This represents the priority load statistics matrix for network element nodes.

[0042] Then, based on the initial node embedding matrix, a preset number of graph convolution propagation layers are sequentially performed, starting from the first... layer to the first The graph convolution update formula for a layer is:

[0043] in, For the first Layer node embedding matrix; The adjacency matrix after adding self-loops, This is the original adjacency matrix corresponding to the network topology graph, used to represent the connection relationship between each network element node. If there is a communication link between two network element nodes, the corresponding element value in the adjacency matrix is ​​1, otherwise it is 0. It is an identity matrix with the same dimensions as the adjacency matrix; for The correspondence degree matrix is ​​used to eliminate feature bias caused by differences in node degree; For the first Layer node embedding matrix; For the first The layer learnable parameter matrix is ​​used to perform linear transformations on node features; For activation functions, such as Sigmoid.

[0044] The final node embedding matrix is ​​obtained after graph convolution propagation with a preset number of layers. The node is embedded in the matrix. The node embedding vector of each network element node in the network not only contains the node's own information, but also incorporates its own information. The topology and resource status information of all neighboring nodes within the jump range.

[0045] Finally, for each candidate transmission path in the candidate transmission path set According to the candidate transmission path Internet access node connection order, from node embedding matrix The node embedding vectors of each network element node are extracted sequentially and aggregated into a path representation vector. , representing the path as a vector With business flow feature vector This candidate transmission path is formed by splicing together. Decision input data :

[0046] Among them, decision input data Includes effective priority Therefore, the near-end policy optimization network can directly perceive the real-time urgency of the service when selecting a path.

[0047] In the policy decision-making phase, the near-end policy optimization network adopts an Actor-Critic structure, including: First, the Actor network uses the decision input data of each candidate transmission path output by the graph neural network. Each candidate transmission path is scored, and the scores of all candidate transmission paths are transformed into a candidate path selection probability distribution using Softmax:

[0048] in, For scoring functions; These are the parameters of the Actor network; Candidate transmission paths Decision input data; Indicates the candidate transmission path index during the summation process; This represents the total number of candidate transmission paths. The probability distribution for selecting candidate paths represents the data in the decision state. Next choice The probability of each candidate transmission path is calculated, and the sum of the probabilities of selecting all candidate paths is 1, to prevent a candidate transmission path from having an excessively high score and monopolizing the selection. It is an exponential function, which is the standard way of writing softmax. Its function is to turn the score value output by the Actor network for each candidate transmission path into a non-negative number, and then normalize it into a probability.

[0049] Based on the candidate path selection probability distribution, the candidate transmission path with the highest selection probability is selected:

[0050] in, Select a probability distribution for candidate paths; The index of the final selected candidate transmission path is used; for example, i=3 indicates that the third candidate transmission path is selected. After the candidate transmission path set is generated, the reinforcement learning action space defines the candidate transmission path index set. .

[0051] The finally selected candidate transmission path is determined as the target transmission path for the service flow to be scheduled. This target transmission path is denoted as... .

[0052] Then, the Critic network analyzes the decision state data. Decision input data and the output of the Actor network Estimate state value This is used for advantage function estimation and network parameter updates during the training phase. In this way, the structured state representation (i.e., decision input data) output by the graph neural network is directly used in the PPO policy and value evaluation process, realizing the unification of network topology awareness and policy learning.

[0053] During the training phase, the process is executed cyclically in the following order: state construction, graph neural network encoding, PPO path decision-making, hop-by-hop time slot allocation and constraint verification, reward calculation, and state update; among which: The reward function adopts a multi-objective joint form:

[0054] in, For the reward function; The value indicates whether the scheduling was successful or not. The value is 1 for successful scheduling and 0 for scheduling failure (scheduling failure means that the constraint verification failed or the time slot allocation failed). This indicator represents the effectiveness of high-priority service assurance and is linked to the effective priority level (e.g., a value of 1 is assigned if the high-priority service is successfully scheduled, otherwise a value of 0 or a score based on the effective priority level). It is achieved through a weighted priority success metric. Calculated; The delay margin is represented by the end-to-end cutoff delay and the actual end-to-end delay of all communication links on the target transmission path. The degree of path congestion is characterized by a weighted average of the congestion levels of all communication links on the target transmission path. These are preset weighting coefficients. Because... Linked with effective priorities, the strategy training process can continuously strengthen the priority protection behavior for high-urgency business.

[0055] PPO updates use a pruning objective function:

[0056] in, The calculation is as follows:

[0057] in, To prune the objective function, For the parameters of the Actor network, In order to make decisions at the current moment Calculate the expected value (batch update, ensuring stability); This is the probability ratio between the new and old strategies (measuring the difference between the current and old strategies); For the current decision-making moment The advantage function estimate is used to measure the performance of decision state data. Select path The degree of superiority or inferiority relative to the average level can be obtained through calculations of actual rewards and state value; This is a clipping function used to clip... Limited to Between, avoid Too large or too small a value will result in excessively large parameter update fluctuations. The hyperparameters are used to prune and limit the magnitude of a single policy update. It is a minimum value function; To update the decision state data for the new strategy Select path The probability of; To update the decision state data of the old strategy before updating Select path The probability of.

[0058] Critic networks use state values Error is minimized when updating parameters. During the online scheduling phase, the trained model parameters are used for path selection and time slot scheduling without changing the algorithm flow; parameter updates are either not performed or performed very infrequently.

[0059] Through the above steps, a complete execution link is formed from network state graph construction, graph neural network encoding, PPO path decision to TSN hop-by-hop time slot allocation and constraint verification. Furthermore, the dynamic priority mapping result runs through the entire process of service feature construction, path decision and scheduling execution, thereby realizing the collaborative optimization of path selection and time slot allocation in the 5G-TSN converged network.

[0060] Step 105: Based on the target transmission path and effective priority, perform hop-by-hop time slot allocation in the TSN domain based on multiple condition constraints for the service flow to be scheduled, and obtain the hop-by-hop time slot allocation result.

[0061] In this embodiment of the application, when performing hop-by-hop slot allocation in the TSN domain based on multiple constraints for the service flow to be scheduled, based on the target transmission path and effective priority, the following methods may be used, but are not limited to: First, based on the effective priority, determine the priority queue that the scheduled service flow will enter on each TSN node included in the target transmission path.

[0062] Then, based on the priority queue that the service flow to be scheduled enters at each TSN node, the time slot search space of each hop communication link contained in the target transmission path is determined.

[0063] Finally, based on multiple constraints, in the time slot search space of each hop communication link included in the target transmission path, the earliest available transmission time slot is searched for each hop communication link to obtain the hop-by-hop time slot allocation result. Among them, the multiple constraints include link time slot mutual exclusion constraints, cross-hop timing constraints, and end-to-end cutoff delay constraints. The link time slot mutual exclusion constraint means that the same transmission time slot of the same priority queue on the same communication link is allocated to only one service flow. The cross-hop timing constraint means that the transmission time of the next hop communication link is not earlier than the sum of the transmission completion time and frame processing delay of the previous hop communication link. The end-to-end cutoff delay constraint means that the actual end-to-end delay of the service flow is less than or equal to the end-to-end cutoff delay.

[0064] Specifically, during the scheduling and execution phase, based on the target transmission path and the effective priority of the service flow to be scheduled Perform hop-by-hop slot allocation in the TSN domain: First, based on First, determine the priority queue that the scheduled service flow enters at each TSN node included in the target transmission path. Then, based on the priority queue that the scheduled service flow enters at each TSN node, determine the time slot search space for each hop communication link included in the target transmission path. Following the order of the communication links in the target transmission path, sequentially perform the earliest feasible time slot search for each hop communication link in the time slot search space of each hop communication link to obtain the hop-by-hop time slot allocation result. Let the th... The arrival time of the service that hops the communication link is (For the first hop communication link, (usually the known schedulable time for the service), then in the... In the priority queue of the hop communication link, that is, in the priority queue that the TSN node acting as the sender enters, the priority is selected based on the priority of the hop communication link. The earliest available transmission slot And based on the earliest available transmission slot Calculate the first The completion time of transmission of the hop communication link is taken as the first... The arrival time of the service in the next hop. This hop-by-hop allocation process is executed sequentially along the target transmission path until the last hop, thereby obtaining the hop-by-hop time slot allocation result of the service flow to be scheduled.

[0065] During hop-by-hop time slot allocation, multi-condition constraint verification is performed synchronously. These constraints include at least link time slot mutual exclusion constraints, hop-crossing timing constraints, and end-to-end deadline constraints. Link time slot mutual exclusion constraints ensure that only one service flow is allocated to the same time slot in the same queue on the same link; hop-crossing timing constraints ensure that the transmission time of the subsequent hop is not earlier than the completion time of the transmission of the previous hop, plus necessary processing delays; and end-to-end deadline constraints require specific end-to-end delays. satisfy:

[0066] in, This is the actual end-to-end delay. This is the end-to-end cutoff delay.

[0067] If any constraint is not met, the scheduling of the service flow to be scheduled is determined to have failed, the temporarily occupied resources are released and the failure result is recorded; if all constraints are met, the scheduling of the service flow to be scheduled is determined to have succeeded.

[0068] Furthermore, in order to quantitatively measure the effectiveness of the scheduling strategy in ensuring high-priority services during the training phase, the set of service flows to be scheduled within the current scheduling window is analyzed. Each business flow to be scheduled Based on effective priority Determine the weighted priority success metric :

[0069] in, A weighted priority success metric; For business flow; The set of service flows to be scheduled within the current scheduling window; Effective priority; For indicator functions, indicator functions Indicates the service flow to be scheduled Whether scheduling was successful within the current scheduling window. When scheduling is successful, ,when When scheduling fails, .

[0070] Weighted Priority Success Metrics This indicator is used to quantitatively evaluate the scheduling results within the current scheduling window, reflecting the scheduling guarantee status of high-priority services. Compared with only counting the overall scheduling success rate, this indicator can highlight whether key services are successfully scheduled with priority.

[0071] Step 106: Based on the target transmission path and hop-by-hop time slot allocation results, generate the actual scheduling configuration of the service flow to be scheduled and distribute it to each network element node on the target transmission path to complete the scheduling of the service flow to be scheduled.

[0072] In this embodiment of the application, after generating the target transmission path and hop-by-hop time slot allocation results for the service flow to be scheduled, based on the target transmission path and hop-by-hop time slot allocation results, a gate control list (GCL) is used to generate the actual scheduling configuration (including the receiver R) of each hop communication link included in the target transmission path for the service flow to be scheduled. X Mapping configuration and sender T X The process involves gating configuration and updating the communication link utilization, available time slot percentage, node queue occupancy, and priority load statistics in the network state graph. Through this execution process, the path selection results output by the near-end policy optimization network are converted into actual scheduling configurations that satisfy TSN constraints.

[0073] In summary, compared with existing 5G-TSN network traffic scheduling technologies, this application adopts a collaborative optimization framework that better reflects the actual operating state of converged networks in the cross-domain QoS mapping and scheduling decision-making process. Based on the 5G-side 5QI to TSN-side PCP basic mapping relationship, it introduces factors such as service information freshness, deadline constraint urgency, and transmission path tension to dynamically correct service priorities. Furthermore, it combines graph neural networks and the PPO algorithm to achieve intelligent decision-making for path selection and time slot scheduling. This not only achieves cross-domain QoS semantic correspondence but also allows service priority expression to be adjusted in real time according to changes in service and network states. This makes the scheduling process more consistent with the actual transmission needs of deterministic services, and helps to improve the problems of static priority mapping and inconsistencies between mapping results and scheduling requirements in existing technologies.

[0074] See Figure 4As shown, after receiving the service flow to be scheduled, the control plane first extracts the service flow parameters and performs cross-domain priority mapping, and then dynamically corrects the basic mapping results to obtain an effective priority that reflects the real-time urgency of the service. Subsequently, it reads the current resource status of the 5G-TSN network, constructs a network state graph containing node status, link resource occupancy status, and priority load information, and uses a graph neural network to extract a structured representation of the topology and resource contention relationships. Based on this, after optimizing the target transmission path output by the network using near-end policies, it combines hop-by-hop slot allocation and constraint verification on the TSN side to complete the actual scheduling configuration generation, and feeds the execution results back to the state update and policy training process. Through the above process, priority mapping, path decision-making, slot execution, and feedback updates can be unified into a single scheduling closed loop, thereby achieving collaborative optimization of cross-domain priority mapping and intelligent scheduling. Moreover, by comprehensively considering the real-time service status, network topology, link resource occupancy, and TSN scheduling constraints, more reasonable path and slot resource allocation can be achieved in complex topology and dynamic load scenarios, which is beneficial for improving the guarantee capability of high-priority services, reducing the probability of priority inversion, and improving the overall scheduling success rate and the stability of the scheduling policy. Meanwhile, the introduction of a hop-by-hop time slot allocation and constraint verification mechanism after the strategy output ensures that the scheduling results are consistent with the actual executable configuration on the TSN side, thereby improving the engineering feasibility and practical deployment value of the solution.

[0075] Based on the above embodiments, this application provides a 5G-TSN network traffic scheduling system applied to the control plane, see below. Figure 5 As shown, the 5G-TSN network traffic scheduling system 500 provided in this application embodiment includes at least: The data acquisition unit 501 is used to acquire the service flow parameters of the service flow to be scheduled, as well as the network topology map and network status data of the 5G-TSN network; wherein, the service flow parameters include the 5G service quality identifier and end-to-end transmission attributes; the network topology map includes each network element node and each communication link between each network element node; The priority mapping unit 502 is used to determine the effective priority of the service flow to be scheduled based on the service flow parameters, network topology map and network status data, using a two-layer priority mapping mechanism. The two-layer priority mapping mechanism includes: mapping the 5G service quality identifier to the basic priority based on a preset mapping relationship, and correcting the basic priority to the effective priority based on end-to-end transmission attributes, network topology map and network status data. The preprocessing unit 503 is used to generate decision status data of the service flow to be scheduled based on the service flow parameters, effective priority, network topology map and network status data. The scheduling decision unit 504 is used to input decision state data into a pre-trained scheduling decision model to obtain the target transmission path of the service flow to be scheduled; wherein, the scheduling decision model adopts a reinforcement learning model based on graph neural network and near-end policy optimization network; The time slot allocation unit 505 is used to perform hop-by-hop time slot allocation in the TSN domain based on multi-condition constraint verification for the service flow to be scheduled, based on the target transmission path and effective priority, and to obtain the hop-by-hop time slot allocation result. The scheduling configuration unit 506 is used to generate the actual scheduling configuration of the service flow to be scheduled based on the target transmission path and the hop-by-hop time slot allocation result, and to send it to each network element node on the target transmission path to complete the scheduling of the service flow to be scheduled.

[0076] In one possible implementation, the priority mapping unit 502 is used to map the 5G service quality identifier to a basic priority represented by the TSN priority code point based on a preset mapping relationship between the 5G service quality identifier and the TSN priority code point; determine the service information freshness, deadline constraint urgency, and transmission path tension of the service flow to be scheduled based on end-to-end transmission attributes, network topology map, and network status data; and modify the basic priority based on the service information freshness, deadline constraint urgency, and transmission path tension to obtain an effective priority.

[0077] In one possible implementation, end-to-end transmission attributes include source node, destination node, traffic flow period, end-to-end cutoff delay, and single-frame message length. Priority mapping unit 502 is used to determine the arrival time of the most recent successful update of the receiving end of the service flow to be scheduled based on the service flow cycle, and to determine the freshness of the service information of the service flow to be scheduled based on the arrival time of the most recent successful update of the receiving end; to calculate the remaining available time of the service flow to be scheduled based on the end-to-end cutoff delay, to estimate the lower bound of the shortest reachable end-to-end delay of the service flow to be scheduled based on the network topology map, and to determine the urgency of the cutoff constraint of the service flow to be scheduled based on the remaining available time and the lower bound of the shortest reachable end-to-end delay; to generate a set of candidate transmission paths for the service flow to be scheduled based on the source node, the destination node and the network topology map, to determine the transmission path congestion of each candidate transmission path in the candidate transmission path set based on network state data, and to determine the transmission path tension of the service flow to be scheduled based on the transmission path congestion of each candidate transmission path.

[0078] In one possible implementation, the priority mapping unit 502 is used to normalize and weight the freshness of business information, the urgency of deadline constraints, and the tension of transmission paths to obtain a priority adjustment amount; and to determine the effective priority based on the priority adjustment amount and the basic priority.

[0079] In one possible implementation, the base preprocessing unit 503 is used to generate a service flow feature vector of the service flow to be scheduled based on service flow parameters and effective priorities; generate a network state diagram of the 5G-TSN network based on the network topology diagram and network state data; and generate decision state data of the service flow to be scheduled based on the service flow feature vector of the service flow to be scheduled, a set of candidate transmission paths, and the network state diagram of the 5G-TSN network; wherein the set of candidate transmission paths is generated based on the network topology diagram and the source node and destination node in the end-to-end transmission attributes.

[0080] In one possible implementation, the decision state data includes a service flow feature vector, a set of candidate transmission paths, and a network state graph. The scheduling decision unit 504 is used to obtain a node embedding matrix by performing state encoding on the network state graph through a graph neural network, extract the path representation vector of each candidate transmission path in the candidate transmission path set based on the node embedding matrix, and concatenate the path representation vector of each candidate transmission path with the service flow feature vector to form the decision input data of each candidate transmission path. Through the near-end policy optimization network, based on the decision input data of each candidate transmission path, the candidate path selection probability of each candidate transmission path is calculated, and the candidate transmission path with the highest candidate path selection probability is determined as the target transmission path of the service flow to be scheduled.

[0081] In one possible implementation, the time slot allocation unit 505 is used to determine, based on effective priority, the priority queue that the scheduled service flow enters at each TSN node included in the target transmission path; based on the priority queue that the scheduled service flow enters at each TSN node, determine the time slot search space for each hop communication link included in the target transmission path; and based on multiple constraints, search for the earliest available transmission time slot for each hop communication link in the time slot search space included in the target transmission path to obtain a hop-by-hop time slot allocation result; wherein, the multiple constraints include link time slot mutual exclusion constraints, hop-crossing timing constraints, and end-to-end cutoff delay constraints; the link time slot mutual exclusion constraint is that the same communication link, the same priority queue, and the same transmission time slot are allocated to only one service flow; the hop-crossing timing constraint is that the transmission time of the subsequent hop communication link is not earlier than the sum of the transmission completion time of the previous hop communication link and the frame processing delay; and the end-to-end cutoff delay constraint is that the actual end-to-end delay of the service flow is less than or equal to the end-to-end cutoff delay.

[0082] It should be noted that the principle of the 5G-TSN network traffic scheduling system 500 provided in this application embodiment to solve the technical problem is similar to that of the 5G-TSN network traffic scheduling method provided in this application embodiment. Therefore, the implementation of the 5G-TSN network traffic scheduling system 500 provided in this application embodiment can refer to the implementation of the 5G-TSN network traffic scheduling method provided in this application embodiment, and the repeated parts will not be described again.

[0083] After introducing the 5G-TSN network traffic scheduling method and system provided in the embodiments of this application, the electronic equipment provided in the embodiments of this application will be briefly introduced next.

[0084] The electronic device provided in this application embodiment may be, but is not limited to, a control plane in a 5G-TSN network, etc. See also Figure 6 As shown, the electronic device 600 provided in this application embodiment includes at least a processor 601, a memory 602, and a computer program stored in the memory 602 and executable on the processor 601. When the processor 601 executes the computer program, it implements the 5G-TSN network traffic scheduling method provided in this application embodiment.

[0085] In one possible implementation, processor 601 can be a single processing element or a collective term for multiple processing elements. For example, processor 601 can be a central processing unit (CPU), or one or more integrated circuits configured to implement the 5G-TSN network traffic scheduling method described in the embodiments of this application. Specifically, processor 601 can be a general-purpose processor, including but not limited to CPUs, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc.

[0086] In one possible implementation, memory 602 may include a readable medium in the form of volatile memory, such as random access memory (RAM) 6021 and / or cache memory 6022, and may further include read-only memory (ROM) 6023; memory 602 may also include a program tool 6025 having a set (at least one) of program modules 6024, including but not limited to: operating subsystem, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0087] In one possible implementation, the electronic device 600 provided in this application embodiment may further include a bus 603 connecting different components (including processor 601 and memory 602). The bus 603 represents one or more types of bus structures, including memory bus, peripheral bus, local area bus, etc.

[0088] In one possible implementation, the electronic device 600 can also communicate with one or more external devices 604, including but not limited to devices that enable user interaction (e.g., mobile phones, computers, etc.), and / or devices that enable it to communicate with one or more other electronic devices (e.g., routers, modems, etc.). This communication can be performed via an input / output (I / O) interface 605. Furthermore, the electronic device 600 can also communicate with one or more networks (e.g., local area networks, wide area networks, public networks, etc.) via a network adapter 606. Figure 6 As shown, network adapter 606 communicates with other modules of electronic device 600 via bus 603. It should be understood that, although... Figure 6 As not shown in the diagram, other hardware and / or software modules may be used in conjunction with the electronic device 600, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, Redundant Arrays of Independent Disks (RAID) subsystems, tape drives, and data backup storage subsystems.

[0089] It should be noted that, Figure 6 The electronic device 600 shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0090] Furthermore, this application also provides a computer-readable storage medium storing computer instructions. When executed by a processor, these computer instructions implement the 5G-TSN network traffic scheduling method described above in this application. Specifically, the computer instructions may be built into or installed in a processor, enabling the processor to implement the 5G-TSN network traffic scheduling method described above in this application by executing the built-in or installed computer instructions.

[0091] Of course, the 5G-TSN network traffic scheduling method provided in the embodiments of this application can also be implemented as a program product, which includes program code. When the program code is executed by a processor, it implements the 5G-TSN network traffic scheduling method provided in the embodiments of this application.

[0092] The program product provided in this application embodiment can be any combination of one or more readable media, wherein the readable media can be a readable signal medium or a readable storage medium, and the readable storage medium can be, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or apparatus, or any combination thereof. Specifically, more specific examples of readable storage media (a non-exhaustive list) include: electrical connections with one or more wires, portable disks, hard disks, RAM, ROM, erasable programmable read-only memory (EPROM), optical fibers, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0093] It should be noted that although several units or sub-units of the device have been mentioned in the detailed description above, this division is merely exemplary and not mandatory. In fact, according to embodiments of this application, the features and functions of two or more units described above can be embodied in one unit. Conversely, the features and functions of one unit described above can be further divided and embodied by multiple units.

[0094] Furthermore, although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0095] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0096] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of the embodiments of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

Claims

1. A 5G-TSN network traffic scheduling method, characterized in that, Applied to the control plane, including: The system acquires the service flow parameters of the service flow to be scheduled, as well as the network topology and network status data of the 5G-TSN network; wherein, the service flow parameters include 5G quality of service identifiers and end-to-end transmission attributes; the network topology includes each network element node and each communication link between the network element nodes; Based on the service flow parameters, the network topology map, and the network status data, a two-layer priority mapping mechanism is adopted to determine the effective priority of the service flow to be scheduled; wherein, the two-layer priority mapping mechanism includes: mapping the 5G service quality identifier to a basic priority based on a preset mapping relationship, and correcting the basic priority to the effective priority based on the end-to-end transmission attributes, the network topology map, and the network status data; Based on the service flow parameters, the effective priority, the network topology, and the network status data, decision status data for the service flow to be scheduled is generated. The decision state data is input into a pre-trained scheduling decision model to obtain the target transmission path of the service flow to be scheduled; wherein, the scheduling decision model adopts a reinforcement learning model based on graph neural network and near-end policy optimization network; Based on the target transmission path and the effective priority, hop-by-hop time slot allocation in the TSN domain based on multiple condition constraints is performed for the service flow to be scheduled, and the hop-by-hop time slot allocation result is obtained. Based on the target transmission path and the hop-by-hop time slot allocation result, the actual scheduling configuration of the service flow to be scheduled is generated and sent to each network element node on the target transmission path to complete the scheduling of the service flow to be scheduled.

2. The 5G-TSN network traffic scheduling method as described in claim 1, characterized in that, Based on the service flow parameters, the network topology map, and the network status data, a two-layer priority mapping mechanism is used to determine the effective priority of the service flow to be scheduled, including: Based on the preset mapping relationship between 5G service quality identifiers and TSN priority code points, the 5G service quality identifiers are mapped to the basic priority represented by TSN priority code points. Based on the end-to-end transmission attributes, the network topology, and the network status data, the freshness of the service information, the urgency of the deadline constraint, and the tension of the transmission path of the service flow to be scheduled are determined. The effective priority is obtained by modifying the basic priority based on the freshness of the business information, the urgency of the deadline constraint, and the tension of the transmission path.

3. The 5G-TSN network traffic scheduling method as described in claim 2, characterized in that, The end-to-end transmission attributes include source node, destination node, service flow period, end-to-end cutoff delay, and single-frame message length. Based on the end-to-end transmission attributes, the network topology, and the network status data, the freshness of the service information, the urgency of the deadline constraint, and the tension of the transmission path for the service flow to be scheduled are determined, including: Based on the business flow cycle, determine the arrival time of the most recent successful update of the receiving end of the business flow to be scheduled, and determine the freshness of the business information of the business flow to be scheduled based on the arrival time of the most recent successful update of the receiving end. The remaining available time of the scheduled service flow is calculated based on the end-to-end cutoff delay. The lower bound of the shortest reachable end-to-end delay of the scheduled service flow is estimated based on the network topology. The urgency of the cutoff constraint of the scheduled service flow is determined based on the remaining available time and the lower bound of the shortest reachable end-to-end delay. A set of candidate transmission paths for the service flow to be scheduled is generated based on the source node, the destination node, and the network topology. The transmission path congestion degree of each candidate transmission path in the set is determined based on the network status data. The transmission path tension of the service flow to be scheduled is determined based on the transmission path congestion degree of each candidate transmission path.

4. The 5G-TSN network traffic scheduling method as described in claim 2, characterized in that, Based on the freshness of the business information, the urgency of the deadline constraint, and the tension of the transmission path, the basic priority is modified to obtain the effective priority, including: The freshness of the business information, the urgency of the deadline constraint, and the tension of the transmission path are normalized and weighted to obtain the priority adjustment amount. The effective priority is determined based on the priority adjustment amount and the base priority.

5. The 5G-TSN network traffic scheduling method as described in claim 1, characterized in that, Based on the service flow parameters, the effective priority, the network topology, and the network status data, decision status data for the service flow to be scheduled is generated, including: Based on the service flow parameters and the effective priority, generate the service flow feature vector of the service flow to be scheduled; Based on the network topology map and the network status data, a network status map of the 5G-TSN network is generated; Based on the service flow feature vector and candidate transmission path set of the service flow to be scheduled, as well as the network state diagram of the 5G-TSN network, decision state data of the service flow to be scheduled is generated; wherein, the candidate transmission path set is generated based on the network topology diagram and the source node and destination node in the end-to-end transmission attributes.

6. The 5G-TSN network traffic scheduling method as described in claim 1, characterized in that, The decision state data includes service flow feature vectors, candidate transmission path sets, and network state diagrams. The decision state data is input into a pre-trained scheduling decision model to obtain the target transmission path of the service flow to be scheduled, including: The graph neural network is used to encode the state graph to obtain a node embedding matrix. Based on the node embedding matrix, the path representation vector of each candidate transmission path in the candidate transmission path set is extracted. The path representation vector of each candidate transmission path is concatenated with the service flow feature vector to form the decision input data of each candidate transmission path. By optimizing the network through a near-end strategy, the candidate path selection probability of each candidate transmission path is calculated based on the decision input data of each candidate transmission path, and the candidate transmission path with the highest candidate path selection probability is determined as the target transmission path of the service flow to be scheduled.

7. The 5G-TSN network traffic scheduling method as described in any one of claims 1-6, characterized in that, Based on the target transmission path and the effective priority, hop-by-hop time slot allocation based on multiple constraints in the TSN domain is performed on the service flow to be scheduled, resulting in hop-by-hop time slot allocation results, including: Based on the effective priority, determine the priority queue that the scheduled service flow enters on each TSN node included in the target transmission path; Based on the priority queue that the scheduled service flow enters at each TSN node, the time slot search space of each hop communication link contained in the target transmission path is determined. Based on the aforementioned multi-condition constraints, in the time slot search space of each hop communication link included in the target transmission path, the earliest available transmission time slot is searched for each hop communication link to obtain the hop-by-hop time slot allocation result; wherein, the multi-condition constraints include link time slot mutual exclusion constraints, hop-crossing timing constraints, and end-to-end cutoff delay constraints; the link time slot mutual exclusion constraint is that the same transmission time slot of the same priority queue on the same communication link is allocated to only one service flow; the hop-crossing timing constraint is that the transmission time of the next hop communication link is not earlier than the sum of the transmission completion time and frame processing delay of the previous hop communication link; the end-to-end cutoff delay constraint is that the actual end-to-end delay of the service flow is less than or equal to the end-to-end cutoff delay.

8. A 5G-TSN network traffic scheduling system, characterized in that, Applied to the control plane, including: The data acquisition unit is used to acquire the service flow parameters of the service flow to be scheduled, as well as the network topology map and network status data of the 5G-TSN network; wherein, the service flow parameters include 5G service quality identifier and end-to-end transmission attributes; the network topology map includes each network element node and each communication link between the network element nodes; The priority mapping unit is used to determine the effective priority of the service flow to be scheduled based on the service flow parameters, the network topology map, and the network status data using a two-layer priority mapping mechanism; wherein, the two-layer priority mapping mechanism includes: mapping the 5G service quality identifier to a basic priority based on a preset mapping relationship, and correcting the basic priority to the effective priority based on the end-to-end transmission attributes, the network topology map, and the network status data; The preprocessing unit is used to generate decision status data for the service flow to be scheduled based on the service flow parameters, the effective priority, the network topology, and the network status data. The scheduling decision unit is used to input the decision state data into a pre-trained scheduling decision model to obtain the target transmission path of the service flow to be scheduled; wherein, the scheduling decision model adopts a reinforcement learning model based on graph neural network and near-end policy optimization network; The time slot allocation unit is used to perform hop-by-hop time slot allocation in the TSN domain based on the target transmission path and the effective priority for the scheduled service flow, and obtain the hop-by-hop time slot allocation result. The scheduling configuration unit is used to generate the actual scheduling configuration of the service flow to be scheduled based on the target transmission path and the hop-by-hop time slot allocation result, and to send it to each network element node on the target transmission path to complete the scheduling of the service flow to be scheduled.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the 5G-TSN network traffic scheduling method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the 5G-TSN network traffic scheduling method as described in any one of claims 1-7.

Images