Cross-domain deterministic data transfer control method, system and apparatus

Abstract

The application provides a cross-domain deterministic data transmission control method, system and device, the method comprising: determining the sending time slot offset value of each data packet in the cross-domain network during cross-domain transmission based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model; generating a configuration table for data stream transmission scheduling based on the sending time slot offset value of each data packet in the cross-domain network during cross-domain transmission and sending the configuration table to the communication nodes in the cross-domain network; the input feature vector corresponding to any data packet is determined based on data packet information and egress port queue state information, and the egress port queue state information comprises the egress port queue state information of the first edge node and the second edge node when the data packet arrives at the first edge node; and the time slot mapping network model is obtained by training the input feature vector sample with the sending time slot offset value label. The automation degree of cross-domain deterministic network time slot mapping is improved.

Description

Technical Field

[0001] This invention relates to the field of network technology, and in particular to a method, system and apparatus for cross-domain deterministic data transmission control. Background Technology

[0002] Building cross-domain industrial internet networks has become an important development trend, and how to provide deterministic quality of service in cross-domain network environments has become a major problem that future industrial internet network systems urgently need to solve.

[0003] Integrated Deterministic Networks (IDNs) employ Time-Sensitive Networking (TSN) for local access and Deterministic Networking (DetNet) for wide-area interconnection. IDNs benefit from both local access and wide-area transmission, thus providing a feasible cross-domain transmission model. Because both TSN and Deterministic Networking rely on time-slot mapping to provide deterministic quality of service, IDNs also depend on a time-slot mapping model to provide deterministic quality of service.

[0004] Existing time slot mapping mechanisms for local-to-wide-area and wide-area-to-local transmission nodes in integrated deterministic network systems are generally designed manually, resulting in insufficient automation in the design process. Furthermore, the design requires trade-offs among multiple performance metrics, leading to high design difficulty and cost. In addition, existing time slot mapping mechanisms lack the ability to generalize and identify unknown data streams. When a new data stream appears, a feasible orchestration table needs to be recalculated by the scheduling algorithm, but since scheduling models are generally nondeterministic polynomial-time complex combinatorial optimization problems, the computation process is quite time-consuming. Summary of the Invention

[0005] To address the problems existing in the prior art, the present invention provides a cross-domain deterministic data transmission control method, system, and apparatus.

[0006] In a first aspect, the present invention provides a cross-domain deterministic data transmission control method, comprising:

[0007] Based on the input feature vector corresponding to each data packet of the data stream to be transmitted, and the time slot mapping network model, the transmission time slot offset value of each data packet during cross-domain transmission in the cross-domain network is determined.

[0008] Based on the transmission time slot offset value of each data packet during cross-domain transmission in the cross-domain network, a configuration table for data stream transmission scheduling is generated, and the configuration table is sent to the communication nodes in the cross-domain network.

[0009] The input feature vector corresponding to any data packet is determined based on the information of the data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first edge node and the second edge node when the data packet arrives at the first edge node. The first edge node and the second edge node belong to different network domains, and the second edge node is the next-hop node of the first edge node.

[0010] The time slot mapping network model is trained based on input feature vector samples with labels for transmission time slot offset values.

[0011] Optionally, the time slot mapping network model is a feedforward network structure, which includes convolutional layers, linear rectified ReLU fully connected layers, and flexible maximum softmax layers;

[0012] The convolutional layer is used to extract local feature information from the input feature vector;

[0013] The ReLU fully connected layer is used to linearly rectify and integrate the local feature information extracted by the convolutional layer to obtain the classification value of the input feature vector corresponding to different transmission time slot offset values.

[0014] The softmax layer is used to calculate the probability value of the input feature vector corresponding to different transmission slot offset value categories based on the classification value, and output the transmission slot offset value corresponding to the maximum probability value.

[0015] Optionally, the cross-domain network includes at least one Time-Sensitive Networking (TSN) network domain and at least one Deterministic Internet Protocol (DIP) network domain.

[0016] Optionally, determining the transmission time slot offset value of each data packet during cross-domain transmission in a cross-domain network based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model includes:

[0017] In the case of cross-domain transmission of target data packets from TSN network domain to DIP network domain, the transmission time slot offset value of the target data packets during cross-domain transmission is determined based on the first input feature vector corresponding to the target data packets and the first time slot mapping network model.

[0018] The first input feature vector is determined based on the information of the target data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first TSN node and the next-hop DIP node when the target data packet arrives at the first TSN node. The first TSN node is an edge TSN node in the TSN network domain that is connected to the DIP network domain.

[0019] The first time slot mapping network model is trained based on a first input feature vector sample with a label for the transmission time slot offset value.

[0020] Optionally, determining the transmission time slot offset value of each data packet during cross-domain transmission in a cross-domain network based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model includes:

[0021] In the case of cross-domain transmission of target data packets from DIP network domain to TSN network domain, the transmission time slot offset value of the target data packets during cross-domain transmission is determined based on the second input feature vector corresponding to the target data packets and the second time slot mapping network model.

[0022] The second input feature vector is determined based on the information of the target data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first DIP node and the next-hop TSN node when the target data packet arrives at the first DIP node. The first DIP node is an edge DIP node in the DIP network domain that is connected to the TSN network domain.

[0023] The second time slot mapping network model is trained based on a second input feature vector sample with a label for the transmission time slot offset value.

[0024] Optionally, the information of the data packet includes the utility value of the data packet, the deadline of the data packet, and the size of the data packet.

[0025] Secondly, the present invention also provides a cross-domain deterministic data transmission control system, comprising:

[0026] The software-defined networking (SDN) controller, the DIP network domain, and multiple TSN network domains connected to the DIP network domain; each TSN network domain contains one or more TSN nodes, and each DIP network domain contains one or more DIP nodes.

[0027] The SDN controller is used to execute the cross-domain deterministic data transmission control method described in the first aspect above;

[0028] The TSN node and the DIP node are used to perform time slot mapping and forwarding operations on the received data packets based on the configuration table sent by the SDN controller, so as to realize the cross-domain transmission of deterministic data streams.

[0029] Thirdly, the present invention also provides a cross-domain deterministic data transmission control device, comprising:

[0030] The determination module is used to determine the transmission time slot offset value of each data packet during cross-domain transmission in a cross-domain network based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model.

[0031] The sending module is used to generate a configuration table for data stream transmission scheduling based on the sending time slot offset value of each data packet during cross-domain transmission in the cross-domain network, and send the configuration table to the communication node in the cross-domain network.

[0032] The input feature vector corresponding to any data packet is determined based on the information of the data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first edge node and the second edge node when the data packet arrives at the first edge node. The first edge node and the second edge node belong to different network domains, and the second edge node is the next-hop node of the first edge node.

[0033] The time slot mapping network model is trained based on input feature vector samples with labels for transmission time slot offset values.

[0034] Fourthly, the present invention also 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 program to implement the cross-domain deterministic data transmission control method described in the first aspect above.

[0035] Fifthly, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the cross-domain deterministic data transmission control method described in the first aspect above.

[0036] The cross-domain deterministic data transmission control method, system, and apparatus provided by this invention determine the transmission time slot offset value of each data packet during cross-domain transmission in a cross-domain network based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model. A configuration table for data stream transmission scheduling is then generated and sent to the communication nodes in the cross-domain network. This allows the communication nodes to determine the time slot for sending data packets when they are transmitted between nodes in different network domains, thereby improving the automation level of cross-domain deterministic network time slot mapping, reducing the design difficulty and cost of time slot mapping, and enhancing the system's generalized recognition and forwarding capabilities for data streams. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0038] Figure 1 This is a flowchart illustrating the cross-domain deterministic data transmission control method provided by the present invention;

[0039] Figure 2 This is a schematic diagram of the time slot mapping network model provided by the present invention;

[0040] Figure 3 This is a schematic diagram of the cross-domain deterministic data transmission control system provided by the present invention;

[0041] Figure 4 This is an architecture diagram of the cross-domain deterministic data transmission system provided by the present invention;

[0042] Figure 5 This is a schematic diagram of time slot mapping for data packet transmission provided by the present invention;

[0043] Figure 6 This is a schematic diagram of the cross-domain deterministic data transmission control device provided by the present invention;

[0044] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0045] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0046] Integrated deterministic networks employ time-sensitive networking (TSN) for local access and deterministic networking (SDN) for wide-area interconnection. Integrated deterministic networks benefit from both local access and wide-area transmission, thus providing a feasible cross-domain transmission model. Since both TSN and SDN rely on time-slot mapping to provide deterministic quality of service (QoS), integrated deterministic networks also depend on a time-slot mapping model to provide deterministic QoS.

[0047] Specifically, in the local portion of an integrated deterministic network, a circular queuing and forwarding mechanism is generally used to implement time slot mapping, i.e. This mapping indicates that a data packet sent by data flow f in time slot i of node A will be forwarded by node B, a direct downstream node of node A (determined by the data flow path), to a direct downstream node of node B in time slot i+1. Meanwhile, in the wide-area portion of integrated deterministic networks, this is generally used... This implements time slot mapping, meaning that a data packet sent by data stream f in time slot i of node A will be sent by node A's direct downstream node B in time slot j to node B's direct downstream node. Furthermore, a similar time slot mapping relationship is implemented at nodes in the local area to wide area connection. and Integrated deterministic networks achieve hop-by-hop forwarding through time slot mapping, providing end-to-end deterministic quality of service, and the determinism of quality of service comes from the deterministic time slot mapping per flow and per packet.

[0048] However, existing integrated deterministic network systems suffer from time slot mapping between local and wide area transmission nodes (i.e., and The existing time-slot mapping mechanism is generally designed manually, and the automation level of the design process is insufficient. Furthermore, the design requires trade-offs among multiple metrics, resulting in high design difficulty and cost. In addition, the existing time-slot mapping mechanism lacks the ability to generalize and identify unknown data streams. When a new data stream appears, a feasible scheduling table needs to be recalculated by the scheduling algorithm. Since the scheduling model is generally a nondeterministic polynomial-time combinatorial optimization problem, the calculation process is quite time-consuming.

[0049] As an artificial intelligence model, neural networks, after being trained by learning algorithms, can be used to improve the design efficiency and generalization ability of network systems. How to transfer neural networks to integrated deterministic network systems, realize neural time slot mapping, thereby ensuring deterministic service quality, while improving the degree of design automation, reducing design difficulty, and improving generalization recognition ability, is the key problem to be solved by this invention.

[0050] Figure 1 This is a flowchart illustrating the cross-domain deterministic data transmission control method provided by the present invention, as shown below. Figure 1 As shown, the method includes the following steps:

[0051] Step 100: Based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model, determine the transmission time slot offset value of each data packet when it is transmitted across domains in the cross-domain network.

[0052] Step 101: Based on the transmission time slot offset value of each data packet during cross-domain transmission in the cross-domain network, generate a configuration table for data stream transmission scheduling, and send the configuration table to the communication nodes in the cross-domain network.

[0053] The input feature vector corresponding to any data packet is determined based on the information of the data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first edge node and the second edge node when the data packet arrives at the first edge node. The first edge node and the second edge node belong to different network domains, and the second edge node is the next-hop node of the first edge node.

[0054] Specifically, the execution entity of this method can be a Software Defined Networking (SDN) controller. The SDN controller can collect cross-domain network information, configure data flow information according to user needs, and implement scheduling signaling functions. In this invention, the SDN controller can determine the inter-domain time slot mapping mechanism based on a neural network model.

[0055] For each data packet in the data stream to be transmitted, the SDN controller can determine the transmission time slot offset value for each data packet during cross-domain transmission in the cross-domain network based on its corresponding input feature vector and the time slot mapping network model. The transmission time slot offset value refers to the offset value of the time slot in which the second edge node transmits a data packet relative to the time slot in which the first edge node transmits the same data packet. Here, the first edge node and the second edge node belong to different network domains, and the second edge node is the next-hop node of the first edge node.

[0056] Optionally, the first edge node and the second edge node can be switches, routers, or other network nodes that can be used for data forwarding, and there is no limitation on the specific form.

[0057] The input feature vector corresponding to any data packet can be determined based on the information of the data packet and the outgoing port queue status information of the first and second edge nodes when the data packet arrives at the first edge node (such as the number of data packets to be sent in the outgoing port queue).

[0058] Optionally, the information of the data packet may include the utility value of the data packet, the deadline of the data packet, and the size of the data packet. The utility value of the data packet refers to the numerical value that characterizes the commercial value of the data packet; the deadline of the data packet refers to the maximum value of the pre-set time difference between the sending and receiving of the data packet.

[0059] For example, the input feature vector can be represented as x = col(s) 1 ,s 2 ,p), where s 1 This is the vector corresponding to the outgoing port queue status information of the first edge node, where the superscript 1 indicates the first network domain to which the first edge node belongs; s 2is the vector corresponding to the outgoing port queue status information of the second edge node, where the superscript 2 indicates the second network domain to which the second edge node belongs; p represents the vector corresponding to the information of the data packet to be transmitted across domains, which may include the utility value of the data packet, the deadline of the data packet, and the size of the data packet.

[0060] In one implementation, the outgoing port queue status information can be determined as follows: before any node receives the data stream to be transmitted, the outgoing port queue status information is set to an initial value (e.g., 0). Each time the node receives a data packet, i.e., the data packet enters the outgoing port queue, the outgoing port queue status information is incremented by one. When the node sends a data packet from the outgoing port queue to the next-hop node, i.e., the number of data packets in the node's outgoing port queue decreases by one, the outgoing port queue status information is decremented by one. Since the time slot for each node to transmit data packets is fixed or pre-calculated, the outgoing port queue status information of the first and second edge nodes when the data packet arrives at the first edge node can be simulated or pre-calculated.

[0061] The time slot mapping network model can be trained based on input feature vector samples with labels for transmission time slot offset values.

[0062] In one implementation, the input feature vector samples with transmission slot offset value labels include a training sample set of input feature vectors with transmission slot offset value labels and a test sample set of input feature vectors with transmission slot offset value labels. These are used to train the weights of the neural network and test the performance of the neural network, respectively. The loss function can be cross-entropy plus a first-order regularization term for the model parameters, and the neural network can be trained in a gradient iterative manner. Training stops when the error of the test set is less than the expected value, thus obtaining the slot mapping network model.

[0063] By using the time slot mapping network model to determine the transmission time slot offset value of each data packet in the data stream to be transmitted across domains, the input feature vector corresponding to each data packet in the data stream to be transmitted can be used to generate a configuration table for data stream transmission scheduling. The configuration table is then sent to communication nodes in the cross-domain network, such as data forwarding devices, data senders, and data receivers in the network.

[0064] After receiving the configuration table, the communication nodes in the cross-domain network can transmit each data packet in the data stream according to the transmission time slot offset value recorded in the configuration table.

[0065] The cross-domain deterministic data transmission control method provided by this invention determines the transmission time slot offset value of each data packet during cross-domain transmission in a cross-domain network based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model. It also generates a configuration table for data stream transmission scheduling and sends it to the communication nodes in the cross-domain network. This allows the communication nodes to determine the time slot for sending data packets when they are transmitted between nodes in different network domains, thereby improving the automation level of cross-domain deterministic network time slot mapping, reducing the design difficulty and cost of time slot mapping, and improving the system's ability to generalize and forward data streams.

[0066] Optionally, the time-slot mapping network model can be a feedforward network structure, which may include convolutional layers, linearly rectified ReLU fully connected layers, and softmax layers.

[0067] Specifically, Figure 2 This is a schematic diagram of the time slot mapping network model provided by the present invention, as shown below. Figure 2 As shown, this time-slot mapping network model contains two convolutional layers, two Rectified Linear Unit (ReLU) fully connected layers, and one softmax layer. However, in practice, the network model can be configured to contain one or more convolutional layers, one or more ReLU fully connected layers, and one softmax layer; the specific number of convolutional and ReLU fully connected layers is not limited.

[0068] Convolutional layers are used to extract local feature information from the input feature vector.

[0069] The ReLU fully connected layer is used to linearly rectify and integrate the local feature information extracted by the convolutional layer to obtain the classification value of the input feature vector corresponding to different transmission slot offset values.

[0070] Specifically, assuming there are N categories of transmission time slot offset values, the ReLU fully connected layer can linearly rectify and integrate the local feature information extracted by the convolutional layer to obtain the classification values ​​of the input feature vector corresponding to these N transmission time slot offset value categories. The N transmission time slot offset value categories correspond one-to-one with the N types of transmission time slot offset values. N is an integer greater than 1, and its specific value can be set according to the actual situation of the ReLU fully connected layer.

[0071] For example, if two transmission slot offset values ​​are set (offset by 1 bit and offset by 2 bits respectively), then N=2 is set, and the ReLU fully connected layer is configured so that the transmission slot offset value obtained by the ReLU fully connected layer of the slot mapping network model has two categories (i.e., the obtained classification value is two).

[0072] The softmax layer is used to calculate the probability values ​​of the input feature vector corresponding to different transmission slot offset values ​​based on the classification value, and outputs the transmission slot offset value corresponding to the maximum probability value.

[0073] Specifically, the softmax layer can normalize the classification value of the transmit slot offset value category obtained by the ReLU fully connected layer to obtain the probability value of the transmit slot offset value category, select the transmit slot offset value category corresponding to the maximum probability, and then output the transmit slot offset value corresponding to the transmit slot offset value category.

[0074] By employing a feedforward network structure, specifically a time-slot mapping network model comprising convolutional layers, ReLU fully connected layers, and softmax layers, various transmission time slot offset values ​​can be set. Based on the input feature vector corresponding to the data packet, the most suitable transmission time slot offset value for cross-domain transmission in a cross-domain network can be calculated. This improves design automation, reduces design complexity, and enhances generalization and recognition capabilities.

[0075] Optionally, the cross-domain network may include at least one Time-Sensitive Networking (TSN) network domain and at least one Deterministic Internet Protocol (DIP) network domain.

[0076] Specifically, a cross-domain network can refer to a network combining a local area network (LAN) and a wide area network (WAN). In one implementation, a cross-domain network may include at least one LAN and at least one WAN, wherein the LAN may be a TSN network and the WAN may be a Deterministic Internet Protocol (DIP) network.

[0077] By using the input feature vector corresponding to each data packet in the data stream to be transmitted, and the time slot mapping network model, the transmission time slot offset value of each data packet when it is transmitted across domains including the TSN network domain and the DIP network domain can be determined. A configuration table for data stream transmission scheduling is generated and sent to the communication nodes in the TSN network and the DIP network. This allows the communication nodes to determine the time slot for sending data packets when they are transmitted between nodes in the TSN network and the DIP network. This improves the automation of deterministic network time slot mapping when local area crosses wide area or wide area crosses local area, reduces the design difficulty and cost of time slot mapping, and enhances the system's ability to generalize and forward time-sensitive data streams.

[0078] Optionally, based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model, the transmission time slot offset value of each data packet during cross-domain transmission in the cross-domain network is determined, including:

[0079] In the case of cross-domain transmission of target data packets from TSN network domain to DIP network domain, the transmission time slot offset value of target data packets during cross-domain transmission is determined based on the first input feature vector corresponding to the target data packet and the first time slot mapping network model.

[0080] The first input feature vector is determined based on the target data packet information and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first TSN node and the next-hop DIP node when the target data packet arrives at the first TSN node. The first TSN node is an edge TSN node in the TSN network domain that is connected to the DIP network domain.

[0081] The first time slot mapping network model is trained based on the first input feature vector sample with the sending time slot offset value label.

[0082] Specifically, when a target data packet is transmitted across domains from the TSN network domain to the DIP network domain (i.e., the target data packet is transmitted from the first TSN node in the TSN network domain to the next-hop DIP node in the DIP network domain, where the first TSN node is an edge TSN node in the TSN network domain connected to the DIP network domain, and the next-hop DIP node is the next-hop node of the first TSN node), the first input feature vector corresponding to the target data packet can be mapped to the first time slot network model to determine the transmission time slot offset value of the data packet during cross-domain transmission (i.e., the offset value between the time slot when the next-hop DIP node sends the data packet and the time slot when the first TSN node sends the data packet).

[0083] Optionally, TSN nodes and DIP nodes can be switches, routers, or other network nodes that can be used for data forwarding, and there is no limitation on their specific form.

[0084] The first input feature vector can be determined based on the information of the target data packet and the outgoing port queue status information of the first TSN node and the next-hop DIP node when the data packet arrives at the first TSN node (such as the number of data packets to be sent in the outgoing port queue).

[0085] For example, the first input feature vector can be represented as in, This is the vector corresponding to the outgoing port queue status information of the first TSN node, with the superscript 1 indicating the TSN network domain to which the first TSN node belongs; This is the vector corresponding to the outgoing port queue status information of the next-hop DIP node. The superscript 2 indicates the DIP network domain to which the next-hop DIP node belongs. p represents the vector corresponding to the information of the target data packet to be transmitted across domains, which may include the utility value of the target data packet, the deadline of the target data packet, and the size of the target data packet.

[0086] The first time slot mapping network model can be trained based on a first input feature vector sample with a label for the transmission time slot offset value.

[0087] In one implementation, the first input feature vector samples labeled with transmission time slot offset values ​​include a training sample set of the first input feature vectors labeled with transmission time slot offset values ​​and a test sample set of the first input feature vectors labeled with transmission time slot offset values. These are used to train the weights of the neural network and test the performance of the neural network, respectively. The loss function can be cross-entropy plus a first regularization term of the model parameters, and the neural network can be trained in a gradient iterative manner. Training stops when the error of the test set is less than the expected value, thus obtaining the first time slot mapping network model.

[0088] When a target data packet is transmitted across domains from the TSN network domain to the DIP network domain, the transmission time slot offset value of the target data packet during the cross-domain transmission from the TSN network domain to the DIP network domain can be determined based on the first input feature vector corresponding to the target data packet and the first time slot mapping network model. This improves the automation level of deterministic network time slot mapping when the local area crosses the wide area, reduces the design difficulty and cost of time slot mapping, and enhances the system's ability to generalize and forward time-sensitive data streams.

[0089] Optionally, based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model, the transmission time slot offset value of each data packet during cross-domain transmission in the cross-domain network is determined, including:

[0090] In the case of cross-domain transmission of target data packets from DIP network domain to TSN network domain, the transmission time slot offset value of the target data packets during cross-domain transmission is determined based on the second input feature vector corresponding to the target data packets and the second time slot mapping network model.

[0091] The second input feature vector is determined based on the target data packet information and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first DIP node and the next-hop TSN node when the target data packet arrives at the first DIP node. The first DIP node is an edge DIP node in the DIP network domain that is connected to the TSN network domain.

[0092] The second time slot mapping network model is trained based on a second input feature vector sample with a label for the transmission time slot offset value.

[0093] Specifically, when a target data packet is transmitted across domains from the DIP network domain to the TSN network domain (i.e., the target data packet is transmitted from the first DIP node in the DIP network domain to the next-hop TNS node in the TSN network domain, where the first DIP node is an edge DIP node in the DIP network domain connected to the TSN network domain, and the next-hop TSN node is the next-hop node of the first DIP node), the second input feature vector corresponding to the target data packet can be mapped through the second time slot mapping network model to determine the transmission time slot offset value of the data packet during cross-domain transmission (i.e., the offset value between the time slot when the next-hop TSN node sends the data packet and the time slot when the first DIP node sends the data packet).

[0094] Optionally, TSN nodes and DIP nodes can be switches, routers, or other network nodes that can be used for data forwarding, and there is no limitation on their specific form.

[0095] The second input feature vector can be determined based on the information of the target data packet and the outgoing port queue status information of the first DIP node and the next-hop TSN node when the data packet arrives at the first DIP node (such as the number of data packets to be sent in the outgoing port queue).

[0096] For example, the second input feature vector can be represented as in, This is the vector corresponding to the outgoing port queue status information of the first DIP node, with the superscript 2 indicating the DIP network domain to which the first DIP node belongs; This is the vector corresponding to the outgoing port queue status information of the next-hop TSN node. The superscript 1 indicates the TSN network domain to which the next-hop TSN node belongs. p represents the vector corresponding to the information of the target data packet to be transmitted across domains, which may include the utility value of the target data packet, the deadline of the target data packet, and the size of the target data packet.

[0097] The second time slot mapping network model can be trained based on a second input feature vector sample with a label for the transmission time slot offset value.

[0098] In one implementation, the second input feature vector samples labeled with transmission time slot offset values ​​include a training sample set of second input feature vectors labeled with transmission time slot offset values ​​and a test sample set of second input feature vectors labeled with transmission time slot offset values. These are used to train the weights of the neural network and test the performance of the neural network, respectively. The loss function can be cross-entropy plus a first regularization term of the model parameters, and the neural network can be trained in a gradient iterative manner. Training stops when the error of the test set is less than the expected value, thus obtaining the second time slot mapping network model.

[0099] When a target data packet is transmitted across domains from the DIP network domain to the TSN network domain, the transmission time slot offset value of the target data packet during cross-domain transmission from the DIP network domain to the TSN network domain can be determined by using the second input feature vector corresponding to the target data packet and the second time slot mapping network model. This improves the automation level of deterministic network time slot mapping in wide-area cross-local domain scenarios, reduces the design difficulty and cost of time slot mapping, and enhances the system's ability to generalize and forward time-sensitive data streams.

[0100] Figure 3 This is a schematic diagram of the cross-domain deterministic data transmission control system provided by the present invention, as shown below. Figure 3 As shown, the system includes:

[0101] Software-defined networking (SDN) controller, DIP network domain, and multiple TSN network domains connected to the DIP network domain; each TSN network domain contains one or more TSN nodes, and each DIP network domain contains one or more DIP nodes.

[0102] The SDN controller is used to execute the cross-domain deterministic data transmission control method described in the above embodiments; the TSN node and DIP node are used to perform time slot mapping and forwarding operations on the received data packets based on the configuration table sent by the SDN controller, so as to realize the cross-domain transmission of deterministic data streams.

[0103] Specifically, a cross-domain deterministic data transmission control system may include an SDN controller, a DIP network domain, and multiple TSN network domains connected to the DIP network domain. Each TSN network domain contains one or more TSN nodes (such as TSN switches), and each DIP network domain contains one or more DIP nodes (such as DIP routers).

[0104] The SDN controller can serve as the execution entity for any of the cross-domain deterministic data transmission control methods described above. It is used to control the network system consisting of the DIP network domain and multiple TSN network domains connected to the DIP network domain. It is responsible for network information collection (e.g., the outgoing port queue status information of the edge nodes at the network connection when data packets cross domains), training the time slot mapping network model, calculating the time slot mapping of the data stream, generating a configuration table for data stream transmission scheduling, and sending the configuration table to the TSN nodes and DIP nodes in each network domain.

[0105] After receiving the configuration table sent by the SDN controller, the TSN node and DIP node can perform time slot mapping and forwarding operations on the received data packets based on the configuration table sent by the SDN controller, thereby realizing cross-domain transmission of deterministic data streams.

[0106] The cross-domain deterministic data transmission control system provided by this invention determines the transmission time slot offset value of each data packet when it is transmitted across domains in a network containing DIP network domains and multiple TSN network domains through an SDN controller, and generates a configuration table for data stream transmission scheduling, which is sent to the communication nodes in the network. This allows the communication nodes to determine the time slot for sending data packets when they are transmitted between nodes in different network domains in the data stream, thereby improving the automation level of cross-domain deterministic network time slot mapping, reducing the design difficulty and cost of time slot mapping, and improving the system's ability to generalize and forward time-sensitive data streams.

[0107] The system provided by the present invention will be described in detail below through an embodiment of a specific application scenario.

[0108] Figure 4 This is an architecture diagram of the cross-domain deterministic data transmission system provided by the present invention, as shown below. Figure 4 As shown, this system is a hardware and software integrated network system, in which the hardware entities include servers (e1), TSN switches (v1, v2, v6, v7), DIP routers (v3, v4, v5), and terminals (e2) industrial systems. Figure 4 (Taking a robotic arm as an example in the Chinese industrial system); the SDN controller is a software entity, including a neural network model, a sample library, a user controller, a scheduler, and a network configurator. The hardware and software entities consist of four main modules: the SDN controller, the TSN A data plane, the DIP network data plane, and the TSN B data plane. The SDN controller performs the control functions of the network system, responsible for network information collection, neural network training, and time-sensitive data stream scheduling calculations and configuration distribution. The TSN A and B data planes realize the transmission of time-sensitive data streams within their respective domains. The neural mapping model uses neural networks to complete time slot mapping between network domains (between TSN A and DIP networks, and between DIP networks and TSN B), achieving inter-domain interconnection and providing deterministic quality of service.

[0109] All entities in the system support their traditional functions, such as: servers providing data stream time slot offsets, co-domain TSN switches supporting time synchronization, TSN switches supporting cyclic queuing forwarding, DIP routers supporting traditional time slot mapping functions, SDN controllers providing functions for collecting network information and distributing scheduling signaling, routers and switches supporting wired connectivity, nodes (TSN switches and DIP routers) having computing resources and the ability to execute computation results, and TSN switches and DIP routers having deterministic time slot mapping functions (see IEEE standard [IEEE Std 802.1Qch-2017] and IEEE standard draft [draft-qiang-detnet-large-scale-detnet-05]).

[0110] In addition to traditional functions, the new features of this cross-domain deterministic network transmission system are as follows:

[0111] (1) Inter-domain time slot mapping based on neural networks

[0112] Inter-domain neural mapping is responsible for allocating time slots for data packets between LANs and WANs. Its calculation relies on two neural network models, NN1 and NN2, which operate at the connections between TSN A and DIP networks, and TSN B and DIP networks, respectively. For each data packet's inter-domain time slot allocation task, the feature vector is input into the neural network, and the output time slot offset value is calculated to complete the inter-domain time slot mapping.

[0113] (2) Neural Network Model

[0114] Input feature vector of neural network model NN1 Specifically, this includes the status of the edge TSN switches in the TSN A and DIP networks, namely the outgoing port queue status vector. The superscript 1 indicates the TSN A domain; TSN A and the state of the edge DIP routers in the DIP network, i.e., the outgoing port queue state vector of the edge DIP routers. The superscript 2 indicates the DIP network domain; the data packet information vector p of the time slot to be calculated includes the data packet utility, deadline, and data packet size.

[0115] The output of the neural network model NN1 is the time slot offset value of the data packet p sent by the DIP edge router at the TSN A and DIP network connection point.

[0116] Input feature vector of neural network model NN2 Specifically, this includes: the state of the TSN switch at the connection point between the TSN B and the DIP network, i.e., the outgoing port queue state vector. The superscript 3 indicates the TSN B domain; the outgoing port queue state vector of the DIP router at the connection point between the TSN B and the DIP network. The data packet information vector p for the time slot to be calculated includes the data packet utility, deadline, and data packet size.

[0117] The output of the neural network model NN2 is the time slot offset value of the data packet p sent by the TSN edge switch at the TSN B and DIP network connection point.

[0118] The neural network uses a feedforward structure and includes two convolutional layers, two ReLU (rectified linear unit) fully connected layers, and one softmax layer.

[0119] (3) Sample library

[0120] The sample database contains four datasets, D1, D2, D3, and D4, used for training and testing neural networks.

[0121] in Two datasets were used to train neural network models NN1 and NN2, respectively.

[0122] Two datasets are used to test neural network models NN1 and NN2, respectively, satisfying the conditions n2>n1>1 and m2>m1>1. Here, x represents the input feature vector sample, and y represents the transmission time slot offset label.

[0123] (4) Neural mapping driven scheduler

[0124] By integrating time slot offset operations with a neural network model, a neural mapping-driven scheduler is formed. The scheduler has the following functions: 1) The scheduler can divide the time slots of switches and routers, with a time slot period of T. 2) The scheduler can perform time slot offset operations on data packets sent by the server. 3) The scheduler has the function of acquiring network information and data packet information and forming neural network input features. 4) The scheduler has the function of inputting features into the neural network and obtaining the output time slot. 5) The scheduler has the function of sending scheduling result signaling to TSN switches and DIP routers, such as the signaling sending function conforming to the Network Configuration Protocol (NETCONF).

[0125] The following describes the operating mechanism of this cross-domain deterministic network transmission system.

[0126] (1) Training of neural network models

[0127] By training a neural network model based on a sample set, the ability of the neural network to identify the optimal time slot allocation can be optimized. This enables the neural network to calculate the corresponding output time slot for data packets based on features such as routers, switches, and data packet information. This output time slot is then used by the scheduler to calculate the scheduling results, supporting deterministic transmission across domains.

[0128] The neural network model training process is as follows:

[0129] 1. Input data: Training sample sets D1 and D2, and test sample sets D3 and D4, which are used to train the weights of the neural network and test the performance of the neural network, respectively.

[0130] 2. Execution process: The loss function is a combination of cross-entropy and a first-order regularization term for the model parameters. The neural network is trained using gradient iteration. Training stops when the error of the test set is less than the expected value.

[0131] 3. Output Model: Transmit the trained NN1 and NN2 model parameters to the scheduler for the network scheduling mechanism to run.

[0132] (2) Scheduler scheduling calculation and configuration table distribution

[0133] Based on time slot offset operations and neural network models, the SDN controller can optimize network throughput, satisfy data flow constraints, calculate scheduling results, and plan deterministic cross-domain transmission patterns for data packets.

[0134] The scheduler's scheduling calculation and configuration table distribution process is as follows:

[0135] 1. Input data: packet utility, start address, destination address, deadline, packet size, period, and path. Planning period T h .

[0136] 2. Execution process in SDN controller: For each data stream, perform (a) sequentially set a time slot offset 0,…,T h -1, prioritize the first feasible timeslot offset; (b) calculate the last hop state of the TSN A domain according to the circular queuing forwarding mechanism (see IEEE standard [IEEE Std 802.1Qch-2017]), and calculate the transmission timeslot offset of the packet in the first hop of the DIP domain through the trained NN1; (c) send the packet according to the transmission timeslot offset of the first hop of the DIP domain in (b), and calculate the state of the packet being transmitted to the end node of the DIP domain according to the DIP timeslot mapping mechanism (see IEEE standard draft [draft-qiang-detnet-large-scale-detnet-05]), calculate the transmission timeslot offset of the packet in the first hop of the TSN B domain through the trained NN2, and send the packet according to the transmission timeslot offset. The TSN B domain sends the packet to the end node according to the circular queuing forwarding mechanism.

[0137] 3. Output variables: After step 2 above is completed, the feasibility of the scheduling results is checked. Infeasible results are deleted, and feasible results are converted into configuration entries with scheduling information. The configuration is then distributed to physical network devices (servers, switches, routers, and terminals) through the SDN controller.

[0138] (3) Deterministic forwarding mechanism of network devices

[0139] After receiving the configuration table entry, a forwarding operation is performed on a certain data packet according to the data table, which can achieve deterministic forwarding of cross-domain data packets.

[0140] The deterministic forwarding mechanism of network devices proceeds as follows:

[0141] 1. Execution process: Set the first-hop timeslot offset of the data packet in the TSN A domain, and transmit the intermediate hops according to the round-robin queuing forwarding mechanism; when the data packet reaches the last hop node in the TSN A domain, the first-hop node in the DIP domain sends the data packet according to the timeslot in the configuration table calculated by neural mapping; in the DIP domain, except for the first-hop node, the data packet is forwarded according to the original timeslot mapping mechanism of the DIP router (see IEEE standard draft [draft-qiang-detnet-large-scale-detnet-05]); the first-hop node in the TSN B domain forwards the data packet according to the timeslot in the configuration table calculated by neural mapping, and in the TSN B domain, except for the first-hop node, the data packet is transmitted to the terminal according to the round-robin queuing forwarding mechanism.

[0142] 2. Output Variables: After forwarding is complete, the output packet delay is calculated, and jitter is calculated based on multiple delays. Because the designed time slot mapping is deterministic, the output packet delay and jitter have definite upper bounds, meaning that deterministic delay and jitter quality of service can be provided.

[0143] This cross-domain deterministic network transmission system can implement any of the method embodiments described above, and the specific implementation process is as follows:

[0144] Step 1: In the SDN controller, train the neural network model using the training sample set and the test sample set of the input feature vector labeled with the transmission slot offset value. The trained neural network models are NN1 and NN2, respectively. The softmax layer of both NN1 and NN2 models has two outputs: the probability of the transmission slot shifting backward by 1 and 2 bits, respectively. Finally, the transmission slot offset value with the highest probability is selected as the model output.

[0145] Step 2: The TSN domain and DIP domain divide the time slots into t0, t1, t2... Each time slot corresponds to a period of T. There is a clock offset between the nodes in the DIP domain and the nodes in the TSN domain.

[0146] Step 3: Consider an example of three data streams, where the packet transmission period of each data stream is 2T. The scheduler in the SDN controller forwards the data streams one by one in the order of data streams 1, 2, and 3, and uses neural network models NN1 and NN2 respectively to calculate the cross-domain transmission slot offset value for cross-domain node pairs (v2,v3) and (v5,v6) according to the maximum probability.

[0147] Figure 5 This is a schematic diagram of the time slot mapping for data packet transmission provided by the present invention, as shown below. Figure 5 As shown, Figure 5The diagonal squares, blank spaces, and grid squares in the diagram all represent data packets, corresponding to the first data packet of data streams 1, 2, and 3 after scheduling, respectively; the position of the square on the coordinate axis indicates the transmission time slot when the node corresponding to that coordinate axis sends the data packet corresponding to that square.

[0148] Step 4: The SDN controller sends the cross-domain transmission slot offset value and other slot mapping scheduling information to the network data plane in the form of a configuration table. The data plane forwards the data packets according to the configuration table and forwarding mechanism, and finally sends the data stream to the terminal to control the robotic arm.

[0149] At this point, the network system has completed the scheduling and forwarding process, and data packets are periodically forwarded to the terminals. Because the forwarding process is deterministic (circular queuing forwarding, DIP time slot mapping determination, and time slot mapping network model determination), the system can provide upper-bounded latency and jitter quality of service for data streams 1, 2, and 3. Furthermore, because the system uses a neural network model for time slot mapping in the cross-domain portion, it improves the scheduler's ability to identify data streams based on network conditions and enhances the degree of automation in the design.

[0150] The cross-domain deterministic data transmission control device provided by the present invention will be described below. The cross-domain deterministic data transmission control device described below can be referred to in correspondence with the cross-domain deterministic data transmission control method described above.

[0151] Figure 6 This is a schematic diagram of the cross-domain deterministic data transmission control device provided by the present invention, as shown below. Figure 6 As shown, the device includes:

[0152] The determination module 600 is used to determine the transmission time slot offset value of each data packet during cross-domain transmission in a cross-domain network based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model.

[0153] The sending module 610 is used to generate a configuration table for data stream transmission scheduling based on the sending time slot offset value of each data packet when it is transmitted across domains in the cross-domain network, and send the configuration table to the communication nodes in the cross-domain network.

[0154] The input feature vector corresponding to any data packet is determined based on the information of the data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first edge node and the second edge node when the data packet arrives at the first edge node. The first edge node and the second edge node belong to different network domains, and the second edge node is the next-hop node of the first edge node.

[0155] The time slot mapping network model is trained based on input feature vector samples with labels for transmission time slot offset values.

[0156] Optionally, the time slot mapping network model is a feedforward network structure, which includes convolutional layers, linear rectified ReLU fully connected layers, and flexible maximum softmax layers;

[0157] Convolutional layers are used to extract local feature information from the input feature vector;

[0158] The ReLU fully connected layer is used to linearly rectify and integrate the local feature information extracted by the convolutional layer to obtain the classification value of the input feature vector corresponding to different transmission slot offset values.

[0159] The softmax layer is used to calculate the probability values ​​of the input feature vector corresponding to different transmission slot offset values ​​based on the classification value, and outputs the transmission slot offset value corresponding to the maximum probability value.

[0160] Optionally, the cross-domain network includes at least one TSN network domain and at least one DIP network domain.

[0161] Optionally, based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model, the transmission time slot offset value of each data packet during cross-domain transmission in the cross-domain network is determined, including:

[0162] In the case of cross-domain transmission of target data packets from TSN network domain to DIP network domain, the transmission time slot offset value of target data packets during cross-domain transmission is determined based on the first input feature vector corresponding to the target data packet and the first time slot mapping network model.

[0163] The first input feature vector is determined based on the target data packet information and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first TSN node and the next-hop TSN node when the target data packet arrives at the first TSN node. The first TSN node is an edge TSN node in the TSN network domain that is connected to the DIP network domain.

[0164] The first time slot mapping network model is trained based on the first input feature vector sample with the sending time slot offset value label.

[0165] Optionally, based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model, the transmission time slot offset value of each data packet during cross-domain transmission in the cross-domain network is determined, including:

[0166] In the case of cross-domain transmission of target data packets from DIP network domain to TSN network domain, the transmission time slot offset value of the target data packets during cross-domain transmission is determined based on the second input feature vector corresponding to the target data packets and the second time slot mapping network model.

[0167] The second input feature vector is determined based on the target data packet information and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first DIP node and the next-hop TSN node when the target data packet arrives at the first DIP node. The first DIP node is an edge DIP node in the DIP network domain that is connected to the TSN network domain.

[0168] The second time slot mapping network model is trained based on a second input feature vector sample with a label for the transmission time slot offset value.

[0169] Optionally, the information in the data packet includes the packet's utility value, the packet's deadline, and the packet's size.

[0170] It should be noted that the device provided by the present invention can implement all the method steps implemented in the above method embodiments and can achieve the same technical effect. Therefore, the parts and beneficial effects that are the same as those in the method embodiments will not be described in detail here.

[0171] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 7 As shown, the electronic device may include a processor 710, a communications interface 720, a memory 730, and a communication bus 740, wherein the processor 710, the communications interface 720, and the memory 730 communicate with each other via the communication bus 740. The processor 710 may call logical instructions in the memory 730 to execute any of the cross-domain deterministic data transmission control methods provided in the above embodiments.

[0172] Furthermore, the logical instructions in the aforementioned memory 730 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0173] It should be noted that the electronic device provided by the present invention can implement all the method steps implemented in the above method embodiments and can achieve the same technical effect. Therefore, the parts and beneficial effects that are the same as those in the method embodiments will not be described in detail here.

[0174] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform any of the cross-domain deterministic data transmission control methods provided in the above embodiments.

[0175] It should be noted that the non-transitory computer-readable storage medium provided by the present invention can implement all the method steps implemented in the above method embodiments and can achieve the same technical effect. Here, the parts that are the same as those in the method embodiments and the beneficial effects will not be described in detail.

[0176] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0177] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0178] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling cross-domain deterministic data transmission, characterized in that, include: Based on the input feature vector corresponding to each data packet of the data stream to be transmitted, and the time slot mapping network model, the transmission time slot offset value of each data packet during cross-domain transmission in the cross-domain network is determined. Based on the transmission time slot offset value of each data packet during cross-domain transmission in the cross-domain network, a configuration table for data stream transmission scheduling is generated, and the configuration table is sent to the communication nodes in the cross-domain network. The input feature vector corresponding to any data packet is determined based on the information of the data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first edge node and the second edge node when the data packet arrives at the first edge node. The first edge node and the second edge node belong to different network domains, and the second edge node is the next-hop node of the first edge node. The time slot mapping network model is trained based on input feature vector samples with sending time slot offset value labels; by setting multiple sending time slot offset values, and according to the input feature vector corresponding to each data packet, the sending time slot offset value of each data packet when transmitting across domains in the cross-domain network is determined.

2. The cross-domain deterministic data transmission control method according to claim 1, characterized in that, The time slot mapping network model is a feedforward network structure, which includes convolutional layers, linear rectified ReLU fully connected layers, and flexible maximum softmax layers; The convolutional layer is used to extract local feature information from the input feature vector; The ReLU fully connected layer is used to linearly rectify and integrate the local feature information extracted by the convolutional layer to obtain the classification value of the input feature vector corresponding to different transmission time slot offset values. The softmax layer is used to calculate the probability value of the input feature vector corresponding to different transmission slot offset value categories based on the classification value, and output the transmission slot offset value corresponding to the maximum probability value.

3. The cross-domain deterministic data transmission control method according to claim 1 or 2, characterized in that, The cross-domain network includes at least one Time-Sensitive Networking (TSN) network domain and at least one Deterministic Internet Protocol (DIP) network domain.

4. The cross-domain deterministic data transmission control method according to claim 3, characterized in that, The determination of the transmission time slot offset value for each data packet during cross-domain transmission in a cross-domain network, based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model, includes: In the case of cross-domain transmission of target data packets from TSN network domain to DIP network domain, the transmission time slot offset value of the target data packets during cross-domain transmission is determined based on the first input feature vector corresponding to the target data packets and the first time slot mapping network model. The first input feature vector is determined based on the information of the target data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first TSN node and the next-hop DIP node when the target data packet arrives at the first TSN node. The first TSN node is an edge TSN node in the TSN network domain that is connected to the DIP network domain. The first time slot mapping network model is trained based on a first input feature vector sample with a label for the transmission time slot offset value.

5. The cross-domain deterministic data transmission control method according to claim 3, characterized in that, The determination of the transmission time slot offset value for each data packet during cross-domain transmission in a cross-domain network, based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model, includes: In the case of cross-domain transmission of target data packets from DIP network domain to TSN network domain, the transmission time slot offset value of the target data packets during cross-domain transmission is determined based on the second input feature vector corresponding to the target data packets and the second time slot mapping network model. The second input feature vector is determined based on the information of the target data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first DIP node and the next-hop TSN node when the target data packet arrives at the first DIP node. The first DIP node is an edge DIP node in the DIP network domain that is connected to the TSN network domain. The second time slot mapping network model is trained based on a second input feature vector sample with a label for the transmission time slot offset value.

6. The cross-domain deterministic data transmission control method according to claim 1, characterized in that, The information in the data packet includes the utility value of the data packet, the expiration time of the data packet, and the size of the data packet.

7. A cross-domain deterministic data transmission control system, characterized in that, include: Software-defined networking (SDN) controller, DIP network domain, and multiple TSN network domains connected to the DIP network domain; The TSN network domain contains one or more TSN nodes, and the DIP network domain contains one or more DIP nodes; The SDN controller is used to execute the cross-domain deterministic data transmission control method according to any one of claims 1 to 6; The TSN node and the DIP node are used to perform time slot mapping and forwarding operations on the received data packets based on the configuration table sent by the SDN controller, so as to realize the cross-domain transmission of deterministic data streams.

8. A cross-domain deterministic data transmission control device, characterized in that, include: The determination module is used to determine the transmission time slot offset value of each data packet during cross-domain transmission in a cross-domain network based on the input feature vector corresponding to each data packet of the data stream to be transmitted and the time slot mapping network model. The sending module is used to generate a configuration table for data stream transmission scheduling based on the sending time slot offset value of each data packet during cross-domain transmission in the cross-domain network, and send the configuration table to the communication node in the cross-domain network. The input feature vector corresponding to any data packet is determined based on the information of the data packet and the outgoing port queue status information. The outgoing port queue status information includes the outgoing port queue status information of the first edge node and the second edge node when the data packet arrives at the first edge node. The first edge node and the second edge node belong to different network domains, and the second edge node is the next-hop node of the first edge node. The time slot mapping network model is trained based on input feature vector samples with sending time slot offset value labels; by setting multiple sending time slot offset values, and according to the input feature vector corresponding to each data packet, the sending time slot offset value of each data packet when transmitting across domains in the cross-domain network is determined.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the cross-domain deterministic data transmission control method as described in any one of claims 1 to 6.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the cross-domain deterministic data transmission control method as described in any one of claims 1 to 6.

Images