A task intention driven data link network dynamic construction method
By constructing task intent, network status, and domain knowledge graphs, and dynamically adjusting the data link network configuration, the response and adaptation problems of traditional networking methods in multi-task environments are solved, achieving efficient and reliable information transmission and task execution.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- 10TH RES INST OF CETC
- Filing Date
- 2026-02-06
- Publication Date
- 2026-07-07
Smart Images
Figure CN121664670B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data communication technology, and more specifically, to a method for dynamically constructing a data link network driven by task intent. Background Technology
[0002] With the continuous development of intelligent technologies, unmanned swarm missions and distributed cross-domain collaborative missions are increasing, the number of platforms connected to data link networks is rising sharply, transmission service types are becoming increasingly diversified, and network situations are changing rapidly. Data link networks need to support rapid access and dynamic networking of large-scale nodes, and have real-time reconfiguration capabilities. Their core objective is to enhance joint command and control capabilities, battlefield situational awareness consistency, and cross-system collaborative efficiency, ultimately achieving real-time, precise control of various access platforms and improving mission initiative. Intent-driven Network (IDN), as a new network architecture, can achieve autonomous driving and planning optimization of data link networks in complex and highly dynamic adversarial environments through the parsing and mapping of high-level mission intents, significantly improving the level of network intelligence. However, in actual combat or emergency scenarios, different tactical missions place differentiated demands on network transmission performance. For example, surveillance and reconnaissance missions emphasize spatial coverage and efficient allocation of communication resources, while search and rescue missions focus more on timeliness and energy efficiency. Traditional static networking methods that rely on manual predefined configurations are difficult to adapt to the dynamic networking requirements under multiple mission intents. Therefore, achieving reasonable allocation of multi-chain network parameters and efficient scheduling of resources based on task intent has become the key to improving task execution efficiency and network reliability.
[0003] Currently, several technologies have attempted to provide adaptive networking capabilities to meet the needs of dynamic networking. One is adaptive self-organizing routing technology, such as the Adaptive Demand Distance Vector (AODV) protocol for wireless ad hoc networks. This protocol uses a distributed route discovery and maintenance mechanism to update the routing table in real time based on network topology changes. This type of method does not require fixed infrastructure support; nodes have routing and relay functions and can autonomously adjust topology connections as the network environment changes. However, it suffers from high routing overhead and insufficient scalability: proactive routing protocols require periodic interactive communication to maintain the topology, and on-demand routing frequently triggers reconstruction in highly dynamic environments, introducing significant latency and control load. Furthermore, performance degrades significantly as the number of nodes increases, making it difficult to support ultra-large-scale node networking. The second is intelligent cross-layer optimization technology. By breaking down the layer barriers of traditional protocol stacks, it shares information such as channel status, link quality, and congestion levels between the physical layer, link layer, and network layer, enabling joint adjustment and optimization decisions of cross-layer parameters and improving network environment adaptability. However, this approach also brings problems such as protocol stack complexity, poor cross-system compatibility and policy conflicts. Frequent cross-layer intervention can easily cause system oscillations or parameter inconsistencies, making actual deployment difficult.
[0004] In summary, existing data link networking methods mainly suffer from the following bottlenecks: First, they fail to build networks and allocate resources based on mission intent, relying mostly on fixed optimization models or general performance indicators. They lack the ability to dynamically respond to the differences in communication requirements brought about by different tactical intents, resulting in limited network transmission efficiency and mission execution effectiveness. Second, static or predefined networking methods are difficult to support real-time adaptive reconstruction in highly dynamic adversarial environments. They cannot quickly adjust network configurations in the face of scenarios such as node mobility, link quality jitter, and sudden tasks, which seriously affects the reliability of networks and mission assurance capabilities in complex environments.
[0005] Therefore, it is necessary to develop a solution that extracts key communication requirements from the task intent, combines real-time network status and resource conditions, dynamically derives network configuration parameters, and achieves on-demand networking and precise resource scheduling. This will ensure efficient and reliable information transmission and optimal overall task performance in heterogeneous, multi-tasking, and highly adversarial environments. Summary of the Invention
[0006] The purpose of this application is to overcome the shortcomings of existing technologies and provide a task intent-driven dynamic construction method for data link networks. By constructing and integrating three major graphs—task intent, network state, and domain expertise—for collaborative reasoning, abstract task intents are transformed into specific network construction parameters, and optimization algorithms are used for real-time solution. This significantly improves the rapid response and adaptive construction capabilities of data link networks for complex and ever-changing tasks.
[0007] The objective of this application is achieved through the following technical solution:
[0008] Firstly, this application proposes a task intent-driven method for dynamically constructing data link networks, including:
[0009] Key information about the task intent is extracted through knowledge extraction, and a task intent knowledge graph is constructed.
[0010] Each node senses its own status information and reports it to the management center to obtain the status information of the entire network. Based on the status information of the entire network, a network status knowledge graph is constructed. The status information of the entire network includes throughput, latency and packet loss rate.
[0011] Construct a domain-specific knowledge graph, which includes the technical terminology, concept set, communication rules, and transmission constraints in the field of data link communication technology;
[0012] Based on the domain knowledge graph and the task intent knowledge graph, the network topology and business message type corresponding to the task intent are inferred, and the communication node type and node communication requirements are determined according to the business message type.
[0013] Based on the node state information in the network state knowledge graph, combined with the communication node type, node communication requirements and domain professional knowledge graph, the source node and destination node for business message transmission are determined.
[0014] By combining the network topology, source nodes, and destination nodes, determine the data link network node connections that are appropriate for the current task intent.
[0015] In one possible implementation, the steps of extracting key information about the task intent through knowledge extraction and constructing a task intent knowledge graph include:
[0016] Extract key information such as task intent type, task time, task location, and task objective;
[0017] The ALBERT-BiLSTM-CRF entity recognition model is used to extract entities, relations, and attributes from the task intent.
[0018] The extracted entities, attributes, and relationships are stored in a graph database to construct a task intent knowledge graph.
[0019] In one possible implementation, the ALBERT-BiLSTM-CRF entity recognition model includes an ALBERT layer, a BiLSTM layer, and a CRF layer.
[0020] The ALBERT layer is used to convert input characters into vector form;
[0021] The BiLSTM layer is used to extract features from the embedding vectors generated by the ALBERT layer;
[0022] The CRF layer is used to learn the constraint relationships between labels and predict entity labels.
[0023] In one possible implementation, the steps of constructing a domain knowledge graph include:
[0024] Construct at least one of the following knowledge graphs: mapping relationship between task intent and network topology configuration, mapping relationship between task intent and business message type, mapping relationship between business message type and communication node type, and mapping relationship between business message type and node communication requirements.
[0025] In one possible implementation, the step of reasoning the network topology and service message type corresponding to the task intent based on the domain knowledge graph and the task intent knowledge graph includes:
[0026] Based on the task intent type, the corresponding business message type is obtained through reasoning using the domain knowledge graph. The business message types include status messages, image messages, video messages, monitoring messages, and evaluation messages.
[0027] Based on the task intent type and / or business message type, the corresponding network topology configuration is obtained through domain knowledge graph reasoning. The network topology configurations include distributed topology, star topology, mesh topology, and hierarchical topology.
[0028] In one possible implementation, the step of determining the communication node type and node communication requirements based on the service message type includes:
[0029] Based on the domain knowledge graph, the communication node types corresponding to the business message types are determined. The communication node types include control nodes, perception nodes, execution nodes, and information processing nodes.
[0030] Determine the node communication requirements corresponding to the business message type. The node communication requirements include tolerable latency, required throughput, and maximum packet loss rate.
[0031] In one possible implementation, the step of determining the source and destination nodes for service message transmission based on node state information in a network state knowledge graph, combined with communication node types, node communication needs, and domain-specific knowledge graphs, includes:
[0032] IF-THEN rules are formulated based on domain-specific knowledge graphs, and the rules define the selection criteria for source and destination nodes.
[0033] Based on the task intent type, business message type, and node communication requirements, and combined with the current network status information, the source node and destination node are obtained through IF-THEN rule matching.
[0034] In one possible implementation, the method further includes:
[0035] When a new business message arrives, retrieve the type of the business message;
[0036] Check the current network status based on the network state knowledge graph;
[0037] Update matching rules by incorporating domain-specific knowledge graphs;
[0038] The inference engine is used to automatically match source and destination nodes.
[0039] In one possible implementation, the steps of determining the data link network node connections suitable for the current task intent, based on the network topology, source nodes, and destination nodes, and constructing the data link network, include:
[0040] The simulated annealing algorithm is used for optimization and solution.
[0041] Set the constraint that the network topology remains unchanged when the network topology changes;
[0042] By exchanging the connection relationships of nodes or edges, a neighborhood solution is generated, and under constraints, the optimization is iteratively sought to determine the connection relationships of network nodes and construct a data link network.
[0043] In one possible implementation, the objective function for optimization is: , Represents the set of all network metric types. Represents the set of all node pairs. This indicates the weights of throughput, latency, and packet loss rate. This indicates the business message communication requirement between a pair of nodes. It represents the actual communication transmission capacity between a pair of nodes.
[0044] Further alternative solutions can be freely combined to form multiple solutions, all of which are applicable and protected by this application; furthermore, the non-conflicting options can also be freely combined with each other and with other options. Those skilled in the art, after understanding the solutions of this application, will realize from the prior art and common general knowledge that there are many combinations, all of which are technical solutions to be protected by this application, and will not be exhaustively listed here.
[0045] This application discloses a task intent-driven dynamic construction method for data link networks. First, key information of the task intent is extracted through knowledge extraction, and a task intent knowledge graph is constructed. Each node perceives its own state information and reports it to the management center to obtain overall network state information. A network state knowledge graph and a domain knowledge knowledge graph are then constructed. The network topology and business message type corresponding to the task intent are inferred. Based on the business message type, the communication node type and node communication requirements are determined. Then, the source and destination nodes for business message transmission, as well as the data link network node connection relationships applicable to the current task intent, are determined. By constructing and fusing the three graphs of task intent, network state, and domain knowledge for collaborative reasoning, the abstract task intent is transformed into specific network construction parameters. An optimization algorithm is used for real-time solution, effectively improving the adaptive capability and task response speed of on-demand data link network construction. Attached Figure Description
[0046] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0047] Figure 1 The diagram illustrates a flowchart of a task intent-driven dynamic construction method for data link networks proposed in an embodiment of this application.
[0048] Figure 2 A schematic diagram of the task intent knowledge graph proposed in an embodiment of this application is shown.
[0049] Figure 3 The model framework diagram of the ALBERT-BiLSTM-CRF entity recognition model is shown.
[0050] Figure 4 A schematic diagram of the network state knowledge graph proposed in an embodiment of this application is shown.
[0051] Figure 5 A flowchart of the simulated annealing algorithm proposed in an embodiment of this application is shown. Detailed Implementation
[0052] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.
[0053] Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0054] The existing technologies have the following problems and shortcomings: First, most methods fail to fully consider the task intent when constructing the network, instead employing fixed optimization strategies and focusing only on general network performance indicators. They ignore the differences in communication requirements due to different task intents, lacking adaptive adjustments to meet these requirements, thus affecting communication efficiency and overall task execution. Second, static networking struggles to achieve real-time adaptive construction of data link networks: traditional static networking methods typically rely on predefined topologies, lacking the ability to quickly respond to changes in task requirements. When faced with sudden tasks, node movement, or changes in link quality, static networking struggles to adjust network configurations in a timely manner, leading to decreased communication efficiency and failing to meet the adaptive networking requirements of highly dynamic scenarios.
[0055] Therefore, in order to solve the above-mentioned technical problems, this application proposes a task intent-driven dynamic construction method for data link networks. By constructing and integrating three major graphs—task intent, network state, and domain expertise—for collaborative reasoning, the abstract task intent is transformed into specific network construction parameters, and an optimization algorithm is used for real-time solution, which significantly improves the rapid response and adaptive construction capability of data link networks for complex and ever-changing tasks.
[0056] Please refer to Figure 1 , Figure 1 This paper presents a flowchart illustrating a task intent-driven dynamic construction method for data link networks, which includes the following steps:
[0057] Step S1: Extract key information about the task intent through knowledge extraction and construct a task intent knowledge graph.
[0058] Commanders input task instructions containing task intent through a front-end interface. These instructions cover key elements such as task intent type, task time, task location, and task objective. The system processes the input task instructions based on an ALBERT-BiLSTM-CRF entity recognition model. The ALBERT layer converts the input character or text sequence into vector form, capturing rich semantic information. The BiLSTM layer, based on the embedded vectors, extracts text features through bidirectional reading, considering contextual information. The CRF layer learns the constraints between labels, predicts entity labels, and uses the Viterbi algorithm to calculate the optimal label sequence, thereby accurately extracting key information such as entities, relationships, and attributes from the task. Subsequently, this key information is stored in the Neo4j graph database in a standardized knowledge graph format, ultimately forming a task intent knowledge graph, achieving accurate characterization of task intent.
[0059] In one possible implementation, the steps of extracting key information about the task intent through knowledge extraction and constructing a task intent knowledge graph include:
[0060] Extract key information such as task intent type, task time, task location, and task objective;
[0061] The ALBERT-BiLSTM-CRF entity recognition model is used to extract entities, relations, and attributes from the task intent.
[0062] The extracted entities, attributes, and relationships are stored in a graph database to construct a task intent knowledge graph.
[0063] Figure 2The diagram illustrates a task intent knowledge graph proposed in this application. This graph provides a machine-readable semantic representation of task intents and related attributes in a structured manner. Its core node, "Task 1," serves as the carrier of the intent and is connected to various parameter nodes through multiple predefined relational attributes (including, conditional, and parameter), thus comprehensively describing the multi-dimensional characteristics of a specific task. As shown in the figure, the task intent is parsed into key elements such as task type (monitoring), task number (01), task subject, task space (area A, air), and task objective (unrestricted). These elements are interconnected through semantic relationships, forming a knowledge network with clear hierarchy and logical connections.
[0064] When a task intent is input from the front end, the system extracts entities, relationships, and attributes from the task based on the ALBERT-BiLSTM-CRF entity recognition model, forming knowledge stored in the Neo4j graph database, thus creating a task intent knowledge graph and characterizing the task intent. Simultaneously, the underlying sensing devices extract node network state information and report it to the management center for preprocessing, also storing it in the form of a knowledge graph.
[0065] In this application implementation example, the task intent can have inputs similar to the following forms:
[0066] "Task 1: Monitor aerial area A. The mission requires covert operation and will commence at 00:00 on January 1, 2024, lasting for 2 hours."
[0067] The ALBERT-BiLSTM-CRF entity recognition model consists of an ALBERT layer, a BiLSTM layer, and a CRF layer.
[0068] The ALBERT layer is used to convert input characters into vector form;
[0069] The BiLSTM layer is used to extract features from the embedding vectors generated by the ALBERT layer;
[0070] The CRF layer is used to learn the constraint relationships between labels and predict entity labels.
[0071] Figure 3 The diagram illustrates the framework of the ALBERT-BiLSTM-CRF entity recognition model, which mainly consists of an ALBERT layer, a BiLSTM layer, and a CRF layer. Regarding the ALBERT layer, BERT is a versatile semantic representation model capable of capturing rich semantic information and complex patterns between sentences. This layer converts the input character data into vector form. The BiLSTM layer, by adding a reverse layer, enables bidirectional reading, fully considering contextual information and improving the embedding vectors generated by the ALBERT layer. Feature extraction is performed; the CRF layer learns the constraint relationships between labels and predicts entity labels; the optimal sequence of labels is calculated using the Viterbi algorithm to achieve global optimization.
[0072] Step S2: Each node senses its own status information and reports it to the management center to obtain the status information of the entire network. Based on the status information of the entire network, a network status knowledge graph is constructed. The status information of the entire network includes throughput, latency and packet loss rate.
[0073] Each node in the network uses its underlying sensing devices to extract and perceive its own state information. This state information includes, but is not limited to, key network performance indicators such as throughput, latency, and packet loss rate. Each node reports the perceived state information to the management center, which is responsible for the centralized processing and analysis of the collected network-wide state information. The processed network state information is stored in the form of a knowledge graph, that is, various types of network state information are written into the Neo4j graph database to construct a network state knowledge graph.
[0074] Figure 4 The diagram illustrates a network state knowledge graph proposed in this application. This knowledge graph uses multiple key network performance indicators as core nodes, including but not limited to "latency," "throughput," "jitter," "packet loss rate," and "node ID," and connects these attributes and their refined dimensions through semantic relationships indicated by arrows. "Latency" can be further subdivided into subcategories such as "transmission latency" and "processing latency"; the "throughput" node is associated with specific performance indicators such as "full-duplex throughput" and "half-duplex throughput"; "jitter" is characterized as "deterministic jitter" and "random jitter"; and "packet loss rate" is also distinguished into extreme states such as "maximum packet loss rate" and "minimum packet loss rate," demonstrating the completeness of a multi-dimensional and multi-granular description of the network state.
[0075] Step S3: Construct a domain knowledge graph, which includes the technical terminology, concept set, communication rules, and transmission constraints in the field of data link communication technology.
[0076] The domain-specific knowledge map encompasses the technical terminology and concept set of data link communication technology, as well as key knowledge such as communication rules and transmission constraints in different scenarios. It provides solid theoretical and professional knowledge support for the subsequent adaptive construction of data link networks, ensuring that the network construction meets the needs of actual applications and matches the task intent and network state.
[0077] In one possible implementation, the steps of constructing a domain knowledge graph include:
[0078] Construct at least one of the following knowledge graphs: mapping relationship between task intent and network topology configuration, mapping relationship between task intent and business message type, mapping relationship between business message type and communication node type, and mapping relationship between business message type and node communication requirements.
[0079] Constructing a domain-specific knowledge graph, encompassing but not limited to the technical terms and concepts of data link communication technology, as well as knowledge of communication rules and transmission constraints in different scenarios, provides professional knowledge support for adaptive network construction. The domain-specific knowledge graph, containing expertise in data link communication technology, supports task intent and network state analysis, ensuring that the construction of the data link network meets actual needs. Specifically, it includes: the relationship between task intent and network topology, the relationship between task intent and business message types, the relationship between business message types and transmission node types, and communication requirements between nodes, reflecting communication rules and transmission constraints in different scenarios.
[0080] Step S4: Based on the domain knowledge graph and task intent knowledge graph, reason out the network topology and business message type corresponding to the task intent, and determine the communication node type and node communication requirements according to the business message type.
[0081] By leveraging the domain-specific knowledge graph's coverage of data link communication technology terminology, concept sets, communication rules, and transmission constraints, and combining this with key information from the task intent knowledge graph, such as task intent type, task time, task location, and task objective, a reasoning mechanism is used to determine the network topology configuration and required service message types that match the task intent. Based on the characteristics and requirements of different service message types, the corresponding communication node types and inter-node communication requirements, such as latency, bandwidth, and packet loss rate, are identified.
[0082] In one possible implementation, the step of reasoning the network topology and service message type corresponding to the task intent based on the domain knowledge graph and the task intent knowledge graph includes:
[0083] Based on the task intent type, the corresponding business message type is obtained through reasoning using the domain knowledge graph. The business message types include status messages, image messages, video messages, monitoring messages, and evaluation messages.
[0084] Based on the task intent type and / or business message type, the corresponding network topology configuration is obtained through domain knowledge graph reasoning. The network topology configurations include distributed topology, star topology, mesh topology, and hierarchical topology.
[0085] Task intents include surveillance, coordination, and reconnaissance missions; business message types include, but are not limited to, status messages, image messages, video messages, monitoring messages, and assessment messages; as shown in Table 1, different task intents... Figure 1 Generally, they correspond to different task objectives and key requirements, and require the use of domain knowledge graphs to reason about the message types needed for the successful execution of the task intent;
[0086] Table 1
[0087]
[0088] Network topology configuration inference based on task intent refers to the fact that different task intents have different focuses on node communication requirements during business message transmission, including throughput, latency, and packet loss rate; at the same time, different network topologies differ in meeting these node communication requirements. As shown in Table 2, appropriate network topology configurations are selected based on different task intent inferences, combined with domain knowledge graphs.
[0089] Table 2
[0090]
[0091] In one possible implementation, the step of determining the communication node type and node communication requirements based on the service message type includes:
[0092] Based on the domain knowledge graph, the communication node types corresponding to the business message types are determined. The communication node types include control nodes, perception nodes, execution nodes, and information processing nodes.
[0093] Determine the node communication requirements corresponding to the business message type. The node communication requirements include tolerable latency, required throughput, and maximum packet loss rate.
[0094] Communication node types include, but are not limited to, control nodes, sensing nodes, execution nodes, and information processing nodes. Different node types have different business message communication requirements. Based on the business message type, the communication node type and node communication requirements are further determined. This means that specific business messages can only be transmitted between specific node types and have specific node communication requirements, as shown in Table 3. The reasoning process can be completed by combining the domain knowledge graph.
[0095] Table 3
[0096]
[0097] Step S5: Based on the node status information in the network state knowledge graph, combined with the communication node type, node communication requirements and domain professional knowledge graph, determine the source node and destination node for business message transmission.
[0098] First, real-time status data of each node is obtained from the network state knowledge graph, including key indicators such as throughput, latency, and packet loss rate. Then, based on the communication node type and communication requirements, the selection rules for source and destination nodes are dynamically updated using rules and constraints from the domain-specific knowledge graph. Finally, through automated processing by the inference engine, source and destination nodes that meet the current business message transmission requirements are accurately matched, ensuring that business messages can be transmitted efficiently and accurately in the network, meeting the dynamic networking requirements of the data link network driven by task intent.
[0099] In one possible implementation, the step of determining the source and destination nodes for service message transmission based on node state information in a network state knowledge graph, combined with communication node types, node communication needs, and domain-specific knowledge graphs, includes:
[0100] IF-THEN rules are formulated based on domain-specific knowledge graphs, and the rules define the selection criteria for source and destination nodes.
[0101] Based on the task intent type, business message type, and node communication requirements, and combined with the current network status information, the source node and destination node are obtained through IF-THEN rule matching.
[0102] By combining professional domain knowledge graphs, IF-THEN rules are formulated to define the selection criteria for source and destination nodes. The rules are defined by a comprehensive judgment of the network status. By classifying the network indicators and communication requirements of network nodes, such as latency being divided into high latency, relatively high latency, medium latency, relatively low latency, and low latency, nodes with latency not lower than the node's communication requirements are matched, which can quickly and accurately select source and destination nodes.
[0103] The selection rules for source and destination nodes are defined as follows:
[0104] RULE x:
[0105] IF Task Intent Type == "Specific Task Intent Type";
[0106] AND Business message types && Inter-node communication requirements;
[0107] THEN source node = "N1, N2, N3, ...";
[0108] AND destination node = "N4, N5, N6, ...";
[0109] For example, rule 1:
[0110] IF Mission Intent Type = "Reconnaissance Mission";
[0111] AND video information && high bandwidth, high latency (bandwidth > 10Mbps, latency < 300ms);
[0112] THEN Source nodes = 4, 5, 7, 10, 17, 24, 39 (scout nodes);
[0113] AND destination node = 1, 6, 11, 23 (control node);
[0114] Rule 2:
[0115] IF Mission Intent Type = "Reconnaissance Mission";
[0116] AND Image Information && Higher bandwidth, higher latency (bandwidth > 7Mbps, latency < 200ms);
[0117] THEN Source nodes = 3, 4, 8, 17, 19 (scout nodes);
[0118] AND destination node = 1, 11 (control node);
[0119] In one possible implementation, the method further includes:
[0120] When a new business message arrives, retrieve the type of the business message;
[0121] Check the current network status based on the network state knowledge graph;
[0122] Update matching rules by incorporating domain-specific knowledge graphs;
[0123] The inference engine is used to automatically match source and destination nodes.
[0124] When a new business message arrives, the system executes the following process to ensure its effective transmission. First, the type of the business message is obtained. Then, based on the network state knowledge graph, the current network state information is checked, including key indicators such as throughput, latency, and packet loss rate. Next, the matching rules are updated using a domain-specific knowledge graph to adapt to the current network conditions and business requirements. Finally, the inference engine automatically matches and determines the source and destination nodes for the business message transmission based on the updated rules, ensuring the rational allocation of network resources and efficient transmission of business messages.
[0125] Step S6: Combining the network topology, source node, and destination node, determine the data link network node connection relationship applicable to the current task intent.
[0126] After determining the network topology based on the task intent, the optimal connection relationships between nodes are determined using specific algorithms or rules, such as simulated annealing, based on the locations of the source and destination nodes for business message transmission. This step aims to ensure that the connection methods of network nodes meet the task intent and network performance requirements, achieving efficient and reliable data link network construction to adapt to the needs of the current task.
[0127] In one possible implementation, the steps of determining the data link network node connections suitable for the current task intent, based on the network topology, source nodes, and destination nodes, and constructing the data link network, include:
[0128] The simulated annealing algorithm is used for optimization and solution.
[0129] Set the constraint that the network topology remains unchanged when the network topology changes;
[0130] By exchanging the connection relationships of nodes or edges, a neighborhood solution is generated, and under constraints, the optimization is iteratively sought to determine the connection relationships of network nodes and construct a data link network.
[0131] By combining the source and destination nodes of the business message transmission with the network topology, a suitable network topology for the current task is determined. This process uses the simulated annealing algorithm to find the optimal topology by continuously changing the network topology. Figure 5 A flowchart of the simulated annealing algorithm proposed in an embodiment of this application is shown. The algorithm focuses on setting the objective function, determining the constraints, and the method for finding neighborhood solutions.
[0132] The network node connectivity is determined using the simulated annealing algorithm. By continuously changing the network topology, the optimal topology is sought. The algorithm focuses on setting the objective function, determining the constraints, and finding the neighborhood solution. Specifically:
[0133] Objective function: Different tasks require different transmission of service messages. The constructed network needs to ensure that the communication transmission capacity of any link between the source and destination nodes exceeds the service message demand. The objective function for optimization is: , Represents the set of all network metric types. Represents the set of all node pairs. This indicates the weights of throughput, latency, and packet loss rate. This indicates the business message communication requirement between a pair of nodes. It represents the actual communication transmission capacity between a pair of nodes.
[0134] Constraints: The task intent determines the network topology, and the network topology remains unchanged during network construction;
[0135] Neighborhood solution: When the result of the objective function does not meet the given requirements, the network topology is kept unchanged, and the network topology is changed by exchanging nodes or edges.
[0136] For distributed network topologies, connections between nodes typically lack a fixed center, allowing for the continuous redistribution of key nodes and the swapping of edges. For star network topologies, characterized by a single central node and multiple peripheral nodes, connections can be achieved by randomly swapping the central node or adjusting the connections of the peripheral nodes. For mesh network topologies, each node may be connected to multiple neighboring nodes, so the communication transmission capacity between each pair of nodes should be considered, placing nodes with weaker capabilities closer to the network edge and those with stronger capabilities closer to the network interior. For hierarchical network topologies, given the layered nature of the structure, methods such as swapping cluster heads and randomly assigning cluster members can be used.
[0137] Compared with the prior art, the embodiments of this application have the following beneficial effects:
[0138] First, by constructing and integrating three major graphs—task intent, network state, and domain knowledge—high-level, abstract task intents are automatically inferred into specific network topologies, service types, and node requirements, overcoming the shortcomings of traditional methods that rely on manual planning and cannot understand the connotation of tasks.
[0139] Secondly, based on real-time network state knowledge graphs and rule-based reasoning, it can dynamically respond to changes in node performance and fluctuations in link quality, quickly match the optimal source / destination nodes and optimize connection relationships, meeting the rapid networking requirements in highly dynamic and adversarial environments.
[0140] Third, optimization is performed with the objective function of meeting the end-to-end communication requirements of specific business messages, ensuring that network resources are accurately allocated to the most needed tasks and links, thereby improving the overall efficiency of network resource utilization and communication reliability.
[0141] Fourth, by embedding domain expertise into the graph, the network construction process not only depends on real-time status but also follows mature communication rules and constraints.
[0142] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.
Claims
1. A method for dynamically constructing a data link network driven by task intent, characterized in that, include: Key information about the task intent is extracted through knowledge extraction, and a task intent knowledge graph is constructed. Each node senses its own status information and reports it to the management center to obtain the status information of the entire network. Based on the status information of the entire network, a network status knowledge graph is constructed, which includes throughput, latency and packet loss rate. Construct a domain-specific knowledge graph, which includes the technical terminology, concept set, communication rules, and transmission constraints in the field of data link communication technology; Based on the domain knowledge graph and the task intent knowledge graph, the network topology and business message type corresponding to the task intent are inferred, and the communication node type and node communication requirements are determined according to the business message type. Based on domain-specific knowledge graphs and task intent knowledge graphs, the steps for reasoning to derive the network topology and business message type corresponding to the task intent include: Based on the task intent type, the corresponding business message type is obtained through reasoning using the domain knowledge graph. The business message types include status messages, image messages, video messages, monitoring messages, and evaluation messages. Based on the task intent type and / or business message type, the corresponding network topology configuration is obtained through reasoning based on the domain knowledge graph. The network topology configuration includes distributed topology, star topology, mesh topology and hierarchical topology. The steps to determine the communication node type and node communication requirements based on the business message type include: Based on the domain knowledge graph, the communication node types corresponding to the business message types are determined. The communication node types include control nodes, perception nodes, execution nodes, and information processing nodes. Determine the node communication requirements corresponding to the business message type. The node communication requirements include tolerable latency, required throughput, and maximum packet loss rate. Based on the node state information in the network state knowledge graph, combined with the communication node type, node communication requirements and domain professional knowledge graph, the source node and destination node for business message transmission are determined. Based on node state information in the network state knowledge graph, combined with communication node types, node communication needs, and domain-specific knowledge graphs, the steps to determine the source and destination nodes for business message transmission include: IF-THEN rules are formulated based on domain-specific knowledge graphs, and the rules define the selection criteria for source and destination nodes. Based on the task intent type, business message type, and node communication requirements, and combined with the current network status information, the source node and destination node are obtained by matching the IF-THEN rule. By combining the network topology, source nodes, and destination nodes, the data link network node connection relationships suitable for the current task intent are determined, and the data link network is constructed. The steps for constructing a data link network, based on the network topology, source nodes, and destination nodes, to determine the data link network node connections suitable for the current task intent include: The simulated annealing algorithm is used for optimization and solution. Set the constraint that the network topology remains unchanged when the network topology changes; By exchanging the connection relationships of nodes or edges, a neighborhood solution is generated, and under constraints, the network node connection relationships are determined to construct a data link network. The objective function for optimization is: , Represents the set of all network metric types. Represents the set of all node pairs. This indicates the weights of throughput, latency, and packet loss rate. This indicates the business message communication requirement between a pair of nodes. It represents the actual communication transmission capacity between a pair of nodes.
2. The method for dynamically constructing a data link network as described in claim 1, characterized in that, The steps involved in extracting key information about the task intent through knowledge extraction and constructing a task intent knowledge graph include: Extract key information such as task intent type, task time, task location, and task objective; The ALBERT-BiLSTM-CRF entity recognition model is used to extract entities, relations, and attributes from the task intent. The extracted entities, attributes, and relationships are stored in a graph database to construct a task intent knowledge graph.
3. The method for dynamically constructing a data link network as described in claim 2, characterized in that, The ALBERT-BiLSTM-CRF entity recognition model consists of an ALBERT layer, a BiLSTM layer, and a CRF layer. The ALBERT layer is used to convert input characters into vector form; The BiLSTM layer is used to extract features from the embedding vectors generated by the ALBERT layer; The CRF layer is used to learn the constraint relationships between labels and predict entity labels.
4. The method for dynamically constructing a data link network as described in claim 1, characterized in that, The steps to construct a domain knowledge graph include: Construct at least one of the following knowledge graphs: mapping relationship between task intent and network topology configuration, mapping relationship between task intent and business message type, mapping relationship between business message type and communication node type, and mapping relationship between business message type and node communication requirements.
5. The method for dynamically constructing a data link network as described in claim 1, characterized in that, The method further includes: When a new business message arrives, retrieve the type of the business message; Check the current network status based on the network state knowledge graph; Update matching rules by incorporating domain-specific knowledge graphs; The inference engine is used to automatically match source and destination nodes.