A method and system for automated generation of TSN network configuration based on reflective agents

Abstract

This invention relates to a method and system for automated generation of TSN network configurations based on a reflective agent. The system comprises a network state perception and intent reflection module, a YANG model configuration parameter and structure decision module, and a YANG model intelligent generation module. The network state perception and intent reflection module is responsible for structurally perceiving network state and resource information. The YANG model configuration parameter and structure decision module integrates multi-source knowledge and contextual information, intelligently generating deployable TSN scheduling parameters and YANG configuration structures based on business needs. The YANG model intelligent generation module automatically generates the corresponding YANG configuration model based on the optimized results, performs syntax and semantic verification, and then distributes it to the target device, achieving automated deployment of network configurations. This invention significantly improves the intelligence level and deployment automation capability of TSN network configuration by introducing reflective reasoning and feedback optimization techniques.

Description

Technical Field

[0001] This invention relates to the field of TSN network configuration, and specifically to a method and system for automatically generating TSN network configurations based on reflective agents. Background Technology

[0002] In the fields of industrial automation, aerospace, and intelligent transportation, TSN networks have become a core supporting technology for ensuring real-time communication due to their excellent determinism and low latency characteristics.

[0003] Currently, traditional methods generally lack a unified mechanism to efficiently and accurately transform optimized scheduling parameters into standardized configurations. Particularly in the configuration structure construction and parameter mapping, there is a high reliance on human experience and difficulty in closed-loop verification. The fragmented and low-intelligence configuration generation and deployment processes make it difficult to achieve end-to-end collaboration in TSN network design, optimization, verification, and actual deployment.

[0004] With the rapid development of artificial intelligence technology, a new path to break through the aforementioned bottlenecks is emerging. Reflective agents built on LLM (Low-Low Mechanics) possess superior language understanding, semantic reasoning, and structure generation capabilities. They can integrate network-aware data with business configuration intent, driving the entire process of automatic transformation from scheduling parameter optimization to standardized configuration generation. By introducing a reflective mechanism, the reflective agent can achieve self-optimization and policy adjustment through continuous feedback, thereby realizing closed-loop management of the entire process from network state awareness, scheduling optimization, configuration structure construction to configuration verification and automatic deployment, significantly improving the intelligence level and system adaptability of TSN (Traffic Safety Network) configuration. Summary of the Invention

[0005] This invention addresses the pain points of fragmented scheduling parameter optimization and configuration deployment in traditional TSN networks, characterized by complex processes and low efficiency. It innovatively proposes a method and system for automated TSN network configuration generation based on a reflective agent. This system integrates the efficient state awareness, semantic understanding, and continuous feedback optimization capabilities of the reflective agent with the powerful language generation and structural reasoning advantages of LLM in scheduling parameter optimization, YANG structure construction, and configuration verification. This achieves closed-loop automated management and control of the entire process, from network state awareness, scheduling parameter optimization, configuration structure generation to automatic verification and deployment. This system bridges the critical link between simulation optimization and actual deployment, significantly reducing manual configuration costs and error risks, improving the intelligent configuration efficiency and deployment accuracy of TSN networks, and providing solid technical support for the implementation of next-generation high-performance, automated network systems.

[0006] In one general aspect, a TSN network configuration automation generation system based on reflective agents is provided, characterized by a network state perception and intent reflection module, a YANG model configuration parameter and structure decision module, and a YANG model intelligent generation module:

[0007] The network state awareness and intent reflection module is used to comprehensively perceive the operating status of various entities in the network, structurally perceive and process device information, topology and link resources, and accurately identify the core optimization intent of the current simulation and configuration task based on context modeling and reflective reasoning mechanism, generate interpretable semantic input, and support intelligent decision-making for subsequent scheduling parameter optimization and YANG model configuration structure.

[0008] The YANG model configuration parameter and structure decision module, designed for specific business needs and network scheduling objectives, integrates multi-source knowledge and contextual information to intelligently generate high-quality, simulable, and deployable TSN scheduling parameters and YANG structure decisions. This module integrates the capabilities of knowledge acquisition, parameter optimization, structure planning, and result reflection, supports a closed-loop generation process from knowledge-driven to data feedback, realizes the parameter optimization and structure planning required for YANG model configuration, ensures the rationality and semantic integrity of the model structure, and supports continuous optimization and iterative improvement based on feedback.

[0009] The YANG model intelligent generation module accurately transforms network configuration intent into standardized YANG model configurations based on optimized scheduling parameters and structural decision results. By calling the MCP interface, it performs rigorous semantic and syntactic verification on the generated configurations to ensure their accuracy and executability. Once the verification is successful, the configuration will be automatically distributed to the target device, thereby achieving efficient and accurate automated deployment and controllable network management.

[0010] Furthermore, the network state awareness and intent reflection module is the foundation of the entire system. It comprehensively perceives the operational status of various entities in the network, structurally perceives and processes device information, topology, and link resources, and, based on contextual modeling and reflective reasoning mechanisms, accurately identifies the core optimization intent of the current simulation and configuration tasks, generating interpretable semantic input to support intelligent decision-making for subsequent scheduling parameter optimization and YANG model configuration structure. The network state awareness unit is responsible for perceiving and acquiring raw state data of various devices, network topology connections, and link resources in the network, providing comprehensive and accurate underlying network information support for subsequent data preprocessing. Through various protocols and interfaces, the network state awareness unit collects operational parameters of network devices, covering switches, routers, terminal devices, and key information on network topology connections and link resources, thereby providing rich and accurate underlying network information for subsequent data preprocessing.

[0011] Furthermore, the data preprocessing unit cleans, transforms, and standardizes the raw network state data. This unit uses a series of algorithms and rules to denoise, detect and repair outliers, and fill in missing data to ensure accuracy and completeness. Simultaneously, this unit converts unstructured or semi-structured data into a unified structured format, such as converting log data, text descriptions, and various sensor data into standardized tabular or JSON formats. In addition, this unit performs semantic annotation and normalization on the data to ensure semantic and structural consistency and usability, thus providing a high-quality, uniformly formatted data foundation for subsequent scheduling and configuration generation modules.

[0012] Furthermore, the intent reflection unit based on context engineering fully integrates multi-source heterogeneous information from the network operating environment, including current network state data, historical configuration records, simulation feedback results, and policy priors and empirical rules provided by the knowledge base. Through feature extraction, representation learning, and semantic modeling mechanisms, it constructs a multi-dimensional context vector representation with temporal sequence, relevance, and semantic consistency. Building upon this, the system utilizes embedded intent modeling and contrastive learning mechanisms, combined with an understanding of the target and constraint domains of the current configuration task, to identify the core business intent, optimization goals, and potential contradictions behind the configuration behavior. This enables proactive reflection on configuration behavior, precise positioning of optimization goals, and structured expression of configuration intent. This identification process not only enhances the perception and understanding of complex configuration tasks but also provides clear, interpretable, and semantically driven guidance for subsequent scheduling parameter optimization and YANG model structure planning.

[0013] Furthermore, the YANG model configuration parameter and structure decision module, oriented towards specific business needs and network scheduling objectives, integrates multi-source knowledge and contextual information to intelligently generate high-quality, simulable, and deployable TSN scheduling parameters and YANG structure decisions. This module integrates knowledge acquisition, parameter optimization, structure planning, and result reflection capabilities, supporting a closed-loop generation process from knowledge-driven to data feedback. It achieves parameter optimization and structure planning required for YANG model configuration, ensuring the rationality and semantic integrity of the model structure, and supports continuous optimization and iterative improvement based on feedback. The RAG-based knowledge base construction unit is used to build an intelligent knowledge base system integrating the RAG mechanism. It can efficiently integrate rich knowledge resources in the field, including simulation configuration knowledge, YANG model knowledge, configuration examples, modeling specifications, and related technical documents, achieving accurate acquisition and dynamic retrieval of the core knowledge required for simulation configuration generation and YANG model generation. This knowledge base not only covers professional content on TSN network scheduling and configuration but also supports context awareness and semantic understanding for specific tasks, ensuring the relevance and accuracy of the knowledge invoked. Furthermore, this unit supports continuous updates and adaptive optimization of knowledge content. It can dynamically adjust and improve the knowledge base content by combining feedback from the evaluation and reflection unit based on few-sample thinking chain technology and the intent reflection unit based on context engineering, thereby achieving closed-loop evolution and intelligent enhancement of knowledge and ensuring continuous optimization and improvement of the generation task in terms of professionalism, completeness and timeliness.

[0014] Furthermore, the configuration parameter optimization decision unit, based on given traffic demands and service characteristics, integrates sensing data, knowledge base content, and the current system context information. It performs joint modeling and efficient scheduling of communication tasks in the TSN network based on multi-strategy optimization algorithms, automatically solving for optimization schemes that satisfy various scheduling constraints. It also supports converting the generated scheduling parameters into structured configuration files that meet the requirements of specific simulation platforms. Configuration verification and performance evaluation are conducted by calling the corresponding simulation platform. Furthermore, based on simulation feedback, adaptive optimization and iterative updates of the scheduling strategy are achieved, thus strongly supporting the accurate generation and continuous optimization of TSN scheduling parameters. The TSN scheduling algorithm subunit, combining given traffic demands and service characteristics and comprehensively considering multi-dimensional network indicators, efficiently schedules communication tasks in the TSN network based on various optimization algorithms. This unit automatically solves for scheduling schemes that satisfy network constraints through joint modeling of sensing data and service characteristics, generating scheduling configuration results suitable for network simulation or actual deployment. During the scheduling process, this subunit can flexibly select and combine different scheduling schemes according to the real-time requirements of traffic, packet size, and transmission priority to ensure that communication tasks are executed efficiently while meeting network constraints. In addition, this subunit can receive results from the feedback optimization subunit and further optimize the scheduling parameters based on these feedback results to improve the accuracy and efficiency of scheduling and ensure that the generated scheduling configuration results are more in line with the actual network operation requirements.

[0015] The simulation configuration generation subunit integrates multi-source semantic information and scheduling parameter outputs, combining cross-source semantic fusion with intelligent modeling, meta-learning-driven context-aware reasoning, and cognitive enhancement generation capabilities based on large language models. This enables the automatic generation of high-precision, highly adaptable simulation configuration files, ensuring that the configuration content is highly consistent with the actual network scenario and scheduling parameters. This subunit uses advanced semantic processing technology to fuse data from different sources, ensuring that the generated configuration file accurately reflects the actual network environment and scheduling requirements. Simultaneously, it leverages advanced meta-learning and large language model technologies to enhance the intelligence level of the generated configuration file, enabling it to better adapt to complex and ever-changing network scenarios. Through this comprehensive processing approach, this unit can generate highly accurate and adaptable simulation configuration files, providing a solid foundation for network simulation and scheduling parameter verification.

[0016] The simulation execution subunit is used to invoke the network simulation platform to execute the configuration file output by the simulation configuration generation subunit. This completes the scenario restoration of the TSN network and the actual operation simulation of scheduling parameters. Based on the input scheduling parameters, it can conduct targeted simulation tests, obtain key network performance indicators and scheduling effect evaluation data, and provide realistic and quantifiable performance evidence for subsequent feedback optimization subunits. This unit has good compatibility and can seamlessly interface with various mainstream simulation tools, ensuring the comprehensiveness and effectiveness of simulation testing. Through precise simulation execution, key performance indicators are obtained, providing reliable data support for subsequent optimization. Furthermore, this unit can dynamically adjust test parameters based on simulation results to optimize the simulation process and improve testing efficiency and accuracy.

[0017] The feedback optimization subunit systematically analyzes the simulation execution results, extracts feedback data related to scheduling quality and configuration effectiveness, and uses this data to reverse-optimize the simulation configuration generation process and knowledge base content, achieving closed-loop adaptive optimization and improved simulation accuracy. This subunit deeply analyzes the simulation results, identifies potential problems in the configuration files, and feeds these problems back to the simulation configuration generation unit and the knowledge base. This enables dynamic updates to the knowledge base and continuous optimization of the configuration generation process, ensuring the accuracy of the simulation configuration is constantly improved. Furthermore, the feedback optimization unit also feeds back the analysis results to the TSN scheduling algorithm subunit, allowing this subunit to further adjust and optimize scheduling parameters based on the actual simulation results, making them more aligned with actual network operation requirements, thereby achieving coordinated optimization and performance improvement of the entire system.

[0018] Furthermore, the YANG structure decision unit is responsible for transforming network semantics and scheduling parameters into a standardized YANG model structure. It supports deep mapping to IEC 61850, automatically plans model modules and field organization, generates type definitions and configuration constraints, and ensures that the model semantics are accurate, the structure is reasonable, and the data is consistent. This provides a solid foundation for the intelligent generation and automatic deployment of efficient and standard-compatible TSN network configuration models.

[0019] The semantic structure mapping subunit is used to accurately map business semantics to the YANG model structure. This subunit not only supports extracting key semantic units from scheduling parameters, network intentions, and configuration goals, and establishing their correspondence with YANG syntax nodes, but also further extends its mapping support to the IEC 61850 standard information model. It can deeply analyze logical nodes, data objects, functional constraints, and service semantics, constructing a systematic semantic and structural conversion mechanism to achieve semantic alignment and structural fusion from the IEC 61850 model to the YANG model. Through this mapping capability, the semantic structure mapping subunit can effectively support the automatic conversion of power communication configurations from standard models to deployable configurations, improving the adaptability and standard consistency of the YANG model in smart grids and multi-protocol collaborative environments, and providing a stable and reliable semantic foundation for subsequent model generation and configuration deployment.

[0020] The structural planning subunit, based on semantic mapping results and combined with the network configuration scope and module partitioning logic provided in the knowledge base, plans the module boundaries, container nesting levels, and field organization of the YANG model. By analyzing business logic and network configuration requirements, it decides which modeling mechanism to adopt, ensuring the model structure possesses logical clarity, scalability, and reusability. Specifically, this subunit can reasonably divide the module boundaries of the YANG model according to the complexity and hierarchical structure of business requirements, design the nesting hierarchy of containers, and optimize the field organization. Through this structured planning, the YANG model not only clearly expresses business logic but also facilitates subsequent expansion and maintenance, while supporting inter-module reuse, improving the model's versatility and flexibility.

[0021] The constraint generation subunit automatically determines the type definitions and configuration constraints that each node field should possess, ensuring that the YANG model has a complete data consistency verification mechanism in subsequent generation stages and supports alignment and adaptation with device capabilities or scenario rules. This subunit analyzes business requirements and device characteristics, combined with device capabilities and scenario rules stored in the knowledge base, to automatically generate field type definitions, value ranges, and default values, ensuring that the generated YANG model conforms to actual application requirements in both syntax and semantics. Furthermore, this subunit can dynamically adjust constraints based on device capabilities and scenario rules, ensuring the model's applicability and consistency across different devices and scenarios. Through this automated constraint generation mechanism, the YANG model can better adapt to actual network environments, improving the accuracy and reliability of configuration.

[0022] Furthermore, the evaluation and reflection unit based on few-sample thinking chain technology, relying on few-sample reasoning and multi-step thinking chain mechanisms, comprehensively and intelligently evaluates the results after the configuration parameters and YANG structure decision generation. This unit not only detects the rationality, completeness, and adaptability of the generated content, but also delves deeper into potential defects and inconsistencies through step-by-step reasoning, forming a structured reflection report. Based on the evaluation results, the unit automatically generates targeted optimization schemes and improvement suggestions, and systematically feeds the reflection process and conclusions back to the knowledge base, achieving dynamic updates and continuous evolution of knowledge. Through this closed-loop mechanism, the system's generalization ability and decision-making accuracy for similar tasks are effectively enhanced, ensuring that subsequent configuration generation processes are more robust and efficient.

[0023] Furthermore, the YANG model intelligent generation module, based on optimized scheduling parameters and structural decision results, accurately transforms network configuration intent into standardized YANG model configurations. By calling the MCP interface, it rigorously verifies the generated configuration at the semantic and syntactic levels to ensure its accuracy and executability. Once verification is successful, the configuration is automatically distributed to the target device, achieving efficient and accurate automated deployment and controllable network management. The intelligent generation unit integrates semantic extraction, structural reasoning, and contextual fine-tuning capabilities, enabling it to extract core elements from scheduling parameters and network scenarios, and intelligently generate configuration files with complete structure and correct semantics. This unit performs deep analysis and intelligent transformation based on the optimized scheduling parameters and structural decision results provided by the YANG model configuration parameter and structural decision module. Specifically, the intelligent generation unit first calls the RAG-based knowledge base construction unit to efficiently integrate domain knowledge documents, configuration examples, and modeling specifications, accurately acquiring knowledge content related to scheduling parameters and network configuration intent. This unit closely relies on the module hierarchy, field organization, and constraint methods provided by the YANG structural decision unit, and through powerful structural reasoning capabilities, intelligently generates configuration files with complete structure and clear logic. This not only ensures the standardization and accuracy of the model but also greatly improves the efficiency and quality of configuration. Furthermore, this unit possesses excellent context-aware capabilities, enabling it to fine-tune the generated configuration files according to different network environments and business needs, thereby ensuring its effectiveness and adaptability in practical applications. This seamless integration from decision-making to generation, along with precise adaptation to real-world application environments, allows the intelligent generation unit to exhibit extremely high flexibility and reliability in complex network scenarios.

[0024] Furthermore, the MCP-based configuration verification unit performs syntax and semantic checks on the generated YANG model configuration by calling the management protocol interface. It relies on the device-side model constraints to perform format and logical consistency checks and receives execution feedback to ensure the accuracy and executability of the configuration. This unit, through a rigorous verification process, ensures that the generated YANG model configuration not only conforms to syntax specifications but also meets the operational requirements of the actual device. Through interaction with the device, this unit can promptly identify and correct potential configuration problems, thereby improving the success rate and reliability of the configuration.

[0025] Furthermore, the configuration distribution unit is responsible for accurately distributing the verified YANG configuration to the target network devices via the management and control protocol, ensuring that the configuration can be correctly received and applied by the devices, thus achieving automated configuration application and closed-loop deployment control. This unit securely transmits the configuration file to the target device through a reliable communication protocol and confirms through a feedback mechanism that the device has correctly received and applied the configuration. This process achieves automated and closed-loop management of network configuration, improving the efficiency and accuracy of network operation and maintenance.

[0026] In another overall aspect, a method for automatically generating TSN network configurations based on reflexive agents is provided, including the following steps:

[0027] S11: Sensing raw data from devices, topology, and links, combined with contextual modeling and reflective reasoning, to identify configuration goals and optimization intentions.

[0028] S12: Based on the knowledge base and scheduling algorithm, generate scheduling parameters that meet business requirements, and plan the YANG model structure and constraints that conform to the standards.

[0029] S13: If the evaluation passes, the YANG model configuration is intelligently generated based on the optimization results and accurately distributed through the management protocol after rigorous verification; if the evaluation fails, return to step S12 to continue optimization.

[0030] Compared with existing technologies, this invention innovatively introduces advanced LLM, Agent, and RAG technologies to construct a fully automated system covering network state awareness, intent reflection, scheduling parameter optimization, verification, and automatic deployment, comprehensively opening up the closed-loop link from TSN network design to implementation. Through the deep integration of the reflection mechanism, the system can dynamically adjust configuration strategies based on context and continuously optimize scheduling schemes through multiple rounds of evaluation and feedback, achieving adaptive intelligent decision-making and continuous improvement. This technology significantly improves the efficiency and accuracy of TSN network configuration, providing a solid guarantee for intelligent network management and efficient deployment. Specifically, it includes:

[0031] 1. It has achieved full automation from network status awareness and configuration parameter optimization to actual deployment, significantly reducing manual intervention, simplifying the configuration process, and effectively reducing configuration complexity and error risk.

[0032] 2. By introducing reflective reasoning and contextual modeling, the level of decision-making intelligence is improved. By adopting a reflective reasoning mechanism and combining contextual information and multi-source data, the intention of configuration optimization is dynamically identified, and interpretable and structured semantic input is generated, which effectively improves the adaptability and intelligence of the configuration scheme.

[0033] 3. Integrate multi-source knowledge-driven parameter optimization and structural decision-making capabilities, integrate RAG-based knowledge acquisition mechanisms, and combine policy knowledge, configuration specifications, and simulation experience to improve the generation quality and deployability of TSN scheduling parameters and YANG structures.

[0034] 4. Achieve cross-platform compatibility and support the use of various mainstream network simulation tools, which can significantly improve the efficiency of scheduling parameter optimization.

[0035] 5. Compared with the traditional manual configuration method, this system significantly shortens the configuration cycle of TSN network, reduces human resource investment, effectively avoids human error, improves configuration efficiency, supports rapid service deployment and iteration, and accelerates the network system launch process. Attached Figure Description

[0036] Figure 1 This is a system structure diagram of the present invention;

[0037] Figure 2 This is an overall flowchart of the method of the present invention;

[0038] Figure 3 A partial Prompt template for generating periodic flow scheduling and gating control configuration files for this invention;

[0039] Figure 4 This is a partial configuration file (for periodic flow scheduling and gating control) generated for this invention. Detailed Implementation

[0040] The following detailed embodiments are provided to aid the reader in gaining a comprehensive understanding of the methods and systems described herein. However, various changes, modifications, and equivalents of the methods and systems described herein will become apparent upon understanding this disclosure. For example, the order of operations described herein is merely illustrative and is not limited to those orders set forth herein, but may be changed as will become apparent upon understanding this disclosure, except for operations that must occur in a specific order. Furthermore, for clarity and conciseness, descriptions of features known in the art may be omitted.

[0041] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings, examples of which are illustrated in the drawings, wherein the same reference numerals always refer to the same parts. The embodiments will now be described with reference to the accompanying drawings in order to explain this disclosure.

[0042] This invention aims to provide a method and system for automated generation of TSN network configurations based on a reflective agent. Addressing the problems of low deployment efficiency and disconnect between optimized scheduling parameters and actual deployment in existing TSN networks, it innovatively introduces a reflective mechanism to achieve dynamic monitoring and multi-round feedback of network status and the configuration process. Through continuous evaluation and adjustment by the reflective agent, it ensures a high degree of consistency and adaptability between the scheduling scheme and deployment execution, ultimately achieving a fully automated closed loop of design, optimization, and deployment, significantly improving the efficiency, accuracy, and robustness of network configuration.

[0043] First, let's introduce and explain several terms involved in this invention:

[0044] Large Language Models (LLMs) are large-scale natural language processing models based on deep learning techniques. They typically employ a Transformer architecture, utilizing multi-head self-attention mechanisms and feedforward neural networks to efficiently model language. Through pre-training and fine-tuning, LLMs learn general knowledge and domain characteristics on massive general corpora and task-specific data, respectively, possessing powerful contextual understanding and language generation capabilities. Their parameter scale usually reaches hundreds of millions to trillions, enabling the model to capture complex semantic and language patterns. They are widely used in text generation, question-answering systems, machine translation, and intelligent assistants, and can also be used to support intelligent decision-making and scheduling optimization in complex scenarios.

[0045] The YANG (Yet Another Next Generation) model is a concrete data model file built based on the YANG data modeling language. It is used to describe the configuration data, status data, RPC operations, and notifications of network devices. The YANG model organizes data in a hierarchical structure, supports rich data types and flexible semantic expression, and can accurately describe the management information of network devices.

[0046] Time-Sensitive Networking (TSN) is a set of communication protocols developed by the IEEE 802.1 standardization organization to support real-time, reliable, and high-bandwidth Ethernet communication. The goal of TSN is to ensure data transmission under strict time constraints through deterministic networking technology, and it is widely used in industrial automation, automotive electronics, and aerospace.

[0047] Retrieval-Augmented Generation (RAG) is a method that combines retrieval and generation techniques. It first retrieves task-relevant information from a large dataset through retrieval, and then uses this information to assist in generating results. In LLM-based systems, RAG can help better understand and generate text, improving accuracy and relevance.

[0048] The system of this invention includes a network state perception and intent reflection module, a YANG model configuration parameter and structure decision module, and a YANG model intelligent generation module, such as... Figure 1 As shown, Figure 1 This is a system architecture diagram of the present invention. The network state perception and intent reflection module, as the fundamental core of the system, is responsible for comprehensively perceiving the operational status of various entities in the network and extracting device information, topology connections, and link resources in a structured manner. The network state perception unit collects real-time operational parameters of switches, routers, and terminal devices, as well as raw state data of network topology and link resources, through various protocols and interfaces, providing complete and accurate basic information for subsequent analysis. The data preprocessing unit cleans, denoises, detects and repairs anomalies in the perceived raw data, and converts unstructured or semi-structured data into a unified structured format. Combined with semantic annotation and normalization processing, this ensures the consistency and usability of the data at both the structural and semantic levels. Based on this, the intent reflection unit, based on context engineering, integrates network state, historical configuration records, simulation feedback, and knowledge base content to construct a multi-dimensional context vector representation. Through embedded intent modeling and comparative learning mechanisms, it accurately identifies the core business intent, optimization goals, and potential conflicts in the current configuration task, achieving proactive reflection and semantically driven expression of configuration behavior. This provides clear and interpretable semantic guidance for scheduling parameter optimization and YANG model structure decisions.

[0049] The YANG model configuration parameter and structure decision module, tailored to specific business needs and network scheduling objectives, integrates multi-source knowledge and contextual information to intelligently generate high-quality, simulable, and deployable TSN scheduling parameters and YANG structure decisions. This module integrates knowledge acquisition, parameter optimization, structure planning, and result reflection capabilities, supporting a closed-loop generation process from knowledge-driven to data feedback. It achieves parameter optimization and structure planning required for YANG model configuration, ensuring the rationality and semantic integrity of the model structure, and supports continuous optimization and iterative improvement based on feedback. The RAG-based knowledge base construction unit integrates multi-source knowledge resources, including simulation configurations, YANG models, configuration examples, and modeling specifications, enabling accurate acquisition and dynamic retrieval of core knowledge. The knowledge base possesses context awareness and semantic understanding capabilities, ensuring the accuracy and relevance of knowledge application. This unit supports the combination of an evaluation and reflection unit based on few-sample thinking chain technology and an intent reflection unit based on context engineering, continuously and dynamically updating and optimizing knowledge content to achieve closed-loop knowledge evolution, ensuring the professionalism, completeness, and timeliness of the generation task.

[0050] The configuration parameter optimization decision unit, oriented towards given traffic demands and service characteristics, integrates perception data, knowledge base content, and system context information. Based on a multi-strategy optimization algorithm, it jointly models and efficiently schedules TSN network communication tasks, automatically generating optimization schemes that meet various scheduling constraints. It also supports converting scheduling parameters into structured configuration files that conform to the simulation platform's requirements. By calling the simulation platform for configuration verification and performance evaluation, and combining the results from the feedback optimization unit, the scheduling parameters are dynamically adjusted to achieve adaptive optimization and continuous iteration of the scheduling strategy. This unit includes a TSN scheduling algorithm subunit, which comprehensively considers multi-dimensional network indicators and service characteristics to flexibly select scheduling schemes and ensure efficient execution of communication tasks; a simulation configuration generation subunit, which automatically generates high-precision, highly adaptable simulation configuration files using cross-source semantic fusion, meta-learning, and large language model technologies; a simulation execution subunit, which calls the simulation platform to complete scenario restoration and runtime simulation, obtaining key performance indicators to provide data support for subsequent optimization; and a feedback optimization subunit, which systematically analyzes the simulation results, identifies potential configuration problems, and feeds them back to the configuration generation and knowledge base, enabling dynamic knowledge updates and continuous optimization of the configuration process, thereby promoting overall system performance improvement and intelligent collaboration.

[0051] The YANG structure decision unit is responsible for transforming network semantics and scheduling parameters into a standardized YANG model structure, supporting deep mapping to the IEC 61850 standard and achieving system integration of semantics and structure. Its semantic structure mapping subunit accurately extracts business semantics and maps them to YANG syntax nodes, specifically parsing logical nodes, data objects, and functional constraints in IEC 61850 to ensure standard consistency of the model in smart grids and multi-protocol environments. The structure planning subunit intelligently plans model module boundaries, container nesting levels, and field organization based on semantic mapping results and configuration rules in the knowledge base, ensuring clear, scalable, and reusable model logic. The constraint generation subunit automatically decides on field type definitions and configuration constraints, dynamically adjusting them based on device capabilities and scenario rules to ensure the accuracy of model syntax and semantics and data consistency, improving the adaptability of the YANG model to actual network environments and the reliability of configuration. Overall, it achieves the construction of a high-quality, standard-compatible, and adaptive TSN network configuration model structure.

[0052] The YANG model intelligent generation module transforms optimized scheduling parameters and structural decision results into standardized YANG model configurations. The intelligent generation unit, through semantic extraction, structural reasoning, and contextual fine-tuning, combined with a RAG-based knowledge base and information provided by the YANG structural decision-making unit, generates standardized and highly adaptable YANG model configurations. The MCP-based configuration verification unit performs syntactic and semantic checks on the generated configurations to ensure their accuracy and executability. The configuration distribution unit then distributes the verified configurations to the target devices, achieving automated deployment and closed-loop management. This process achieves seamless integration from decision-making to deployment, improving the automation level of network configuration and operational efficiency.

[0053] The specific implementation process includes building a network state perception and intention reflection module, a YANG model configuration parameter and structural decision module, and a YANG model intelligent generation module.

[0054] Taking a smart substation as an example, the network status perception and intent reflection module first perceives the information of various intelligent devices (such as IEDs, protection devices, and monitoring and control terminals) and their interconnection status in real time, acquiring network topology, port connection status, link bandwidth occupancy, and transmission latency. Subsequently, based on contextual modeling and reflective reasoning mechanisms, it deeply analyzes the differences between the current network operation characteristics and configuration goals, identifies potential bottlenecks and optimization directions, and constructs an interpretable intent semantic structure. This provides accurate input for scheduling strategy optimization and YANG model configuration generation, realizing closed-loop linkage and intelligent adaptation between network status perception and configuration intent.

[0055] Based on RAG, a knowledge base building unit integrates the RAG mechanism to construct an intelligent knowledge base system. This system integrates core knowledge resources in the field of TSN network scheduling and configuration, including simulation configuration schemes, YANG model structure examples, modeling specifications, device capability models, and the IEC 61850 standard. For example, ... Figure 3 As shown, Figure 3 This invention generates a partial Prompt template for periodic flow scheduling and gating control configuration files. This unit supports contextual semantic retrieval based on specific configuration tasks, dynamically invoking knowledge fragments highly relevant to the task intent fed back by the network state awareness and intent reflection modules, becoming a key support for parameter optimization and model structure generation. Simultaneously, through linkage with the reflection mechanism, it continuously receives optimization suggestions, evaluation results, and configuration verification feedback, dynamically adjusting knowledge content and indexing strategies to achieve self-evolution and precision improvement of the knowledge base, providing a consistently fresh, accurate, and task-adaptive knowledge guarantee for subsequent parameter reasoning and structural modeling.

[0056] The configuration parameter optimization decision unit relies on multi-dimensional operational data provided by the network state awareness and intent reflection module. Combined with specific traffic requirements and service characteristics, it invokes protocol specifications and configuration instances from the RAG knowledge base to conduct joint modeling and intelligent analysis. This unit employs a multi-strategy optimization algorithm, integrating link load, latency tolerance, and service priority to model communication tasks and reason about scheduling strategies. It generates optimized scheduling parameters that satisfy timing consistency and quality of service, and automatically converts them into a structured configuration format adapted to the simulation platform. For example, it can generate configuration files for OMNeT++ simulations, such as... Figure 4 As shown, Figure 4 This is a partial configuration file for periodic flow scheduling and gating control generated by this invention. Furthermore, the TSN scheduling algorithm subunit, simulation configuration generation subunit, simulation execution subunit, and feedback optimization subunit work together to complete performance verification and feedback evaluation, and continuously optimize the scheduling strategy based on feedback, achieving efficient iteration and accurate convergence of the scheme. Through closed-loop optimization, it ensures that scheduling parameters dynamically adapt to network state and configuration intent, and continuously strengthens the generalization and adaptive capabilities under complex scenarios based on a knowledge base, providing a solid parameter foundation and intelligent decision support for subsequent YANG structure modeling.

[0057] In the YANG model configuration parameter and structure decision module, the YANG structure decision unit is responsible for making scientific decisions and planning the YANG model structure based on optimized configuration parameters and semantic intent results, combined with domain knowledge and standard specifications. The semantic structure mapping subunit maps scheduling parameters and power business logic (such as logical nodes and data objects in IEC 61850) to the YANG syntax structure, constructing a systematic semantic and structure conversion mechanism. The structure planning subunit, based on module functional boundaries, data hierarchical relationships, and configuration granularity, clarifies the model container organization, hierarchical nesting, and node partitioning strategies. The constraint generation subunit, combining scenario constraints and equipment capabilities, automatically determines field types, configuration ranges, and constraint rules, ensuring the model's clarity, rigor, and adaptability in logic, semantics, and application. This unit lays a solid foundation for model generation and configuration applications, improving the standard compatibility and practicality of configuration.

[0058] The evaluation and reflection unit, based on few-shot thinking chain technology, intelligently evaluates the generated configuration parameters and YANG structure after configuration parameter optimization and YANG structure decision generation. This is achieved by combining contextual information from the network state perception and intent reflection modules with domain knowledge from the RAG knowledge base, employing few-shot reasoning and a multi-step thinking chain mechanism. This unit progressively identifies potential defects and inconsistencies, proposes specific optimization suggestions, and feeds the reflection results back to the knowledge base, driving dynamic updates and closed-loop evolution of knowledge content. Through continuous learning and reflection, the system's generalization ability and decision accuracy in complex environments are improved, ensuring higher accuracy and applicability of the configuration model in smart grids and multi-protocol environments, providing a solid guarantee for intelligent modeling and automated deployment.

[0059] Finally, the YANG model intelligent generation module transforms the optimized scheduling parameters and structural decision results into standardized YANG configurations. The intelligent generation unit, relying on semantic extraction, structural reasoning, and contextual fine-tuning, combined with a RAG-based knowledge base and information from the YANG structural decision-making unit, generates standardized and highly adaptable YANG model configurations. Subsequently, the MCP-based configuration verification unit calls the management protocol interface to perform syntax and semantic verification, relies on device-side model constraints to check format and logical consistency, and receives execution feedback to ensure accurate and reliable configuration. Finally, the configuration distribution unit accurately distributes the verified YANG configuration to the target device through the management and control protocol, achieving efficient and automated deployment and network management.

[0060] A method for automatically generating TSN network configurations based on reflective agents, such as Figure 2 As shown, Figure 2 The overall flowchart of the method includes the following steps:

[0061] S11: Sensing raw data from devices, topology, and links, combined with contextual modeling and reflective reasoning, to identify configuration goals and optimization intentions.

[0062] S12: Based on the knowledge base and scheduling algorithm, generate scheduling parameters that meet business requirements, and plan the YANG model structure and constraints that conform to the standards.

[0063] S13: If the evaluation passes, the YANG model configuration is intelligently generated based on the optimization results and accurately distributed through the management protocol after rigorous verification; if the evaluation fails, return to step S12 to continue optimization.

[0064] This invention introduces a reflection mechanism to achieve efficient integration and automated closed-loop between TSN network scheduling parameter optimization and actual deployment. By combining a dynamic knowledge base, multi-strategy optimization, and feedback iteration, it intelligently evaluates and optimizes parameter generation results, continuously improving the system's adaptability and decision-making accuracy. This ensures that scheduling parameters precisely match network requirements, achieving efficient and intelligent automatic deployment.

[0065] The contents not described in detail in this specification are existing technologies known to those skilled in the art, and the above description is only a preferred embodiment of the present invention.

[0066] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A TSN network configuration automated generation system based on reflective agents, characterized in that, It includes a network state perception and intent reflection module, a YANG model configuration parameter and structure decision module, and a YANG model intelligent generation module. The network state perception and intent reflection module is used to comprehensively perceive the operational state of various entities in the network and generate interpretable semantic input YANG model configuration parameters and structural decision module. The YANG model configuration parameter and structure decision module integrates multi-source knowledge and contextual information to intelligently generate deployable TSN scheduling parameters and YANG structure decisions, and transmits them to the YANG model intelligent generation module. The YANG model intelligent generation module, based on the YANG model configuration parameters and the optimized scheduling parameters and structure decision results of the structure decision module, transforms the network configuration intent into a standardized YANG model configuration. By calling the MCP interface, it performs rigorous semantic and syntactic verification on the generated configuration to ensure its accuracy and executability. Once the verification is successful, the configuration will be automatically distributed to the target device, thereby achieving efficient and accurate automated deployment and controllable network management. The network state awareness and intent reflection module includes a network state awareness unit, a data preprocessing unit, and an intent reflection unit based on context engineering. The network status awareness unit is responsible for sensing and acquiring information on various devices in the network, network topology connections, and raw status data of link resources. The data preprocessing unit is used to clean, structure, and standardize the perceived raw network state data to ensure the semantic and structural integrity, consistency, and usability of the network data. The intent reflection unit based on context engineering integrates the current network state, historical configuration records, simulation feedback data, and knowledge base content to construct a multi-dimensional context vector representation. On this basis, it identifies the core business intent and optimization goals in the current configuration task, realizing proactive reflection on configuration behavior and intent-driven expression, providing clear semantic direction guidance for scheduling parameters and YANG structure planning.

2. The TSN network configuration automated generation system based on reflective agent as described in claim 1, characterized in that, The YANG model configuration parameter and structure decision module includes a RAG-based knowledge base construction unit, a configuration parameter optimization decision unit, a YANG structure configuration decision unit, and an evaluation and reflection unit based on small-sample thinking chain technology. The RAG-based knowledge base construction unit is used to build an intelligent knowledge base system that integrates the RAG mechanism. It can efficiently integrate domain knowledge documents, configuration examples and modeling specifications, and realize the accurate acquisition and dynamic calling of the core knowledge required for simulation configuration generation and YANG model generation. The configuration parameter optimization decision unit, oriented towards given traffic demands and business characteristics, integrates perception data, knowledge base content, and the current context information of the system, and performs joint modeling and efficient scheduling of communication tasks in the TSN network based on a multi-strategy optimization algorithm, automatically solving optimization schemes that satisfy multiple scheduling constraints. The configured YANG structure decision unit is responsible for transforming network semantics and scheduling parameters into a standardized YANG model structure; The evaluation and reflection unit based on few-sample thinking chain technology is used to intelligently evaluate and structurally reflect on the rationality, completeness and adaptability of the generated results after the configuration parameters and YANG structure decision are generated, based on the few-sample reasoning and thinking chain mechanism, thereby improving the system's generalization ability and decision accuracy in similar tasks.

3. The TSN network configuration automated generation system based on reflective agent as described in claim 2, characterized in that, The configuration parameter optimization decision unit further includes a TSN scheduling algorithm subunit, a simulation configuration generation subunit, a simulation execution subunit, and a feedback optimization subunit. The TSN scheduling algorithm subunit is used to efficiently schedule communication tasks in the TSN network by combining given traffic requirements and service characteristics, and to receive the results given by the feedback optimization subunit and further optimize the scheduling parameters based on these feedback results. The simulation configuration generation subunit automatically generates simulation configuration files, providing a solid foundation for network simulation and verification of scheduling parameters. The simulation execution subunit is used to call the network simulation platform to execute the configuration file output by the simulation configuration generation subunit, and complete the scene restoration and actual operation simulation of the scheduling parameters of the TSN network. The feedback optimization subunit systematically analyzes the simulation execution results, extracts feedback data related to scheduling quality and configuration effectiveness, and uses it to reverse-optimize the simulation configuration generation process and knowledge base content, thereby achieving closed-loop adaptive optimization and improving simulation accuracy.

4. The TSN network configuration automated generation system based on reflective agent as described in claim 2, characterized in that, The configuration YANG structure decision unit includes a semantic structure mapping subunit, a structure planning subunit, and a constraint generation subunit. The semantic structure mapping subunit is used to accurately map business semantics to the YANG model structure; The structural planning subunit is based on semantic mapping results, combined with the network configuration scope and module division logic provided in the knowledge base, to plan the module boundaries, container nesting levels and field organization of the YANG model; The constraint generation subunit automatically determines the type definition and configuration constraints that each node field should have, ensuring that the YANG model has a complete data consistency verification mechanism in the subsequent generation stage and supports alignment and adaptation with device capabilities or scenario rules.

5. The TSN network configuration automated generation system based on reflective agent as described in claim 1, characterized in that, The YANG model intelligent generation module includes an intelligent generation unit, an MCP-based configuration verification unit, and a configuration distribution unit. The intelligent generation unit, based on optimized scheduling parameters and structural decision results, integrates semantic extraction and structural mapping capabilities to automatically generate model configuration files that conform to the YANG specification. The MCP-based configuration verification unit performs syntax and semantic verification on the generated YANG model configuration by calling the management protocol interface; The configuration distribution unit is responsible for accurately distributing the verified YANG configuration to the target network device through the management and control protocol, ensuring that the configuration can be correctly received and applied by the device, and realizing automated configuration application and closed-loop deployment control.

Images