Multi-agent automatic operation method and equipment of optical transmission system and medium

Abstract

The invention provides a multi-agent automatic operation method and device for an optical transmission system and a medium, and the method achieves the full-chain autonomous operation of the optical transmission system through constructing a framework of the cooperative work of four agents, namely execution, monitoring, fault detection and repair. The execution agent analyzes the natural language instruction of the user and calls a tool function to complete experiment configuration; the monitoring agent analyzes the operation state of the system in real time; the fault detection intelligent agent adopts a multi-mode large model for fine adjustment of an optical transmission scene, system images and data are fused and analyzed, and accurate fault positioning and severity evaluation are achieved; and the repairing agent automatically calls a tool to execute parameter adjustment or hardware optimization according to the diagnosis result. The problems that traditional manual operation and maintenance is high in learning cost, response lags behind, diagnosis depends on experience and closed-loop control cannot be achieved are solved, and high-automation and intelligent operation and maintenance from system configuration, state monitoring, intelligent diagnosis to automatic repairing are achieved.

Description

Technical Field

[0001] This invention relates to the field of optical transmission system technology, and more specifically, to a method, device, and medium for automated operation of multiple agents in an optical transmission system. Background Technology

[0002] Optical transmission systems are the core of modern communication networks, and their operation and maintenance face many challenges. Traditional methods rely heavily on manual operation, resulting in high learning costs, slow response times, reliance on personal experience for fault diagnosis, and the inability to form an automated closed loop from problem discovery to repair, making it difficult to meet the needs of efficient and intelligent operation and maintenance.

[0003] Existing automation technologies often focus on single steps, such as using scripts to configure parameters or using rules to provide simple alarm responses. These solutions lack flexibility and adaptability, especially when detecting system faults. They typically only analyze structured data such as logs and metrics, while ignoring crucial unstructured visual information such as spectral images. This failure to achieve deep fusion of multimodal data limits the accuracy and comprehensiveness of fault location.

[0004] In recent years, advancements in large language models for task planning and multimodal large models for text and image understanding have offered new possibilities for building intelligent operation and maintenance systems. However, how to deeply integrate these capabilities with knowledge from the optical transmission domain to design a fully automated framework capable of collaboratively completing the entire chain from configuration, monitoring, multimodal fault diagnosis to autonomous repair remains a pressing technical challenge. Currently, there is a lack of a unified multi-agent collaborative architecture to achieve professional division of labor and organic collaboration among various stages, ultimately achieving a high degree of autonomy in system operation and maintenance. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a method, device, and medium for automated operation of multiple agents in an optical transmission system.

[0006] A multi-agent automated operation method for an optical transmission system according to the present invention includes the following steps: Construct four collaborative intelligent agents: an execution agent, a monitoring agent, a fault detection agent, and a repair agent. The execution agent is used to interact with the user in natural language, parse the experimental intent, and call pre-built utility functions to complete the configuration of the optical transmission link and start the experiment. The monitoring agent is used to monitor the system's operating status and experimental logs after the experiment is started, and to determine whether the experiment is normal or to trigger a fault handling process based on the operating status. The fault detection agent uses a finely tuned multimodal large language model as its core, receives images and data generated during system operation, performs multimodal analysis to locate faults and predict their severity; The repair agent is used to call corresponding tool functions to adjust hardware devices or experimental parameters based on the output results of the fault detection agent, so as to repair system faults and restore performance. The execution agent, monitoring agent, fault detection agent, and repair agent interact and perform task flow according to a preset collaborative logic, thereby achieving full-chain automation of the optical transmission system from configuration, operation, monitoring, fault detection, and repair.

[0007] Preferably, the executing agent uses a plain text large language model as its core and combines it with the ReAct prompt word framework to realize the automated link configuration and experimental operation of thinking and execution.

[0008] Preferably, the monitoring agent uses a plain text large language model as its core, and judges the running status based on experimental logs and historical experimental results, guided by prompt words.

[0009] Preferably, the finely tuned multimodal large language model used by the fault detection agent is obtained by artificially adding faults to a normally operating optical transmission system and recording the corresponding images and data, and then training the model so that it can predict fault labels based on the multimodal information during fault operation.

[0010] Preferably, the repair agent uses a plain text large language model as its core and automatically calls tool functions to complete hardware optimization or parameter adjustment based on fault information.

[0011] Preferably, the collaborative logic specifically includes: The executing agent wakes up the monitoring agent after the experiment is started; The monitoring agent wakes up the fault detection agent when it detects a system anomaly. The fault detection agent wakes up the repair agent after completing the fault location and assessment; After the repair agent completes the repair, it returns control to the executing agent or the monitoring agent to continue or re-execute the experiment.

[0012] An electronic device according to the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the multi-agent automated operation method of the optical transmission system.

[0013] According to the present invention, a computer-readable storage medium is provided thereon storing a computer program, which, when executed by a processor, implements the multi-agent automated operation method of the optical transmission system.

[0014] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention employs intelligent agents with natural language understanding capabilities, allowing users to configure the system and start experiments through natural language interaction without needing to master complex device commands or script knowledge, greatly reducing the operational threshold. Simultaneously, each intelligent agent automatically performs monitoring, diagnosis, and repair tasks, improving overall operational efficiency and solving the problems of high learning costs and low efficiency in existing technologies.

[0015] 2. In this invention, the monitoring agent can continuously analyze experimental logs and operational data, replacing manual judgment of the status. Once an anomaly is detected, it can immediately trigger the subsequent fault handling process, overcoming the lag and negligence of manual monitoring, and effectively improving the real-time perception and response speed of the system status.

[0016] 3. This invention uses a multimodal large model finely tuned for optical transmission scenarios as the core of the fault detection agent. This agent can comprehensively analyze the image and numerical data generated during system operation and perform multimodal fusion analysis, thereby achieving more accurate location and severity assessment of complex and latent faults. Attached Figure Description

[0017] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This invention provides a multi-agent automation method for an optical transmission system. Figure 2 This is a diagram illustrating an example of operation in an embodiment of the present invention. Detailed Implementation

[0018] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the protection scope of the present invention.

[0019] This invention discloses a multi-agent automated operation method for an optical transmission system, based on Figure 1The overall framework is shown below. First, a hardware and software integrated environment needs to be built: an optical transmission experimental system containing key equipment such as lasers, modulators, fiber optic channels, spectrometers, and optical power meters. All key parameters can be remotely controlled and data acquired through software interfaces. Simultaneously, a series of tool functions are developed and encapsulated for the control and data reading functions of the aforementioned equipment, forming a unified tool library that serves as a bridge for interaction between intelligent agents and the physical world.

[0020] Construction and configuration of each intelligent agent 1. Construction of the execution agent This intelligent agent uses a plain text large language model as its core engine. During implementation, system prompts based on the ReAct (Think-Act) framework need to be designed and embedded. These prompts guide the model to perform structured parsing of the user's input natural language commands. Specifically, the model first identifies the implicit experimental configuration intent in the command through semantic understanding, such as "set modulation format" or "configure fiber link length." Then, it decomposes the overall command into several sequentially executed sub-operation steps. For each sub-step, the model matches the corresponding callable function interface from a pre-defined tool function library based on its semantic features (such as operation type, target module, and parameter values). This tool function library predefines standardized operation interfaces that correspond one-to-one with the hardware or simulation modules of the optical communication experimental platform. Simultaneously, the model extracts structured parameters from the natural language and performs validity checks according to preset parameter constraint rules. Finally, the model outputs an ordered sequence of tool calls, which the execution engine sequentially calls to complete the automatic configuration and startup of the experimental link.

[0021] 2. Construction of the monitoring agent This intelligent agent is also based on a plain text large language model. By providing it with real-time, structured experimental operation data streams (such as bit error rate, optical power value, and device status code) and historical normal operating condition data ranges, specific analysis prompts are designed to guide the model to continuously evaluate whether the current data is within the normal range, that is, whether the current experimental results differ significantly from the data in similar experimental scenarios in the past, or even directly generate alarm information. Once abnormal indicators are detected or a preset fault mode is met, the system is judged to be malfunctioning.

[0022] 3. Construction of Fault Detection Intelligent Agent This intelligent agent is the core of this invention for achieving accurate diagnosis. It employs a specially fine-tuned multimodal large language model. The fine-tuning process is as follows: Under the normal operation of the optical transmission experimental system, various preset faults are artificially introduced, such as loose connectors, laser wavelength shift, and amplifier saturation. Simultaneously, multimodal information generated by the system at this time is collected, including equipment status images (such as screenshots of the spectrometer display interface), panel indicator light images, and corresponding structured performance data. These "multimodal data-fault label" paired samples are used to supervise the fine-tuning of the basic multimodal large model, enabling the model to learn to associate specific graphical and textual feature patterns with specific fault types and severity. After deployment, this intelligent agent can perform end-to-end analysis of the input real-time system snapshots and output structured diagnostic conclusions.

[0023] 4. Repair the construction of intelligent agents This intelligent agent is based on a plain text large language model, and its knowledge base contains pre-built repair strategy knowledge for different fault types. When invoked, it receives the fault diagnosis results, parses out the sequence of repair actions to be performed (e.g., "first fine-tune the laser temperature, then recalibrate the bias point"), and sequentially calls the corresponding tool functions to complete the system tuning.

[0024] Multi-agent collaborative workflow The automated operation of the entire system follows Figure 1 The defined collaborative logic, combined with Figure 2 Here is a running example, a complete workflow implementation, with the following specific process: 1. Task Initiation: The user issues experimental commands to the executing agent via a natural language interface. After the executing agent completes the configuration and starts the experiment, the monitoring agent is automatically activated.

[0025] 2. Status Monitoring and Anomaly Triggering: The monitoring agent begins real-time analysis of the experiment's logs. If the status is determined to be normal, the results are fed back after the experiment ends; if an anomaly is determined (e.g., ...), the system detects an anomaly. Figure 2 As shown in the running example, if a sudden increase in the bit error rate is detected, the fault handling process is immediately triggered, and a diagnostic request is sent to the fault detection agent, along with a snapshot of the system's multimodal data at the current moment.

[0026] 3. Fault Diagnosis: The fault detection agent is awakened and uses its fine-tuned multimodal analysis capabilities to comprehensively analyze the received images and data, and output accurate diagnostic information such as "Fault location: Abnormal loss at the third fiber optic connection; Severity level: Medium".

[0027] 4. Automatic Repair: The diagnostic information described above is sent to the repair agent. The repair agent generates a specific repair operation sequence based on the fault type, such as calling a series of tool functions like "clean fiber optic connector interface" or "re-optimize modulator bias voltage" to automatically execute the repair actions.

[0028] 5. Closed-loop verification: After the repair action is completed, system control returns to the monitoring agent. The monitoring agent reassesses the system status. If the performance indicators return to the normal range, the experimental process continues; if not, the diagnosis-repair process can be iteratively executed, or a high-level alarm can be triggered.

[0029] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. A multi-agent automated operation method for an optical transmission system, characterized in that, Includes the following steps: Construct four collaborative intelligent agents: an execution agent, a monitoring agent, a fault detection agent, and a repair agent. The execution agent is used to interact with the user in natural language, parse the experimental intent, and call pre-built utility functions to complete the configuration of the optical transmission link and start the experiment. The monitoring agent is used to monitor the system's operating status and experimental logs after the experiment is started, and to determine whether the experiment is normal or to trigger a fault handling process based on the operating status. The fault detection agent uses a finely tuned multimodal large language model as its core, receives images and data generated during system operation, performs multimodal analysis to locate faults and predict their severity; The repair agent is used to call corresponding tool functions to adjust hardware devices or experimental parameters based on the output results of the fault detection agent, so as to repair system faults and restore performance. The execution agent, monitoring agent, fault detection agent, and repair agent interact and perform task flow according to a preset collaborative logic, thereby achieving full-chain automation of the optical transmission system from configuration, operation, monitoring, fault detection, and repair.

2. The multi-agent automated operation method for an optical transmission system according to claim 1, characterized in that, The execution agent uses a plain text large language model as its core and combines it with the ReAct prompt word framework to realize the automated link configuration and experimental operation of thinking and execution.

3. The multi-agent automated operation method for an optical transmission system according to claim 1, characterized in that, The monitoring agent uses a plain text large language model as its core, and judges the running status based on experimental logs and historical experimental results, guided by prompt words.

4. The multi-agent automated operation method for an optical transmission system according to claim 1, characterized in that, The fault detection agent uses a finely tuned multimodal large language model, which is trained by artificially adding faults to a normally operating optical transmission system and recording the corresponding images and data. This enables the model to predict fault labels based on the multimodal information during fault operation.

5. The multi-agent automated operation method for an optical transmission system according to claim 1, characterized in that, The repair agent uses a plain text large language model as its core and automatically calls tool functions to complete hardware optimization or parameter adjustment based on fault information.

6. The multi-agent automated operation method for an optical transmission system according to any one of claims 1 to 5, characterized in that, The collaborative logic specifically includes: The executing agent wakes up the monitoring agent after the experiment is started; The monitoring agent wakes up the fault detection agent when it detects a system anomaly. The fault detection agent wakes up the repair agent after completing the fault location and assessment; After the repair agent completes the repair, it returns control to the executing agent or the monitoring agent to continue or re-execute the experiment.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the multi-agent automated operation method of the optical transmission system as described in any one of claims 1 to 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the multi-agent automated operation method of the optical transmission system as described in any one of claims 1 to 6.

Images