Fiber touch tool optical fiber positioning system and identification method

Abstract

The invention relates to the technical field of optical fiber positioning, and provides a fiber touch tool optical fiber positioning system and an identification method, which not only can accurately identify the type of an optical fiber, but also can capture the physical change of the surface of the optical fiber by comprehensively sensing the state of the optical fiber and performing feature extraction and mapping through a deep neural network. The accuracy and the reliability of judging the state of the optical fiber are improved; an optical fiber topological structure is generated based on space-time correlation analysis, and real-time correction is performed in combination with environmental data, so that the morphological change of the optical fiber caused by environmental change or operation can be quickly and accurately reflected; an obstacle avoidance path is generated according to the compensation topological structure and the compensation parameter set, and track errors are monitored in real time to judge whether the optical fiber topological structure needs to be reconstructed or not, so that in the operation process of the fiber touch tool, a safe and efficient motion path can be planned, potential dangerous areas are avoided, the optical fiber topological structure can be adjusted in time according to the actual operation condition, and the operation efficiency is improved. Therefore, the safety and the efficiency of optical fiber operation are remarkably improved.

Description

Technical Field

[0001] The present application relates to the field of optical fiber positioning technology, and in particular to an optical fiber positioning system and identification method for a fiber-feeding tool. Background Art

[0002] In many fields such as modern communication networks, data centers, and medical equipment, fiber-optic communication has become a key information transmission method due to its significant advantages such as high bandwidth, low loss, and strong anti-interference. In this context, fiber optic positioning, as a basic link in operations such as fiber connection and fiber patching, plays a decisive role in the performance and stability of the entire communication system due to its accuracy and efficiency.

[0003] Current traditional fiber optic positioning methods have exposed many problems in practical applications. On the one hand, manual operation is extremely dependent on manpower. Operators not only have to manually connect the optical fiber to the correct port, but the operation process is cumbersome and complicated, consuming a lot of time and energy, and extremely inefficient. Moreover, human factors can easily interfere with the positioning results. In a complex environment with numerous optical cable cores and dense ports, the probability of error increases significantly, making it difficult to ensure the accuracy of positioning. On the other hand, with the continuous expansion of the scale and increasing complexity of communication networks, manual positioning can no longer meet the urgent needs of fast and accurate wiring.

[0004] In recent years, while artificial intelligence technologies such as image recognition have been gradually applied to fiber optic positioning, traditional image recognition algorithms struggle to effectively capture key features due to the large number of fiber distribution ports and dense fiber layouts, resulting in poor positioning accuracy. Furthermore, existing fiber optic positioning systems lack effective monitoring and dynamic correction mechanisms for real-time changes in fiber topology when faced with complex environmental factors. This causes the positioning system to become disconnected from the actual fiber status, further reducing positioning accuracy. Most approaches have not addressed how to monitor fiber status to achieve topology correction and accurate fiber positioning in complex application scenarios. Summary of the Invention

[0005] In response to the deficiencies of the existing technology, the present application provides a fiber optic positioning system and identification method for a fiber optic tool.

[0006] In a first aspect, the present application provides a fiber positioning system for a fiber-feeding tool, the system comprising: monitoring physical signals on the surface of an optical fiber, extracting features of the physical signals to generate a data tensor including a vibration spectrum, pressure distribution, and infrared reflectivity, performing feature mapping on the data tensor based on a deep neural network, identifying the optical fiber type, and generating a feature vector;

[0007] Performing spatiotemporal correlation analysis on the eigenvectors to generate a fiber topology, identifying bending nodes, axial deviations, and stress distributions of the fiber topology, monitoring environmental data to correct the fiber topology, outputting a compensation topology, generating a compensation parameter set including an angle offset, a pressure correction value, and a vibration suppression strength based on the compensation topology, and incrementally updating a data tensor based on the compensation parameter set to regenerate the eigenvectors;

[0008] The compensation topology and compensation parameter set are received, and an obstacle avoidance path is generated through a path planning algorithm. The obstacle avoidance path is converted into a control instruction sequence of the fiber-feeding tool, and the trajectory error between the actual trajectory of the fiber-feeding tool and the obstacle avoidance path is monitored to determine whether the optical fiber topology should be reconstructed.

[0009] As an optional implementation, the logic for generating the feature vector includes:

[0010] Based on a dual-branch convolutional neural network, the spatial features and dependency features of the data tensor are extracted, and the spatial features and dependency features are spliced and fused to obtain a multi-scale feature map;

[0011] The dual-channel attention mechanism is used to weight the multi-scale feature map from the channel dimension and spatial dimension respectively. The weighted multi-scale feature map is input into the dynamic threshold-based classification network, and the classification score is output. The configured threshold parameter is compared with the classification score to identify the fiber type.

[0012] The fiber type is integrated with the spatial position and confidence of the fiber to generate a feature vector.

[0013] As an optional implementation, the data tensor generation sub-logic includes:

[0014] Monitor physical signals on the optical fiber surface, including vibration signals, pressure signals, and infrared signals;

[0015] The vibration signal is decomposed using the empirical mode decomposition algorithm, and the vibration spectrum of the vibration signal is extracted using the Hilbert-Huang transform. At the same time, the pressure signal is processed based on the morphological algorithm and the pressure area boundary is identified to extract the pressure distribution of the pressure signal.

[0016] The infrared signal is normalized by piecewise linear interpolation, and the infrared reflectivity of the infrared signal is extracted by principal component analysis. The vibration spectrum, pressure distribution and infrared reflectivity are fused according to timestamp and spatial position to generate a data tensor.

[0017] As an optional implementation manner, the generation logic of the compensation parameter set includes:

[0018] Extract the compensation topology, calculate the bending angle change of each bending node in the optical fiber topology and the compensation topology, and obtain the angle offset;

[0019] Determine pressure-sensitive areas based on stress distribution, calculate pressure correction values for these areas using Hooke's law, and configure pressure thresholds based on fiber type and fiber operation scenarios. Compare the pressure distribution on the fiber surface with the thresholds to adjust the update frequency of the pressure correction values.

[0020] Perform time-frequency analysis on the vibration spectrum to extract the vibration frequency components and vibration energy, and determine the vibration suppression intensity based on the vibration frequency components and vibration energy;

[0021] The angle offset, pressure correction value and vibration suppression strength are normalized, an identifier and a timestamp are added to each compensation parameter, and the processed compensation parameters are integrated to generate a compensation parameter set.

[0022] As an optional implementation manner, the generation sub-logic of the optical fiber topology structure includes:

[0023] The spatial position and timestamp of the feature vector are extracted and combined into a four-dimensional index. The four-dimensional index is used as the data point, and a spatiotemporal R-tree index is constructed based on the data point. The time window is dynamically set according to the fiber operation scenario, and the data points within each time window are filtered through the spatiotemporal R-tree index.

[0024] The selected data points are processed by spatiotemporal density clustering algorithm to perform spatiotemporal correlation analysis on the feature vectors and obtain spatiotemporal clustering results;

[0025] Based on the spatiotemporal clustering results, each cluster center is extracted as a topological node. The connection relationship of the topological nodes is established through the triangulation algorithm to form topological edges to generate the optical fiber topology structure. A multidimensional attribute vector including bending angle, curvature radius and stress value is attached to each topological node.

[0026] As an optional implementation manner, the optical fiber topology identification sub-logic includes:

[0027] Calculate the space vector angle between adjacent topological nodes in the optical fiber topology structure, dynamically set the angle threshold according to the optical fiber type, compare the space vector angle with the angle threshold to determine the candidate bending node, and detect the curvature continuity of the candidate bending node to identify the bending node of the optical fiber topology structure;

[0028] The theoretical axis of the optical fiber is constructed based on the standard parameters of the optical fiber type. The topological nodes are fitted using the least squares method, and the vertical distance from each topological node to the theoretical axis is calculated to identify the axial deviation of the optical fiber topology structure.

[0029] Combining pressure distribution, fiber material properties and topological nodes, the pressure distribution is converted into node loads through finite element analysis, and the stress distribution of the fiber topological structure is calculated using the conjugate gradient method.

[0030] Based on the topological isomorphism detection algorithm, the graph edit distance is calculated to measure the difference between the fiber topology and the standard fiber topology in topological nodes and topological edges. The difference degree and stress distribution are combined to identify the deformed topology.

[0031] As an optional implementation manner, the output sub-logic of the compensation topology structure includes:

[0032] Monitor environmental data, including temperature, humidity, air pressure, and external vibration. Normalize the environmental data into environmental factors. Combine these environmental factors with the fiber material properties to update the coordinates of each topological node, reestablish the connection between the topological nodes, and generate new topological edges.

[0033] The stress distribution, bending nodes, updated topological node coordinates, and new topological edges are fused through a spatiotemporal alignment algorithm to obtain enhanced properties. Reinforcement learning is then used to dynamically modify the fiber topology based on the enhanced properties.

[0034] The modified optical fiber topology is evaluated based on geometric accuracy, physical rationality and environmental adaptability to obtain the reliability of the optical fiber topology. Based on the reliability judgment of the optical fiber topology, a compensation topology is output.

[0035] As an optional implementation manner, the reconstruction judgment sub-logic of the optical fiber topology structure includes:

[0036] The obstacle avoidance path is converted into a control instruction sequence for the fiber-feeding tool. The spatial position error between the actual trajectory of the fiber-feeding tool and the obstacle avoidance path is calculated through the root mean square error to determine the position error. The actual posture of the fiber-feeding tool during the control process is monitored and compared with the posture preset in the obstacle avoidance path to determine the posture error.

[0037] Analyze the actual motion parameters of the fiber-sensing tool during the control process and compare them with the preset motion parameters of the obstacle avoidance path to determine the parameter error. The position error, attitude error, and parameter error are used as trajectory errors. Combined with environmental data and the reliability of the compensation topology structure, an evidence set is constructed.

[0038] Based on Bayesian reasoning, the probability of compensation topology structure changes under different evidence sets is learned according to historical data, and the posterior probability that the compensation topology structure needs to be reconstructed is calculated through Bayesian reasoning. Based on the posterior probability, it is determined whether the optical fiber topology structure should be reconstructed.

[0039] As an optional implementation manner, the obstacle avoidance path generation sub-logic includes:

[0040] Convert the bending nodes, axial deviations, and stress distribution of the compensation topology into physical constraints, and build an environmental risk map based on environmental data, while setting task priorities based on task types.

[0041] Based on physical constraints, environmental risk maps, and task priorities, an initial path is generated on the compensation topology using a fast-exploring random tree algorithm.

[0042] The initial path is optimized based on a deep reinforcement learning network to output optimized local path segments. The optimized local path segments are smoothed using a spline curve fitting algorithm, and the speed of the smoothed local path segments is planned according to the motion performance parameters of the fiber-feeding tool to generate an obstacle avoidance path.

[0043] In a second aspect, the present application provides a method for optical fiber positioning and identification using a fiber-feeding tool, the method comprising: monitoring physical signals on the surface of an optical fiber, extracting features of the physical signals, and generating a data tensor including a vibration spectrum, pressure distribution, and infrared reflectivity;

[0044] Perform feature mapping on data tensors based on deep neural networks to identify fiber types and generate feature vectors;

[0045] Performing spatiotemporal correlation analysis on the feature vectors to generate the fiber topology and simultaneously identify the bending nodes, axial deviation and stress distribution of the fiber topology;

[0046] Monitoring environmental data to correct the optical fiber topology, outputting a compensation topology, and generating a compensation parameter set including an angle offset, a pressure correction value, and a vibration suppression strength according to the compensation topology;

[0047] Incrementally updating the data tensor based on the compensation parameter set to regenerate the feature vector;

[0048] Receive the compensation topology and compensation parameter set, generate an obstacle avoidance path through a path planning algorithm, and convert the obstacle avoidance path into a control instruction sequence for the fiber optic tool;

[0049] Monitor the trajectory error between the actual trajectory of the fiber-touching tool and the obstacle avoidance path to determine whether to reconstruct the optical fiber topology.

[0050] Compared with the existing technology, the beneficial effects of the present application are: through comprehensive perception of the optical fiber status, breaking through the limitations of traditional single signal monitoring, integrating the physical signals of vibration signals, pressure signals and infrared signals, and performing feature extraction and mapping through deep neural networks, it can not only accurately identify the optical fiber type, but also capture the physical changes on the optical fiber surface, greatly improving the accuracy and reliability of optical fiber status judgment; and generating the optical fiber topology based on spatiotemporal correlation analysis, and combining it with environmental data for real-time correction, which can quickly and accurately reflect the morphological changes of the optical fiber caused by environmental changes or operations; generating an obstacle avoidance path based on the compensation topology and compensation parameter set, and converting it into a control instruction for the fiber-touching tool, monitoring the actual trajectory of the tool and the trajectory error of the obstacle avoidance path in real time to determine whether the optical fiber topology needs to be reconstructed. This process enables the fiber-touching tool to plan a safe and efficient movement path during operation, avoid potential danger areas, and adjust the optical fiber topology in time according to actual operating conditions, thereby significantly improving the safety and efficiency of optical fiber operation and reducing the cost of manual intervention.

[0051] The positioning system comprehensively monitors physical signals such as vibration, pressure, and infrared on the optical fiber surface, and extracts features such as vibration spectrum, pressure distribution, and infrared reflectivity to generate data tensors. It cooperates with deep neural networks for feature mapping to achieve accurate identification of optical fiber types and generate feature vectors. This process changes the previous limitation of single signal monitoring that was difficult to accurately judge the optical fiber status. The fusion analysis of multimodal signals can capture subtle changes on the optical fiber surface, providing a rich and accurate data foundation for subsequent operations.

[0052] The positioning system conducts spatiotemporal correlation analysis based on eigenvectors, generates optical fiber topology, and identifies key information such as bending nodes, axial deviations, and stress distribution. At the same time, it corrects the topology in combination with environmental data, outputs a compensated topology, and then generates a compensation parameter set for updating the data tensor and eigenvector. This process realizes the dynamic modeling of the optical fiber topology and can reflect in real time the morphological changes of the optical fiber caused by environmental changes or human operations. For example, in a high-temperature environment, the positioning system can correct the coordinates of the topological nodes according to the environmental data and adjust the optical fiber morphology simulation results to ensure that the optical fiber topology is always in line with reality, thereby improving the timeliness and accuracy of the optical fiber status assessment.

[0053] The positioning system receives the compensation topology and compensation parameter set, generates an obstacle avoidance path through a path planning algorithm, and converts it into a control instruction sequence for the fiber-feeding tool. At the same time, it monitors the error between the actual trajectory of the tool and the planned path to determine whether the optical fiber topology needs to be reconstructed. During operation, the fiber-feeding tool can plan a safe and efficient movement path based on the real-time topology of the optical fiber and environmental constraints, avoiding dangerous areas such as bending nodes and stress concentration areas. Moreover, through trajectory error feedback, the positioning system can promptly detect potential changes in the optical fiber topology. For example, abnormal tool movement indicates optical fiber position offset, which in turn triggers the reconstruction of the optical fiber topology, realizing adaptive adjustment of the positioning system, ensuring the safety and accuracy of optical fiber operation, and improving the stability and reliability of the entire optical fiber positioning system in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be derived from these drawings without inventive work. Among them:

[0055] Figure 1 This is a system flow chart of the optical fiber positioning system for the fiber-touching tool provided in an embodiment of the present application;

[0056] Figure 2 A sub-logic diagram for generating the optical fiber topology structure of the optical fiber positioning system of the fiber-touching tool provided in an embodiment of the present application;

[0057] Figure 3 A sub-logic diagram for generating an obstacle avoidance path of the optical fiber positioning system of the fiber-touching tool provided in an embodiment of the present application;

[0058] Figure 4 This is a flow chart of the method for optical fiber positioning and identification of the fiber-touching tool provided in an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the objectives, technical solutions and advantages of the embodiments of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application are clearly and completely described below in conjunction with the drawings in the specification. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments.

[0060] Example 1

[0061] like Figure 1 As shown, a system flow chart of a fiber optic positioning system for a fiber optic tool is provided for an embodiment of the present application. The system includes a modal perception module, a topology construction module and a path planning module.

[0062] The modal perception module is used to monitor the physical signals on the surface of the optical fiber, extract the features of the physical signals to generate data tensors including vibration spectrum, pressure distribution and infrared reflectivity, perform feature mapping on the data tensor based on a deep neural network, identify the optical fiber type and generate a feature vector.

[0063] Furthermore, the data tensor generation logic includes:

[0064] Monitor physical signals on the optical fiber surface, including vibration signals, pressure signals, and infrared signals;

[0065] The vibration signal is decomposed using the empirical mode decomposition algorithm, and the vibration spectrum of the vibration signal is extracted using the Hilbert-Huang transform. At the same time, the pressure signal is processed based on the morphological algorithm and the pressure area boundary is identified to extract the pressure distribution of the pressure signal.

[0066] The infrared signal is normalized by piecewise linear interpolation, and the infrared reflectivity of the infrared signal is extracted by principal component analysis. The vibration spectrum, pressure distribution and infrared reflectivity are fused according to timestamp and spatial position to generate a data tensor.

[0067] In actual use, the physical signals on the surface of optical fibers contain a wealth of status information, including vibrations that reflect external interference and touch conditions, pressure that reflects the stress state, and infrared signals that can identify material and temperature changes. Acquiring these signals is the basis for subsequent analysis. A multi-type sensor array is integrated on the head of the fiber-touching tool, where vibration signals are acquired by a piezoelectric sensor that can capture weak vibrations, pressure signals are acquired by a pressure sensor, and infrared signals are acquired by an infrared sensor with adjustable wavelength that can operate at multiple wavelengths to fully acquire infrared reflection information from optical fibers of different materials. The three sensors work together to continuously acquire physical signals from the surface of the optical fiber. The integration and collaborative acquisition of multiple types of sensors achieves comprehensive coverage of the physical signals on the surface of the optical fiber, avoids information loss caused by single signal acquisition, and provides a sufficient data basis for subsequent analysis. The acquired original physical signal will serve as input and enter the subsequent signal processing link. Its quality and integrity directly affect the accuracy of extracting features such as vibration spectrum, pressure distribution, and infrared reflectivity.

[0068] The raw vibration and pressure signals contain noise and interference and lack a characteristic form suitable for analysis, requiring specific algorithms for processing and feature extraction. For vibration signals, the Empirical Mode Decomposition (EMD) algorithm is first used to decompose the complex vibration signal into multiple intrinsic mode functions (IMFs), each representing the vibration characteristics of a different frequency component, thereby removing noise interference. The Hilbert-Huang transform is then used to convert the decomposed vibration signal from the time domain to the time-frequency domain, clearly extracting the vibration spectrum. For pressure signals, a morphological algorithm is used to first binarize the pressure signal obtained by the pressure sensor to highlight the pressure region. Then, through dilation and erosion operations, the pressure region boundaries are accurately identified to extract the pressure distribution. The combination of EMD and the Hilbert-Huang transform effectively separates the noise and effective components in the vibration signal and accurately extracts the vibration spectrum. The morphological algorithm processes the pressure signal, accurately defines the pressure region boundaries, and obtains a reliable pressure distribution, improving the quality of signal processing and the accuracy of feature extraction. The extracted vibration spectrum and pressure distribution, together with the subsequently obtained infrared reflectivity, are used to generate a data tensor, the accuracy of which determines the accuracy of the data tensor's description of the fiber state.

[0069] The original infrared signal has data differences due to factors such as sensor characteristics and needs to be normalized. At the same time, in order to reduce the data dimension and extract key features, the principal component analysis method is used to finally fuse the three signal features into a data tensor to achieve a unified expression of multimodal information. The infrared signal is first normalized using the piecewise linear interpolation method to make the infrared reflectivity data at different wavelengths in the same scale range. Then, the principal component analysis method is used to reduce the dimensionality of the normalized data and extract the component that best represents the infrared signal characteristics to obtain the infrared reflectivity. Finally, the vibration spectrum, pressure distribution and infrared reflectivity are combined according to the timestamp and spatial position. The data are fused to construct a three-dimensional data tensor. During the fusion process, time is used as the timestamp axis and the sensor array position corresponds to the spatial position axis. The piecewise linear interpolation method and principal component analysis method are used to process the infrared signal, eliminating data differences, extracting key features, and reducing data redundancy. The fusion of multiple features generates a data tensor, realizing a unified multimodal description of the physical signal on the optical fiber surface, which is convenient for subsequent feature mapping and analysis based on deep neural networks. The generated data tensor serves as the input of the logical step of generating feature vectors. The multimodal information it contains provides a rich data foundation for deep neural networks to extract spatial features and dependent features.

[0070] Specifically, the feature vector generation logic includes:

[0071] Based on a dual-branch convolutional neural network, the spatial features and dependency features of the data tensor are extracted, and the spatial features and dependency features are spliced and fused to obtain a multi-scale feature map;

[0072] The dual-channel attention mechanism is used to weight the multi-scale feature map from the channel dimension and spatial dimension respectively. The weighted multi-scale feature map is input into the dynamic threshold-based classification network, and the classification score is output. The configured threshold parameter is compared with the classification score to identify the fiber type.

[0073] The fiber type is integrated with the spatial position and confidence of the fiber to generate a feature vector.

[0074] The information in the data tensor has multi-scale characteristics, and feature extraction at a single scale cannot fully describe the fiber characteristics. The dual-branch convolutional neural network can extract spatial features and dependency features at different scales, providing richer feature expressions for accurate fiber type identification. A dual-branch convolutional neural network is constructed. The main branch consists of multiple convolutional layers with different convolutional kernel sizes. Small convolutional kernels are used to extract local and subtle spatial features and capture detailed information on the fiber surface, while large convolutional kernels are used to extract macroscopic spatial features and grasp the overall morphology of the fiber. The auxiliary branch uses a dilated convolutional layer to expand the network's receptive field without increasing excessive computational effort, extracting long-range dependency features in the data tensor, namely the correlation between signals at different locations. The outputs of the two branches are spliced and fused to obtain a multi-scale feature map containing multi-scale spatial and dependency features. The design of the dual-branch convolutional neural network fully utilizes the characteristics of different convolutional layers and dilated convolutional layers to extract features from the data tensor at multiple scales, avoiding the limitations of single-scale feature extraction, enriching the feature expression, and improving the accuracy and comprehensiveness of the fiber feature description. The multi-scale feature map provides rich feature input for the dual-channel attention mechanism, which helps to further enhance key features and suppress irrelevant information.

[0075] The importance of features in each channel and spatial dimension in the multi-scale feature map is different. The dual-channel attention mechanism can adaptively give higher weights to key features. The classification network based on dynamic threshold can achieve more accurate classification according to the feature distribution of different fiber types. The multi-scale feature map is input into the dual-channel attention mechanism, which is divided into a channel attention module and a spatial attention module. In the channel attention module, feature information is extracted from the channel dimension through global average pooling and global maximum pooling operations, and then the importance weight of each channel is calculated by the multi-layer perceptron to weight the channels of the multi-scale feature map. In the spatial attention module, the convolution operation is used to generate a spatial attention map to highlight the multi-scale features. The features of the key spatial positions in the feature map are used to weight the spatial positions of the multi-scale feature map. The weighted multi-scale feature map is input into the classification network based on the dynamic threshold. The classification network learns the distribution pattern of different fiber types in the feature space, dynamically adjusts the classification threshold, and compares the classification score output by the classification network with the configured threshold parameter to identify the fiber type. The dual-channel attention mechanism’s weighted processing of the multi-scale feature map effectively enhances the key features, suppresses redundant information, and improves the expressiveness of the features. The classification network based on the dynamic threshold can adapt to the feature differences of different fiber types, achieve more accurate fiber type identification, and improve the accuracy and reliability of identification.

[0076] To fully describe the fiber status, it is necessary to integrate information such as fiber type, spatial position and confidence into a feature vector to provide comprehensive and structured data input for the topology construction module; after determining the fiber type, the fiber type is combined with the spatial position of the fiber, where the spatial position is determined by the sensor array position. At the same time, the confidence is calculated based on the classification score output by the classification network, reflecting the reliability of fiber type identification. Finally, the fiber type, spatial position and confidence are integrated in a specific order and format to generate a feature vector; the feature vector integrates key information such as fiber type, spatial position and confidence, and comprehensively describes the fiber status in a structured form, providing clear, accurate and complete data input for the topology construction module, which helps to generate the fiber topology structure more accurately in the future. The generated feature vector serves as the input of the topology construction module, and the information it contains will be used for spatiotemporal correlation analysis to generate the fiber topology structure, which is an important data foundation for topology construction.

[0077] The topology construction module is used to perform spatiotemporal correlation analysis on the eigenvectors to generate the optical fiber topology, identify the bending nodes, axial deviations and stress distribution of the optical fiber topology, monitor environmental data to correct the optical fiber topology, output the compensation topology, generate a compensation parameter set including angle offset, pressure correction value and vibration suppression strength according to the compensation topology, and incrementally update the data tensor based on the compensation parameter set to regenerate the eigenvector.

[0078] Furthermore, if Figure 2 As shown, the generation sub-logic of the fiber topology structure includes:

[0079] The spatial position and timestamp of the feature vector are extracted and combined into a four-dimensional index. The four-dimensional index is used as the data point, and a spatiotemporal R-tree index is constructed based on the data point. The time window is dynamically set according to the fiber operation scenario, and the data points within each time window are filtered through the spatiotemporal R-tree index.

[0080] The selected data points are processed by spatiotemporal density clustering algorithm to perform spatiotemporal correlation analysis on the feature vectors and obtain spatiotemporal clustering results;

[0081] Based on the spatiotemporal clustering results, each cluster center is extracted as a topological node. The connection relationship of the topological nodes is established through the triangulation algorithm to form topological edges to generate the optical fiber topology structure. A multidimensional attribute vector including bending angle, curvature radius and stress value is attached to each topological node.

[0082] The feature vector contains the spatial position and timestamp of the optical fiber. Single-dimensional data is difficult to accurately reflect the changing rules of the optical fiber in the time and space dimensions. Constructing a four-dimensional index and generating a spatiotemporal R-tree index can efficiently organize and manage these data, facilitate the subsequent rapid screening and query of optical fiber data within a specific time and space range, and provide a basis for spatiotemporal correlation analysis; extract the spatial position (x, y, z) and timestamp (t) from the feature vector, combine them into a four-dimensional index (x, y, z, t), use these four-dimensional indexes as data points, and organize the data points in layers according to the construction rules of the R-tree index. First, data points with similar spatial positions and close timestamps are aggregated into bottom-level nodes, and then the bottom-level nodes are further aggregated to form higher-level nodes until a complete spatiotemporal correlation analysis is constructed. During the construction process, the empty R-tree index optimizes the node partitioning strategy and balances the tree hierarchy to ensure the query efficiency of the index. The construction of the spatiotemporal R-tree index realizes the orderly management of fiber data in the spatiotemporal dimension, greatly improving the speed and accuracy of data retrieval. Whether querying the fiber data in a specific area at a certain point in time or obtaining the change information of the fiber over a period of time, it can be quickly located and extracted through the index, saving a lot of time and computing resources for the subsequent spatiotemporal clustering analysis. The generated spatiotemporal R-tree index provides an efficient data query tool for dynamic time window screening data, enabling the positioning system to quickly and accurately obtain data points in each time window, providing input data for the spatiotemporal density clustering algorithm and promoting the spatiotemporal correlation analysis.

[0083] In actual operation scenarios, the state changes of optical fibers have different time scales. When the optical fiber is subjected to instantaneous touch, the vibration changes rapidly and requires a short time window to capture. However, the slow deformation caused by changes in ambient temperature requires long time window monitoring. Dynamically setting the time window can adapt to different types of changes in optical fibers and ensure that comprehensive and effective data is obtained for spatiotemporal correlation analysis. According to the optical fiber operation scenario and task requirements, the positioning system pre-sets time windows of different lengths, including short-time windows and long-time windows. During operation, it automatically selects the appropriate time window by analyzing the current state change trend and environmental conditions of the optical fiber. The spatiotemporal R-tree index is used to filter out the corresponding data points in each time window. When a rapid vibration signal of the optical fiber is detected, the short-time window is enabled to quickly extract all relevant data points in the short time from the spatiotemporal R-tree index. If the optical fiber state changes slowly, the long-time window is used for data screening.

[0084] The setting of dynamic time windows enables the positioning system to flexibly respond to changes in optical fibers in different scenarios, avoiding data loss or redundancy problems caused by fixed time windows. By accurately screening data, it ensures that the data processed by the spatiotemporal density clustering algorithm can accurately reflect the spatiotemporal change characteristics of the optical fiber, thereby improving the accuracy and effectiveness of spatiotemporal correlation analysis. The data points in each time window screened out serve as the input of the spatiotemporal density clustering algorithm, laying the foundation for accurate spatiotemporal correlation analysis and obtaining reliable spatiotemporal clustering results, which directly affects the determination of subsequent topological nodes and the generation of optical fiber topology structure.

[0085] The filtered data points are distributed in a disorderly manner in the space-time dimension. It is necessary to use a clustering algorithm to aggregate data points with similar space-time characteristics to identify the distribution pattern and change law of optical fiber in different space-time regions, thereby realizing the space-time correlation analysis of the characteristic vector and providing key clustering information for constructing the optical fiber topology structure; the filtered data points are processed by the space-time density clustering algorithm. When calculating the distance between data points, the space-time density algorithm considers both spatial distance and time interval, and introduces a time decay factor to give recent data points a higher weight in the clustering process. The space-time density algorithm starts from a certain data point and searches for data points with reachable density around it, continuously expanding the cluster until no new data points can be added. By setting appropriate density thresholds and distance parameters, the data points are divided into different space-time clusters to obtain space-time clustering results. Each cluster represents a state set of the optical fiber in a specific space-time region.

[0086] The spatiotemporal density clustering algorithm fully considers the spatiotemporal characteristics of optical fiber data, can accurately cluster data points with similar spatiotemporal characteristics, and effectively identify the spatiotemporal change pattern of optical fiber. Compared with traditional clustering algorithms, this algorithm highlights the importance of recent data through the time attenuation factor, is more in line with the actual changes of optical fiber, improves the accuracy and reliability of clustering results, and provides accurate clustering information for subsequent optical fiber topology generation. The obtained spatiotemporal clustering results are the core basis for generating optical fiber topology structure. Based on these clustering results, the positioning system can determine the location of topological nodes and then construct the optical fiber topology structure. Its accuracy directly determines whether the generated optical fiber topology structure can truly reflect the actual shape and distribution of the optical fiber.

[0087] The spatiotemporal clustering results only represent the distribution of optical fibers in the spatiotemporal region, and need to be converted into an intuitive optical fiber topology structure in order to clearly display the spatial direction and connection relationship of the optical fiber. At the same time, adding a multidimensional attribute vector to each topological node can enrich the information of the topological structure so that it not only contains geometric position information, but also reflects the physical properties of the optical fiber, providing more comprehensive data support for subsequent structural analysis and application; based on the spatiotemporal clustering results, each cluster center is extracted as a topological node, and a connection relationship is established between these topological nodes through the triangulation algorithm to form a topological edge, thereby constructing the initial topological structure of the optical fiber. Then, according to the position and connection relationship of the topological nodes, the bending angle and curvature radius of each node are calculated. The calculation of the stress value will be explained in detail in the identification sub-logic of the optical fiber topology structure. The information such as the bending angle, curvature radius and stress value is integrated into a multidimensional attribute vector and attached to the corresponding topological node to complete the generation of the optical fiber topology structure.

[0088] The topological structure generated by the triangulation algorithm can intuitively present the spatial layout of the optical fiber, clearly show the direction and connection relationship of the optical fiber, and add multi-dimensional attribute vectors to the topological nodes, so that the optical fiber topological structure contains rich physical information, which is conducive to in-depth analysis of the stress conditions and morphological characteristics of the optical fiber. This topological structure containing geometric and physical properties provides a comprehensive and accurate data foundation for subsequent bending node identification, axial deviation calculation, and stress distribution analysis, thereby enhancing the application value of the optical fiber topological structure. The generated optical fiber topology structure and its node attributes are the basic data for subsequent steps such as optical fiber topology structure identification, compensation topology structure output, and compensation parameter set generation. Subsequent identification and analysis work will be carried out based on this structure and attributes. Its accuracy and completeness directly affect the performance of the entire topology construction module and the working effect of the subsequent path planning module.

[0089] Furthermore, the identification sub-logic of the optical fiber topology structure includes:

[0090] Calculate the space vector angle between adjacent topological nodes in the optical fiber topology structure, dynamically set the angle threshold according to the optical fiber type, compare the space vector angle with the angle threshold to determine the candidate bending node, and detect the curvature continuity of the candidate bending node to identify the bending node of the optical fiber topology structure;

[0091] The theoretical axis of the optical fiber is constructed based on the standard parameters of the optical fiber type. The topological nodes are fitted using the least squares method, and the vertical distance from each topological node to the theoretical axis is calculated to identify the axial deviation of the optical fiber topology structure.

[0092] Combining pressure distribution, fiber material properties and topological nodes, the pressure distribution is converted into node loads through finite element analysis, and the stress distribution of the fiber topological structure is calculated using the conjugate gradient method.

[0093] Based on the topological isomorphism detection algorithm, the graph edit distance is calculated to measure the difference between the fiber topology and the standard fiber topology in topological nodes and topological edges. The difference degree and stress distribution are combined to identify the deformed topology.

[0094] Bend nodes are an important feature of optical fiber topology. Accurately identifying bend nodes helps understand the morphological changes of optical fibers and determine whether the optical fibers are in a normal state. It also provides key information for subsequent path planning to prevent fiber damage caused by fiber-touching tools at bend nodes. The spatial vector angle between adjacent topological nodes in the optical fiber topology is calculated, and the corresponding angle threshold is dynamically set according to the different optical fiber types. For single-mode optical fibers, which are more sensitive to bending, a smaller angle threshold is set. For multi-mode optical fibers, the threshold can be appropriately relaxed. The calculated spatial vector angle is compared with the set angle threshold. If the angle is greater than the angle threshold, the topological node is marked as a candidate bend node. In order to further accurately identify bend nodes, the curvature continuity of the candidate bend node is detected. The curvature change near the candidate bend node is calculated. If the curvature suddenly changes or is discontinuous at the candidate bend node, the candidate bend node is determined to be a true bend node, and its position and related parameters are recorded.

[0095] By dynamically setting the angle threshold, the bending characteristics of different types of optical fibers are fully considered, and the accuracy and specificity of bend node identification are improved. Combined with curvature continuity detection, it effectively avoids misjudgments caused by noise or local fluctuations, ensuring that the identified bend nodes can truly reflect the actual bending conditions of the optical fiber. Accurate identification of bend nodes provides a reliable basis for subsequent analysis of the stress distribution of the optical fiber, assessment of the health status of the optical fiber, and planning of obstacle avoidance paths for fiber-feeding tools. The identified bend node information will be used for subsequent steps such as stress distribution calculation, identification of deformed topological structures, and correction of compensatory topological structures. In stress distribution calculation, bend nodes are usually areas of stress concentration, and their accurate position and parameters are crucial for accurate calculation of stress distribution. In deformed topological structure identification, the distribution and characteristics of bend nodes are important bases for judging whether the optical fiber has deformities such as entanglement and knotting.

[0096] Axial deviation reflects the degree of deviation between the actual position of the optical fiber and the ideal position. Identifying axial deviation helps to evaluate the installation and laying quality of the optical fiber, determine whether the optical fiber meets the use requirements, and provide an important reference for the maintenance and adjustment of the optical fiber. According to the standard parameters of the optical fiber type, a theoretical optical fiber axis is constructed. The theoretical optical fiber axis represents the position and direction of the optical fiber under ideal conditions. The topological nodes are fitted by the least squares method, and the data corresponding to the actual topological nodes are matched with the theoretical optical fiber axis to minimize the sum of the squares of the distances from the topological nodes to the theoretical optical fiber axis. The vertical distance from each topological node to the theoretical optical fiber axis is calculated. This vertical distance is the axial deviation of the topological node. By analyzing the axial deviation data of all topological nodes, the overall axial deviation of the optical fiber can be understood.

[0097] The application of the least squares method can accurately fit the topological node and the theoretical axis of the optical fiber, so that the calculated axial deviation can accurately reflect the difference between the actual position of the optical fiber and the ideal position. By identifying and analyzing the axial deviation, the offset problem of the optical fiber during installation or use can be discovered in time, providing data support for the calibration and adjustment of the optical fiber, ensuring the normal operation and stable performance of the optical fiber. The axial deviation will participate in the process of identifying the deformed topological structure and correcting the compensatory topological structure. In the identification of the deformed topological structure, a large axial deviation is a manifestation of deformities such as twisting and deformation of the optical fiber. When correcting the compensatory topological structure, the axial deviation can be used to adjust the position of the topological node, so that the corrected optical fiber topology is closer to the actual state of the optical fiber, thereby improving the accuracy and reliability of the optical fiber topology.

[0098] Understanding the stress distribution of the optical fiber topology structure helps to evaluate the stress state of the optical fiber, predict the failure and damage that may occur in the optical fiber, and provide an important mechanical basis for the maintenance, repair and path planning of the optical fiber, so as to avoid further damage to the optical fiber caused by the operation of the fiber-touching tool in the stress concentration area. Combining the obtained pressure distribution data, optical fiber material properties and topological nodes, the pressure distribution is converted into a node load through the finite element analysis method and applied to the corresponding topological nodes of the optical fiber topology structure. Then, the mechanical parameters of the optical fiber material are defined according to the optical fiber material properties. The conjugate gradient method is used to solve the finite element equation, and the stress magnitude and stress direction of each topological node in the optical fiber topology structure are calculated to obtain the stress distribution of the optical fiber. Through visualization technology, the stress distribution is displayed in the form of a cloud map, which intuitively presents the stress concentration area and stress change trend.

[0099] The finite element analysis method can accurately simulate the stress distribution of optical fibers under actual stress conditions, taking into account the influence of multiple factors, including pressure distribution, optical fiber material properties and topological structure. Through stress distribution identification, it can accurately locate stress concentration areas, discover potential failure risks of optical fibers in advance, and provide a scientific basis for preventive maintenance of optical fibers. At the same time, stress distribution is of great significance for optimizing the obstacle avoidance path of fiber-touching tools, avoiding operations in high-stress areas, and protecting the integrity of optical fibers. Stress distribution is an important basis for identifying deformed topological structures and generating compensation parameter sets. In the identification of deformed topological structures, stress concentration areas are often associated with abnormal topological morphologies. Combining stress distribution can more accurately determine whether the optical fiber is deformed. In the generation of compensation parameter sets, stress distribution is used to calculate pressure correction values and determine vibration suppression strength to adjust the stress state of the optical fiber and reduce the impact of vibration, ensuring that the optical fiber operates within a safe stress range.

[0100] Abnormal topology can lead to degraded optical fiber performance, signal transmission failures, and even damage. Timely identification of abnormal topology helps to take appropriate measures to repair or adjust it, ensuring the normal operation of the optical fiber. At the same time, it provides accurate environmental information for path planning of fiber-feeding tools to avoid dangerous operations of fiber-feeding tools in abnormal areas. Based on the topological isomorphism detection algorithm, the degree of difference between the optical fiber topology and the standard optical fiber topology in topological nodes and topological edges is calculated. This is specifically measured by calculating the graph edit distance. The graph edit distance represents the minimum number of node and edge edit operations required to convert one topology graph into another topology graph, where editing operations include insertion, deletion, and modification. At the same time, the relationship between the stress concentration area and the topological morphology in the topological structure is analyzed in combination with the stress distribution. If the graph edit distance is greater than the set number threshold and there is obvious stress concentration in the corresponding area, the optical fiber topology is determined to be abnormal, including entanglement, knotting, or crossing, and the type and location of the abnormality are recorded.

[0101] Combining the topological isomorphism detection algorithm with stress distribution analysis can more comprehensively and accurately identify deformed topological structures. Traditional detection methods that rely solely on topological structure morphology will miss some potential deformities caused by stress. Combining stress distribution can further verify the abnormalities of the topological structure from a mechanical perspective, improving the accuracy and reliability of deformity identification. Timely discovery of deformed topological structures can provide timely guidance for optical fiber maintenance and repair, avoiding failures caused by deformed structures.

[0102] Furthermore, the output sub-logic of the compensation topology structure includes:

[0103] Monitor environmental data, including temperature, humidity, air pressure, and external vibration. Normalize the environmental data into environmental factors. Combine these environmental factors with the fiber material properties to update the coordinates of each topological node, reestablish the connection between the topological nodes, and generate new topological edges.

[0104] The stress distribution, bending nodes, updated topological node coordinates, and new topological edges are fused through a spatiotemporal alignment algorithm to obtain enhanced properties. Reinforcement learning is then used to dynamically modify the fiber topology based on the enhanced properties.

[0105] The modified optical fiber topology is evaluated based on geometric accuracy, physical rationality and environmental adaptability to obtain the reliability of the optical fiber topology. Based on the reliability judgment of the optical fiber topology, a compensation topology is output.

[0106] In the real world, optical fibers are subject to changes in their physical form due to environmental factors such as temperature, humidity, and air pressure. For example, temperature changes cause the fiber to expand and contract, while external vibrations can cause slight displacements. If these environmental influences are not accounted for, the generated topology will deviate from the actual situation. Therefore, the fiber topology needs to be updated based on environmental data to better reflect the actual situation and provide a reliable foundation for subsequent precise adjustments. The positioning system uses integrated environmental sensors to collect real-time environmental data such as temperature, humidity, air pressure, and external vibration, and normalizes this data into environmental factors. The positioning system also includes a built-in database of the physical properties of different optical fiber materials, including quartz and plastic, under various environmental conditions. Based on the current environmental factors and optical fiber material properties, the system uses historical experimental data and machine learning methods to train the relationship between the environment and the optical fiber to calculate the theoretical displacement of each topological node due to environmental influences. When the temperature rises, the elongation of each optical fiber component is determined based on the thermal expansion characteristics of quartz optical fiber, and the coordinates of the topological nodes are then updated. After the coordinates are updated, the connections between the topological nodes are reassessed. If the coordinate position of a topological node changes, making the original connection no longer reasonable, a new topological edge is generated, completing the update of the optical fiber topology.

[0107] Through updates driven by environmental data, the interference of environmental factors on the accuracy of the fiber topology structure is effectively reduced, making the fiber topology structure closer to the actual form of the fiber from the beginning, reducing the workload and complexity of subsequent corrections, and improving the efficiency and accuracy of the entire topology construction process. The updated fiber topology structure provides a more realistic initial state for the dynamic correction of multi-source data fusion. Further corrections based on this can more accurately adjust the fiber topology structure, so that the final compensated topology structure better reflects the actual state of the fiber in the current environment.

[0108] Relying solely on environmental data for updates cannot fully consider the complex situations in the actual use of optical fibers, such as the actual pressure on the optical fiber and its own bending deformation. Therefore, it is necessary to integrate multi-source data such as stress distribution and bending nodes, and dynamically correct the optical fiber topology structure in combination with actual task requirements to ensure that it can accurately guide the operation of the fiber-feeding tool and avoid damaging the optical fiber; the stress distribution, bending nodes and updated topology structure are integrated through the time-space alignment algorithm to ensure that data from different sources correspond one-to-one in time and space dimensions, and the stress distribution at a specific time point is accurately mapped to the corresponding topological node to obtain enhanced attributes. The enhanced attributes of all topological nodes are summarized to form a complete enhanced attribute data set; the optical fiber containing enhanced attributes The topological structure is defined as the input of reinforcement learning, and the topological adjustment action is output through the policy network, that is, moving a node or changing the connection relationship between two edges. After executing the adjustment action, reward feedback is obtained based on the completion of the task, that is, whether other optical fibers are touched and damaged. Then the policy network is optimized and this process is repeated continuously, gradually making the topological structure more in line with actual operational requirements and physical constraints; multi-source data fusion and dynamic correction of reinforcement learning enable the optical fiber topology structure to fully adapt to various complex situations. It not only takes environmental factors into account, but also combines the actual force and morphological characteristics of the optical fiber, and optimizes it based on the task objectives. The generated topological structure is more accurate and reliable, and can effectively guide the fiber-touching tool to complete the operation task safely and efficiently.

[0109] Although the dynamically corrected topology is more realistic, it still needs to be evaluated for reliability to determine whether it meets the requirements of fiber optic operation tasks. Only a reliable fiber optic topology can be output as a compensating topology to provide an accurate basis for path planning and fiber-touching tool control. Otherwise, operation failure or fiber damage will occur. An evaluation index system for the reliability of the topology is constructed, and evaluation is performed from three dimensions: geometric accuracy, physical rationality, and environmental adaptability. In terms of geometric accuracy, the corrected fiber optic topology is compared with the standard topology to evaluate the accuracy of its shape and position, that is, to check whether the bending angle and axial deviation of the fiber are within a reasonable range. In terms of physical rationality, the fiber optic topology is judged to be in compliance with physical laws based on whether the stress distribution is within the stress range of the fiber material, so as to avoid stress concentration that causes fiber damage. In terms of environmental adaptability, changes in different environmental conditions are simulated, the stability of the fiber optic topology is observed, and its reliability in various environments is evaluated.

[0110] A dynamic weight is assigned to each evaluation indicator, and the weight is determined according to the current environmental conditions and operation tasks. When performing fiber optic docking tasks in a high-temperature environment, the weight of physical rationality will increase. The scores of various indicators are combined to calculate the reliability score of the fiber optic topology. If the score is greater than the set score threshold, the fiber optic topology is output as a compensation topology; otherwise, the dynamic correction step is returned to continue adjustment and optimization; through a comprehensive and systematic reliability evaluation, it is ensured that the output compensation topology can function stably and reliably in practical applications, avoiding the risks brought by the use of unreliable topologies, improving the success rate and safety of fiber optic operations, and ensuring the normal operation of the fiber. The output compensation topology is used as important input data to generate a compensation parameter set, providing specific parameter guidance for the operation of the fiber optic touch tool, and is passed to the path planning module as the basis for generating an obstacle avoidance path, so that path planning can be based on accurate fiber optic topology information, improving the accuracy and effectiveness of path planning.

[0111] Specifically, the compensation parameter set generation logic includes:

[0112] Extract the compensation topology, calculate the bending angle change of each bending node in the optical fiber topology and the compensation topology, and obtain the angle offset;

[0113] Determine pressure-sensitive areas based on stress distribution, calculate pressure correction values for these areas using Hooke's law, and configure pressure thresholds based on fiber type and fiber operation scenarios. Compare the pressure distribution on the fiber surface with the thresholds to adjust the update frequency of the pressure correction values.

[0114] Perform time-frequency analysis on the vibration spectrum to extract the vibration frequency components and vibration energy, and determine the vibration suppression intensity based on the vibration frequency components and vibration energy;

[0115] The angle offset, pressure correction value and vibration suppression strength are normalized, an identifier and a timestamp are added to each compensation parameter, and the processed compensation parameters are integrated to generate a compensation parameter set. The data tensor is incrementally updated based on the compensation parameter set to regenerate the feature vector.

[0116] The compensation topology adjusts the bending shape of the optical fiber. Compared with the optical fiber topology, the angle of the bending node has changed. In order for the fiber-touching tool to operate accurately according to the compensation topology, it is necessary to calculate the angle change of each bending node in the two topologies to obtain the angle offset, thereby providing precise angle adjustment parameters for the motion control of the fiber-touching tool; the compensation topology and the optical fiber topology are extracted, and the bending nodes therein are compared. For each bending node, its bending angle in the two topologies is calculated respectively, and the angle offset is obtained by the difference between the two bending angles; the obtained angle offset can more accurately reflect the change in the bending shape of the optical fiber, so that the fiber-touching tool can accurately adjust the angle during operation, reduce the risk of optical fiber damage caused by inaccurate angles, and improve the accuracy and safety of the operation. The calculated angle offset, together with parameters such as pressure correction value and vibration suppression strength, enters the subsequent normalization and integration process to finally form a compensation parameter set.

[0117] Pressure-sensitive areas identified by the stress distribution require special attention during fiber operation. The pressure exerted on these areas can significantly impact fiber performance and even cause damage. Therefore, pressure correction values must be calculated for these pressure-sensitive areas to adjust and control the pressure in these areas during operation, protecting the fiber from damage caused by excessive pressure. Based on the stress distribution of the fiber topology calculated previously, pressure-sensitive areas are determined. For these pressure-sensitive areas, pressure correction values are calculated using Hooke's law, combined with the elastic modulus and stress-strain relationship of the fiber material. Furthermore, corresponding pressure thresholds are configured based on the fiber type and specific fiber operation scenarios. The pressure distribution on the fiber surface is monitored in real time and compared with the pressure threshold. When the pressure approaches the threshold, the pressure correction value update frequency is increased to enable more timely pressure adjustments. When the pressure moves away from the threshold, the update frequency is appropriately reduced to reduce the computational effort. By calculating the pressure correction value and adjusting the update frequency based on the pressure situation, the pressure in the pressure-sensitive areas can be effectively controlled, protecting the fiber from damage caused by excessive pressure while avoiding unnecessary frequent adjustments. This improves operational stability and efficiency, ensures that the fiber remains within a safe pressure range during operation, and safeguards the performance and service life of the fiber.

[0118] Optical fibers are subject to various vibrations in real-world environments, which can affect their performance and the accuracy of fiber-patch tools. Therefore, it's necessary to determine the appropriate vibration suppression strength based on vibration spectrum analysis to minimize the adverse effects of vibration on the fiber, ensuring stable signal transmission and precise fiber-patch tool operation. Time-frequency analysis of the vibration spectrum extracts the primary frequency components and energy distribution. Based on this information, combined with the fiber's natural frequency and operational requirements, the vibration suppression strength is determined. If a frequency component in the vibration spectrum has high energy and is close to the fiber's natural frequency, the vibration suppression strength is increased to prevent resonance damage. Furthermore, given the varying vibration suppression requirements in different fiber-patch operation scenarios, such as fiber-patch docking, the positioning system automatically adjusts the vibration suppression strength based on the specific operation scenario. By determining the vibration suppression strength based on the vibration spectrum, the system can effectively suppress vibrations that are harmful to the fiber, effectively reducing vibration interference with signal transmission and improving the stability of the positioning system. This ensures the smooth operation of the fiber-patch tool, enhancing operational accuracy and reliability.

[0119] Compensation parameters such as angle offset, pressure correction value and vibration suppression strength each adjust and control the optical fiber operation from different aspects. Only by integrating these parameters to form a unified compensation parameter set can we provide comprehensive and systematic parameter guidance for the operation of the fiber-touching tool, ensuring that the fiber-touching tool can complete the task accurately and safely; the calculated angle offset, pressure correction value and vibration suppression strength are normalized so that they have the same dimension and value range, which is convenient for subsequent processing and use. An identifier and timestamp are added to each compensation parameter. The identifier is used to distinguish parameters of different optical fiber types, and the timestamp records the generation time of the parameter for traceability and management. The processed compensation parameters are integrated in a specific format and order to generate a compensation parameter set. Finally, the data tensor is incrementally updated based on the compensation parameter set, triggering the modal perception module to regenerate the feature vector, realizing closed-loop optimization of the positioning system, and at the same time outputting the compensation parameter set and passing it to the path planning module and other related modules.

[0120] The integrated compensation parameter set provides a complete parameter basis for the operation of the fiber-feeding tool, enabling the fiber-feeding tool to comprehensively consider various factors and perform precise operations. By updating the data tensor and regenerating the eigenvector, the self-optimization and adaptability of the positioning system are realized, and the performance and reliability of the entire fiber optic positioning system are improved. The output compensation parameter set provides key parameter support for the path planning module to generate obstacle avoidance paths, enabling the path planning to fully consider the various states and operating requirements of the optical fiber and generate a more reasonable and safe path. At the same time, it provides a data basis for subsequent system optimization and adjustment, which helps to continuously improve the performance and adaptability of the positioning system.

[0121] The path planning module is used to receive the compensation topology and compensation parameter set, generate an obstacle avoidance path through the path planning algorithm, convert the obstacle avoidance path into a control instruction sequence of the fiber-feeding tool, and monitor the trajectory error between the actual trajectory of the fiber-feeding tool and the obstacle avoidance path to determine whether to reconstruct the optical fiber topology.

[0122] Furthermore, if Figure 3 As shown, the generation logic of the obstacle avoidance path includes:

[0123] Convert the bending nodes, axial deviations, and stress distribution of the compensation topology into physical constraints, and build an environmental risk map based on environmental data, while setting task priorities based on task types.

[0124] Based on physical constraints, environmental risk maps, and task priorities, an initial path is generated on the compensation topology using a fast-exploring random tree algorithm.

[0125] The initial path is optimized based on a deep reinforcement learning network to output optimized local path segments. The optimized local path segments are smoothed using a spline curve fitting algorithm, and the speed of the smoothed local path segments is planned according to the motion performance parameters of the fiber-feeding tool to generate an obstacle avoidance path.

[0126] The compensation topology reflects the actual shape and physical state of the optical fiber. The bend nodes, axial deviations, and stress distribution areas within it restrict the movement of the fiber-feeding tool. Forcing the tool through can damage the fiber. Environmental factors also affect operational safety and fiber performance. In addition, different task types have different emphases on path planning, including troubleshooting and fiber docking. Therefore, this information needs to be converted into specific constraints and a visual environmental risk map, and task priorities are set to provide an accurate basis for path planning. Bending nodes, axial deviations, and stress distribution are extracted from the compensation topology. For bending nodes, path turning restrictions are set based on their bending angle and curvature radius. Sharp turns are prohibited in areas with excessive curvature. Areas with large axial deviations are designated as cautious passage zones, limiting the tool's movement speed, and areas with concentrated stress distribution are marked as no-pass zones. At the same time, combined with real-time environmental data, namely high-temperature areas detected by temperature sensors and high-vibration areas identified by vibration sensors, these areas are annotated on a three-dimensional spatial map to construct an environmental risk map.

[0127] According to the current operation task type, the task priority is set through the human-computer interaction interface or preset rules. Among them, the fiber optic docking task prioritizes the accuracy of the path, and the troubleshooting task prioritizes the coverage of the path. By converting the physical characteristics of the optical fiber and environmental factors into intuitive constraints and risk maps, and clarifying the task priority, the path planning can fully consider various actual limitations and needs, avoiding the blindness of path planning, improving the feasibility and safety of the obstacle avoidance path, ensuring that the fiber optic touch tool will not damage the optical fiber during operation, and meeting the requirements of different tasks.

[0128] After clarifying the physical constraints, environmental risk map, and task priorities, an algorithm is needed to quickly find a feasible path from the starting point to the target point in the complex fiber topology space. As the basis for subsequent optimization, the rapid exploration random tree algorithm has the ability to quickly search in high-dimensional space and is suitable for solving such problems. The compensation topology structure is abstracted into a three-dimensional spatial graph, where the topological nodes are the vertices of the graph and the topological edges are the edges of the graph. The starting position of the fiber tracing tool is used as the root node. The target node is set according to the task objective, and the rapid exploration random tree algorithm is run. In each iteration, the rapid exploration random tree algorithm randomly samples a point in space and then finds the node closest to the sampled point from the current tree. It attempts to grow a new edge to the sampled point while satisfying the physical constraints and avoiding the dangerous areas in the environmental risk map. The new node is added to the tree and this process is repeated until the generated tree contains the target node. At this time, the path from the root node to the target node is the initial path, where the physical constraints include not entering the prohibited area and the turning angle meets the limit.

[0129] During the generation process, the path search direction is guided according to task priority. For fiber optic docking tasks that prioritize accuracy, the tree tends to grow in the direction close to the target fiber. The rapid exploration random tree algorithm can efficiently generate a feasible initial path in a complex fiber topology environment, greatly shortening the path planning time. At the same time, combined with the guidance of physical constraints and task priority, the initial path is more in line with actual operational needs while meeting basic feasibility, providing a good starting point for subsequent path optimization.

[0130] The initial path is generated by random sampling, and there will be problems such as tortuous paths and abrupt turns, which will not only increase the movement time and energy consumption of the fiber-feeding tool, but also affect the operation accuracy, and may even be impossible to achieve in actual movement. Therefore, it is necessary to optimize and smooth the initial path, and set a reasonable speed according to the movement performance of the tool to generate an obstacle avoidance path that meets actual needs; the initial path is input into the deep reinforcement learning network, and the deep reinforcement learning network takes reducing the path length, reducing the proximity of the path to the dangerous area, and improving the path smoothness as the optimization goals. The deep reinforcement learning network takes path information and environmental status as input, where the path information includes the node sequence and the distance to the dangerous area. The environmental status includes a real-time environmental risk map. By continuously simulating and interacting with the environment, the optimal path adjustment strategy is learned according to the preset reward mechanism, and the optimized local path segment is output. The reward mechanism includes positive rewards for shortening the path and negative rewards for approaching the dangerous area. Then, the spline curve fitting algorithm is used to smooth the optimized local path segment, converting the original broken line path into a smooth curve and eliminating sharp turns. Finally, according to the maximum acceleration and maximum speed of the fiber-feeding tool and other motion performance parameters, the speed of the smoothed path is planned, and reasonable speeds are set at different stages of the path, that is, the speed is reduced at the turns and increased appropriately in the straight sections.

[0131] The optimization of the deep reinforcement learning network makes the path more reasonable and efficient, reducing unnecessary path length and contact with dangerous areas. Spline curve fitting and smoothing ensure the smoothness of the initial path, which conforms to the motion characteristics of the fiber-feeding tool and improves the stability and accuracy of the operation. Speed planning fully considers the motion performance of the tool, avoiding operational errors or equipment damage caused by unreasonable speed. The final generated obstacle avoidance path enables the fiber-feeding tool to complete the operation task safely and efficiently. The generated obstacle avoidance path will be converted into a control instruction sequence of the fiber-feeding tool for actual motion control. At the same time, during the movement of the fiber-feeding tool, the obstacle avoidance path will be used as a reference. By monitoring the error between the actual trajectory and the obstacle avoidance path, it is determined whether the optical fiber topology structure needs to be reconstructed to ensure the accuracy and reliability of the operation.

[0132] Specifically, the reconstruction judgment sub-logic of the optical fiber topology structure includes:

[0133] The obstacle avoidance path is converted into a control instruction sequence for the fiber-feeding tool. The spatial position error between the actual trajectory of the fiber-feeding tool and the obstacle avoidance path is calculated through the root mean square error to determine the position error. The actual posture of the fiber-feeding tool during the control process is monitored and compared with the posture preset in the obstacle avoidance path to determine the posture error.

[0134] Analyze the actual motion parameters of the fiber-sensing tool during the control process and compare them with the preset motion parameters of the obstacle avoidance path to determine the parameter error. The position error, attitude error, and parameter error are used as trajectory errors. Combined with environmental data and the reliability of the compensation topology structure, an evidence set is constructed.

[0135] Based on Bayesian reasoning, the probability of compensation topology structure changes under different evidence sets is learned according to historical data, and the posterior probability that the compensation topology structure needs to be reconstructed is calculated through Bayesian reasoning. Based on the posterior probability, it is determined whether the optical fiber topology structure should be reconstructed.

[0136] During the actual movement of the fiber-touching tool, due to the influence of various factors, its actual trajectory will deviate from the preset obstacle avoidance path. By calculating and analyzing the trajectory error, it is possible to determine whether the current optical fiber topology still accurately reflects the actual situation, providing a basis for whether the topology needs to be reconstructed; the obstacle avoidance path is converted into a series of precise control instruction sequences and sent to the motion control system of the fiber-touching tool. During the movement of the tool, the actual position of the fiber-touching tool is monitored in real time by a laser locator and compared with the preset position on the obstacle avoidance path. The position deviation between the two is calculated by the root mean square error to obtain the position error. At the same time, the inertial measurement unit is used to monitor the actual posture of the fiber-touching tool. , including pitch angle and yaw angle, are compared with the preset posture of the obstacle avoidance path to determine the posture error. In addition, the actual motion parameters of the fiber-feeding tool during movement, including speed and acceleration, are analyzed and compared with the motion parameters preset in the obstacle avoidance path to determine the parameter error. The position error, posture error and parameter error are integrated to comprehensively evaluate the trajectory error. Through multi-dimensional trajectory error calculation and analysis, the trajectory deviation of the fiber-feeding tool during movement can be discovered in a timely and accurate manner, which helps operators understand the accuracy of the current operation, determine whether there are potential problems, and provide important data support for subsequent decision-making, avoiding optical fiber operation failure or damage due to trajectory deviation.

[0137] Relying solely on trajectory error to determine whether to reconstruct the fiber topology is not accurate and comprehensive. It is necessary to combine current environmental data and the reliability assessment results of the compensating topology to comprehensively construct an evidence set. Learning the probability of topology changes under different evidence sets based on historical data allows for a more scientific judgment on whether the current topology needs to be reconstructed. By integrating various trajectory error indicators with real-time environmental data and the reliability scores of the compensating topology, an evidence set containing multiple aspects of information is constructed. The positioning system pre-collects and stores a large amount of historical operation data, including records of actual fiber topology changes under different evidence sets. Using this historical data, the probabilistic relationship between different evidence sets and changes in the compensating topology is learned through Bayesian inference. In other words, the prior probability of topology changes under the current situation can be calculated based on the input evidence set. By constructing a comprehensive evidence set and learning the probabilistic relationship based on historical data, the judgment of fiber topology reconstruction is made more scientific and reasonable. This fully considers multiple influencing factors, avoids the limitations of single-factor judgments, improves the accuracy and reliability of judgments, and can more promptly and accurately detect potential topology changes, providing strong support for subsequent decision-making.

[0138] On the basis of constructing the evidence set and learning the probability relationship, it is necessary to calculate the posterior probability that the compensatory topology structure needs to be reconstructed through Bayesian reasoning, quantify the probability into a specific decision-making basis, and thus determine whether the current optical fiber topology structure needs to be reconstructed to ensure that it is consistent with the actual situation and to ensure the accuracy and safety of the subsequent operation of the fiber-feeding tool; the constructed evidence set is processed by Bayesian reasoning, combined with the prior probability obtained by previous learning, and based on Bayesian reasoning, the influence of various factors in the evidence set on the topology structure change is comprehensively considered to calculate the posterior probability that the compensatory topology structure needs to be reconstructed, and a reasonable probability threshold is set. When the calculated posterior probability is greater than the probability threshold, it is determined that the current optical fiber topology structure is likely to have changed, and the optical fiber topology structure reconstruction process is triggered; if the posterior probability is less than or equal to the probability threshold, the topology structure is considered to be still reliable, and operations continue to be performed according to the current topology structure, and continuous monitoring of the trajectory error is maintained.

[0139] Through posterior probability calculation and clear probability threshold setting, an objective and quantitative decision-making standard is provided for the reconstruction of the optical fiber topology structure, avoiding the subjectivity and uncertainty of human judgment. It can make accurate and timely decisions on whether to reconstruct the topology structure based on actual conditions, ensuring that the optical fiber positioning system always operates based on accurate topology information, improving the reliability and stability of the positioning system. If it is determined that the optical fiber topology structure needs to be reconstructed, the optical fiber topology structure reconstruction process is started to update and optimize the topology structure; if reconstruction is not required, the current path planning and operation tasks are continued, and the trajectory error is continuously monitored to provide data support for subsequent judgments.

[0140] When it is determined that the optical fiber topology needs to be reconstructed, since the overall structure of the optical fiber will be large, comprehensive data re-acquisition is inefficient and unnecessary. Therefore, it is necessary to accurately locate the key areas where the topology changes will occur based on the trajectory error analysis results, and then perform data enhancement acquisition in these areas to obtain more detailed information and provide accurate data support for topology reconstruction; analyze the trajectory error data, combine environmental data and the historical change rules of the compensated topology structure, and determine the key areas where the topology changes will occur. If the position error of the tool continues to increase in a certain area, and there is environmental interference near the area, including high temperature or vibration sources, this area will be marked as a critical area, and the fiber optic tool will be controlled to perform data enhancement acquisition in the key area. The task is to improve the sampling frequency and sampling range of the modal perception module. The vibration signal, which was originally sampled once per second, is increased to five times per second in the key area. At the same time, the detection range of the pressure sensor and the infrared sensor is expanded to obtain more comprehensive physical signals, including vibration signals, pressure signals and infrared signals, to supplement the data tensor information of the key area. By accurately locating the key area and performing data enhancement acquisition, unnecessary data collection of the entire optical fiber structure is avoided, which greatly improves the efficiency of data acquisition. At the same time, the more detailed data obtained can more accurately reflect the actual situation of the key area, providing a rich and accurate data basis for the reconstruction of the topology structure, which helps to improve the accuracy and reliability of the reconstructed optical fiber topology structure.

[0141] Based on the new data acquired from the key areas, the fiber topology needs to be locally updated to reflect the actual changes in that area. To ensure the consistency and rationality of the entire fiber topology, a global optimization is required to ensure that the updated fiber topology better adapts to the actual situation and meets the operational requirements of the fiber-feeding tool. The data acquired from the key areas is input into the modal perception module, which regenerates the feature vector of the area and passes it to the topology construction module. Based on the new feature vector, the topology construction module locally updates the topology of the key areas, including the location and connection relationships of the topological nodes, as well as node attributes such as bend angles, curvature radii, and stress values. After the local update is completed, the entire fiber topology is globally optimized, integrating the locally updated fiber topology with the original fiber topology. Graph theory algorithms are used to detect node conflicts or edge inconsistencies, including node duplication and unreasonable connections. If conflicts are found, they are corrected. Then, based on the current environmental data and operational task requirements, the compensation topology output and compensation parameter set generation processes are executed again. The entire fiber topology is comprehensively evaluated and optimized, and parameters such as angle offset, pressure correction value, and vibration suppression strength are updated to make the fiber topology more accurate and reasonable.

[0142] The combination of local update and global optimization not only ensures timely and accurate correction of the changed areas of the optical fiber topology, but also ensures the integrity and consistency of the entire optical fiber topology. By re-executing the compensation topology output and compensation parameter set generation process, the optical fiber topology can better adapt to environmental changes and operational task requirements, improve the reliability and practicality of the optical fiber topology, and provide more accurate guidance for the operation of the fiber-feeding tool. The optimized optical fiber topology will be passed to the path planning module as the new basic data. The path planning module regenerates the obstacle avoidance path and control instruction sequence according to the new optical fiber topology, so that the fiber-feeding tool can operate based on the updated topology information, ensuring the accuracy and safety of the operation.

[0143] Example 2

[0144] like Figure 4 As shown, a flow chart of a method for positioning and identifying an optical fiber using a fiber-touching tool is provided for an embodiment of the present application. The method includes:

[0145] Monitor the physical signals on the optical fiber surface and extract the characteristics of the physical signals to generate data tensors including vibration spectrum, pressure distribution and infrared reflectivity;

[0146] Perform feature mapping on data tensors based on deep neural networks to identify fiber types and generate feature vectors;

[0147] Performing spatiotemporal correlation analysis on the feature vectors to generate the fiber topology and simultaneously identify the bending nodes, axial deviation and stress distribution of the fiber topology;

[0148] Monitoring environmental data to correct the optical fiber topology, outputting a compensation topology, and generating a compensation parameter set including an angle offset, a pressure correction value, and a vibration suppression strength according to the compensation topology;

[0149] Incrementally updating the data tensor based on the compensation parameter set to regenerate the feature vector;

[0150] Receive the compensation topology and compensation parameter set, generate an obstacle avoidance path through a path planning algorithm, and convert the obstacle avoidance path into a control instruction sequence for the fiber optic tool;

[0151] Monitor the trajectory error between the actual trajectory of the fiber-touching tool and the obstacle avoidance path to determine whether to reconstruct the optical fiber topology.

[0152] Since the principle of solving the problem by the method in the embodiment of the present application is similar to that of the system described above in the embodiment of the present application, the implementation of the method refers to the implementation of the system, and the repeated parts will not be repeated.

Claims

1. The fiber optic positioning system of the fiber optic tool is characterized by: include: Monitor the physical signals on the optical fiber surface, extract the features of the physical signals to generate data tensors including vibration spectrum, pressure distribution, and infrared reflectivity, perform feature mapping on the data tensors based on deep neural networks, identify the optical fiber type, and generate feature vectors; Performing spatiotemporal correlation analysis on the eigenvectors to generate a fiber topology, identifying bending nodes, axial deviations, and stress distributions of the fiber topology, monitoring environmental data to correct the fiber topology, outputting a compensation topology, generating a compensation parameter set including an angle offset, a pressure correction value, and a vibration suppression strength based on the compensation topology, and incrementally updating a data tensor based on the compensation parameter set to regenerate the eigenvectors; The compensation topology and compensation parameter set are received, and an obstacle avoidance path is generated through a path planning algorithm. The obstacle avoidance path is converted into a control instruction sequence of the fiber-feeding tool, and the trajectory error between the actual trajectory of the fiber-feeding tool and the obstacle avoidance path is monitored to determine whether the optical fiber topology should be reconstructed.

2. The optical fiber positioning system of the fiber-touching tool according to claim 1, characterized in that: The generation logic of the feature vector includes: Based on a dual-branch convolutional neural network, the spatial features and dependency features of the data tensor are extracted, and the spatial features and dependency features are spliced and fused to obtain a multi-scale feature map; The dual-channel attention mechanism is used to weight the multi-scale feature map from the channel dimension and spatial dimension respectively. The weighted multi-scale feature map is input into the dynamic threshold-based classification network, and the classification score is output. The configured threshold parameter is compared with the classification score to identify the fiber type. The fiber type is integrated with the spatial position and confidence of the fiber to generate a feature vector.

3. The optical fiber positioning system of the fiber-touching tool according to claim 2, characterized in that: The data tensor generation sub-logic includes: Monitor physical signals on the optical fiber surface, including vibration signals, pressure signals, and infrared signals; The vibration signal is decomposed using the empirical mode decomposition algorithm, and the vibration spectrum of the vibration signal is extracted using the Hilbert-Huang transform. At the same time, the pressure signal is processed based on the morphological algorithm and the pressure area boundary is identified to extract the pressure distribution of the pressure signal. The infrared signal is normalized by piecewise linear interpolation, and the infrared reflectivity of the infrared signal is extracted by principal component analysis. The vibration spectrum, pressure distribution and infrared reflectivity are fused according to timestamp and spatial position to generate a data tensor.

4. The optical fiber positioning system of the fiber-touching tool according to claim 3, characterized in that: The generation logic of the compensation parameter set includes: Extract the compensation topology, calculate the bending angle change of each bending node in the optical fiber topology and the compensation topology, and obtain the angle offset; Determine pressure-sensitive areas based on stress distribution, calculate pressure correction values for these areas using Hooke's law, and configure pressure thresholds based on fiber type and fiber operation scenarios. Compare the pressure distribution on the fiber surface with the thresholds to adjust the update frequency of the pressure correction values. Perform time-frequency analysis on the vibration spectrum to extract the vibration frequency components and vibration energy, and determine the vibration suppression intensity based on the vibration frequency components and vibration energy; The angle offset, pressure correction value and vibration suppression strength are normalized, an identifier and a timestamp are added to each compensation parameter, and the processed compensation parameters are integrated to generate a compensation parameter set.

5. The optical fiber positioning system of the fiber-touching tool according to claim 4, characterized in that: The generation sub-logic of the optical fiber topology structure includes: The spatial position and timestamp of the feature vector are extracted and combined into a four-dimensional index. The four-dimensional index is used as the data point, and a spatiotemporal R-tree index is constructed based on the data point. The time window is dynamically set according to the fiber operation scenario, and the data points within each time window are filtered through the spatiotemporal R-tree index. The selected data points are processed by spatiotemporal density clustering algorithm to perform spatiotemporal correlation analysis on the feature vectors and obtain spatiotemporal clustering results; Based on the spatiotemporal clustering results, each cluster center is extracted as a topological node. The connection relationship of the topological nodes is established through the triangulation algorithm to form topological edges to generate the optical fiber topology structure. A multidimensional attribute vector including bending angle, curvature radius and stress value is attached to each topological node.

6. The optical fiber positioning system of the fiber-touching tool according to claim 5, characterized in that: The identification sub-logic of the optical fiber topology structure includes: Calculate the space vector angle between adjacent topological nodes in the optical fiber topology structure, dynamically set the angle threshold according to the optical fiber type, compare the space vector angle with the angle threshold to determine the candidate bending node, and detect the curvature continuity of the candidate bending node to identify the bending node of the optical fiber topology structure; The theoretical axis of the optical fiber is constructed based on the standard parameters of the optical fiber type. The topological nodes are fitted using the least squares method, and the vertical distance from each topological node to the theoretical axis is calculated to identify the axial deviation of the optical fiber topology structure. Combining pressure distribution, fiber material properties and topological nodes, the pressure distribution is converted into node loads through finite element analysis, and the stress distribution of the fiber topological structure is calculated using the conjugate gradient method. Based on the topological isomorphism detection algorithm, the graph edit distance is calculated to measure the difference between the fiber topology and the standard fiber topology in topological nodes and topological edges. The difference degree and stress distribution are combined to identify the deformed topology.

7. The optical fiber positioning system of the fiber-touching tool according to claim 6, characterized in that: The output sub-logic of the compensation topology structure includes: Monitor environmental data, including temperature, humidity, air pressure, and external vibration. Normalize the environmental data into environmental factors. Combine these environmental factors with the fiber material properties to update the coordinates of each topological node, reestablish the connection between the topological nodes, and generate new topological edges. The stress distribution, bending nodes, updated topological node coordinates, and new topological edges are fused through a spatiotemporal alignment algorithm to obtain enhanced properties. Reinforcement learning is then used to dynamically modify the fiber topology based on the enhanced properties. The modified optical fiber topology is evaluated based on geometric accuracy, physical rationality and environmental adaptability to obtain the reliability of the optical fiber topology. Based on the reliability judgment of the optical fiber topology, a compensation topology is output.

8. The optical fiber positioning system of the fiber-touching tool according to claim 7, characterized in that: The reconstruction judgment sub-logic of the optical fiber topology structure includes: The obstacle avoidance path is converted into a control instruction sequence for the fiber-feeding tool. The spatial position error between the actual trajectory of the fiber-feeding tool and the obstacle avoidance path is calculated through the root mean square error to determine the position error. The actual posture of the fiber-feeding tool during the control process is monitored and compared with the posture preset in the obstacle avoidance path to determine the posture error. Analyze the actual motion parameters of the fiber-sensing tool during the control process and compare them with the preset motion parameters of the obstacle avoidance path to determine the parameter error. The position error, attitude error, and parameter error are used as trajectory errors. Combined with environmental data and the reliability of the compensation topology structure, an evidence set is constructed. Based on Bayesian reasoning, the probability of compensation topology structure changes under different evidence sets is learned according to historical data, and the posterior probability that the compensation topology structure needs to be reconstructed is calculated through Bayesian reasoning. Based on the posterior probability, it is determined whether the optical fiber topology structure should be reconstructed.

9. The optical fiber positioning system of the fiber-touching tool according to claim 8, characterized in that: The generation sub-logic of the obstacle avoidance path includes: Convert the bending nodes, axial deviations, and stress distribution of the compensation topology into physical constraints, and build an environmental risk map based on environmental data, while setting task priorities based on task types. Based on physical constraints, environmental risk maps, and task priorities, an initial path is generated on the compensation topology using a fast-exploring random tree algorithm. The initial path is optimized based on a deep reinforcement learning network to output optimized local path segments. The optimized local path segments are smoothed using a spline curve fitting algorithm, and the speed of the smoothed local path segments is planned according to the motion performance parameters of the fiber-feeding tool to generate an obstacle avoidance path.

10. A method for positioning and identifying an optical fiber of a fiber-touching tool, implemented based on the optical fiber positioning system of a fiber-touching tool according to any one of claims 1 to 9, characterized in that: include: Monitor the physical signals on the optical fiber surface and extract the characteristics of the physical signals to generate data tensors including vibration spectrum, pressure distribution and infrared reflectivity; Perform feature mapping on data tensors based on deep neural networks to identify fiber types and generate feature vectors; Performing spatiotemporal correlation analysis on the feature vectors to generate the fiber topology and simultaneously identify the bending nodes, axial deviation and stress distribution of the fiber topology; Monitoring environmental data to correct the optical fiber topology, outputting a compensation topology, and generating a compensation parameter set including an angle offset, a pressure correction value, and a vibration suppression strength according to the compensation topology; Incrementally updating the data tensor based on the compensation parameter set to regenerate the feature vector; Receive the compensation topology and compensation parameter set, generate an obstacle avoidance path through a path planning algorithm, and convert the obstacle avoidance path into a control instruction sequence for the fiber optic tool; Monitor the trajectory error between the actual trajectory of the fiber-touching tool and the obstacle avoidance path to determine whether to reconstruct the optical fiber topology.

Images