A method and system for cable tunnel route planning

Abstract

This invention provides a method and system for cable tunnel route planning. The method includes: acquiring and integrating geological survey data, existing pipeline network data, structured specifications, and unstructured documents of the planning area; parsing the unstructured documents using a visual language model to extract textual semantic constraints and associate them with spatial entities to generate a semantic constraint layer; integrating the semantic constraint layer, geological data, and pipeline data to construct a unified three-dimensional mesh constraint field through quantitative encoding; based on this constraint field, employing a hierarchical decision-making strategy, first performing a global path search that integrates multi-dimensional states, and then using a local optimization algorithm based on physical motion constraints to smooth and verify the feasibility of the path, generating a continuous and constructable three-dimensional path; finally, outputting an interpretable report of the path and all associated constraints. This invention realizes automated and intelligent planning of cable tunnel routes, moving from reliance on manual experience to data and model-driven approaches.

Description

Technical Field

[0001] This invention relates to the field of urban underground space planning and power facility design technology, and in particular to a method and system for cable tunnel route planning. Background Technology

[0002] Cable tunnel route planning is a crucial aspect of urban underground engineering construction. Its core is determining a cable laying route that simultaneously meets the requirements of safety, economy, compliance, and constructability within the complex three-dimensional underground space. Currently, as this field evolves towards intelligent systems, it still faces several pressing technical bottlenecks that need to be addressed.

[0003] Current technologies primarily rely on Geographic Information Systems (GIS) and Building Information Modeling (BIM) for spatial analysis, and then apply pathfinding algorithms for route searching. However, this approach has inherent limitations: First, the input data upon which planning depends suffers from "information silos." Data from multiple sources, such as geological exploration reports, existing pipeline as-built drawings, and design code clauses, exhibits heterogeneous formats. In particular, many key constraints exist in unstructured text and drawings, making it difficult for the system to automatically understand semantically rich descriptions such as "must avoid cultural relic burial areas" and "net distance from high-pressure gas pipelines not less than three meters," hindering the direct conversion of this crucial information into calculable planning parameters. Second, the constraint modeling approach is crude. Existing methods typically simplify complex engineering constraints into binary barriers or single-cost grids, failing to accurately express the strength, priority, and logical relationships between constraints, such as "in soft soil sections, additional support is required if the open-cut method is adopted," making it difficult for planning algorithms to make precise trade-offs. Finally, the interpretability of the planning process is insufficient. The path generated by the algorithm is often a "black box" result, lacking a direct connection with the original specifications and geological conditions. When the plan needs to be demonstrated and reviewed, it is difficult to quickly provide convincing technical evidence, resulting in repeated modifications to the planning scheme and low efficiency.

[0004] Existing technologies have failed to automatically extract semantic constraints from multimodal data, construct refined and computable models that fuse semantics, and drive the generation of interpretable and verifiable optimization path problems. Therefore, there is an urgent need for a cable tunnel path planning method and system that can deeply integrate multi-dimensional constraints, achieve multi-objective dynamic intelligent optimization, and support forward-looking simulation verification of solutions. Summary of the Invention

[0005] To address the shortcomings of the existing technologies, this invention proposes a cable tunnel route planning method and system.

[0006] The present invention specifically provides the following technical solution: A method for cable tunnel route planning, comprising the following steps: S1. Acquire and integrate multi-source constraint data for the planning area. The multi-source constraint data shall include at least geological and survey data of underground space structures, existing pipeline network topology data, structured design code provisions, and unstructured engineering documents. S2. Based on the MinerU2.5 visual language model, unstructured engineering documents are parsed to extract textual semantic constraint information about regional taboo conditions, safety distance requirements and construction technology requirements. The textual semantic constraint information is then geographically registered and spatially indexed with spatial entities in geological and survey data and existing pipeline network data to generate a semantic constraint layer associated with spatial index. S3. Construct a computable three-dimensional constraint field that integrates semantic information, including spatially overlaying and uniformly quantizing and encoding geological and survey data, existing pipeline network topology data, and semantic constraint layers associated with spatial indexes to generate a unified three-dimensional grid constraint field data structure covering the planning area. Each three-dimensional grid cell corresponds to a multi-dimensional state vector, which is used to characterize the comprehensive state of the cell in terms of obstacle avoidance level, construction cost coefficient, geological risk level, and standard compliance score. S4. A hierarchical decision-making strategy is used to generate a cable laying path sequence, including: Global path search phase: Based on a three-dimensional mesh constraint field, a weighted cost path search algorithm that integrates multi-dimensional state vectors is used to generate a global coarse path sequence consisting of multiple key waypoints between the preset path start and end points. Local path optimization stage: Guided by the global coarse path sequence, in the three-dimensional mesh constraint field, a local trajectory optimization algorithm based on the physical motion constraint model is used to smooth the global coarse path sequence, verify its construction feasibility, and make precise obstacle avoidance adjustments to generate a continuous and constructable three-dimensional spatial path curve. S5. Output planning results and semantic basis, including generating the final three-dimensional path of the cable tunnel, and simultaneously generating a structured interpretable report. The report automatically marks the specific engineering specifications, geological conditions and construction constraints that are satisfied or avoided in key sections of the three-dimensional path by associating the multi-dimensional state vectors of relevant grid cells in the three-dimensional grid constraint field with the extracted text semantic constraint information.

[0007] Optionally, in S1, multi-source constraint data of the planning area is acquired and integrated, specifically including: Data Acquisition: Obtain raw data for the planning area. The raw data must include at least: Drilling data, geological profiles, and geophysical exploration data obtained through geological surveys and geophysical exploration; existing pipeline as-built drawings, attribute tables, and spatial coordinate data retrieved from municipal pipeline management departments; mandatory clauses and recommended parameters related to cable tunnel design read from the structured specification knowledge base; and scanned copies of paper design drawings, construction log texts, and environmental assessment report PDF documents as unstructured engineering documents. S12. Data Standardization and Spatial Registration: The spatial information in the original data is uniformly converted to the same geographic coordinate system, and the attribute data is standardized and encoded to form a basic data layer with a unified spatial reference and format.

[0008] Optionally, in S2, unstructured engineering documents are parsed based on the MinerU2.5 visual language model, specifically including: S21. Convert the scanned copies of paper design drawings, construction log text, and environmental assessment report PDF document into image and text formats that can be processed by the MinerU2.5 model, respectively, while preserving the document layout and structure information; S22. Input the preprocessed document into the MinerU2.5 model. The model uses its visual encoder to parse the graphic symbols, annotations, and legends in the drawings, and its text encoder to understand the narrative text in the construction log and environmental assessment report. Combining visual and text features, the model identifies and structures the key semantic constraint entities. The key semantic constraint entities include area prohibition entities, safety distance entities, and construction process entities. Area prohibition entities include "cultural relic protection area", "chemical plant protection distance", "existing important pipeline protection area" and their corresponding boundary descriptions. Safety distance entities include "parallel clearance between cables and heating pipelines", "minimum vertical distance for cable crossing", "minimum burial depth for cable laying" and their corresponding values. Construction process entities include "sections requiring pipe jacking", "areas where open excavation is prohibited", and "recommended locations for cable joint wells".

[0009] Optionally, in S2, the text semantic constraint information is geographically registered and spatially indexed with spatial entities, specifically including: S23. For semantic constraint entities with clearly defined locations parsed from design drawings, use the known control point coordinates or scale information in the drawings to map them to the geographic coordinate system and directly use them as spatial elements in the semantic constraint layer. S24. For semantically constrained entities whose location information is described in words and parsed from construction logs or environmental assessment reports, geocoding technology is used in conjunction with geological and survey data and the distribution of ground features in existing pipeline network data to perform spatial relationship reasoning, determine the spatial range or associated spatial entities corresponding to the semantically constrained entities, and associate the semantically constrained entities with the determined corresponding locations. S25. Organize and manage the spatial elements corresponding to all mapped or associated semantic constraint entities using spatial database indexing technology to construct a semantic constraint layer. The semantic constraint layer supports fast querying and retrieval based on spatial location.

[0010] Optionally, in S3, a computable three-dimensional constraint field incorporating semantic information is constructed, specifically including: S31. Based on the three-dimensional boundary of the planning area and the undulation of the stratigraphic interface in the geological and survey data, a non-uniform three-dimensional grid is adaptively generated, wherein the grid cell size is dynamically adjusted according to the stratigraphic complexity and the density of existing pipelines; based on the geotechnical engineering geological parameters, the geological stability index of each grid cell is calculated. The geological stability index is obtained by a weighted entropy model that integrates the stratigraphic dip angle, soil cohesion and groundwater level, and is used to quantify the construction disturbance risk. S32. For each grid cell, perform coupled quantization of multi-source constraints: Based on existing pipeline network topology data, calculate the three-dimensional distance from the cell to the nearest pipeline, and generate a dynamic obstacle avoidance potential field according to the pipeline type, including high-voltage cables, water supply trunk pipes, and corresponding safety protection specifications; query the semantic constraint layer associated with the spatial index, and if the cell falls into a "no open-cut zone" or "cultural relic protection zone", set the construction permit flag to a prohibited state; if it is adjacent to a "safe distance entity", calculate the constraint compliance deviation value; encode the geological stability index, dynamic obstacle avoidance potential field value, constraint compliance deviation value, and construction cost coefficient based on cell burial depth and stratum properties into a five-dimensional state vector; S33. Associate the spatial coordinates of all grid cells with the corresponding five-dimensional state vectors to construct a sparse voxelized three-dimensional constraint field data structure; calculate the gradient field of the constraint field in three-dimensional space to indicate the direction of the fastest decrease in comprehensive cost during path search.

[0011] Optionally, in S4, during the global path search phase, a weighted cost path search algorithm incorporating multi-dimensional state vectors is used. A phased cost evaluation strategy is employed to search for a coarse global path sequence within the three-dimensional constraint field. The specific process includes: S41. Starting from the origin, expand the nodes to candidate mesh cells in the 3D spatial neighborhood according to the 3D constraint field data structure; for each candidate mesh cell, calculate the first evaluation cost. The formula is:

[0012] Where D is the three-dimensional Euclidean distance from the previous node to the center point of the candidate grid cell, and R is the geological stability index of the candidate grid cell. To estimate the foundation construction cost based on unit stratigraphic properties and burial depth, , , The preset weighting coefficients; select The node with the smallest value is expanded until the destination is reached, thus generating the initial path; S42. Based on the initial path, perform secondary optimization by combining the semantic constraint layer and gradient field; calculate the second evaluation cost for each grid cell on the path. The formula is:

[0013] in, The gradient of the dynamic obstacle avoidance potential field at the location of the grid cell is θ, where θ is the direction of the path segment. The angle between the directions is used to guide the path along the downward direction of the potential field to avoid obstacles; It is a function of the constraint compliance deviation value of the unit, used to penalize path segments that violate semantic safety distance; This is a Boolean penalty term; if the unit construction permission flag is in a prohibited state, then... Take a maximum positive value, otherwise set to 0; iterate through the path nodes to minimize the overall path. The value generates the final global coarse path sequence.

[0014] Optionally, in the local path optimization stage of S4, the specific execution process of the local trajectory optimization algorithm based on the physical motion constraint model includes: S43. Initial deformation guided by gradient: Taking each key waypoint in the global coarse path sequence as the center, in the neighborhood of each key waypoint, according to the gradient field direction pre-calculated in the three-dimensional constraint field data structure, local sampling is performed along the downward direction of the dynamic obstacle avoidance potential field to generate a series of candidate fine-tuning path segments that satisfy the minimum bending radius constraint of the cable. S44. Establish a mechanically feasible elastic spline optimization model: Using the global coarse path sequence and candidate fine-tuning path segments as initial control points, generate a parameterized 3D B-spline curve; establish a nonlinear optimization problem with the control point coordinates of the 3D B-spline curve as the optimization variables, and the objective function of the nonlinear optimization problem is... Defined as:

[0015] Among them, the integral variable Let be the arc length parameter along the 3D B-spline curve. This represents the total length of the three-dimensional B-spline curve; For a 3D B-spline curve in arc length The curvature at the point, the integral term of the square of the curvature is used to ensure the smoothness of the path and meet the physical constraint of the minimum bending radius of the cable; Points on a 3D B-spline curve The dynamic obstacle avoidance potential field gradient at the location; For a 3D B-spline curve in arc length The unit tangent vector at point , the dot product term Used to align the path direction with the downward direction of the dynamic obstacle avoidance potential field to enhance obstacle avoidance performance; The weighting coefficient is used to guide the potential field. As a constraint penalty term, when the 3D B-spline curve passes through an area marked as "no open excavation" in the semantic constraint layer or passes through a cell in the 3D mesh constraint field where the construction permit flag is in a prohibited state, the constraint penalty term takes a preset maximum positive value. S45. Iterative Solution and Feasibility Verification: The nonlinear optimization problem is solved using a sequential quadratic programming algorithm, iteratively updating the coordinates of the control points of the 3D B-spline curve. After each iteration, it is verified whether the curvature of each sampling point on the updated path curve is less than the maximum allowable curvature calculated from the minimum bending radius constraint of the cable, and whether the updated path curve avoids all cells where the construction permit flag is in a prohibited state. Finally, a continuous and constructable 3D spatial path curve that satisfies all constraints is output.

[0016] Optionally, constraint penalty items The specific calculation method is as follows:

[0017] in, To uniformly sample the 3D B-spline curve during the optimization process. Coordinates of a spatial point, This represents the total number of sampling points; For the indicator function, if the sampling point If a cell falls into any construction permit flag that is in a prohibited state or falls into an area in the semantic constraint layer where "open excavation is prohibited", the indicator function value is 1; otherwise, it is 0. Indicates the first The center coordinates of a known obstacle, which is defined by existing pipeline network topology data or characterized by high obstacle avoidance level cells in a three-dimensional mesh constraint field; Sampling points The distance to the nearest obstacle in the known set of obstacles; The minimum safe distance defined by the "safe distance entity" in the semantic constraint layer and required by the obstacle type; The penalty coefficient for entering the prohibited area is set to a value that is much larger than other positive numbers to ensure that the path strictly avoids the prohibited area; To mitigate the penalty for violating the safety distance, a quadratic penalty term is used to allow for smooth path adjustments near the safety distance boundary.

[0018] Optionally, the method also includes a network-wide evaluation and scheduling step S6 following S5: S61. Perform a three-dimensional spatial topology analysis on the cable tunnel three-dimensional path generated in S5 and the existing cable tunnel path extracted from the existing pipeline network topology data to identify the connection points, crossing points and parallel segments between the new path and the existing network. S62. At connection points and intersections, based on the requirements of the structured design code regarding the cross-sectional dimensions, layered layout, and safety distances of the utility tunnel, verify the feasibility of accommodating the new cables in the existing utility tunnel cross-section; if there is a space conflict or insufficient capacity, generate an early warning in the structured interpretability report and identify the specific three-dimensional location of the conflict and the violated code clauses. S63. When multiple cable routes need to be planned in the planning area, execute S1 to S5 sequentially or in parallel, and perform overall coupling analysis on all generated routes; based on the construction period, equipment resources and dynamic construction cost coefficients in the three-dimensional mesh constraint field, simulate the impact of the construction sequence of different routes on the overall cost and construction period, and output recommended construction sequence and resource scheduling scheme.

[0019] The present invention also provides a cable tunnel route planning system, the system comprising: The data fusion module is used to acquire and fuse multi-source constraint data of the planning area to form a fused data foundation for path planning; The semantic parsing and association module communicates with the data fusion module and is used to parse unstructured engineering documents based on the MinerU2.5 visual language model to extract text semantic constraint information, and associate the text semantic constraint information with geological and pipeline spatial entities to generate a semantic constraint layer. The 3D constraint field construction module communicates with the semantic parsing and association module to integrate geological data, pipeline data and semantic constraint layers, and constructs a unified 3D mesh constraint field through quantization encoding. The hierarchical path planning module communicates with the three-dimensional constraint field construction module and is used to generate continuous and constructable three-dimensional spatial path curves based on the three-dimensional mesh constraint field through a hierarchical decision-making strategy that combines global search and local optimization. The results generation and reporting module communicates with the hierarchical path planning module to generate a structured and interpretable report of the final 3D path and associated 3D mesh constraint field and textual semantic constraint information.

[0020] This invention offers the following beneficial technical effects: It provides a method and system for cable tunnel route planning. This invention utilizes a visual language model to automatically extract engineering semantics, such as prohibited areas and safety distances, from unstructured documents and accurately correlates them with spatial data such as geological models and existing pipelines. This constructs a machine-understandable and computable "semantic constraint field," solving the problem of scattered constraint information and difficulty in quantifying its application. Within a unified three-dimensional semantic constraint field, this invention employs a hierarchical optimization strategy, combining global semantic-guided search with local physical feasibility optimization, to optimize the path. This ensures that the output path automatically and synchronously meets multiple complex constraints, including safety regulations, geological risks, construction costs, and engineering feasibility, outputting a comprehensive optimal solution. The system not only outputs the three-dimensional path but also automatically generates an interpretable report linking all constraint data, making design decisions verifiable and improving review efficiency and result credibility. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart of a cable tunnel route planning method provided for an embodiment of the present invention.

[0023] Figure 2 This is a schematic diagram illustrating the parsing of unstructured documents and the generation of semantic constraint layers based on a visual language model, as provided in an embodiment of the present invention.

[0024] Figure 3 This is a schematic diagram illustrating the construction of a three-dimensional mesh constraint field that integrates semantic information, as provided in an embodiment of the present invention.

[0025] Figure 4a This is a schematic diagram illustrating the generation of a coarse path sequence through global path search, as provided in an embodiment of the present invention.

[0026] Figure 4b This is a schematic diagram illustrating the generation of smooth, constructible B-spline curves through local path optimization, as provided in an embodiment of the present invention.

[0027] Figure 5 A schematic diagram of the interface for the planning results and structured interpretability report provided for embodiments of the present invention. Detailed Implementation

[0028] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0029] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0030] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and the structure and / or function of any markings described herein are merely illustrative. Based on this application, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number and aspects set forth herein can be used to implement the apparatus and / or practical methods.

[0031] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this application.

[0032] Additionally, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that practice can be carried out without these marked details.

[0033] The purpose of this invention is to provide a cable tunnel route planning method and system, which aims to construct a computable semantic constraint field by fusing visual language models with multi-source engineering data, and to adopt a hierarchical intelligent decision-making technical solution to realize the transformation from traditional planning that relies on human experience to automated, multi-objective optimization intelligent planning, thereby improving design efficiency, scheme reliability and result interpretability.

[0034] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0035] In this invention, all the formula calculations are presented after parameter normalization and dimension removal.

[0036] Reference Figure 1 This paper illustrates a cable tunnel route planning method according to an embodiment of the present application. The method includes the following steps: S1. Acquire and integrate multi-source constraint data for the planning area. The multi-source constraint data shall include at least geological and survey data of underground space structures, topology data of existing pipeline networks, structured design code provisions, and unstructured engineering documents.

[0037] In S1, multi-source constraint data of the planning area is acquired and integrated, specifically including: S11. Data Acquisition: Obtain raw data for the planning area. The raw data shall include at least the following: Drilling data, geological profiles, and geophysical exploration data obtained through geological surveys and geophysical exploration; existing pipeline as-built drawings, attribute tables, and spatial coordinate data retrieved from municipal pipeline management departments; mandatory clauses and recommended parameters related to cable tunnel design read from the structured specification knowledge base; and scanned copies of paper design drawings, construction log texts, and environmental assessment report PDF documents as unstructured engineering documents. S12. Data Standardization and Spatial Registration: The spatial information in the original data is uniformly converted to the same geographic coordinate system, and the attribute data is standardized and encoded to form a basic data layer with a unified spatial reference and format.

[0038] The task is to plan the optimal laying path for a 110kV high-voltage cable that is about 1.2 kilometers long, using the cable corridor project supporting the newly built data center in a high-tech industrial development zone as the task requirement.

[0039] S11. For the planned area, the system initiated a multi-source data acquisition process: Geological and survey data: The project commissioned a surveying unit to drill 15 geological boreholes, obtaining borehole data tables containing information such as soil layering, lithological description, SPT blow count, and groundwater level; based on the borehole data, three geological profile maps were generated; simultaneously, ground-penetrating radar was used to detect fill areas that the route might traverse and key suspected karst areas, obtaining geophysical exploration data. Existing pipeline network data: The latest integrated pipeline as-built drawings for the area, corresponding to the CAD format and the corresponding attribute database, were retrieved from the Municipal Urban Construction Archives. The attribute database contains information on the material, diameter, burial depth, ownership unit, and laying year of pipelines such as water supply, drainage, gas, communication, and electricity. Additionally, BIM model files of recently approved and under-construction utility tunnel projects were obtained from the park management committee. Structured design code provisions: Mandatory provisions and recommended parameters directly related to this design were automatically retrieved and extracted from the integrated code knowledge base. Examples include: the minimum clearance requirements for parallel crossings of cables with heating pipes or other pipelines in the *Design Standard for Cables in Power Engineering* (GB 50217-2018); and the regulations on the cross-sectional layout of utility tunnels and the spacing of cable supports in the *Technical Standard for Urban Integrated Utility Tunnel Engineering* (GB / T50838-2015). Unstructured engineering documents: The following key documents were collected and digitized: the latest 1:500 paper-scanned topographic map of the park, the PDF report of the construction organization design and completion summary of the existing utility tunnel, the PDF report of the preliminary site environmental assessment, and JPG image diagrams of the surrounding pipeline survey provided by the developers of adjacent plots.

[0040] S12. Import the collected raw data into the system's data preprocessing platform and perform the following operations: Spatial benchmark unification: Perform geometric correction on the topographic map scans using known control points; uniformly convert the measurement coordinates of all pipeline entities in the CAD as-built drawings, components in the BIM model, and boreholes to the project-agreed "City Independent Coordinate System" and "1985 National Elevation Datum". Attribute data standardization coding: Establish a unified attribute coding dictionary. For example, uniformly map pipeline material descriptions "MPP" and "Modified Polypropylene" to the standard code "C-PP"; uniformly associate geological descriptions such as "Silty Clay" and "Clayey Silt" with standard geotechnical classification codes; bind specific clauses in the specification with unique specification IDs and clause numbers. Construction of the basic data layer: After the above processing, a basic data layer with consistent spatial references and unified attribute definitions is formed. This data layer is organized in the system as a combination of spatial database and file index. Boreholes, pipelines, etc. are stored as spatial features with attributes, while documents are indexed and managed through metadata such as document type, associated spatial range, and key page number, providing standardized input for subsequent semantic parsing and constraint field construction.

[0041] Figure 2 This paper illustrates a three-stage process for intelligently parsing unstructured documents to generate semantic constraint layers. Specifically, S2. Based on the MinerU2.5 visual language model, the unstructured engineering documents are parsed to extract textual semantic constraint information regarding regional taboos, safety distance requirements, and construction process requirements. This textual semantic constraint information is then geographically registered and spatially indexed with spatial entities in geological and survey data as well as existing pipeline network data to generate a spatially indexed semantic constraint layer.

[0042] In S2, unstructured engineering documents are parsed based on the MinerU2.5 visual language model, specifically including: S21. Convert the scanned copies of paper design drawings, construction log text, and environmental assessment report PDF document into image and text formats that can be processed by the MinerU2.5 model, respectively, while preserving the document layout and structure information; S22. Input the preprocessed document into the MinerU2.5 model. The model uses its visual encoder to parse the graphic symbols, annotations, and legends in the drawings, and its text encoder to understand the narrative text in the construction log and environmental assessment report. Combining visual and text features, the model identifies and structures the key semantic constraint entities. The key semantic constraint entities include area prohibition entities, safety distance entities, and construction process entities. Area prohibition entities include "cultural relic protection area", "chemical plant protection distance", "existing important pipeline protection area" and their corresponding boundary descriptions. Safety distance entities include "parallel clearance between cables and heating pipelines", "minimum vertical distance for cable crossing", "minimum burial depth for cable laying" and their corresponding values. Construction process entities include "sections requiring pipe jacking", "areas where open excavation is prohibited", and "recommended locations for cable joint wells".

[0043] In S2, textual semantic constraint information is geographically registered and spatially indexed with spatial entities, specifically including: S23. For semantic constraint entities with clearly defined locations parsed from design drawings, use the known control point coordinates or scale information in the drawings to map them to the geographic coordinate system and directly use them as spatial elements in the semantic constraint layer. S24. For semantically constrained entities whose location information is described in words and parsed from construction logs or environmental assessment reports, geocoding technology is used in conjunction with geological and survey data and the distribution of ground features in existing pipeline network data to perform spatial relationship reasoning, determine the spatial range or associated spatial entities corresponding to the semantically constrained entities, and associate the semantically constrained entities with the determined corresponding locations. S25. Organize and manage the spatial elements corresponding to all mapped or associated semantic constraint entities using spatial database indexing technology to construct a semantic constraint layer. The semantic constraint layer supports fast querying and retrieval based on spatial location.

[0044] For example, based on the completion of S1 data fusion, the system performs deep semantic parsing and spatial association on key unstructured engineering documents.

[0045] S21. For the documents collected in this embodiment, the system performs the following preprocessing to adapt to the MinerU2.5 model: Scanned drawings: Scanned copies of 1:500 topographic maps and surrounding pipeline survey diagrams are converted into PNG format images with a resolution of 300 DPI, while retaining their original size scale and spatial reference information. PDF documents: For construction organization designs, completion summaries, and environmental assessment reports, PDF parsing tools are used for content separation and recognition. Extracted plain text, such as construction process descriptions and problem records, is saved as UTF-8 encoded TXT files; pages containing charts and diagrams are converted into JPG format images. Layout structure preservation: During processing, by recording the page order, page numbers of images in the original document, and the position information of text paragraphs, the layout logic of the original document is ensured to be preserved, providing support for subsequent contextual understanding.

[0046] S22. Batch input the preprocessed images and text files into the pre-trained MinerU2.5 model. The model initiates multimodal joint inference: First, drawing parsing: The model's visual encoder recognizes contour lines and landmark symbols such as building outlines and road edges in the topographic map, as well as legends in the pipeline survey diagram, such as different line types representing gas pipes and communication pipes. Simultaneously, it parses the annotation text in the drawings, such as "planned road red line" and "existing high-voltage tower protection range 50 meters." Text understanding: The model's text encoder analyzes the descriptions in the construction log, such as "Unknown obstacle encountered at chainage K0+350, further exploration recommended"; and interprets the conclusions in the environmental assessment report, such as "The area on the north side of the site was historically a pond backfill, with poor geological conditions; large-scale excavation is not recommended." Structured output: The model integrates visual and textual features, ultimately outputting a set of semantically constrained entities with clear types, attributes, and location references. In this embodiment, the model successfully identified and structured the following typical entities: **Prohibited Area Entity:** Type: "Existing Important Pipeline Protection Zone"; Boundary Description: "A strip-shaped area extending 5 meters outwards on each side from the pipeline centerline in the as-built drawing"; Associated Drawing ID: G-2023-001. **Safety Distance Entity:** Type: "Parallel Clearance Between Cables and Heating Pipelines"; Value: "2.0 meters"; Source Standard Clause: "GB50217-2018, Clause 5.3.5". **Construction Technology Entity:** Type: "Area Where Open Excavation is Prohibited"; Location Description: "From Weisan Road in the north to Weiwu Road in the south, bordered by the planned green belt on the east side of Jingshi Road in the west"; Source Document: "Environmental Assessment Report, Page 15".

[0047] S23-S25. The system precisely associates the above-mentioned structured semantic information with the standardized spatial data formed in S1: Direct coordinate mapping: For entities with clear coordinate references parsed from drawings, such as "protected areas" generated based on specific pipeline coordinates, the system uses the known control point coordinates from the drawings to directly map the entity boundary coordinates to a unified geographic coordinate system through affine transformation, generating polygonal spatial features. Descriptive location reasoning and matching: For entities with textual descriptions of location parsed from text reports, such as "north side area of ​​the site," the system activates the geocoding engine. This engine combines the road network and known feature boundaries in the basic data layer to perform spatial reasoning on the descriptive text. For example, "from Weisan Road in the north to Weiwu Road in the south" is converted into the area between these two roads, and then spatially overlaid with "fill area" in the geological data to finally determine a suggested rectangular range to avoid, and associates the "no open excavation" constraint with this range. Semantic constraint layer generation: All generated spatial features and their associated semantic attributes are imported into the PostGIS spatial database and organized using R-Tree spatial indexing technology to create a vector layer named `semantic_constraints`. This layer supports efficient spatial queries, such as "querying all forbidden entities within a 10-meter buffer of the planned path," providing a directly accessible semantic rule base for subsequent intelligent planning.

[0048] Figure 3 The process of constructing a three-dimensional grid constraint field is illustrated, including non-uniform grid partitioning, state vector assignment, and the orientation of the integrated cost gradient field. Specifically, S3. Constructing a computable three-dimensional constraint field that integrates semantic information includes spatially overlaying and uniformly quantizing and encoding geological and survey data, existing pipeline network topology data, and semantic constraint layers associated with spatial indexes to generate a unified three-dimensional grid constraint field data structure covering the planning area. Each three-dimensional grid cell corresponds to a multi-dimensional state vector, which is used to characterize the comprehensive state of the cell in terms of obstacle avoidance level, construction cost coefficient, geological risk level, and specification compliance score.

[0049] In S3, a computable three-dimensional constrained field incorporating semantic information is constructed, specifically including: S31. Based on the three-dimensional boundary of the planning area and the undulation of the stratigraphic interface in the geological and survey data, a non-uniform three-dimensional grid is adaptively generated, wherein the grid cell size is dynamically adjusted according to the stratigraphic complexity and the density of existing pipelines; based on the geotechnical engineering geological parameters, the geological stability index of each grid cell is calculated. The geological stability index is obtained by a weighted entropy model that integrates the stratigraphic dip angle, soil cohesion and groundwater level, and is used to quantify the construction disturbance risk. S32. For each grid cell, perform coupled quantization of multi-source constraints: Based on existing pipeline network topology data, calculate the three-dimensional distance from the cell to the nearest pipeline, and generate a dynamic obstacle avoidance potential field according to the pipeline type, including high-voltage cables, water supply trunk pipes, and corresponding safety protection specifications; query the semantic constraint layer associated with the spatial index, and if the cell falls into a "no open-cut zone" or "cultural relic protection zone", set the construction permit flag to a prohibited state; if it is adjacent to a "safe distance entity", calculate the constraint compliance deviation value; encode the geological stability index, dynamic obstacle avoidance potential field value, constraint compliance deviation value, and construction cost coefficient based on cell burial depth and stratum properties into a five-dimensional state vector; S33. Associate the spatial coordinates of all grid cells with the corresponding five-dimensional state vectors to construct a sparse voxelized three-dimensional constraint field data structure; calculate the gradient field of the constraint field in three-dimensional space to indicate the direction of the fastest decrease in comprehensive cost during path search.

[0050] For example, in S31, the system constructs a three-dimensional modeling space based on the boundary of the planning area, a strip-shaped region approximately 1200 meters long east-west and 100 meters wide north-south, and a three-dimensional stratigraphic interface model generated from geological profile interpolation. Considering the complexity of the stratigraphic distribution and the density of pipelines, the system uses a non-uniform octree mesh for spatial discretization: in areas with dense boreholes and drastic stratigraphic changes, such as the interface between soft and hard soil layers or suspected karst areas, the minimum mesh size is set to 1m*1m*0.5m. In open areas with uniform stratigraphy and no existing pipelines, the mesh size adaptively increases to 5m*5m*2m. Within a 5-meter buffer zone around existing pipelines, especially high-voltage cables and main water supply pipes, the mesh is automatically refined to the minimum size to accurately represent the obstacle avoidance potential field gradient. Ultimately, approximately 1.2 million three-dimensional mesh cells of varying sizes are generated, covering the entire planning space. For each grid cell, the system queries the geological model based on its center point coordinates to obtain its corresponding stratigraphic unit and geotechnical parameters, such as internal friction angle φ, cohesion c, and groundwater level hw. The geological stability index R is calculated using the following weighted entropy model:

[0051] Where θ is the dip angle of the formation, c is the cohesion, and h w Let R be the groundwater level depth, and D be the unit burial depth. Weights w1, w2, and w3 are set based on regional experience, and a is the dip angle influence coefficient. A lower R value indicates a higher risk of construction disturbance. For example, in silty sand layers with shallow groundwater levels and low cohesion, the R value is significantly lower than that of units located at greater burial depths in stiff plastic clay.

[0052] S32. For each grid cell, the system performs parallel quantitative evaluation of multi-dimensional constraints and synthesizes a five-dimensional state vector [C]. avoid Ccost , R, S comp [, Flag], where obstacle avoidance level C avoid To calculate the three-dimensional Euclidean distance d from the center of the calculation unit to the surface of all existing pipelines. min Different potential field functions are set according to the pipeline type and specifications. For example, for a 110kV high-voltage cable, a safety distance d is set. safe =3.0m, then C avoid =e -λ*(d min / d safe ) λ is the attenuation coefficient. C avoid The value is in the range (0, 1], and the closer it is to the pipeline, the higher the value. Construction cost coefficient C cost Based on the geological strata type, burial depth, and groundwater level of the unit, the local construction quota database is consulted to estimate the comprehensive construction cost per unit length required if the unit were traversed by a route, including earthwork, support, and dewatering, and then normalization is performed. The geological risk level is the calculated R value. Standard compliance score S. comp To query the semantic constraint layer. If the cell falls into the "No Open Cut Zone", S comp Set directly to the lowest value of 0. If it is adjacent to a "safe distance entity", then calculate the compliance deviation Δd = d min - d required S comp =1 / (1 +e -k*Δd ), where k is the sensitivity coefficient. Otherwise S comp The default value is 0.5. The construction permit flag (Flag) is a Boolean value. If the unit falls into any "No Open Cut Zone" or "Cultural Heritage Protection Zone", then Flag = 1, representing a prohibited state; otherwise, it is 0.

[0053] S33. The system corresponds the spatial coordinates of all grid cells to the lower left corner coordinates and dimensions, along with their five-dimensional state vectors [C]. avoid C cost , R, S comp The system uses [Flag] to construct a sparse voxel octree data structure. This structure stores only non-empty grid cells (i.e., those within the planning space) and records their parent-child relationships, significantly compressing storage space and supporting fast neighbor lookup. Subsequently, the system calculates the comprehensive cost field U for each non-forbidden state cell on this data structure. U is C avoid C cost (1-R) ​​represents the risk cost and (1-S) represents the risk cost. comp This is a weighted sum that does not conform to the cost. Based on this scalar field U, its gradient field in three-dimensional space is calculated using the central difference method. U. The vector direction of U at each cell indicates the local direction in which the overall cost decreases the fastest during path search, providing efficient heuristic guidance information for subsequent global path search.

[0054] S4. A hierarchical decision-making strategy is used to generate a cable laying path sequence, including: Global path search phase: Based on a three-dimensional mesh constraint field, a weighted cost path search algorithm that integrates multi-dimensional state vectors is used to generate a global coarse path sequence consisting of multiple key waypoints between the preset path start and end points. Local path optimization stage: Guided by the global coarse path sequence, in the three-dimensional mesh constraint field, a local trajectory optimization algorithm based on the physical motion constraint model is used to smooth the global coarse path sequence, verify its construction feasibility, and make precise obstacle avoidance adjustments, generating a continuous and constructable three-dimensional spatial path curve.

[0055] In S4, during the global path search phase, a weighted cost path search algorithm incorporating multi-dimensional state vectors is used. A phased cost evaluation strategy is employed to search for a coarse global path sequence within the three-dimensional constraint field. The specific process includes: S41. Starting from the origin, expand the nodes to candidate mesh cells in the 3D spatial neighborhood according to the 3D constraint field data structure; for each candidate mesh cell, calculate the first evaluation cost. The formula is:

[0056] Where D is the three-dimensional Euclidean distance from the previous node to the center point of the candidate grid cell, and R is the geological stability index of the candidate grid cell. To estimate the foundation construction cost based on unit stratigraphic properties and burial depth, , , The preset weighting coefficients; select The node with the smallest value is expanded until the destination is reached, thus generating the initial path; S42. Based on the initial path, perform secondary optimization by combining the semantic constraint layer and gradient field; calculate the second evaluation cost for each grid cell on the path. The formula is:

[0057] in, The gradient of the dynamic obstacle avoidance potential field at the location of the grid cell is θ, where θ is the direction of the path segment. The angle between the directions is used to guide the path along the downward direction of the potential field to avoid obstacles; It is a function of the constraint compliance deviation value of the unit, used to penalize path segments that violate semantic safety distance; This is a Boolean penalty term; if the unit construction permission flag is in a prohibited state, then... Take a maximum positive value, otherwise set to 0; iterate through the path nodes to minimize the overall path. The value generates the final global coarse path sequence.

[0058] In the local path optimization stage of S4, the specific execution process of the local trajectory optimization algorithm based on the physical motion constraint model includes: S43. Initial deformation guided by gradient: Taking each key waypoint in the global coarse path sequence as the center, in the neighborhood of each key waypoint, according to the gradient field direction pre-calculated in the three-dimensional constraint field data structure, local sampling is performed along the downward direction of the dynamic obstacle avoidance potential field to generate a series of candidate fine-tuning path segments that satisfy the minimum bending radius constraint of the cable. S44. Establish a mechanically feasible elastic spline optimization model: Using the global coarse path sequence and candidate fine-tuning path segments as initial control points, generate a parameterized 3D B-spline curve; establish a nonlinear optimization problem with the control point coordinates of the 3D B-spline curve as the optimization variables, and the objective function of the nonlinear optimization problem is... Defined as:

[0059] Among them, the integral variable Let be the arc length parameter along the 3D B-spline curve. This represents the total length of the three-dimensional B-spline curve; For a 3D B-spline curve in arc length The curvature at the point, the integral term of the square of the curvature is used to ensure the smoothness of the path and meet the physical constraint of the minimum bending radius of the cable; Points on a 3D B-spline curve The dynamic obstacle avoidance potential field gradient at the location; For a 3D B-spline curve in arc length The unit tangent vector at point , the dot product term Used to align the path direction with the downward direction of the dynamic obstacle avoidance potential field to enhance obstacle avoidance performance; The weighting coefficient is used to guide the potential field. As a constraint penalty term, when the 3D B-spline curve passes through an area marked as "no open excavation" in the semantic constraint layer or passes through a cell in the 3D mesh constraint field where the construction permit flag is in a prohibited state, the constraint penalty term takes a preset maximum positive value. S45. Iterative Solution and Feasibility Verification: The nonlinear optimization problem is solved using a sequential quadratic programming algorithm, iteratively updating the coordinates of the control points of the 3D B-spline curve. After each iteration, it is verified whether the curvature of each sampling point on the updated path curve is less than the maximum allowable curvature calculated from the minimum bending radius constraint of the cable, and whether the updated path curve avoids all cells where the construction permit flag is in a prohibited state. Finally, a continuous and constructable 3D spatial path curve that satisfies all constraints is output.

[0060] Preferably, in this embodiment, the following is set: =0.5, =2, =1, not an arbitrary value, but based on the project's priority analysis. Through reviewing historical project data and expert experience, it was found that soft soil layers are widespread in the project area, and construction risks and cost overruns caused by adverse geological conditions are the main challenges. Therefore, geological risk (1-R) ​​is assigned the highest weight. =2, designed primarily to avoid high-risk strata such as quicksand layers and highly compressible soft soil. Construction cost C cost Secondary correspondence =1, reflecting the requirement of economy. Geometric distance D is assigned a relatively low weight. The weight value of 0.5 is set to allow for appropriate detours when necessary, in exchange for lower risk and cost. The system supports multiple preset weight templates or interactive adjustments based on different engineering scenarios, such as rock formations or densely populated urban pipe network areas.

[0061] The cable is simulated as having a minimum bending radius R. min The constrained elastic body is the core of this stage. Cubic uniform B-spline curves are used for parameterization because they possess inherent C-values. 2 Continuity, meaning the continuity of the curve, the first derivative, and the second derivative, directly guarantees the continuity of curvature, providing an ideal mathematical expression for meeting the requirements of smooth cable laying. Choosing Sequential Quadratic Programming (SQP) as the solver is based on the characteristics of this optimization problem, with the objective function F... local This problem includes nonlinear terms such as squared curvature and gradient dot product, and constraints include inequality constraints, upper bounds on curvature, and obstacle avoidance penalties. The SQP algorithm efficiently handles this type of nonlinear, constrained optimization problem by constructing and solving an approximate quadratic programming subproblem in each iteration, exhibiting good convergence. Curvature and obstacle avoidance verification performed after each iteration ensures the physical feasibility of intermediate and final solutions. If verification fails, the system automatically adjusts the coefficient of the penalty term Φ_penalty or introduces additional constraints and re-optimizes. (Constraint penalty term) The specific calculation method is as follows:

[0062] in, To uniformly sample the 3D B-spline curve during the optimization process. Coordinates of a spatial point, This represents the total number of sampling points; For the indicator function, if the sampling point If a cell falls into any construction permit flag that is in a prohibited state or falls into an area in the semantic constraint layer where "open excavation is prohibited", the indicator function value is 1; otherwise, it is 0. Indicates the first The center coordinates of a known obstacle, which is defined by existing pipeline network topology data or characterized by high obstacle avoidance level cells in a three-dimensional mesh constraint field; Sampling points The distance to the nearest obstacle in the known set of obstacles; The minimum safe distance defined by the "safe distance entity" in the semantic constraint layer and required by the obstacle type; The penalty coefficient for entering the prohibited area is set to a value that is much larger than other positive numbers to ensure that the path strictly avoids the prohibited area; To mitigate the penalty for violating the safety distance, a quadratic penalty term is used to allow for smooth path adjustments near the safety distance boundary.

[0063] Regarding the generation of "candidate fine-tuning path segments" and the minimum bending radius R... min The pre-satisfaction, in step S43, when generating candidate fine-tuning path segments, is not simply achieved through random or grid sampling. The system uses the waypoints of the global coarse path as base points and, along the gradient descent direction, uses R... min To minimize the turning radius, several candidate curve segments are generated by calculating the transition mode of the tangent circle or spiral curve, smoothly transitioning from the current waypoint to the next adjacent region. This ensures that the initial solution for subsequent B-spline curve optimization already basically satisfies R. min Constraints significantly improve optimization efficiency and reduce the risk of getting trapped in local inferior solutions. Figure 4a The diagram shows a coarse path sequence generated by global path search. The background represents the XZ profile of the 3D constraint field, with dark areas indicating high-cost regions that need to be avoided and light areas indicating low-cost regions. The broken lines represent the global coarse path sequence composed of key waypoints, clearly avoiding high-cost regions. Figure 4b This demonstrates how local path optimization generates smooth, constructable B-spline curves. The global coarse path (dashed line) serves as a guide, leading to the generation of smooth 3D B-spline curves (solid lines). The curvature κ and minimum bending radius R are verified at curve bends. min Check the safe distance when near the obstacle.

[0064] Figure 5The interface for displaying the planning results is shown. The left side displays the interaction between the path and the environment in a 3D scene, while the right side provides a structured report with detailed explanations and warnings. Specifically, S5 outputs the planning results and semantic basis, including generating the final 3D path of the cable tunnel and simultaneously generating a structured interpretability report. The report automatically marks the specific engineering specifications, geological conditions, and construction constraints that are satisfied or avoided in key sections of the 3D path by associating the multidimensional state vectors of relevant grid cells in the 3D grid constraint field with the extracted textual semantic constraint information.

[0065] The method also includes a network-wide evaluation and scheduling step S6 following S5: S61. Perform a three-dimensional spatial topology analysis on the cable tunnel three-dimensional path generated in S5 and the existing cable tunnel path extracted from the existing pipeline network topology data to identify the connection points, crossing points and parallel segments between the new path and the existing network. S62. At connection points and intersections, based on the requirements of the structured design code regarding the cross-sectional dimensions, layered layout, and safety distances of the utility tunnel, verify the feasibility of accommodating the new cables in the existing utility tunnel cross-section; if there is a space conflict or insufficient capacity, generate an early warning in the structured interpretability report and identify the specific three-dimensional location of the conflict and the violated code clauses. S63. When multiple cable routes need to be planned in the planning area, execute S1 to S5 sequentially or in parallel, and perform overall coupling analysis on all generated routes; based on the construction period, equipment resources and dynamic construction cost coefficients in the three-dimensional mesh constraint field, simulate the impact of the construction sequence of different routes on the overall cost and construction period, and output recommended construction sequence and resource scheduling scheme.

[0066] For example, in S4, a continuous three-dimensional path curve C that satisfies the constraints is generated. final(t) After that, the system enters the stage of generating, interpreting and expanding the evaluation of the planning results.

[0067] S5. The system does not merely output path coordinates, but simultaneously generates a structured report that deeply integrates planning logic and rationale. 3D Path and Basic Report Generation: The system first... final(t) The data is discretized into a high-density sequence of three-dimensional coordinate points, with geodetic coordinates and mileage information appended, generating a cable centerline .dxf file that can be directly imported into CAD / BIM software. Simultaneously, the core sections of the report are automatically generated: Route Overview: Displaying the total length of 1185 meters, start and end coordinates, and calculating the total turning angles and estimated total construction cost based on the route curve integral. Segmented Technical Details Table: The system divides the route into 12 segments per 100-meter interval. For each segment, the following are calculated and listed: average burial depth, main strata traversed, and average geological stability index of all sampling points within the segment. <r>Average obstacle avoidance level <C avoid For example, the section from K0+800 to K0+900 is shown as "crossing a silty clay layer, with an average burial depth of -4.2m." <r>=0.75 (low risk) <C avoid >=0.1 (far from existing pipelines)". Automatic annotation and tracing of constraint compliance: The system traverses dense sampling points on the path and performs the following association and annotation: Association semantic constraints: For each sampling point, in semantic constraints Spatial queries are performed within the layers. For example, if a point near K0+350 is found to be within the influence range of a "suspected obstacle" identified from the construction log, the report automatically notes next to that point: "Note: The suggested exploration area for the 'unknown obstacle at chainage K0+350' identified in the adjacent S2 stage analysis has been avoided by increasing the burial depth." Related regulatory clauses: When the calculated distance at K0+520, where the closest parallel distance between the path and the existing heating pipeline is 2.1 meters, the system automatically links to the previously extracted regulatory entity "Parallel clearance between cable and heating pipeline ≥ 2.0m," and notes at that location in the report: "Complies with regulation: GB50217-2018, Clause 5.3.5 (actual measured distance to heating pipe 2.1m)." Geological risk warning: For the path... <r>The report highlights the lowest-valued continuous segments, such as K0+150 ~ K0+200, which traverse areas of high-water-content soft soil, in the corresponding sections and links them to data from the geological profile (Profile-2) and borehole ZK-07 for explanation. Finally, the report is output in two formats: a PDF report with images and text, and structured JSON data (including path geometry, attributes, and indexes of all annotations), achieving full traceability of planning decisions.

[0068] S6. Considering that actual engineering projects rarely plan single paths in isolation, the system provides network analysis capabilities after S5. Integration analysis with existing utility tunnel networks: The system loads the path model of the existing 110kV utility tunnel-A extracted from the as-built drawings. Through 3D spatial calculations, it automatically identifies a planar intersection between the planned new path and utility tunnel-A at K0+900, and a 100-meter parallel section from K1+000 to K1+100. Simultaneously, it accurately calculates the 3D coordinates of the intersection point and the minimum clearance of the parallel section. Cross-sectional capacity and conflict verification: At the intersection point K0+900, the system retrieves the BIM model of utility tunnel-A to obtain its standard cross-sectional dimensions of 2.5m wide * 3.0m high and the current occupancy of cable supports on each floor. Based on the spacing requirements for layered cable laying in standard GB 50838, the system attempts to "place" the simulated outer diameter of the planned cables in the cross-sectional model. Verification revealed that if the cable were to cross at the planned height, it would violate the minimum installation space requirement between the top-level cable and the top slab of the utility tunnel. Therefore, the system generated a warning in the report: "Conflict Warning: When a new cable crosses the existing utility tunnel-A at the intersection, the distance between it and the top slab of the utility tunnel does not meet the requirements of the specific standard item. It is recommended to adjust the elevation of this path or negotiate a partial modification with the operating unit." Multi-path collaborative planning simulation: In the extended scenario of this embodiment, it is assumed that a second independent 10kV power path needs to be planned for the data center. The system can reuse the constructed three-dimensional constraint field to quickly generate an optimized path for the second cable. Subsequently, the system initiates coupling analysis: Spatial conflict check: Ensures that there are no conflicts in three-dimensional space between the two new paths and between them and existing pipelines. Construction sequence simulation: The system is based on the cost coefficient C in the constraint field. cost The simulation included two construction sequences: 1) constructing path 1 first, then path 2; 2) constructing both paths concurrently and adjacent to each other. The simulation considered equipment relocation costs, discount factors for repeated excavation in the same geological section, and schedule constraints. The recommended solution was output: Simulation results showed that since both paths traverse the same high-risk soft soil area between K0+300 and K0+500, adopting the "concurrent construction, shared foundation pit and support" solution, although requiring slightly higher initial investment, could save 15% of the overall construction period and reduce total costs by approximately 8%. Based on this, the system output a recommended construction sequence and resource scheduling plan diagram, which was included in the final report.

[0069] One embodiment of this application discloses a cable tunnel route planning system, which includes: a data fusion module for acquiring and fusing multi-source constraint data of the planning area to form a fused data foundation for route planning; The semantic parsing and association module communicates with the data fusion module and is used to parse unstructured engineering documents based on the MinerU2.5 visual language model to extract text semantic constraint information, and associate the text semantic constraint information with geological and pipeline spatial entities to generate a semantic constraint layer. The 3D constraint field construction module communicates with the semantic parsing and association module to integrate geological data, pipeline data and semantic constraint layers, and constructs a unified 3D mesh constraint field through quantization encoding. The hierarchical path planning module communicates with the three-dimensional constraint field construction module and is used to generate continuous and constructable three-dimensional spatial path curves based on the three-dimensional mesh constraint field through a hierarchical decision-making strategy that combines global search and local optimization. The results generation and reporting module communicates with the hierarchical path planning module to generate a structured and interpretable report of the final 3D path and associated 3D mesh constraint field and textual semantic constraint information.

[0070] Based on the same inventive concept, according to another aspect of the present invention, embodiments of the present invention also provide a computer device, comprising: At least one processor; and The memory stores a computer program that can run on a processor. When the processor executes the program, it performs the steps of the cable tunnel path planning method described above.

[0071] Based on the same inventive concept, according to another aspect of the present invention, embodiments of the present invention also provide a computer-readable storage medium storing computer program instructions, which, when executed by a processor, perform the steps of the cable tunnel path planning method described above.

[0072] Finally, it should be noted that those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc. The above computer program embodiments can achieve the same or similar effects as any of the corresponding foregoing method embodiments.

[0073] Furthermore, typically, the devices and equipment disclosed in the embodiments of this invention can be various electronic terminal devices, such as mobile phones, personal digital assistants (PDAs), tablet computers (PADs), smart TVs, etc., or they can be large terminal devices, such as servers. Therefore, the scope of protection disclosed in the embodiments of this invention should not be limited to a specific type of device or equipment. The client disclosed in the embodiments of this invention can be applied to any of the above-mentioned electronic terminal devices in the form of electronic hardware, computer software, or a combination of both.

[0074] Furthermore, the method disclosed in the embodiments of the present invention can also be implemented as a computer program executed by a CPU, which may be stored in a computer-readable storage medium. When the computer program is executed by the CPU, it performs the functions defined in the method disclosed in the embodiments of the present invention.

[0075] Furthermore, the above-described method steps and system units can also be implemented using a controller and a computer-readable storage medium for storing a computer program that enables the controller to perform the functions of the above-described steps or units.

[0076] Furthermore, it should be understood that the computer-readable storage medium (e.g., memory) described herein can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. By way of example, and not limitation, non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM), which can act as external cache memory. By way of example, and not limitation, RAM may be available in various forms, such as synchronous RAM (DRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct Rambus RAM (DRRAM). The storage devices disclosed herein are intended to include, but are not limited to, these and other suitable types of memory.

[0077] The above are exemplary embodiments disclosed in this invention. However, it should be noted that various changes and modifications can be made without departing from the scope of the embodiments of this invention as defined by the claims. The functions, steps, and / or actions of the methods according to the disclosed embodiments described herein do not need to be performed in any marked order. Furthermore, although the elements disclosed in the embodiments of this invention may be described or claimed individually, they may be understood as multiple unless explicitly limited to a singular.

[0078] In this specification, the same or similar parts between the various embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the descriptions of the embodiments described later are relatively simple, and relevant parts can be referred to the descriptions of the foregoing embodiments.

[0079] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.< / r> < / r> < / r>

Claims

1. A method for planning the route of a cable tunnel, characterized in that, The method includes the following steps: S1. Acquire and integrate multi-source constraint data for the planning area. The multi-source constraint data shall include at least geological and survey data of underground space structures, existing pipeline network topology data, structured design code provisions, and unstructured engineering documents. S2. Based on the MinerU2.5 visual language model, unstructured engineering documents are parsed to extract textual semantic constraint information about regional taboo conditions, safety distance requirements and construction technology requirements. The textual semantic constraint information is then geographically registered and spatially indexed with spatial entities in geological and survey data and existing pipeline network data to generate a semantic constraint layer associated with spatial index. S3. Construct a computable three-dimensional constraint field that integrates semantic information, including spatially overlaying and uniformly quantizing and encoding geological and survey data, existing pipeline network topology data, and semantic constraint layers associated with spatial indexes to generate a unified three-dimensional grid constraint field data structure covering the planning area. Each three-dimensional grid cell corresponds to a multi-dimensional state vector, which is used to characterize the comprehensive state of the cell in terms of obstacle avoidance level, construction cost coefficient, geological risk level, and standard compliance score. S4. A hierarchical decision-making strategy is used to generate a cable laying path sequence, including: Global path search phase: Based on a three-dimensional mesh constraint field, a weighted cost path search algorithm that integrates multi-dimensional state vectors is used to generate a global coarse path sequence consisting of multiple key waypoints between the preset path start and end points. Local path optimization stage: Guided by the global coarse path sequence, within a 3D mesh constraint field, a local trajectory optimization algorithm based on a physical motion constraint model is used to smooth the global coarse path sequence, verify its construction feasibility, and precisely adjust obstacle avoidance, generating a continuous and constructable 3D spatial path curve; the local trajectory optimization algorithm includes: Centered on each key waypoint in the global coarse path sequence, in the neighborhood of each key waypoint, local sampling is performed along the descent direction of the dynamic obstacle avoidance potential field based on the gradient field direction pre-calculated in the three-dimensional constraint field data structure to generate a series of candidate fine-tuning path segments that satisfy the minimum bending radius constraint of the cable. The global coarse path sequence and candidate fine-tuning path segments are used together as initial control points to generate a parameterized three-dimensional B-spline curve; a nonlinear optimization problem is established with the coordinates of the control points of the three-dimensional B-spline curve as optimization variables. The nonlinear optimization problem is solved by using a sequential quadratic programming algorithm to iteratively update the coordinates of the control points of the three-dimensional B-spline curve. After each iteration, it is verified whether the curvature of each sampling point on the updated path curve is less than the maximum allowable curvature calculated by the minimum bending radius constraint of the cable, and whether the updated path curve avoids all cells where the construction permission flag is in a prohibited state. Finally, a continuous and constructable three-dimensional spatial path curve that satisfies all constraints is output. S5. Output planning results and semantic basis, including generating the final three-dimensional path of the cable tunnel, and simultaneously generating a structured interpretable report. The report automatically marks the specific engineering specifications, geological conditions and construction constraints that are satisfied or avoided in key sections of the three-dimensional path by associating the multi-dimensional state vectors of relevant grid cells in the three-dimensional grid constraint field with the extracted text semantic constraint information.

2. The cable tunnel route planning method according to claim 1, characterized in that, In S1, multi-source constraint data of the planning area is acquired and integrated, specifically including: S11. Data Acquisition: Obtain raw data for the planning area. The raw data shall include at least the following: Drilling data, geological profiles, and geophysical exploration data obtained through geological surveys and geophysical exploration; existing pipeline as-built drawings, attribute tables, and spatial coordinate data retrieved from municipal pipeline management departments; mandatory clauses and recommended parameters related to cable tunnel design read from the structured specification knowledge base; and scanned copies of paper design drawings, construction log texts, and environmental assessment report PDF documents as unstructured engineering documents. S12. Data Standardization and Spatial Registration: The spatial information in the original data is uniformly converted to the same geographic coordinate system, and the attribute data is standardized and encoded to form a basic data layer with a unified spatial reference and format.

3. The cable tunnel route planning method according to claim 2, characterized in that, In S2, unstructured engineering documents are parsed based on the MinerU2.5 visual language model, specifically including: S21. Convert the scanned copies of paper design drawings, construction log text, and environmental assessment report PDF document into image and text formats that can be processed by the MinerU2.5 model, respectively, while preserving the document layout and structure information; S22. Input the preprocessed document into the MinerU2.5 model. The model uses its visual encoder to parse the graphic symbols, annotation text, and legends in the drawings, and its text encoder to understand the narrative text in the construction log and environmental assessment report. Combining visual and text features, the model identifies and outputs key semantic constraint entities in a structured manner. Key semantic constraint entities include area prohibition entities, safety distance entities, and construction process entities. Area prohibition entities include "cultural relic protection area", "chemical plant protection distance", "existing important pipeline protection area" and their corresponding boundary descriptions. Safety distance entities include "parallel clearance between cables and heating pipelines", "minimum vertical distance for cable crossing", "minimum burial depth for cable laying" and their corresponding values. Construction process entities include "sections requiring pipe jacking", "areas where open excavation is prohibited", and "recommended locations for cable joint wells".

4. The cable tunnel route planning method according to claim 3, characterized in that, In S2, textual semantic constraint information is geographically registered and spatially indexed with spatial entities, specifically including: S23. For semantic constraint entities with clearly defined locations parsed from design drawings, use the known control point coordinates or scale information in the drawings to map them to the geographic coordinate system and directly use them as spatial elements in the semantic constraint layer. S24. For semantically constrained entities whose location information is described in words and parsed from construction logs or environmental assessment reports, geocoding technology is used in conjunction with geological and survey data and the distribution of ground features in existing pipeline network data to perform spatial relationship reasoning, determine the spatial range or associated spatial entities corresponding to the semantically constrained entities, and associate the semantically constrained entities with the determined corresponding locations. S25. Organize and manage the spatial elements corresponding to all mapped or associated semantic constraint entities using spatial database indexing technology to construct a semantic constraint layer. The semantic constraint layer supports fast querying and retrieval based on spatial location.

5. The cable tunnel route planning method according to claim 1, characterized in that, In S3, a computable three-dimensional constrained field incorporating semantic information is constructed, specifically including: S31. Based on the three-dimensional boundary of the planning area and the undulation of the stratigraphic interface in the geological survey data, a non-uniform three-dimensional grid is adaptively generated, wherein the grid cell size is dynamically adjusted according to the stratigraphic complexity and the density of existing pipelines; based on the geotechnical engineering geological parameters, the geological stability index of each grid cell is calculated. The geological stability index is obtained by a weighted entropy model that integrates the stratigraphic dip angle, soil cohesion and groundwater level, and is used to quantify the construction disturbance risk. S32. For each grid cell, perform coupled quantization of multi-source constraints: Based on existing pipeline network topology data, calculate the three-dimensional distance from the cell to the nearest pipeline, and generate a dynamic obstacle avoidance potential field according to the pipeline type, including high-voltage cables, water supply trunk pipes, and corresponding safety protection specifications; query the semantic constraint layer associated with the spatial index, and if the cell falls into a "no open-cut zone" or "cultural relic protection zone", set the construction permit flag to a prohibited state; if it is adjacent to a "safe distance entity", calculate the constraint compliance deviation value; encode the geological stability index, dynamic obstacle avoidance potential field value, constraint compliance deviation value, and construction cost coefficient based on cell burial depth and stratum properties into a five-dimensional state vector; S33. Associate the spatial coordinates of all grid cells with the corresponding five-dimensional state vectors to construct a sparse voxelized three-dimensional constraint field data structure; calculate the gradient field of the constraint field in three-dimensional space to indicate the direction of the fastest decrease in comprehensive cost during path search.

6. The cable tunnel route planning method according to claim 5, characterized in that, In S4, during the global path search phase, a weighted cost path search algorithm incorporating multi-dimensional state vectors is used. A phased cost evaluation strategy is employed to search for a coarse global path sequence within the three-dimensional constraint field. The specific process includes: S41. Starting from the origin, expand the nodes to candidate mesh cells in the 3D spatial neighborhood according to the 3D constraint field data structure; for each candidate mesh cell, calculate the first evaluation cost. The formula is: Where D is the three-dimensional Euclidean distance from the previous node to the center point of the candidate grid cell, and R is the geological stability index of the candidate grid cell. To estimate the foundation construction cost based on unit stratigraphic properties and burial depth, , , The preset weighting coefficients; select The node with the smallest value is expanded until the destination is reached, thus generating the initial path; S42. Based on the initial path, perform secondary optimization by combining the semantic constraint layer and gradient field; calculate the second evaluation cost for each grid cell on the path. The formula is: in, The gradient of the dynamic obstacle avoidance potential field at the location of the grid cell is θ, where θ is the direction of the path segment. The angle between the directions is used to guide the path along the downward direction of the potential field to avoid obstacles; It is a function of the constraint compliance deviation value of the unit, used to penalize path segments that violate semantic safety distance; This is a Boolean penalty term; if the unit construction permission flag is in a prohibited state, then... Take a maximum positive value, otherwise set to 0; iterate through the path nodes to minimize the overall path. The value generates the final global coarse path sequence.

7. The cable tunnel route planning method according to claim 6, characterized in that, In the local path optimization stage of S4, the specific execution process of the local trajectory optimization algorithm based on the physical motion constraint model also includes: Objective function of nonlinear optimization problem Defined as: Among them, the integral variable Let be the arc length parameter along the 3D B-spline curve. This represents the total length of the three-dimensional B-spline curve. For a 3D B-spline curve in arc length The curvature at the point, the integral term of the square of the curvature is used to ensure the smoothness of the path and meet the physical constraint of the minimum bending radius of the cable; Points on a 3D B-spline curve The dynamic obstacle avoidance potential field gradient at the location; For a 3D B-spline curve in arc length The unit tangent vector at point , the dot product term Used to align the path direction with the downward direction of the dynamic obstacle avoidance potential field to enhance obstacle avoidance performance; The weighting coefficient is used to guide the potential field. As a constraint penalty term, when the 3D B-spline curve passes through an area marked as "no open excavation" in the semantic constraint layer or passes through a cell in the 3D mesh constraint field where the construction permit flag is in a prohibited state, the constraint penalty term takes a preset maximum positive value.

8. The cable tunnel route planning method according to claim 7, characterized in that, Constraints and penalties The specific calculation method is as follows: in, To uniformly sample the 3D B-spline curve during the optimization process. Coordinates of a spatial point, This represents the total number of sampling points; For the indicator function, if the sampling point If a cell falls into any construction permit flag that is in a prohibited state or falls into an area where open excavation is prohibited in the semantic constraint layer, the indicator function value is 1; otherwise, it is 0. Indicates the first The center coordinates of a known obstacle, which is defined by existing pipeline network topology data or characterized by high obstacle avoidance level cells in a three-dimensional mesh constraint field; Sampling points The distance to the nearest obstacle in the known set of obstacles; The minimum safe distance defined by the "safe distance entity" in the semantic constraint layer and required by the obstacle type; The penalty coefficient for entering the prohibited area is set to a positive number greater than other items to ensure that the path strictly avoids the prohibited area; To mitigate the penalty for violating the safety distance, a quadratic penalty term is used to allow for smooth path adjustments near the safety distance boundary.

9. The cable tunnel route planning method according to claim 1, characterized in that, The method also includes a network-wide evaluation and scheduling step S6 following S5: S61. Perform a three-dimensional spatial topology analysis on the cable tunnel three-dimensional path generated in S5 and the existing cable tunnel path extracted from the existing pipeline network topology data to identify the connection points, crossing points and parallel segments between the new path and the existing network. S62. At connection points and intersections, based on the requirements of the structural design code regarding the cross-sectional dimensions, layered layout, and safety distances of the utility tunnel, verify the feasibility of accommodating the new cables in the existing utility tunnel cross-section. If there is a spatial conflict or insufficient capacity, an alert will be generated in the structured interpretability report, and the specific three-dimensional location of the conflict and the violated normative clauses will be identified. S63. When multiple cable routes need to be planned in the planning area, execute S1 to S5 sequentially or in parallel, and perform overall coupling analysis on all generated routes; based on the construction period, equipment resources and dynamic construction cost coefficients in the three-dimensional mesh constraint field, simulate the impact of the construction sequence of different routes on the overall cost and construction period, and output recommended construction sequence and resource scheduling scheme.

10. A cable tunnel route planning system for implementing the cable tunnel route planning method according to any one of claims 1-9, characterized in that, The system includes: The data fusion module is used to acquire and fuse multi-source constraint data of the planning area to form a fused data foundation for path planning; The semantic parsing and association module communicates with the data fusion module and is used to parse unstructured engineering documents based on the MinerU2.5 visual language model to extract text semantic constraint information, and associate the text semantic constraint information with geological and pipeline spatial entities to generate a semantic constraint layer. The 3D constraint field construction module communicates with the semantic parsing and association module to integrate geological data, pipeline data and semantic constraint layers, and constructs a unified 3D mesh constraint field through quantization encoding. The hierarchical path planning module communicates with the three-dimensional constraint field construction module and is used to generate continuous and constructable three-dimensional spatial path curves based on the three-dimensional mesh constraint field through a hierarchical decision-making strategy that combines global search and local optimization. The results generation and reporting module communicates with the hierarchical path planning module to generate a structured and interpretable report of the final 3D path and associated 3D mesh constraint field and textual semantic constraint information.

Images