Whole vehicle pipeline design method and electronic equipment

Abstract

The invention provides a whole vehicle pipeline design method and electronic equipment, and the method comprises the steps: constructing a discrete space according to design parameters of a vehicle; according to pipeline position parameters in the design parameters, a plurality of vectors corresponding to first pipelines to be designed in the vehicle are defined in the discrete space, and each pipeline position parameter corresponds to one first pipeline; based on the coordinate corresponding to the first vector in the plurality of vectors, utilizing a preset pipeline trend prediction model to predict coordinates corresponding to other vectors of the first pipeline; determining a trend path of the first pipeline based on the coordinate set corresponding to all the vectors of the first pipeline; and generating a pipeline design model of the vehicle based on the trend paths of all the first pipelines of the vehicle. According to the method, automatic design of the pipeline can be achieved, labor cost can be reduced, and design efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of vehicles, and in particular to a vehicle pipeline design method and electronic equipment. Background Art

[0002] In vehicle design, the design of pipelines (for example, electrical wiring harnesses, fuel lines, cooling lines, air conditioning lines, air ducts, etc.) is a core link in ensuring the safety, reliability, and maintainability of the entire vehicle. In the process of designing vehicle pipelines, traditional design methods rely heavily on manual experience and computer-aided design tools. These methods suffer from cumbersome processes, high cross-departmental communication costs, and time-consuming repeated modifications, resulting in a serious waste of human resources and difficulty in shortening the design cycle. In related technologies, optimization solutions for pipeline design typically focus on improvements at the design tool level, making it difficult to overcome the inherent limitations of manual dominance. This results in a long design cycle for vehicle pipelines, low design efficiency, and high R&D costs. Summary of the Invention

[0003] The embodiments of the present application disclose a vehicle pipeline design method and an electronic device storage medium, which can autonomously generate feasible design solutions for vehicle pipelines, thereby solving problems such as long vehicle design cycles and high R&D costs caused by the difficulty in breaking through manually dominated design methods.

[0004] The present application provides a vehicle pipeline design method, the method comprising: constructing a discrete space based on vehicle design parameters; defining, in the discrete space, multiple vectors corresponding to a first pipeline to be designed in the vehicle based on pipeline position parameters in the design parameters, wherein each pipeline position parameter corresponds to a first pipeline; based on the coordinates corresponding to a first vector in the multiple vectors, using a preset pipeline direction prediction model to predict the coordinates corresponding to the remaining vectors of the first pipeline; based on the coordinate set corresponding to all vectors of the first pipeline, determining the direction path of the first pipeline; and generating a pipeline design model for the vehicle based on the direction paths of all first pipelines in the vehicle.

[0005] In some embodiments of the present application, the design parameters include the dimensions of the vehicle and data of non-pipeline physical components of the vehicle. Constructing a discrete space based on the design parameters of the vehicle includes: constructing a discrete space having a corresponding size relationship with the dimensions, and determining a plurality of voxels of the discrete space based on a preset unit size; determining the coordinates and label of each voxel in the discrete space, the label being used to indicate whether the corresponding voxel includes data of non-pipeline physical components in the vehicle.

[0006] In some embodiments of the present application, after constructing the discrete space, the method further includes: assigning label values ​​to corresponding voxels according to the labels, so that each voxel has a corresponding label value, wherein if the label indicates that the corresponding voxel includes the data of the physical component, the label value corresponding to the voxel is a first value; if the label indicates that the corresponding voxel does not include any data of the physical component, the label value corresponding to the voxel is a second value; based on the label value corresponding to each voxel, constructing a first spatial matrix corresponding to the discrete space, the elements in the first spatial matrix include the first value and the second value.

[0007] In some embodiments of the present application, the pipeline position parameters include the first position coordinates and the second position coordinates corresponding to the first pipeline in the discrete space, and the multiple vectors corresponding to the first pipeline to be designed in the vehicle are defined in the discrete space according to the pipeline position parameters in the design parameters, including: defining the first vector of the first pipeline based on the first position coordinates and the second position coordinates. 0, the first position coordinate is the first vector 0, the second position coordinate is the first vector 0 end coordinates; for the two adjacent vectors of the first pipeline, the vector coordinates of the rear vector are determined according to the vector coordinates of the front vector, including: according to the i-th vector of the first pipeline The end point coordinates of (i-1) define the i+1th vector of the first pipeline i, the i+1th vector The starting point coordinates of i are the i-th vector (i-1), where i represents an integer greater than or equal to 1 and less than j+1, and j+1 represents the number of vectors corresponding to the first pipeline.

[0008] In some embodiments of the present application, based on the coordinates corresponding to the first vector of the plurality of vectors, using a preset pipeline direction prediction model to predict the coordinates corresponding to the remaining vectors of the first pipeline includes: based on the i-th vector of the first pipeline (i-1) corresponding starting point coordinates and end point coordinates, use the pipeline direction prediction model to obtain the i+1th vector of the first pipeline The starting point coordinates and the ending point coordinates of i, where i represents an integer greater than or equal to 1 and less than j+1, and j+1 represents the number of the multiple vectors.

[0009] In some embodiments of the present application, the pipeline direction prediction model is used to obtain the i+1th vector of the first pipeline. The starting point coordinates and end point coordinates of i include: the i+1th vector defined based on The starting point coordinates and end point coordinates of i determine the i-th vector i is the local area range corresponding to the discrete space; obtain the first space matrix corresponding to the discrete space, update the first space matrix based on the local area range, and obtain the i+1th vector i corresponding to the second spatial matrix; adjusting the i+1th vector based on the second spatial matrix i, get the adjusted i+1th vector i1; Determine the i+1th vector i and the x+1th vector of the first pipeline The correlation coefficient of x, based on the correlation coefficient and the x+1th vector x updates the adjusted i+1th vector i1, get the updated i+1th vector i1', where x represents an integer in the range [0, i-1]; using the updated i+1th vector The starting point coordinates of i1' update the i+1th vector The starting point coordinates of i, using the updated i+1th vector The end point coordinates of i1' update the i+1th vector The end point coordinates of i.

[0010] In some embodiments of the present application, the i+1th vector defined based on The starting point coordinates and end point coordinates of i determine the i-th vector The local area range corresponding to i in the discrete space includes: based on the preset area range, the i+1th vector The area corresponding to the starting point coordinates and the end point coordinates of i is expanded to obtain the i+1th vector i corresponds to the local area range in the discrete space.

[0011] In some embodiments of the present application, the elements in the first spatial matrix include a first value and a second value; the first spatial matrix corresponding to the discrete space is obtained, and the first spatial matrix is ​​updated based on the local area range to obtain the i+1th vector i corresponds to the second spatial matrix, including: selecting voxels in the discrete space and outside the local area, updating the values ​​of the elements corresponding to the selected voxels in the first spatial matrix to third values, and obtaining the second spatial matrix, wherein the elements in the second spatial matrix include the first value, the second value and the third value.

[0012] In some embodiments of the present application, the adjusting the i+1th vector based on the second spatial matrix i, get the adjusted i+1th vector i1, including: determining the i+1th vector based on the preset first constraint matrix, the second constraint matrix and the second space matrix i adjustment vector; adjust the i+1th vector according to the adjustment vector i, get the adjusted i+1th vector i1.

[0013] In some embodiments of the present application, the determination of the i+1th vector i and the x+1th vector of the first pipeline The correlation coefficient of x, based on the correlation coefficient and the x+1th vector x updates the adjusted i+1th vector i1, get the updated i+1th vector i1', including: determining the correlation coefficient according to the preset query matrix and key matrix; determining the x+1th vector according to the preset value matrix x value vector; take the correlation coefficient as the corresponding x+1th vector The weight of the value vector of x is used to determine the i+1th vector of the first pipeline based on the weight. The weighted sum of the values ​​of all vectors before i, the i+1th vector of the first pipeline All vectors before i including the first vector 0 to the i-th vector (i-1); Update the adjusted i+1th vector according to the weighted sum i1, get the updated i+1th vector i1'.

[0014] In some embodiments of the present application, the method further includes: taking the designed pipeline as the second pipeline, and training the pipeline direction prediction model based on the standard design model of the second pipeline, including: defining multiple vectors corresponding to the second pipeline in the discrete space according to the pipeline position parameters in the design parameters, including the first vector 0 and i+1th vectors i, i represents an integer greater than or equal to 1 and less than j+1, j+1 represents the number of the plurality of vectors; the i-th vector among the plurality of vectors based on the second pipeline (i-1) corresponding coordinates, use the preset model to predict the i+1th vector The coordinate corresponding to i, get the i+1th vector i corresponding prediction vector; according to the standard design model of the second pipeline, determine the i+1th vector The actual vector of i; the i+1th vector based on the second pipeline The difference between the predicted vector and the actual vector corresponding to i is used to determine the loss value of the preset model; and the preset model is trained according to the loss value until the preset model meets the preset requirements.

[0015] In some embodiments of the present application, the pipeline position parameters include the third position coordinates and the fourth position coordinates corresponding to the second pipeline in the discrete space, and the multiple vectors corresponding to the second pipeline are defined in the discrete space according to the pipeline position parameters in the design parameters, including: defining the first vector of the second pipeline based on the third position coordinates and the fourth position coordinates 0, the first vector The starting point coordinate of 0 is the third position coordinate, and the first vector The end point coordinate of 0 is the fourth position coordinate; for two adjacent vectors of the second pipeline, the vector coordinate of the rear vector is determined according to the vector coordinate of the front vector.

[0016] In some embodiments of the present application, for two adjacent vectors of the second pipeline, determining the vector coordinates of the subsequent vector according to the vector coordinates of the preceding vector includes: determining the vector coordinates of the subsequent vector according to the i-th vector The end point coordinates of (i-1) define the i+1th vector The starting coordinates of i, randomly initialize the i+1th vector i’s end point coordinates; randomly initialize multiple implicit parameters; based on the i+1th vector The starting coordinates, end coordinates and the multiple implicit parameters of i define the i+1th vector of the second pipeline i.

[0017] In some embodiments of the present application, training the preset model according to the loss value until the preset model meets the preset requirements includes: updating the model parameters of the preset model according to the loss value until the loss value is less than a preset loss threshold, wherein the model parameters include multiple implicit parameters, a first constraint matrix, a second constraint matrix, a query matrix, a key matrix, and a value matrix.

[0018] The present application also provides an electronic device, which includes a processor and a memory, and the processor is used to implement the vehicle pipeline design method when executing a computer program stored in the memory.

[0019] The present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the vehicle pipeline design method is implemented.

[0020] In the vehicle pipeline design method provided in this application, a discrete space is constructed based on vehicle design parameters to provide a standardized digital environment for pipeline design and facilitate calculation and analysis. In the discrete space, multiple vectors of the first pipeline are defined according to the pipeline position parameters to refine the pipeline design, improve the flexibility of the design and facilitate data management. The preset model is used to predict the remaining vector coordinates based on the first vector coordinate to achieve automated design. The direction path is determined based on all vector coordinate sets, which can intuitively determine the overall layout and direction of the pipeline, laying the foundation for subsequent design. By integrating all the first pipeline direction paths to generate a design model, the layout and direction of all pipelines can be viewed and managed in a unified model, which is convenient for overall design evaluation and optimization. This method uses the existing project vehicle model layout data to train a self-generated model, which can liberate manpower, reduce cross-departmental communication costs, reduce repetitive work, effectively save project development costs, and significantly shorten the vehicle development cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of the vehicle pipeline design method provided in an embodiment of the present application.

[0022] Figure 2 The pipeline direction prediction model provided in the embodiment of the present application is used to obtain the i+1th vector of the first pipeline Flowchart of the starting and ending coordinates of i.

[0023] Figure 3 This is a flowchart of the training pipeline direction prediction model provided in an embodiment of the present application.

[0024] Figure 4 This is an example of a training pipeline direction prediction model provided in the embodiment of the present application. Figure 1 .

[0025] Figure 5 This is an example of a training pipeline direction prediction model provided in the embodiment of the present application. Figure 2 .

[0026] Figure 6 It is a structural diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0027] To facilitate understanding, some illustrations of concepts related to the embodiments of the present application are given for reference.

[0028] It should be noted that, in this application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can mean: A alone, A and B together, and B alone, where A and B can be singular or plural. The terms "first," "second," "third," "fourth," and so on (if any) in the specification, claims, and drawings of this application are used to distinguish similar objects, not to describe a specific order or precedence.

[0029] In vehicle design, piping design is crucial. This includes electrical wiring harnesses, fuel lines, cooling lines, air conditioning lines, and air ducts. These lines act as the vehicle's "nerves" and "blood vessels," ensuring its proper function. For example, electrical wiring harnesses transmit power and signals, connecting numerous electronic devices; fuel lines power the engine; cooling lines maintain engine operating temperature; air conditioning lines regulate the interior environment; and air ducts power several systems.

[0030] With the development of intelligent vehicles, the number of vehicle functions and electronic equipment has increased, making pipeline layout an increasingly complex task. During design, pipelines are distributed throughout the vehicle, and components such as the body often change due to factors such as layout location and process. This makes pipeline layout cumbersome and requires continuous evaluation, modification, and optimization.

[0031] Furthermore, pipeline design involves multiple specialized areas and departments, including vehicle body, electrical, and powertrain. Interdepartmental communication often involves untimely and inaccurate information transfer, resulting in extremely high communication costs. Furthermore, pipelines require repeated revisions throughout the design cycle, often involving repetitive tasks. For example, changes to components require rechecking and adjusting related pipelines, which consumes time and effort and can easily lead to quality issues due to fatigue and negligence.

[0032] Traditional design methods rely on manual experience and computer-aided design tools. Designers manually draw 3D pipeline models, perform interference checks, and optimize them. This cumbersome process, from initial planning to final verification, is time-consuming and labor-intensive. Furthermore, it's difficult to synchronize design data between departments in real time. Changes made by one department go unnoticed by the other, leading to design inconsistencies and duplication of work.

[0033] To solve the above problems, an embodiment of the present application provides a vehicle pipeline design method and electronic equipment. The embodiment of the present application constructs a discrete space based on vehicle design parameters, provides a standardized digital environment for pipeline design, and facilitates calculation and analysis. In the discrete space, multiple vectors of the first pipeline are defined according to the pipeline position parameters to refine the pipeline design, improve the flexibility of the design and facilitate data management. The preset model is used to predict the remaining vector coordinates based on the first vector coordinate to achieve automated design. The direction path is determined based on all vector coordinate sets, and the overall layout and direction of the pipeline can be intuitively determined, laying the foundation for subsequent design. By integrating all the first pipeline direction paths to generate a design model, the layout and direction of all pipelines can be viewed and managed in a unified model, which is convenient for overall design evaluation and optimization. This method uses the existing project vehicle model layout data to train a self-generated model, which can liberate manpower, reduce cross-departmental communication costs, reduce repetitive work, effectively save project development costs, and significantly shorten the vehicle development cycle.

[0034] Figure 1 This is a flow chart of the vehicle pipeline design method provided by the embodiment of the present application, which is applied to electronic equipment (such as Figure 6 According to different requirements, the order of the steps in the flowchart can be changed, and some steps can be omitted.

[0035] Step S101: construct a discrete space according to the design parameters of the vehicle.

[0036] In some embodiments of the present application, vehicle design parameters include vehicle dimensions and data regarding the vehicle's non-pipeline physical components. In one example, the vehicle's dimensions may be 3000 mm × 4000 mm × 2000 mm, indicating the vehicle's length, width, and height. The vehicle's non-pipeline physical components include, but are not limited to, key components such as electronic equipment, the engine, and air conditioning. These non-pipeline physical components occupy specific spatial locations within the vehicle, directly limiting and impacting the layout and routing of pipelines.

[0037] In one example, a vehicle's computer-aided design (CAD) model can be used to obtain vehicle design parameters. A CAD model includes a digital representation of the vehicle design. Key data, such as the vehicle's dimensions and the specific locations of physical components within the CAD model, can be extracted from the CAD model and used as design parameters. This CAD-based parameter acquisition method not only improves data accuracy and consistency but also shortens parameter acquisition time, providing strong support for subsequent design work.

[0038] In some embodiments of the present application, when constructing a discrete space based on vehicle design parameters, a discrete space having a size relationship corresponding to the vehicle dimensions can be constructed, and a plurality of voxels in the discrete space can be determined based on a preset unit size. The coordinates and label of each voxel in the discrete space are determined, with the label indicating whether the corresponding voxel includes data for a physical component other than a pipeline in the vehicle.

[0039] For example, if a vehicle's dimensions are 3000 mm × 4000 mm × 2000 mm, a discrete space of 3000 mm × 4000 mm × 2000 mm can be constructed. The discrete space can be considered a 3000 mm × 4000 mm × 2000 mm rectangular parallelepiped. This ensures that the size and shape of the discrete space match the vehicle's actual physical dimensions, ensuring that the discrete space accurately reflects the vehicle's actual physical range.

[0040] To accurately locate and describe objects in discrete space, a spatial coordinate system can be established. The discrete space's cuboid consists of eight vertices. Any vertex can be selected as the origin, O, of the spatial coordinate system. Three coordinate axes are constructed based on the length, width, and height of the cuboid, yielding the corresponding spatial coordinate system, Oxyz. This coordinate system provides a unified reference framework for subsequently determining the position and coordinates of voxels.

[0041] In one example, the preset unit size can be 1mm×1mm×1mm, and each voxel can be regarded as a 1mm×1mm×1mm cube. In this way, these voxels are like "building blocks" for constructing discrete spaces, and voxels can be used to achieve detailed division of discrete spaces.

[0042] Each voxel consists of eight vertices, and its 3D coordinates (x, y, z) in the spatial coordinate system can be determined. The minimum value (x_min, y_min, z_min) of the 3D coordinates of all vertices of each voxel is selected as the voxel's coordinates. This accurately describes the voxel's position in discrete space, providing precise spatial information for subsequent pipeline design and analysis.

[0043] In one example, a label indicates whether a corresponding voxel includes data for a non-pipeline entity in a vehicle. For example, if the position of a voxel in discrete space corresponds to a position in a CAD model that includes data for a non-pipeline entity, the label for that voxel indicates that the corresponding voxel includes data for the non-pipeline entity. Alternatively, if the position of a voxel in discrete space does not correspond to a position in the CAD model that does not include data for a non-pipeline entity, the label for that voxel indicates that the corresponding voxel does not include data for a non-pipeline entity.

[0044] Based on the above embodiment, by constructing a discrete space corresponding to the vehicle's dimensions and dividing it into multiple voxels based on a preset unit size, and simultaneously determining the label of each voxel and its precise coordinates in the discrete space, clear spatial information and constraints can be provided for subsequent pipeline design and layout. The label clearly indicates whether the voxel contains data for non-pipeline physical components, allowing designers to avoid these physical components when designing pipelines, preventing interference between pipelines and physical components. This improves the rationality and reliability of pipeline design, laying a solid foundation for efficient overall vehicle pipeline design.

[0045] In some embodiments of the present application, after constructing the discrete space, the method further includes: assigning label values ​​to corresponding voxels based on the labels, such that each voxel has a corresponding label value, wherein if the label indicates that the corresponding voxel includes data of a physical component, the label value corresponding to the voxel is a first value (e.g., -1). If the label indicates that the corresponding voxel does not include data of any physical component, the label value corresponding to the voxel is a second value (e.g., 1). Based on the label values ​​corresponding to each voxel, a first spatial matrix corresponding to the discrete space is constructed, where the elements of the first spatial matrix include the first value and the second value.

[0046] In one example, voxel values ​​can be assigned based on the vehicle's CAD data and the labels of individual voxels (x_min, y_min, z_min). If the label indicates that the corresponding voxel contains data for a physical component, for example, if the location corresponding to the voxel has information about a non-pipeline physical component in the CAD model, the voxel can be assigned a label value of -1. Conversely, if the label indicates that the corresponding voxel does not contain data for a physical component, for such blank areas without data, for example, if the location corresponding to the voxel has no information about a non-pipeline physical component in the CAD model, the voxel can be assigned a label value of 1.

[0047] In one example, a first spatial matrix is ​​constructed based on the label value corresponding to the voxel and the coordinates of the voxel in the discrete space. For example, the first spatial matrix can be a spatial vector, and the values ​​in the spatial vector are the first and second values ​​defined above. Here, all voxels can be arranged into a sequence according to the coordinates of the voxels in the discrete space, and the first or second values ​​corresponding to each voxel in the sequence can be arranged into a numerical sequence. A spatial vector corresponding to the vector of the numerical sequence is obtained as the first spatial matrix.

[0048] For example, when all voxels are arranged into a sequence according to their coordinates in discrete space, the following arrangement principle can be followed: (1) First, set y and z to 0, and arrange x from the smallest coordinate value to the largest (from x = 0 to x = 3000). This step ensures that the voxels are preliminarily arranged in the ascending order of the x coordinate when the y and z coordinates are fixed, laying the foundation for the subsequent vector construction. (2) Then set y to 1 and z to 0, and arrange x from the smallest coordinate value to the largest coordinate value (from x = 0 to x = 3000). When the y coordinate increases by 1, the voxels are arranged again in the ascending order of the x coordinate, so that the vector gradually expands in the y direction. (3) Repeat the above (1) and (2), and arrange y from the smallest coordinate value to the largest coordinate value (from y = 0 to y = 4000). By repeating the first two steps, the value of the y coordinate is gradually increased, and all possible permutations and combinations in the y direction are completed, so that the vector completely covers the voxels in the y direction. (4) Repeat (1)-(3) above, arranging the z coordinates from small to large (from z = 0 to z = 2000). After completing the arrangement in the y direction, repeat the previous arrangement process in the ascending order of the z coordinates so that the vectors also completely cover the voxels in the z direction.

[0049] Based on the above embodiment, each voxel has a label value of 1 or -1, so the arranged voxels form a high-dimensional vector arranged by 1 and -1 _n, The dimension of _n is n rows and 1 column, n = 3000×4000×2000. This high-dimensional vector _n stores and expresses voxel information in discrete space in an orderly manner in the form of a matrix, which contains the information of all voxels in the discrete space, including their position and whether they contain solid components. It can provide rich input data for subsequent pipeline design algorithms, helping to achieve more accurate and efficient pipeline design and layout.

[0050] Step S102 : defining a plurality of vectors corresponding to a first pipeline to be designed in the vehicle in a discrete space according to the pipeline position parameters in the design parameters.

[0051] In some embodiments of the present application, to define a first pipeline to be designed in a vehicle in a discrete space, the first pipeline can be decomposed into multiple end-to-end vectors, and the design position of the first pipeline can be determined by defining multiple vectors corresponding to the first pipeline. The multiple vectors corresponding to the first pipeline can be defined based on pipeline position parameters in the design parameters.

[0052] Each pipeline position parameter corresponds to a first pipeline. The pipeline position parameter includes a first position coordinate and a second position coordinate corresponding to the first pipeline in discrete space. These two position coordinates are key to defining the pipeline vector. In one example, the first position coordinate and the second position coordinate are known data and can be obtained by receiving user input.

[0053] The first vector of the first pipeline can be defined based on the first position coordinate and the second position coordinate 0. For example, the first position coordinate is set to the first vector The starting point coordinate of 0, the second position coordinate is set to the first vector The end point coordinates of 0. The definition of this initial vector can lay the foundation for the subsequent vector decomposition of the entire pipeline.

[0054] For two adjacent vectors of the first pipeline, the vector coordinates of the subsequent vector can be determined based on the vector coordinates of the previous vector. In one example, the i-th vector of the first pipeline can be used to determine the vector coordinates of the subsequent vector. The end point coordinate of (i-1) defines the i+1th vector of the first pipeline i, the i+1th vector The starting point coordinates of i are the i-th vector The endpoint coordinates of (i-1), where i represents an integer greater than or equal to 1 and less than j+1, and j+1 represents the number of vectors corresponding to the first pipeline. j+1 can be set based on past experience, for example, j+1 can be the maximum number of vectors included in a historically designed pipeline, and this application does not impose any specific restrictions on this.

[0055] For example, the first pipeline's multiple vectors can be represented as { }. The i-th vector (i-1) has the starting point coordinates and the end point coordinates , the i+1th vector i has the starting point coordinates and the end point coordinates ,in, = .

[0056] In one example, when defining multiple vectors of a first pipeline, the first vector of the first pipeline The starting and ending coordinates of 0 are known data, and the second vector The starting coordinate of 1 is the first vector The end point coordinates of 0 are also known. From the second vector 1, the starting and ending coordinates of each vector are unknown. 1, and the voxel labels and coordinates, initialize the subsequent vectors as follows: (a) with the second vector 1 as the starting point coordinate and set a search radius r . Search Radius r The size of can be determined according to the complexity of the vehicle's interior space and the layout requirements of the pipelines. For example, in areas with relatively open space and relatively flexible pipeline layout, the search radius can be appropriately increased. r In areas with narrow spaces and dense physical components, the search radius needs to be reduced. r , to ensure the efficiency and accuracy of the search process. (b) Within the determined search range, filter according to the label value of the voxel in the discrete space. Only voxels with a label value of 1 (i.e., no non-pipeline entity component data) are retained as candidate voxels. The area where these candidate voxels are located is the potential space where the pipeline can continue to extend. (c) For each candidate voxel, calculate the center coordinates and the second vector 1's Euclidean distance between the starting coordinates d (d) From all candidate voxels, select a candidate voxel that is a preset distance away from the current starting point coordinates, where the preset distance can be set according to actual needs and is not specifically limited in this application. The coordinates of the candidate voxel are used as the second vector 1's end point coordinates, thus completing the second vector 1. (e) According to the above method, the second vector 1's end point coordinates as the third vector 2, repeat steps (a) to (d) to initialize subsequent vectors in turn. In this way, through step-by-step recursion, we can ensure that the starting and ending coordinates of each vector are within the legal spatial area and extend the pipeline along a reasonable path as much as possible.

[0057] Based on the above embodiment, the first vector can be determined based on the pipeline position parameters in the design parameters using the known first and second position coordinates. The starting coordinates of subsequent vectors can then be determined recursively based on the end coordinates of the previous vector. For unknown vector end coordinates, the voxel labels and coordinates are cleverly used to screen candidate voxels with a reasonable search radius and calculate the distance. Appropriate candidate voxels are selected to determine the end coordinates, and subsequent vectors are gradually recursively initialized. This process not only ensures that each vector coordinate is in a legal spatial region, but also allows the pipeline to extend along a reasonable path as much as possible, providing a solid and reliable foundation for subsequent pipeline design, analysis, and optimization, effectively improving the accuracy and rationality of pipeline design, and reducing the complexity and uncertainty in the design process.

[0058] In another example, in addition to determining the second vector In addition to the method of selecting the end point coordinates, a random selection method can also be used. For example, in the above step (d), a candidate voxel can be randomly selected from all candidate voxels, and the coordinates of the selected candidate voxel are determined as the second vector 1's end point coordinates, completing the second vector in this way 1 initialization operation.

[0059] In yet another example, to make the second vector The definition of the end point coordinates is more reference-oriented and reasonable, and the design model of the same type of pipeline corresponding to the first pipeline can also be obtained. By analyzing and learning from these existing mature design models, we can draw on the useful experience and design ideas therein, and provide more diverse references for the vector initialization and subsequent design of the current pipeline. There are no specific restrictions on the endpoint coordinates of 1 and the specific method of obtaining the reference design model. Designers can flexibly select the appropriate method based on actual needs and scenarios.

[0060] In some embodiments of the present application, when defining each vector of the first pipeline, multiple implicit parameters can be added based on the above starting point coordinates and end point coordinates. For example, for the i+1th vector For i, it can be expressed as a vector ,in represents the starting point coordinates, Indicates the end point coordinates. On the basis of adding multiple implicit parameters, such as implicit parameters {a, b, c, d}, then the i+1th vector i can be represented as a column vector of dimension 10×1 (10 rows and 1 column) , T represents transpose.

[0061] Among them, implicit parameters are preset values, such as the values ​​learned through model training, which can be used to help more carefully describe the posture of the vector in space, the trend of direction change, and the potential relationship with the surrounding environment. a It can be used to express the degree of curvature of a vector in a specific direction. b It can be used to reflect the impact of the density characteristics of the local space where the vector resides on its direction. When the pipeline direction prediction model is subsequently used for vector coordinate prediction and pipeline direction analysis, implicit parameters can be used as important feature dimensions in the calculation, helping the model better capture the inherent laws and complex patterns of pipeline direction, thereby improving the accuracy and reliability of the model's predictions, making the predictions more consistent with the actual pipeline layout, and providing more valuable reference information for vehicle pipeline design.

[0062] Step S103 : Based on the coordinate corresponding to the first vector among the multiple vectors, the coordinates corresponding to the remaining vectors of the first pipeline are predicted using a preset pipeline direction prediction model.

[0063] In some embodiments of the present application, the i-th vector of the first pipeline may be based on The starting point coordinates and end point coordinates corresponding to (i-1) are used to obtain the i+1th vector of the first pipeline using the pipeline direction prediction model. The starting point coordinates and the ending point coordinates of i, where i represents an integer greater than or equal to 1 and less than j+1, and j+1 represents the number of multiple vectors.

[0064] In one example, the first vector of the first pipeline can be The starting point coordinates and end point coordinates corresponding to 0 are used to obtain the second vector of the first pipeline using the pipeline direction prediction model. 1's starting and ending coordinates. Then, according to the second vector of the first pipeline 1, and the starting and ending coordinates of the pipeline are obtained using the pipeline direction prediction model to obtain the third vector of the first pipeline. 2, and so on, to obtain the coordinates corresponding to the remaining vectors of the first pipeline.

[0065] In some embodiments of the present application, the pipeline direction prediction model is used to obtain the i+1th vector of the first pipeline. The starting coordinates and ending coordinates of vector i can be expanded according to the preset range to obtain the local area range of the i+1th vector. The first spatial matrix of the discrete space is obtained, and the values ​​of the elements corresponding to the voxels outside the local area range are updated to a third value (for example, 0) to obtain the second spatial matrix. Using the preset first constraint matrix, second constraint matrix and second spatial matrix, the adjustment vector of the i+1th vector is determined. i, and then get the adjusted i+1th vector i1. Considering the i+1th vector and its previous x+1th vector The correlation of x, using the query matrix and key matrix to determine the i + 1th vector i and x+1th vector The correlation coefficient of x, using the value matrix to determine the x+1th vector The value vector of x. Calculate the weighted sum based on the correlation coefficient and the value vector and update the adjusted i+1th vector i1, get the updated i+1th vector i1'. Use vector Update the i+1th vector with the starting and ending coordinates of i1' In one example, the pipeline direction prediction model is used to obtain the i+1th vector of the first pipeline. The starting point coordinates and end point coordinates of i can be found as follows: Figure 2 The embodiment shown.

[0066] Based on the above embodiment, the coordinates of the remaining vectors of the first pipeline can be automatically and gradually predicted using the pipeline direction prediction model based on the coordinates of the first vector. During the prediction process, the local range is first determined by expanding the corresponding area of ​​the coordinates, and the spatial matrix is ​​updated. The constraint matrix is ​​then used to determine the adjustment vector. The correlation between vectors is also fully considered, and the vector is updated by calculating the weighted sum of multiple matrices. This achieves accurate prediction of the coordinates of each vector of the first pipeline, providing a reliable basis for the rational design and layout of the pipeline, and effectively improving the scientific nature and accuracy of pipeline design.

[0067] In some embodiments of the present application, the pipeline direction prediction model is a pre-trained model for predicting pipeline directions. When training the pipeline direction prediction model, a previously designed pipeline can be used as the second pipeline. For example, the second pipeline can include multiple pipelines for multiple vehicle models. Based on the pipeline position parameters (including the third and fourth position coordinates in discrete space) in the design parameters of the second pipeline, multiple vectors corresponding to the second pipeline are defined in discrete space. Similar to the definition of multiple vectors for the first pipeline, the first vector of the second pipeline has its starting and ending points determined by the third and fourth position coordinates. Adjacent vectors determine the coordinates of the subsequent vector based on the coordinates of the preceding vector. The coordinates of the subsequent vector's ending point and multiple implicit parameters can be randomly initialized to complete the definition. Next, based on the coordinates of the i-th vector of the second pipeline, the preset model is used to predict the i+1-th vector. The actual vector is then obtained from the standard design model, and a loss value is determined by the difference between the predicted and actual vectors. Finally, the model parameters of the preset model (including multiple implicit parameters, various constraints, and matrices) are updated based on the loss value until the preset loss value is less than a preset loss threshold, ensuring that the model meets preset requirements. In one example, the pipeline direction prediction model can be trained as follows: Figure 3 The embodiment shown.

[0068] Based on the above embodiment, it is possible to define the vector of the designed second pipeline in discrete space, use the model to make predictions based on the coordinates of the defined vector, and compare the predicted coordinates with the actual coordinates in the standard design model of the designed second pipeline to obtain a loss value. The model parameters are continuously updated and optimized based on the loss value, so that the model can deeply learn the inherent laws and characteristics of the pipeline direction, significantly improving the accuracy of the model prediction, allowing the trained pipeline direction prediction model to more reliably and accurately predict the direction of the entire vehicle pipeline, providing strong support for the efficient and reasonable design of the entire vehicle pipeline, effectively reducing the trial and error costs in the design process, and improving design efficiency and quality.

[0069] Step S104 : determining a path of the first pipeline based on a coordinate set corresponding to all vectors of the first pipeline.

[0070] In some embodiments of the present application, after predicting the coordinate sets corresponding to all vectors of the first pipeline, the coordinate sets can be preprocessed first, for example, to check whether there are abnormal coordinate points (such as coordinate values ​​outside a reasonable range, abnormal distances between coordinate points, etc.). If so, corrections or eliminations are performed to ensure the accuracy and rationality of the coordinate set. Then, according to the order of the vectors in the pipeline, the starting and ending coordinates corresponding to each vector are connected in sequence. For example, starting from the starting coordinate of the first vector, connect to its ending coordinate, and this ending coordinate is used as the starting coordinate of the next vector, and then connected to the ending coordinate of the next vector, until all vectors are connected. In this way, the discrete vector coordinates are connected into a continuous curve, which is the preliminarily determined direction path of the first pipeline.

[0071] In one example, the initially determined path can be further smoothed. Considering that actual pipeline paths are typically smooth and avoid sharp turns, mathematical methods such as spline interpolation can be used to smoothly optimize the connected path. This ensures that the path of the first pipeline better meets actual engineering requirements and design specifications, providing a reliable basis for subsequent vehicle pipeline layout and design.

[0072] Based on the above embodiment, by determining the direction of the first pipeline based on the coordinate set corresponding to all vectors of the first pipeline and taking a series of optimization measures, the scientificity and rationality of the pipeline design can be significantly improved, providing a solid basis for the pipeline layout of the entire vehicle, helping to improve the overall quality and efficiency of the vehicle design, and reducing subsequent production and maintenance costs.

[0073] Step S105 : generating a pipeline design model of the vehicle based on the paths of all first pipelines of the vehicle.

[0074] In some embodiments of the present application, after obtaining the paths of all first pipelines in a vehicle, these paths can be integrated and analyzed. Considering the limited space within a vehicle and the potential complex interactions between pipelines, conflicts between different pipeline paths can be checked. For example, if pipelines are too close together, they may cause installation difficulties or interfere with each other during operation. If conflicts are found, the conflicting pipeline paths can be adjusted and optimized based on preset priority rules (e.g., priority is determined based on factors such as pipeline importance and diameter) to ensure that the pipelines are properly arranged within the vehicle and do not interfere with each other. Subsequently, the optimized paths of each first pipeline are converted into a three-dimensional model using professional three-dimensional modeling software. During the modeling process, corresponding attributes such as material, color, and texture can be assigned based on the actual pipeline type (e.g., hydraulic pipeline, electrical pipeline, etc.), making the generated three-dimensional model more realistic and easier for designers to intuitively view and evaluate.

[0075] In one example, to facilitate subsequent design modifications and collaborative work, a detailed attribute information library can be established for the generated pipeline design model, recording key information such as each pipeline's number, name, specifications, and route coordinates. This attribute information is associated with the 3D model, allowing designers to quickly understand the pipeline's details by querying the attribute information, improving design efficiency.

[0076] In one example, the generated pipeline design model can be checked, for example, to check the model's geometric accuracy, topological structure rationality, etc., to ensure that the model can accurately reflect the actual design of the vehicle pipeline and provide a reliable design basis for subsequent vehicle production and manufacturing.

[0077] Based on the above embodiments, a leap from an abstract design path for pipelines to an intuitive three-dimensional model has been achieved. By integrating, analyzing, and optimizing the pipeline path, the conflict problem between pipelines is effectively resolved, ensuring the rational use of the vehicle's interior space and the stable operation of the pipelines. By giving the model real attributes with the help of three-dimensional modeling software, designers can grasp the characteristics of the pipelines intuitively and accurately, and improve the efficiency and accuracy of design evaluation. The establishment of a detailed attribute information library provides great convenience for design modifications and collaborative work, and designers can quickly obtain key information about the pipelines. Comprehensive model verification checks ensure the geometric accuracy and topological structure rationality of the model, so that it can accurately reflect the actual design situation, provide a highly reliable design blueprint for vehicle production and manufacturing, and effectively promote the efficient and accurate development of the vehicle design and manufacturing process.

[0078] The vehicle pipeline design method provided in the embodiment of the present application constructs a discrete space based on vehicle design parameters, provides a standardized digital environment for pipeline design, and facilitates calculation and analysis. In the discrete space, multiple vectors of the first pipeline are defined according to the pipeline position parameters to refine the pipeline design, improve the flexibility of the design and facilitate data management. The preset model is used to predict the remaining vector coordinates based on the first vector coordinate to achieve automated design. The direction path is determined based on all vector coordinate sets, and the overall layout and direction of the pipeline can be intuitively determined, laying the foundation for subsequent design. By integrating all the first pipeline direction paths to generate a design model, the layout and direction of all pipelines can be viewed and managed in a unified model, which is convenient for overall design evaluation and optimization. This method uses the existing project vehicle model layout data to train a self-generated model, which can liberate manpower, reduce cross-departmental communication costs, reduce repetitive work, effectively save project development costs, and significantly shorten the vehicle development cycle.

[0079] Figure 2 The pipeline direction prediction model provided in the embodiment of the present application is used to obtain the i+1th vector of the first pipeline The flowchart of the starting point coordinates and the end point coordinates of i includes the following steps.

[0080] Step S201, based on the defined i+1th vector The starting and ending coordinates of i determine the i-th vector The local area range corresponding to i in the discrete space.

[0081] In some embodiments of the present application, the i-th vector is determined When i is in the local area corresponding to the discrete space, the i+1th vector can be calculated based on the preset area. The area corresponding to the starting point coordinates and the end point coordinates of i is expanded to obtain the i+1th vector The local area range corresponding to i in the discrete space.

[0082] In one example, the i+1th vector i can be represented as a vector ,in, represents the starting point coordinates, Indicates the end point coordinates. You can determine min( )= ,min( = ,min( = ,min() represents the minimum value function. Determine max( )= ,max( = ,max( = , max() represents the maximum value function. Get the i+1th vector The area range corresponding to the starting point coordinate and the end point coordinate of i[ ].

[0083] In one example, based on a preset area range, the i+1th vector The area corresponding to the starting point coordinates and the end point coordinates of i can be expanded as follows: Any coordinate in the range of the preset area is reduced in the respective direction, for example, If we reduce any coordinate in by 15, we get .Will Any coordinate in the increases the preset area range in the respective direction, for example, Increment any coordinate in by 15, and we get . Among them, the preset area range can be set according to actual needs, and this application does not impose any specific restrictions on this.

[0084] After the above expansion, we get the i+1th vector i corresponds to the local area range in the discrete space. For example, [ ].

[0085] Based on the above embodiment, the coordinates of the starting and ending points of the i+1th vector are used as the basis. Extreme values ​​are obtained through coordinate parameter comparison and then the range of extreme values ​​is rationally expanded. This fully considers the constraint sufficiency requirements in pipeline design and avoids omissions in subsequent calculations or designs due to an overly narrow range. The resulting local discrete space, represented by a high-dimensional vector, provides a clear and appropriate operating area for subsequent refined spatial constraint calculations. This helps to improve the rationality of local space utilization in vehicle pipeline design, reduce the risk of interference between pipelines and other components, and improve the quality and efficiency of the overall design.

[0086] Step S202: Obtain the first spatial matrix corresponding to the discrete space, update the first spatial matrix based on the local area range, and obtain the i+1th vector The second spatial matrix corresponding to i.

[0087] In some embodiments of the present application, the elements in the first spatial matrix include a first value and a second value. For example, referring to the embodiment in step S101, the first spatial matrix _n includes a first value -1 and a second value 1. The first value indicates that the corresponding voxel includes data of a physical component, and the second value indicates that the corresponding voxel does not include data of a physical component.

[0088] In some embodiments of the present application, the first space matrix corresponding to the discrete space is obtained, and the first space matrix is ​​updated based on the local area range to obtain the i+1th vector The second spatial matrix corresponding to i includes: selecting voxels in the discrete space and located outside the local area, updating the values ​​of the elements corresponding to the selected voxels in the first spatial matrix to a third value (for example, 0), and obtaining the second spatial matrix, wherein the elements in the second spatial matrix include the first value, the second value, and the third value.

[0089] In one example, the coordinate position of each voxel in the discrete space is compared with the local area range determined above, and the voxels whose coordinates are not within the local area range are screened out. The value of the element corresponding to the selected voxel in the first spatial matrix is ​​updated to a third value, such as 0, to obtain the second spatial matrix _i. The elements in the second spatial matrix include a first value of -1, a second value of 1, and a third value of 0. The third value is used to identify voxels outside the local area. The dimension of _i is the same as the first spatial matrix The dimensions of _n are consistent, with n rows and 1 column, n = 3000×4000×2000.

[0090] Based on the above embodiment, the second spatial matrix can more accurately reflect the association between each voxel and the physical component in the discrete space after considering the local area range, providing more accurate data support for subsequent pipeline design and spatial analysis.

[0091] Step S203: Adjust the i+1th vector based on the second spatial matrix i, get the adjusted i+1th vector i1.

[0092] In some embodiments of the present application, based on the preset first constraint matrix, the second constraint matrix and the second space matrix, the i+1th vector is determined Adjustment vector of i. Adjust the i+1th vector according to the adjustment vector i, get the adjusted i+1th vector i1.

[0093] In one example, the first constraint matrix can be a spatial constraint matrix C1 with 10 rows and 5 columns, and the second constraint matrix can be a spatial constraint matrix C2 with 5 rows and n columns, where n = 3000×4000×2000. The first constraint matrix and the second constraint matrix can be preset matrices, such as matrices learned through model training. Determine the i+1th vector The formula used for the adjustment vector of i can be exemplarily expressed as: _i, where Represents the i+1th vector The adjustment vector corresponding to i has a dimension of 10 rows and 1 column.

[0094] In one example, the i+1th vector is adjusted according to the adjustment vector The formula used can be exemplified as: i1= i. i1 represents the adjusted i+1th vector i1.

[0095] Based on the above embodiment, the direction of the pipeline vector in the local discrete space is adjusted through the constraint matrix and equation to ensure that it meets the spatial constraint conditions. Through a series of matrix operations, the complex local discrete space information is converted into _i is converted into a pipeline vector The specific adjustment amount of i , thereby achieving precise control of the pipeline direction under the space constraints of the entire vehicle.

[0096] Step S204, determine the i+1th vector i and the x+1th vector of the first pipeline The correlation coefficient of x, according to the correlation coefficient and the x+1th vector x updates the adjusted i+1th vector i1, get the updated i+1th vector i1'.

[0097] In some embodiments of the present application, x represents an integer in the range [0, i-1]. Determine the i+1th vector i and the x+1th vector of the first pipeline The correlation coefficient of x, according to the correlation coefficient and the x+1th vector x updates the adjusted i+1th vector i1, get the updated i+1th vector i1', including: determining the correlation coefficient according to the preset query matrix and key matrix. Determining the x+1th vector according to the preset value matrix The value vector of x. The correlation coefficient is taken as the corresponding x+1th vector The weight of the value vector of x, based on which the i+1th vector of the first pipeline is determined The weighted sum of the value vectors of all vectors before i, the i+1th vector of the first pipeline All vectors before i including the first vector 0 to the i-th vector (i-1). The i+1th vector adjusted according to the weighted sum update i1, get the updated i+1th vector i1'.

[0098] In one example, the query matrix can be a 5-row 10-column spatial constraint matrix Q, which is used to transform the i+1th vector i is mapped to a low-dimensional space, such as 5 rows and 1 column, for subsequent correlation calculations. The key matrix can be a 5-row 10-column spatial constraint matrix K, which is used to map the x+1th vector x is mapped to a low-dimensional space, such as 5 rows and 1 column, for subsequent correlation calculations.

[0099] In one example, the correlation coefficient is determined based on a preset query matrix and a key matrix, and the formula used can be exemplarily expressed as follows: i, x, , where Q represents the query matrix, Represents the i+1th vector The query value of i. represents the bond matrix, Represents the x+1th vector The key value of x. express and The dot product is a constant used to represent the i+1th vector i and x+1th vector The correlation coefficient of x.

[0100] In one example, the softmax function can be used to compute the i+1th vector i and x+1th vector The correlation coefficient of x is normalized, and the formula used can be exemplarily expressed as: , in, Represents the i+1th vector i and x+1th vector The normalized correlation coefficient of x, ∈[0,1].

[0101] In one example, the value matrix can be a 10-row 10-column spatial constraint matrix V, which is used to determine the x+1th vector The value vector of x can be represented by the formula: x, express A vector of x values.

[0102] In one example, determine the i+1th vector of the first pipeline The weighted sum of the value vectors of all vectors before i , according to the weighted sum update adjusted i+1th vector i1, get the updated i+1th vector i1'= i1+ .

[0103] Based on the above embodiment, the preset query matrix, key matrix and value matrix are used to determine the i+1th vector i and the x+1th vector in the first pipeline The correlation coefficient of x (x is an integer in the range [0, i -1]). First, the vector is mapped to a low-dimensional space with the help of the query matrix and the key matrix and the dot product is calculated to obtain the preliminary correlation coefficient. Then, it is normalized by the softmax function so that the coefficient is in the range [0,1] to ensure the rationality and comparability of the value. At the same time, the x+1th vector is determined by the value matrix The value vector of x. Using the normalized correlation coefficient as the weight, calculate the i+1th vector of the first pipeline All vectors before i ( 0 to (i-1)) weighted sum of the value vectors. Update the adjusted i+1th vector based on this weighted sum i1, get the updated vector This process fully considers the relationship between pipeline vectors, so that the updated vectors can better adapt to the overall pipeline layout, improve the rationality and coordination of pipeline design, and reduce the mismatch between local design and the overall one.

[0104] Step S205: using the updated i+1th vector The starting point coordinates of i1' update the i+1th vector The starting point coordinates of i, using the updated i+1th vector The end point coordinates of i1' update the i+1th vector The end point coordinates of i.

[0105] In some embodiments of the present application, the updated i+1th vector is used The start and end coordinates of i1' serve as accurate coordinate information, ensuring that subsequent pipeline design and spatial analysis operations based on this vector are based on accurate and up-to-date vector information. This not only improves the accuracy and consistency of pipeline design but also provides a reliable data foundation for the optimization and coordination of the entire pipeline system, helping to reduce design errors and conflicts caused by inconsistent vector information and improving overall design efficiency and quality.

[0106] Figure 3 This is a flowchart of the training pipeline direction prediction model provided in an embodiment of the present application, which includes the following steps.

[0107] Step S301 : defining a plurality of vectors corresponding to the second pipeline in a discrete space according to the pipeline position parameters in the design parameters.

[0108] In some embodiments of the present application, the plurality of vectors corresponding to the second pipeline include the first vector 0 and i+1th vectors i, i represents an integer greater than or equal to 1 and less than j+1, and j+1 represents the number of multiple vectors. The pipeline position parameters include the third position coordinates and the fourth position coordinates corresponding to the second pipeline in the discrete space. According to the pipeline position parameters in the design parameters, multiple vectors corresponding to the second pipeline are defined in the discrete space, including: defining the first vector of the second pipeline based on the third position coordinates and the fourth position coordinates 0, the first vector The starting point coordinate of 0 is the third position coordinate, the first vector The end point coordinate of 0 is the fourth position coordinate. For two adjacent vectors of the second pipeline, the vector coordinate of the rear vector is determined according to the vector coordinate of the front vector.

[0109] In some embodiments of the present application, for two adjacent vectors of the second pipeline, the vector coordinates of the subsequent vector are determined according to the vector coordinates of the preceding vector, including: according to the i-th vector The end point coordinates of (i-1) define the i+1th vector The starting coordinate of i, randomly initialize the i+1th vector The end point coordinates of i. Randomly initialize multiple implicit parameters. Based on the i+1th vector The starting coordinates, end coordinates and multiple implicit parameters of i define the i+1th vector of the second pipeline i.

[0110] In one example, the i+1th vector is randomly initialized The method of determining the end point coordinates of i can refer to the embodiment of initializing multiple vectors of the first pipeline in step S102, and will not be described again. i is a column vector with 10 rows and 1 column, where the first 6 rows include the i+1th vector The starting and ending coordinates of i are shown in the following six lines. The last six lines contain multiple implicit parameters. Compared to the known implicit parameters learned through model training used to define the multiple vectors of the first pipeline, during model training, multiple implicit parameters are randomly initialized to define the multiple vectors of the second pipeline, and these implicit parameters are continuously optimized during subsequent model training.

[0111] Step S302: based on the i-th vector among the plurality of vectors in the second pipeline (i-1) corresponding coordinates, using the preset model to predict the i+1th vector The coordinate corresponding to i, get the i+1th vector The prediction vector corresponding to i.

[0112] In some embodiments of the present application, the specific implementation of step S302 can refer to steps S103 and Figure 2 In the embodiment of the process shown, the i+1th vector The prediction vector corresponding to i can be, for example, the i+1th vector i1' (referring to the updated i+1th vector of the first pipeline) i1'). Compared with the known first constraint matrix, second constraint matrix, query matrix, key matrix, and value matrix learned through model training in step S103, the randomly initialized first constraint matrix, second constraint matrix, query matrix, key matrix, and value matrix are used during model training to predict the i+1th vector The prediction vector corresponding to i is continuously optimized in the subsequent model training process for the first constraint matrix, the second constraint matrix, the query matrix, the key matrix, and the value matrix.

[0113] In one example, the preset model can be a combination of one or more of the following models: support vector machine model, decision tree and random forest model, multi-layer perceptron model, recurrent neural network model, and convolutional neural network model. This application does not impose specific restrictions on the model structure of the preset model.

[0114] Step S303: Determine the i+1th vector according to the standard design model of the second pipeline. The actual vector of i.

[0115] In one example, the i+1th vector The actual vector of i can be expressed as ia.

[0116] Step S304: based on the i+1th vector of the second pipeline The difference between the predicted vector and the actual vector corresponding to i determines the loss value of the preset model.

[0117] In some embodiments of the present application, a preset model is trained according to the loss value until the preset model meets the preset requirements, including: updating the model parameters of the preset model according to the loss value until the preset loss value is less than a preset loss threshold, wherein the model parameters include multiple implicit parameters, a first constraint matrix, a second constraint matrix, a query matrix, a key matrix, and a value matrix.

[0118] In one example, for the i+1th vector of the second pipeline i, get its corresponding prediction vector i1' and the actual vector The difference of ia . Based on the preset loss function , and get the loss value . The difference between the j+1 predicted vectors of the second pipeline predicted by the model and the j+1 actual pipeline direction of the second pipeline is used to indicate the position difference. The average loss value of all the second pipelines in the model training set is determined. The formula used can be exemplarily expressed as: , Where z represents the total number of the second pipeline, the loss value The larger it is, the lower the predictive ability of the model.

[0119] Step S305: train the preset model according to the loss value until the preset model meets the preset requirements.

[0120] In some embodiments of the present application, the model parameters of the preset model are updated according to the loss value until the loss value is less than a preset loss threshold, wherein the model parameters include multiple implicit parameters, a first constraint matrix, a second constraint matrix, a query matrix, a key matrix, and a value matrix.

[0121] In one example, based on the optimization function To update the model parameters of the preset model, W Represents the pipeline matrix involved in the pre-training prediction model P , the first constraint matrix C 1. The column vector constructed from all parameters in the second constraint matrix C2, the query matrix Q, the key matrix K, and the value matrix V. The pipeline matrix P represents a matrix composed of multiple vectors of the defined second pipeline, with a dimension of 10 rows and j+1 columns. express The gradient function , construct a series of gradient values ​​as a column vector , whose dimension is consistent with the dimension of the model parameter column vector W to ensure correct parameter update operations.

[0122] In one example, the model parameters can be updated using the negative gradient value as the convergence step (the negative gradient value defines the learning rate), that is, the gradient vector is subtracted from the current parameter vector W. , and obtain the optimized model parameter vector Wop. In this way, the model parameters are adjusted in the direction of reducing the loss value until the loss value is less than the preset loss threshold, so as to gradually improve the prediction accuracy of the model and obtain a model that meets the preset requirements. At this time, all matrices in the model (including the pipeline matrix P , spatial constraint matrix C 1. Spatial constraint matrix C 2. Inquiry Matrix Q , key matrix K , value matrix V ) have been adjusted to a more reasonable state, which can accurately predict and give the pipeline direction based on the known initial input vector parameters and vehicle space constraints.

[0123] Based on the above embodiment, the loss value is determined by calculating the difference between the predicted coordinates and the actual coordinates, and the model parameters are updated based on the loss value, enabling the model to continuously adjust its parameters based on feedback from the training data, thereby improving prediction accuracy. Through iterative training and judgment based on preset requirements, the model is ensured to gradually converge during the training process, ultimately obtaining a prediction model that meets actual needs. This model can accurately predict the direction of the pipeline given initial conditions and spatial constraints, providing a reliable basis for pipeline design and layout, effectively reducing design errors and costs, and improving design efficiency and quality.

[0124] Figure 4 This is an example of a training pipeline direction prediction model provided in the embodiment of the present application. Figure 1 , Figure 5 This is an example of a training pipeline direction prediction model provided in the embodiment of the present application. Figure 2 .in, Figure 4 yes Figure 5 The example diagram of the red box area is shown. The pipeline direction prediction model training includes: Step 1 (refer to Figure 4 、 5 ): 1. Select the first column of pipeline matrix P Fill in the first six parameters (x1, y1, z1, x2, y2, z2) (pipeline position parameters) with the known starting and ending coordinates; 2. Each column of the pipeline matrix P ( ) respectively input the local discrete space selection module, and perform the local discrete space constraints, and output ; 3. Series constrained vector ( ) Input pipeline direction prediction generation module and output series of assigned prediction vectors ( ). Step 2 (reference Figure 4 、5 ): 1. Prediction vector after series assignment ( ) respectively input the local discrete space selection module, and perform the local discrete space constraints, and output ; 2. Series constrained vector ( ) Input pipeline direction prediction generation module and output series of assigned prediction vectors ( ); 3. (reference Figure 5 ) Output pipeline matrix P second column and a second vector with the real pipeline At the same time, input the cost function and obtain the cost value C1; 4. Repeat steps 1 and 2 in points 1 and 2 to obtain the third column of the output pipeline matrix P and a third vector with the real pipeline At the same time, input the cost function to obtain the cost value C2; ​​5. Repeat the above steps until all vectors of the real single pipeline in the training parameters are used up, obtain the series cost values ​​C1, C2, ..., Cj, and input the average cost function; 6. Evaluate the cost function and input the optimization function. After updating the model, repeat the above steps until the optimization converges.

[0125] Based on the above embodiment, in step 1, from the first column of the pipeline matrix P p 0. Select the first six parameters and fill in the known start and end coordinates. Then, input each column vector of the matrix into the local discrete space selection module for constraint. Then, input the constrained vector into the pipeline direction prediction generation module to obtain the assigned prediction vector. Next, proceed to step two. The prediction vector obtained in step one is again input into the local discrete space selection module for constraint. Then, input into the prediction generation module to obtain a new assigned prediction vector. The second column of the pipeline matrix P is output and combined with the real pipeline vector into the cost function to obtain a cost value. This process is repeated until all real single pipeline vectors in the training parameters are exhausted. A series of cost values ​​are obtained and input into the average cost function. Finally, the model is updated by evaluating the cost function and inputting it into the optimization function. The above steps are repeated until the optimization converges. This series of operations enables the model to continuously adjust and optimize its parameters based on the training data, gradually improving the accuracy of pipeline direction prediction. Ultimately, a high-quality model that can reliably predict pipeline direction is obtained, providing strong support for pipeline design and layout, effectively improving design efficiency and quality.

[0126] Figure 6is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The above-mentioned vehicle pipeline design method is applied to an electronic device 120, which can specifically be an onboard device, controller, etc. of a vehicle. The electronic device 120 can also be a cloud server, physical server, etc. that is communicatively connected to the vehicle. The electronic device 120 may include: a memory 1201, a processor 1202, and a computer program stored in the memory 1201 and executable on the processor 1202. When the processor 1202 executes the program, the vehicle pipeline design method provided in the above-mentioned embodiment is implemented.

[0127] Furthermore, the electronic device further includes: a communication interface 1203 for communication between the memory 1201 and the processor 1202 .

[0128] The memory 1201 is used to store computer programs that can be run on the processor 1202 .

[0129] The memory 1201 may include a high-speed RAM memory, and may also include a non-volatile memory (non volatile memory), such as at least one disk storage.

[0130] If the memory 1201, processor 1202, and communication interface 1203 are implemented independently, the communication interface 1203, memory 1201, and processor 1202 can be interconnected via a bus to enable communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus. Buses can be categorized as address buses, data buses, and control buses.

[0131] Optionally, in a specific implementation, if the memory 1201, the processor 1202 and the communication interface 1203 are integrated on a chip, the memory 1201, the processor 1202 and the communication interface 1203 can communicate with each other through an internal interface.

[0132] The processor 1202 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0133] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, which implements the above vehicle pipeline design method when executed by a processor.

[0134] This embodiment also provides a computer program product, including a computer program, which is executed to implement the vehicle pipeline design method of the above embodiment.

[0135] Those skilled in the art will appreciate that all or part of the steps in the method for implementing the above-mentioned embodiment can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiment.

[0136] In addition, the functional units in the various embodiments of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into a module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0137] The storage medium mentioned above may be a read-only memory, a magnetic disk, or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present application. Persons skilled in the art may make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present application.

Claims

1. A vehicle pipeline design method, characterized in that: The method comprises: Construct a discrete space based on the vehicle's design parameters; defining, in the discrete space, a plurality of vectors corresponding to the first pipeline to be designed in the vehicle according to the pipeline position parameters in the design parameters, wherein each pipeline position parameter corresponds to one first pipeline; Based on the coordinates corresponding to the first vector among the multiple vectors, using a preset pipeline direction prediction model to predict the coordinates corresponding to the remaining vectors of the first pipeline; Determining a path of the first pipeline based on a coordinate set corresponding to all vectors of the first pipeline; A pipeline design model of the vehicle is generated based on the paths of all first pipelines of the vehicle.

2. The vehicle pipeline design method according to claim 1, characterized in that: The design parameters include the dimensions of the vehicle and data of non-pipeline physical components of the vehicle. Constructing the discrete space according to the design parameters of the vehicle includes: Constructing a discrete space having a size relationship corresponding to the size, and determining a plurality of voxels of the discrete space based on a preset unit size; The coordinates and a label of each voxel in the discrete space are determined, where the label is used to indicate whether the corresponding voxel includes data of a non-pipeline physical component in the vehicle.

3. The vehicle pipeline design method according to claim 2, characterized in that: After constructing the discrete space, the method further includes: Assigning a label value to the corresponding voxel according to the label, so that each voxel has a corresponding label value, wherein if the label indicates that the corresponding voxel includes the data of the physical component, the label value corresponding to the voxel is a first value; if the label indicates that the corresponding voxel does not include any data of the physical component, the label value corresponding to the voxel is a second value; Based on the label value corresponding to each voxel, a first spatial matrix corresponding to the discrete space is constructed, where elements in the first spatial matrix include the first numerical value and the second numerical value.

4. The vehicle pipeline design method according to claim 1, characterized in that: The pipeline position parameters include first position coordinates and second position coordinates corresponding to the first pipeline in the discrete space. The defining, in the discrete space, a plurality of vectors corresponding to the first pipeline to be designed in the vehicle according to the pipeline position parameters in the design parameters includes: Based on the first position coordinate and the second position coordinate, define the first vector of the first pipeline 0, the first position coordinate is the first vector 0, the second position coordinate is the first vector 0 is the end point coordinate; For two adjacent vectors of the first pipeline, determining the vector coordinates of the subsequent vector according to the vector coordinates of the preceding vector includes: determining the vector coordinates of the subsequent vector according to the i-th vector of the first pipeline The end point coordinates of (i-1) define the i+1th vector of the first pipeline i, the i+1th vector The starting point coordinates of i are the i-th vector (i-1), where i represents an integer greater than or equal to 1 and less than j+1, and j+1 represents the number of vectors corresponding to the first pipeline.

5. The vehicle pipeline design method according to claim 1, characterized in that: The method of predicting coordinates corresponding to remaining vectors of the first pipeline based on the coordinates corresponding to the first vector among the multiple vectors using a preset pipeline direction prediction model includes: Based on the i-th vector of the first pipeline (i-1) corresponding starting point coordinates and end point coordinates, use the pipeline direction prediction model to obtain the i+1th vector of the first pipeline The starting point coordinates and the ending point coordinates of i, where i represents an integer greater than or equal to 1 and less than j+1, and j+1 represents the number of the multiple vectors.

6. The vehicle pipeline design method according to claim 5, characterized in that: The pipeline direction prediction model is used to obtain the i+1th vector of the first pipeline The starting and ending coordinates of i include: Based on the defined i+1th vector The starting point coordinates and end point coordinates of i determine the i-th vector i corresponds to the local area range in the discrete space; Obtain the first spatial matrix corresponding to the discrete space, update the first spatial matrix based on the local area range, and obtain the i+1th vector The second space matrix corresponding to i; Adjust the i+1th vector based on the second spatial matrix i, get the adjusted i+1th vector i1; Determine the i+1th vector i and the x+1th vector of the first pipeline The correlation coefficient of x, based on the correlation coefficient and the x+1th vector x updates the adjusted i+1th vector i1, get the updated i+1th vector i1', where x represents an integer in the range [0, i-1]; Using the updated i+1th vector The starting point coordinates of i1' update the i+1th vector The starting point coordinates of i, using the updated i+1th vector The end point coordinates of i1' update the i+1th vector The end point coordinates of i.

7. The vehicle pipeline design method according to claim 6, characterized in that: The i+1th vector defined based on The starting point coordinates and end point coordinates of i determine the i-th vector The local area range corresponding to i in the discrete space includes: Based on the preset area range, the i+1th vector The area corresponding to the starting point coordinates and the end point coordinates of i is expanded to obtain the i+1th vector i corresponds to the local area range in the discrete space.

8. The vehicle pipeline design method according to claim 6, characterized in that: The elements in the first space matrix include a first value and a second value; the first space matrix corresponding to the discrete space is obtained, and the first space matrix is ​​updated based on the local area range to obtain the i+1th vector The second space matrix corresponding to i includes: Select a voxel in the discrete space and outside the local area, update the value of the element corresponding to the selected voxel in the first spatial matrix to a third value, and obtain the second spatial matrix, wherein the elements in the second spatial matrix include the first value, the second value and the third value.

9. The vehicle pipeline design method according to claim 6, characterized in that: The adjusting the i+1th vector based on the second spatial matrix i, get the adjusted i+1th vector i1, including: Based on the preset first constraint matrix, the second constraint matrix and the second space matrix, the i+1th vector is determined The adjustment vector of i; Adjust the i+1th vector according to the adjustment vector i, get the adjusted i+1th vector i1.

10. The vehicle pipeline design method according to claim 6, characterized in that: The determination of the i+1th vector i and the x+1th vector of the first pipeline The correlation coefficient of x, based on the correlation coefficient and the x+1th vector x updates the adjusted i+1th vector i1, get the updated i+1th vector i1', including: Determining the correlation coefficient according to a preset query matrix and a key matrix; According to the preset value matrix, determine the x+1th vector vector of x values; The correlation coefficient is used as the corresponding x+1th vector The weight of the value vector of x is used to determine the i+1th vector of the first pipeline based on the weight. The weighted sum of the values ​​of all vectors before i, the i+1th vector of the first pipeline All vectors before i including the first vector 0 to the i-th vector ( i-1); Update the adjusted i+1th vector according to the weighted sum i1, get the updated i+1th vector i1'.

11. The vehicle pipeline design method according to claim 1, characterized in that: The method further includes: using the designed pipeline as a second pipeline, and training the pipeline direction prediction model based on a standard design model of the second pipeline, including: According to the pipeline position parameters in the design parameters, multiple vectors corresponding to the second pipeline are defined in the discrete space, including the first vector 0 and i+1th vectors i, i represents an integer greater than or equal to 1 and less than j+1, and j+1 represents the number of the plurality of vectors; Based on the i-th vector among the multiple vectors of the second pipeline (i-1) corresponding coordinates, use the preset model to predict the i+1th vector The coordinate corresponding to i, get the i+1th vector The prediction vector corresponding to i; According to the standard design model of the second pipeline, the i+1th vector is determined The actual vector of i; Based on the i+1th vector of the second pipeline The difference between the predicted vector and the actual vector corresponding to i is used to determine the loss value of the preset model; The preset model is trained according to the loss value until the preset model meets the preset requirements.

12. The vehicle pipeline design method according to claim 11, characterized in that: The pipeline position parameters include third position coordinates and fourth position coordinates corresponding to the second pipeline in the discrete space. The defining of multiple vectors corresponding to the second pipeline in the discrete space according to the pipeline position parameters in the design parameters includes: Define a first vector of the second pipeline based on the third position coordinate and the fourth position coordinate 0, the first vector The starting point coordinate of 0 is the third position coordinate, and the first vector The end point coordinate of 0 is the fourth position coordinate; For two adjacent vectors of the second pipeline, the vector coordinates of the subsequent vector are determined according to the vector coordinates of the preceding vector.

13. The vehicle pipeline design method according to claim 12, characterized in that: For two adjacent vectors of the second pipeline, determining the vector coordinates of the subsequent vector according to the vector coordinates of the preceding vector includes: According to the i-th vector The end point coordinates of (i-1) define the i+1th vector The starting coordinates of i, randomly initialize the i+1th vector The end point coordinates of i; Randomly initialize multiple implicit parameters; Based on the i+1th vector The starting coordinates, end coordinates and the multiple implicit parameters of i define the i+1th vector of the second pipeline i.

14. The vehicle pipeline design method according to claim 12, characterized in that: The step of training the preset model according to the loss value until the preset model meets preset requirements includes: The model parameters of the preset model are updated according to the loss value until the loss value is less than a preset loss threshold, wherein the model parameters include multiple implicit parameters, a first constraint matrix, a second constraint matrix, a query matrix, a key matrix, and a value matrix.

15. An electronic device, characterized in that: The electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the vehicle pipeline design method according to any one of claims 1 to 14 is implemented.

Images