Commercial vehicle product development method based on logistics scenario library and computer readable storage medium
By constructing a value-scenario mapping matrix and a functional module decoupling model, the static nature of value requirements and scenario matching and the coupling of functional modules in commercial vehicle product development are solved, enabling dynamic customization and efficient development of commercial vehicle products.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHERY COMMERCIAL VEHICLE (ANHUI) CO LTD
- Filing Date
- 2026-01-29
- Publication Date
- 2026-05-29
AI Technical Summary
Traditional commercial vehicle product development suffers from static issues in matching value requirements with scenarios and strong coupling in functional module design, resulting in low customization efficiency and existing tools being unable to adapt to the needs of complex logistics scenarios.
We employ text embedding based on the BERT model and the GNN-LSTM model for semantic association mining, and combine K-means clustering and MLP model to decouple functional modules. We achieve dynamic and intelligent matching through a two-layer heterogeneous network of value-scenario-function.
It enables dynamic recommendations between enterprise value goals and specific logistics scenarios, reduces redundant design, and improves the customization efficiency and matching accuracy of commercial vehicle product development.
Smart Images
Figure CN122115058A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of intelligent connected vehicle technology. More specifically, this invention relates to a commercial vehicle product development method based on a logistics scenario library and a computer-readable storage medium. Background Technology
[0002] In the field of early-stage customized development and logistics scenario adaptation for commercial vehicle products, the main limitations of current technology can be summarized as follows:
[0003] The static problem of matching value needs with scenarios: Traditional systems rely on human experience or simple rules to match value goals with logistics scenarios, lacking the ability to dynamically mine semantic associations. For example, "low carbon emission reduction" may only be mechanically mapped to "new energy vehicle models", ignoring the correlation analysis of dynamic parameters of the scenario (such as delivery routes and cargo weight).
[0004] The strong coupling of functional modules: Functional modules (such as "temperature control system" and "path planning") are highly dependent on each other, which can easily lead to redundant design during development, resulting in low customization efficiency. Although existing tools support functional decomposition, their clustering methods are simplistic and cannot adapt to specific scenario requirements. For example, when a logistics company needs to customize lightweight trucks, traditional systems cannot independently optimize the "body material" module without affecting other functions.
[0005] These limitations restrict the customization and optimization capabilities of commercial vehicle products in complex logistics scenarios, and technological innovation is urgently needed to improve matching accuracy and customization efficiency. Summary of the Invention
[0006] In view of this, this application provides a commercial vehicle product development method based on a logistics scenario library, which aims to improve at least one of the above-mentioned problems.
[0007] Specifically, the following technical solutions are included:
[0008] On the one hand, embodiments of this application provide a method for developing commercial vehicle products based on a logistics scenario library, the method being as follows:
[0009] Match the scenarios corresponding to the target value in the logistics scenario library, match the functions corresponding to the scenarios and the product feature combinations corresponding to each function, and form a product solution for the service scenario driven by the target value.
[0010] In some embodiments of the present invention, the target value of the input is embedded in text based on the BERT model to form a value vector, and the value vector is input into the GNN-LSTM model. The GNN-LSTM model outputs the scene with the highest recommendation probability. The GNN-LSTM model is composed of a GNN model and an LSTM model.
[0011] In some embodiments of the present invention, a GNN model is used to extract scene features and value features, and the extracted scene features and value features are input into an LSTM model. The LSTM model outputs the recommendation probability of each scene under the target value, and selects the scene with the highest recommendation probability.
[0012] In some embodiments of the present invention, the target value adaptation scenario is input into the MLP model, and the MLP model outputs the function of the adaptation scenario and the product features corresponding to the function block.
[0013] In some embodiments of the present invention, the functional requirements of the corresponding functions in the adapted scenario are embedded in text based on the BERT model to form functional embedding vectors for each function. The functional requirement similarity of each function is calculated based on the functional embedding vectors to form a functional requirement correlation matrix (DSM). The functions are decoupled based on the functional requirement correlation matrix (DSM).
[0014] In some embodiments of the present invention, the K-means clustering algorithm clusters the functions in the Functional Requirements Association Matrix (DSM) and groups the functions belonging to the same cluster into a module.
[0015] In some embodiments of the present invention, the elbow method analysis function requires the clustering cost curve under the correlation matrix DSM to determine the optimal number of clusters K.
[0016] In some embodiments of the present invention, the scene feature update based on the scene network layer is represented as follows:
[0017] ;
[0018] in, This represents the connection matrix between scene nodes. If two scenes intersect or are related, then this is indicated in the connection matrix. The corresponding element in the array is 1; otherwise, the element is 0. This represents the scene features corresponding to each scene node. Perform a linear transformation; Indicates the connection matrix Aggregate scene features of related scenes; This represents the activation function. This indicates the scene features corresponding to the updated scene node.
[0019] In some embodiments of the present invention, the value feature update based on value nodes within the value network layer is represented as follows:
[0020] ;
[0021] in, This represents the connection matrix between value nodes. If there is an intersection or association between two value nodes, then this is indicated in the connection matrix. The corresponding element in the array is 1; otherwise, the element is 0. Represents the value characteristics corresponding to each value node. Perform a linear transformation; Indicates the connection matrix Aggregate the value characteristics of related values; This represents the activation function. This represents the value characteristics corresponding to the updated value node.
[0022] On the other hand, embodiments of this application provide a computer-readable storage medium storing a computer program thereon, characterized in that the program, when executed by a processor, implements a commercial vehicle product development method as described above for the logistics scenario library.
[0023] The commercial vehicle product development method based on a logistics scenario database provided by this invention has the following beneficial technical effects:
[0024] (1) Value-Scenario Dual-Layer Heterogeneous Network: The GNN-LSTM model is used to realize the semantic association mining between enterprise value goals and specific logistics scenarios, dynamically recommend the optimal scenario combination, and solve the scenario adaptation problem under value-driven;
[0025] (2) Functional module decoupling and intelligent matching: By designing the functional requirement correlation matrix DSM clustering and MLP bidirectional mapping model, the functional modules are decoupled to achieve bidirectional intelligent matching of functions and product features, significantly reducing redundant design and improving development efficiency. Attached Figure Description
[0026] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0027] Figure 1 A flowchart of a commercial vehicle product development method based on a logistics scenario library provided in an embodiment of the present invention;
[0028] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation
[0029] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0030] Figure 1 The flowchart of the commercial vehicle product development method based on a logistics scenario library provided in this embodiment of the invention is as follows:
[0031] (1) Match the scenario corresponding to the target value in the logistics scenario library;
[0032] (2) Match the functions corresponding to the scenarios and the product feature combinations corresponding to each function to form a product solution for the service scenario driven by the target value.
[0033] In this embodiment of the invention, the logistics scenario library is used to store scenarios. Each scenario corresponds to a scenario description. Scenarios include: traffic planning, logistics service process, low-carbon energy, covering urban delivery, cold chain transportation, last-mile sorting, etc. The input value is "reduce carbon emissions", and the applicable scenarios include low-carbon energy logistics scenarios. The input value is "low-carbon", and the applicable scenarios include cold chain transportation logistics scenarios.
[0034] In this embodiment of the invention, the target value is embedded into text using the BERT model to form a value vector. This value vector is then input into a GNN-LSTM model, which outputs the scene with the highest recommendation probability. The GNN-LSTM model consists of a GNN model and an LSTM model. The output of the GNN model is input into the LSTM model. The GNN model extracts scene features and value features, which are then input into the LSTM model. The LSTM model outputs the recommendation probability of each scene under the target value, and selects the scene with the highest recommendation probability.
[0035] In this embodiment of the invention, a value-scenario mapping matrix is constructed, and a GNN-LSTM model is trained based on the value-scenario mapping matrix. The trained GNN-LSTM model is used to predict the appropriate scenario for the target value.
[0036] The GNN model learns the embedding representations of value nodes and the dependencies between them in the value network, as well as the embedding representations of scene nodes and the dependencies between them in the scene network. The scene feature update representation based on the scene network layer is as follows:
[0037] ;
[0038] in, This represents the connection matrix between scene nodes. If two scenes intersect or are related, then this is indicated in the connection matrix. The corresponding element in the array is 1; otherwise, the element is 0. This represents the scene features corresponding to each scene node. Perform a linear transformation; Indicates the connection matrix Aggregate scene features of related scenes; This represents the activation function. This indicates the scene features corresponding to the updated scene node. In the logistics scene library, updating scene nodes (such as "urban delivery" receiving information from "transportation planning") enhances the implicit connections between scenes and supports dynamic recommendations (such as recommending "autonomous driving loop scene").
[0039] In this embodiment of the invention, the value feature update based on the value nodes within the value network layer is represented as follows:
[0040] ;
[0041] in, This represents the connection matrix between value nodes. If there is an intersection or association between two value nodes, then this is indicated in the connection matrix. The corresponding element in the array is 1; otherwise, the element is 0. Represents the value characteristics corresponding to each value node. Perform a linear transformation; Indicates the connection matrix Aggregate the value characteristics of related values; This represents the activation function. This represents the value characteristics corresponding to the updated value node. It ensures that the value description (such as "last-mile delivery efficiency optimization") is adaptively updated within the network, improving the accuracy of matching with the scenario.
[0042] In this embodiment of the invention, the information propagation from scene nodes to value nodes, and the value feature update based on scene features, are specifically as follows:
[0043] ;
[0044] in, This represents the connection matrix between value and scenario. If there is a correlation between value and scenario, then the connection matrix... The corresponding element in the array is 1; otherwise, the element is 0. This represents the scene features corresponding to each scene node. Perform a linear transformation; Indicates the connection matrix Aggregate scene features of related scenes; This indicates the activation function, and += indicates the value feature updates within the stacked value network layers. ) and cross-layer feature updates from the scene network layer to the value network layer ( This enables scenarios to "influence" value, such as "green energy scenarios" enhancing the representation of "low-carbon emission reduction" value and supporting value-driven scenario recommendations.
[0045] In this embodiment of the invention, the information propagation from value nodes to scene nodes, and the scene feature update based on value features, are as follows:
[0046] ;
[0047] in, This represents the value characteristics corresponding to each value node. Perform a linear transformation; Indicates the connection matrix Aggregate the value characteristics of related values; This indicates the activation function, and += indicates the overlay of scene feature updates within the network layer. ) and cross-layer feature updates from the value network layer to the scene network layer ( It allows value to be fed back to the scenario, such as the value of "cost reduction and efficiency improvement" optimizing the "path planning scenario", forming a two-way dynamic mapping.
[0048] The GNN model outputs the final value features and scene features as input to the LSTM model. The LSTM model predicts the recommendation probability for each scene. The GNN-LSTM model realizes the connection of "value input → GNN embedding → LSTM prediction", realizing the matching process of target value and scene.
[0049] In this embodiment of the invention, the target value-adapted scenario is input into the MLP model, and the MLP model outputs the functions of the adapted scenario and the product features corresponding to the function blocks. The MLP model realizes a two-way mapping between scenarios and functions, and between functions and product features. Taking the cold chain transportation scenario as an example, the cold chain transportation is input into the MLP model, and the MLP model outputs the functions recommended for cold chain transportation. The functions include low-temperature warehouses and dynamic temperature control systems, as well as the specific product features that match the functions of low-temperature warehouses and dynamic temperature control systems, such as "dual refrigeration compressors" and "heat insulation coating".
[0050] In this embodiment of the invention, the functional requirements of the corresponding functions in the adapted scenario are embedded in text based on the BERT model to form functional embedding vectors for each function. The functional requirement similarity of each function is calculated based on the functional embedding vectors to form a functional requirement correlation matrix (DSM). The functions are decoupled based on the functional requirement correlation matrix (DSM) to reduce functional coupling between scenarios.
[0051] Since the Functional Requirements Association Matrix (DSM) provides the relationships between functions, this invention employs the K-means clustering algorithm to cluster the functions in the DSM. Functions belonging to the same cluster are grouped into a module, strongly coupled functions are grouped into modules, while weakly coupled or uncoupled functions are separated, thereby eliminating redundant dependencies between scenarios and improving customization efficiency. For example, in the "cold chain transportation" scenario, the warehouse temperature control function may be clustered into an independent functional module, while route planning is assigned to another functional module, avoiding unnecessary interaction effects.
[0052] In this embodiment of the invention, before clustering the functional requirement correlation matrix (DSM) using the K-means clustering algorithm, it is necessary to determine the number of clusters in the current clustering process. This invention uses the elbow method (based on the Kneed algorithm) to analyze the clustering cost curve (e.g., the sum of squares within clusters as a function of K) under the functional requirement correlation matrix DSM to determine the optimal number of clusters K, such as K=11. This ensures that the clustering results are neither overly subdivided (avoiding fragmentation) nor overly merged (preserving coupling).
[0053] This invention follows a closed-loop logic of "value-driven → scenario recommendation → functional module → product features → solution". For example: input value "reduce carbon emissions" → recommend scenario "low carbon energy" → match "hydrogen fuel cell module" → generate "hydrogen VAN vehicle configuration scheme".
[0054] In this embodiment of the invention, a computer-readable storage medium is also provided. This storage medium can store a program suitable for processor execution. During program execution, the device containing the computer-readable storage medium executes the aforementioned commercial vehicle product development method based on a logistics scenario library. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0055] The commercial vehicle product development method based on a logistics scenario database provided by this invention has the following beneficial technical effects:
[0056] (1) Value-Scenario Dual-Layer Heterogeneous Network: The GNN-LSTM model is used to realize the semantic association mining between enterprise value goals and specific logistics scenarios, dynamically recommend the optimal scenario combination, and solve the scenario adaptation problem under value-driven;
[0057] (2) Functional module decoupling and intelligent matching: By designing the functional requirement correlation matrix DSM clustering and MLP bidirectional mapping model, the functional modules are decoupled to achieve bidirectional intelligent matching of functions and product features, significantly reducing redundant design and improving development efficiency.
[0058] Unless otherwise defined, all technical terms used in the embodiments of this application have the same meaning as commonly understood by one of ordinary skill in the art.
[0059] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the application disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only.
[0060] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.
Claims
1. A commercial vehicle product development method based on a logistics scenario library, characterized in that, The method is as follows: Match the scenarios corresponding to the target value in the logistics scenario library, match the functions corresponding to the scenarios and the product feature combinations corresponding to each function, and form a product solution for the service scenario driven by the target value.
2. The commercial vehicle product development method based on a logistics scenario database as described in claim 1, characterized in that, The BERT model embeds the target value of the input into text to form a value vector. The value vector is then input into the GNN-LSTM model, which outputs the scene with the highest recommendation probability. The GNN-LSTM model is composed of a GNN model and an LSTM model.
3. The commercial vehicle product development method based on a logistics scenario database as described in claim 2, characterized in that, The GNN model is used to extract scene features and value features. The extracted scene features and value features are then input into the LSTM model. The LSTM model outputs the recommendation probability of each scene under the target value and selects the scene with the highest recommendation probability.
4. The commercial vehicle product development method based on a logistics scenario database as described in claim 1, characterized in that, The target value is adapted to the scenario input into the MLP model, and the MLP model outputs the functions adapted to the scenario and the product features corresponding to the function blocks.
5. The commercial vehicle product development method based on a logistics scenario database as described in claim 1, characterized in that, Based on the BERT model, the functional requirements of the corresponding functions in the adapted scenarios are embedded into text to form functional embedding vectors for each function. The functional requirement similarity of each function is calculated based on the functional embedding vectors to form a functional requirement association matrix (DSM). The functions are then decoupled based on the functional requirement association matrix (DSM).
6. The commercial vehicle product development method based on a logistics scenario library as described in claim 1, characterized in that, The K-means clustering algorithm clusters the functions in the Functional Requirements Association Matrix (DSM), grouping functions belonging to the same cluster into a module.
7. The commercial vehicle product development method based on a logistics scenario library as described in claim 1, characterized in that, The elbow method analysis function requires the clustering cost curve under the correlation matrix DSM to determine the optimal number of clusters K.
8. The commercial vehicle product development method based on a logistics scenario database as described in claim 3, characterized in that, The scene feature update representation based on scene nodes in the scene network layer is as follows: ； in, This represents the connection matrix between scene nodes. If two scenes intersect or are related, then this is indicated in the connection matrix. The corresponding element in the array is 1; otherwise, the element is 0. This represents the scene features corresponding to each scene node. Perform a linear transformation; Indicates the connection matrix Aggregate scene features of related scenes; This represents the activation function. This indicates the scene features corresponding to the updated scene node.
9. The commercial vehicle product development method based on a logistics scenario database as described in claim 3, characterized in that, The value feature update representation based on value nodes within the value network layer is as follows: ； in, This represents the connection matrix between value nodes. If there is an intersection or association between two value nodes, then this is indicated in the connection matrix. The corresponding element in the array is 1; otherwise, the element is 0. Represents the value characteristics corresponding to each value node. Perform a linear transformation; Indicates the connection matrix Aggregate the value characteristics of related values; This represents the activation function. This represents the value characteristics corresponding to the updated value node.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the commercial vehicle product development method of the logistics scenario library as described in any one of claims 1 to 9.