Modular design method and system for automotive interior based on generative engine

Abstract

The application provides a kind of based on generative engine's automotive interior modular design method and system, the method includes: reading the basic cabin data of target vehicle type, candidate interior module set and authentication witness token, establish target design input set;Based on multi-agent system extraction module to be combined, collaboratively generate target interior combination scheme draft, and establish combination correlation data package;Authentication witness token is hung to the node and edge corresponding to combination correlation graph, along combination correlation graph executes propagation superposition, generates combination authentication tensor graph, and identifies authentication distortion area, generates authentication distortion diagnosis set;According to authentication distortion diagnosis set, determine the hidden compensation structure reconstruction range, execute hidden compensation structure reconstruction to the target interior module or associated adjacent interior module that causes distortion, generate hidden compensation structure parameter set;Hidden compensation structure parameter set is written back to target interior combination scheme, generate reconstructed target interior combination scheme.

Description

Technical Field

[0001] This invention relates to the field of automotive intelligent design technology, and in particular to a modular design method and system for automotive interiors based on a generative engine. Background Technology

[0002] As automotive cockpits evolve towards platformization and digitalization, interior components are increasingly adopting modular designs, and generative engines are also being used to generate automotive interior solutions. By reading cabin boundaries, module parameters, and installation relationships, various interior combination results can be generated, improving design efficiency.

[0003] In existing technologies, when handling module replacement, the dimensional matching relationship between the candidate module and the target installation area is usually checked first, and then the connection positions, adjacent positions, and clearance positions are checked to see if they meet the assembly requirements in order to determine whether the generated interior trim assembly can be installed. However, automotive interior trim is not a collection of independent decorative parts, but an integrated system coupled with airbag deployment paths, occupant contact areas, structural clearance paths, and trim fracture boundaries.

[0004] Most existing technologies still focus on single-module matching or local interface coordination. Even if they can solve problems such as local assembly interference, boundary fitting, and installation positioning, it is difficult to determine whether multiple modules still meet the overall cabin safety requirements when combined. In particular, when instrument panel modules, screen bezel modules, trim panel modules, and buffer modules are freely spliced ​​together, local geometric changes or changes in support relationships in one module may be transmitted to other modules through connection and adjacency relationships, thereby changing the airbag deployment envelope, occupant contact sequence, structural clearance process, and trim fracture boundaries. Such changes may not be exposed in single-module verification, but they are prone to superimposed distortions after multi-module combination.

[0005] Therefore, this invention proposes a modular design method and system for automotive interiors based on a generative engine. The information disclosed in the background section is only for enhancing understanding of the background of this disclosure and may therefore contain prior art information that is not common knowledge to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to address the shortcomings of existing technologies by providing a modular design method and system for automotive interiors based on a generative engine, thereby solving the technical problems mentioned in the background section.

[0007] To achieve the above objectives, the present invention provides the following technical solution: The first part of this invention provides a modular design method for automotive interiors based on a generative engine, comprising the following steps: S1. Read the basic cabin data of the target vehicle, the candidate interior module set and the authentication witness token, establish the corresponding multi-agent system, perform unified semantic parsing, role assignment, integrity verification, interface compatibility verification and authentication witness token matching verification on each candidate interior module, and generate the target design input set. S2. Based on the target design input set, extract the modules to be combined, and the multi-agent system collaboratively generates a draft of the target interior combination scheme, and establishes a combination association data package containing the spatial location, connection relationship, adjacency relationship and authentication witness token attachment relationship of the modules to be combined; S3. Based on the target design input set and combined association data package, attach the authentication witness token to the corresponding node and edge of the combined association graph, perform propagation superposition along the combined association graph to generate a combined authentication tensor graph, and identify the authentication distortion area based on the target vehicle model authentication boundary to generate an authentication distortion diagnosis set. S4. Based on the authentication distortion diagnosis set, target design input set and combined associated data package, determine the scope of hidden compensation structure reconstruction, call the generative engine to perform hidden compensation structure reconstruction on the target interior module or associated adjacent interior module that caused the distortion, and generate hidden compensation structure parameter set. S5. Write the hidden compensation structure parameter set back to the target interior module or adjacent interior module in the target interior combination scheme, generate the reconstructed target interior combination scheme, and perform certification consistency verification on it. If the verification passes, output the modular design result of the automotive interior. If the verification fails, send back the target interior module corresponding to the remaining certification distortion area or the newly added certification distortion area to perform hidden compensation structure reconstruction.

[0008] The second part of this invention provides a modular design system for automotive interiors based on a generative engine, comprising: The target design input set generation module is used to read the basic cabin data of the target vehicle model, the candidate interior module set, and the authentication witness token to establish a multi-agent system; The combined associated data packet generation module is used to extract the modules to be combined based on the target design input set, and the multi-agent system collaboratively generates a draft of the target interior combination scheme; The authentication distortion diagnostic set generation module is used to design the input set and composite association data packet according to the target, and attach the authentication witness token to the corresponding node and edge of the composite association graph; The hidden compensation structure parameter set generation module is used to determine the scope of hidden compensation structure reconstruction based on the authentication distortion diagnosis set, the target design input set, and the combined associated data packets. The design result output module is used to write back the hidden compensation structure parameter set to the corresponding target interior module or adjacent interior module in the target interior combination scheme, and generate the reconstructed target interior combination scheme.

[0009] The beneficial effects of this invention are as follows: This invention establishes a target design input set for the target vehicle model by reading basic cabin data, candidate interior module sets, and authentication witness tokens. This enables unified constraints on installation benchmarks, connection relationships, and safety action relationships at the starting stage of modular design, avoiding the problem in existing technologies where candidate modules only match local dimensions to enter the assembly process, leading to distortion in subsequent safety verification.

[0010] This invention uses a multi-agent system to collaboratively generate modules to be combined and establish combination association data packets. It can uniformly express the spatial location, connection relationship, adjacency relationship and authentication witness token attachment relationship in the target interior combination scheme. This makes the interior combination result no longer stay at the geometric splicing level, but form a structured combination basis that can be directly called by subsequent authentication reasoning, thereby improving the verifiability of modular design results.

[0011] This invention generates a combined authentication tensor graph by attaching authentication witness tokens to the corresponding nodes and edges of the combined association graph and performing propagation and superposition along the combined association graph. This graph can identify authentication distortion areas formed by the linkage of multiple modules, solving the problem that existing technologies can only detect single-module assembly problems and have difficulty identifying overall cabin safety distortion in the combined state. It also improves the ability to identify deployment interference distortion, contact sequence distortion, insufficient clearance distortion, and fracture boundary distortion.

[0012] This invention determines the scope of hidden compensation structure reconstruction based on the authentication distortion diagnosis set, and performs hidden compensation structure reconstruction on the target interior module or related adjacent interior modules that cause distortion. This can extend the safety correction object from a single module to a cross-module linkage structure, avoiding the previous technology's approach of relying solely on manual adjustments to local modules after discovering problems, thereby improving the pertinence and automation of authentication distortion correction.

[0013] This invention generates a set of hidden compensation structural parameters by filtering and writing them back to the corresponding target interior module or adjacent interior module in the target interior combination scheme. It can complete local structural compensation while maintaining the compatibility relationship between the target installation area, the set of connection positions, the set of adjacent positions, and the set of yield positions, so that the freedom of module replacement and the safety consistency of the whole cabin can be taken into account at the same time.

[0014] This invention performs a certification consistency check on the reconstructed target interior combination scheme, and when the check fails, sends back the target interior module corresponding to the remaining or newly added certification distortion area for hidden compensation structure reconstruction, forming a closed-loop processing mechanism of parameter writing back, re-inference, verification judgment and local feedback, thereby improving the stability, repeatability and mass production applicability of the modular design results of automotive interiors. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the modular design method for automotive interiors based on a generative engine, as described in this invention. Figure 2 This is a schematic diagram of the automotive interior modular design system framework based on a generative engine, as described in this invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1: As Figure 1 As shown, this embodiment provides a modular design method for automotive interiors based on a generative engine, including the following steps: S1. Read the basic cabin data of the target vehicle, the candidate interior module set and the authentication witness token, establish the corresponding multi-agent system, perform unified semantic parsing, role assignment, integrity verification, interface compatibility verification and authentication witness token matching verification on each candidate interior module, and generate the target design input set. S2. Based on the target design input set, extract the modules to be combined, and the multi-agent system collaboratively generates a draft of the target interior combination scheme, and establishes a combination association data package containing the spatial location, connection relationship, adjacency relationship and authentication witness token attachment relationship of the modules to be combined; S3. Based on the target design input set and combined association data package, attach the authentication witness token to the corresponding node and edge of the combined association graph, perform propagation superposition along the combined association graph to generate a combined authentication tensor graph, and identify the authentication distortion area based on the target vehicle model authentication boundary to generate an authentication distortion diagnosis set. S4. Based on the authentication distortion diagnosis set, target design input set and combined associated data package, determine the scope of hidden compensation structure reconstruction, call the generative engine to perform hidden compensation structure reconstruction on the target interior module or associated adjacent interior module that caused the distortion, and generate hidden compensation structure parameter set. S5. Write the hidden compensation structure parameter set back to the target interior module or adjacent interior module in the target interior combination scheme, generate the reconstructed target interior combination scheme, and perform certification consistency verification on it. If the verification passes, output the modular design result of the automotive interior. If the verification fails, send back the target interior module corresponding to the remaining certification distortion area or the newly added certification distortion area to perform hidden compensation structure reconstruction.

[0018] Before performing step S1, there is also a step of pre-training the multi-agent system.

[0019] The specific pre-training process includes: Obtain the interior combination dataset of the target vehicle model in history. The dataset includes positive samples (CAD data of interior combination and token attachment data that have passed the whole cabin safety certification) and negative samples (interior combination data with safety defects such as unfolding interference and contact sequence distortion). Construct a deep neural network model comprising a module generation agent, an interface coordination agent, and an authentication reasoning agent. The module generation agent can employ a 3D structure generation architecture based on conditional generative adversarial networks (cGANs) or variational autoencoders (VAEs); the authentication reasoning agent can employ a graph neural network (GNN) architecture. To minimize the reconstruction error between the generated interior and the positive sample, and to maximize the recognition accuracy of the distorted region in the negative sample as the joint loss function, the backpropagation algorithm and gradient descent method are used to iteratively update the network weights of the multi-agent system until the joint loss function converges, thus obtaining the trained multi-agent system.

[0020] S1 specifically includes the following sub-steps: S110: Read the basic cabin data, candidate interior module set, and authentication witness token of the target vehicle model to generate the original design record set.

[0021] During implementation, a unified reading standard is first established for the target vehicle model, and files are created and read separately for three types of objects: basic cabin data, candidate interior module sets, and authentication witness tokens. The basic cabin data is not general vehicle body information, but cabin reference data directly used for subsequent combination generation and authentication reasoning. It includes at least the vehicle model identification, installation reference surface, instrument panel installation boundary, steering mechanism envelope, airbag placement area, occupant contact area, frame connection position, and interior installation reference. Among them, the installation reference surface is used to unify the positioning coordinates of each candidate interior module, the airbag placement area is used to limit the effective area of ​​the subsequent authentication witness token, and the occupant contact area is used to constrain the contact order in the subsequent target interior combination scheme.

[0022] Each candidate interior module in the candidate interior module set corresponds to at least one module record. Each module record includes at least the module identifier, module category, target installation area, connection position set, adjacent position set, yield position set, structural parameter set, and shape parameter set. Among them, the structural parameter set includes at least the wall thickness, rib position, opening position, and mounting ear position, and the shape parameter set includes at least the outer contour dimensions, edge curvature, and surface partition.

[0023] The yielding position set refers to a set of local position constraints on the candidate interior module used to characterize its spatial avoidance of the airbag arrangement area, steering mechanism envelope and occupant contact area; each yielding position is used to define at least the avoidance object, avoidance boundary and minimum reserved space of the corresponding local position.

[0024] Certification witness tokens are used to characterize the impact contribution of the corresponding candidate interior trim module on the airbag deployment envelope, occupant contact area, structural clearance path, and trim fracture boundary. Each certification witness token includes at least a token identifier, associated module identifier, area of ​​effect, type of effect, applicable boundary, and safety impact level. The type of effect includes at least deployment guidance, contact restriction, clearance correction, and fracture constraint. During reading, the three types of objects are uniformly converted into structured records with complete fields. A one-to-one link is established between the candidate interior trim module record and the corresponding certification witness token record, and a reference relationship is established between the candidate interior trim module record and the basic cabin data.

[0025] For example, when the module category is a screen bezel module and the safety impact level is 2, its authentication witness token must correspond to at least three types of action records: the airbag placement area boundary, the occupant contact area boundary, and the adjacent trim panel fracture boundary. After reading is complete, the original design record set is output, which includes at least the cabin baseline record, module record, and token record, and is available for S120 to call.

[0026] S120. Based on the original design record set, establish a multi-agent system corresponding to the generative engine, perform unified semantic parsing and role assignment on each candidate interior module, and generate a module token mapping set.

[0027] During implementation, a multi-agent system is established using the original design record set output by S110 as the sole input source. The multi-agent system includes at least a module generation agent, an interface coordination agent, an authentication reasoning agent, a compensation and reconstruction agent, and a verification and adjudication agent. However, in S120, the module generation agent, the interface coordination agent, and the authentication reasoning agent are actually involved in the processing.

[0028] The module generation agent reads the module records one by one, calls the pre-trained natural language processing model (such as a text embedding model based on the Transformer architecture), and converts the text labels such as module category and target installation area into high-dimensional word vector representations. At the same time, it extracts the numerical features from the set of structural parameters and the set of shape parameters, and concatenates and fuses the word vectors with the numerical features to complete the unified semantic parsing and form a computer-recognizable module structure description.

[0029] The module structure description should at least provide the positioning of the candidate interior module on the mounting reference plane, its occupancy within the target installation area, and its safety function corresponding to the module category.

[0030] The interface coordination agent reads the module structure description and cabin baseline record, assigns roles to the connection bit set, adjacency bit set, and yield bit set, forming an interface relationship description. Connection bits characterize the assembly relationship with the frame connection bit or adjacent candidate interior modules; adjacency bits characterize the adjacent boundary fit relationship; and yield bits characterize the spatial avoidance relationship with the airbag placement area, steering mechanism envelope, and occupant contact area. The authentication inference agent reads the token record and interface relationship description, attaching each authentication witness token to the module identifier, area of ​​action, and type of action of the corresponding candidate interior module, forming a token attachment description.

[0031] Subsequently, the module structure description, interface relationship description, and token attachment description are summarized by module identifier to generate a module token mapping set. This module token mapping set is not a simple index table, but a unified basis for subsequent verification and generation. It includes at least the module identifier, module category, target installation area, connection bit set, adjacent bit set, yield bit set, associated token identifier, scope, and scope type.

[0032] To prevent duplicate references of the same candidate interior module or incorrect token attachments, a uniqueness check is performed on associated tokens under the same module identifier during the aggregation process. If the same token identifier is found to be attached to different module identifiers simultaneously, it is marked as an attachment exception and written to the exception field in the module token mapping set for verification by S130. After completion, the module token mapping set is output for use by S130, S220, and S310.

[0033] S130. Perform integrity verification, interface compatibility verification, and authentication witness token matching verification on the candidate interior modules based on the module token mapping set, delete the candidate interior modules that do not meet the conditions, and generate the target design input set.

[0034] During implementation, an integrity check is performed first. This integrity check is not an abstract judgment, but rather a field-by-field check: if a module record lacks any of the following key fields: module category, target installation area, connection bit set, adjacency bit set, yield bit set, or structural parameter set, the candidate interior module is deemed incomplete; if a token record lacks any of the following key fields: token identifier, associated module identifier, scope, function type, or applicable boundary, the corresponding authentication witness token is deemed incomplete; if any field is incomplete, the corresponding candidate interior module is directly deleted.

[0035] Next, an interface compatibility verification is performed. Interface compatibility verification is determined based on the interface deviation. in, Indicates the interface deviation; This indicates the positioning deviation of the candidate interior module relative to the mounting reference surface; This indicates the deviation in the number of connection bits, the spacing between connection bits, or the correspondence between connection bits; This indicates the boundary deviation of the occupant relative to the airbag placement area, the steering mechanism envelope, or the occupant contact area. , , These represent the weighting coefficients for the position deviation term, connection deviation term, and clearance deviation term, respectively.

[0036] Can be adopted during implementation , , As a set of execution examples, when If the value exceeds a preset interface threshold, an interface conflict is identified, and the corresponding candidate interior module is deleted; for example, the preset interface threshold can be set to 0.8. Subsequently, authentication witness token matching verification is performed, specifically using the token integrity rate for determination. in, Indicates the token integrity rate; This indicates the number of authentication witness tokens that have been successfully attached and have complete fields. This indicates the number of authentication witness tokens that should be configured according to module category and security impact level.

[0037] when When the authentication witness token is found to be incompletely matched, the corresponding candidate interior module is deleted; when In this case, the corresponding candidate interior trim module is retained. For example, when the module category is dashboard module and the safety impact level is 3, It can be set to 4, corresponding to the airbag deployment envelope, occupant contact area, structural clearance path and trim fracture boundary, respectively.

[0038] After completing the three types of verifications, the module identifier, connection bit set, adjacent bit set, yield bit set, and associated token identifier corresponding to the candidate interior modules to be deleted are simultaneously removed from the module token mapping set. Only the candidate interior modules that have passed the integrity verification, interface compatibility verification, and authentication witness token matching verification, and their corresponding mapping relationships are retained, generating the target design input set. The target design input set includes at least the verified basic cabin data, the filtered candidate interior module set, and the confirmed module token mapping set, and is available for use by S210, S310, and S410.

[0039] S2 specifically includes the following sub-steps: S210. Read the target design input set, and extract the modules to be combined from the selected candidate interior modules based on the cabin layout requirements, module replacement requirements and installation boundary requirements of the target vehicle model, and generate the initial combination task set.

[0040] During implementation, the target design input set output by S130 is used as the sole input source, and no additional independent data sources are introduced to ensure consistency in the objects used for subsequent combination generation, authentication inference, and compensation reconstruction. The target design input set includes at least verified basic cabin data, a filtered set of candidate interior modules, and a confirmed set of module token maps.

[0041] After reading, the cabin layout constraints corresponding to the target vehicle model are first established based on the target installation area, installation reference surface, airbag placement area, occupant contact area and frame connection position in the basic cabin data; then, the allowed replacement module categories, allowed replacement target installation areas and safety boundaries that are not allowed to be crossed are determined according to the module replacement requirements; then, the positioning range, connection range and clearance range of each candidate interior module in the corresponding target installation area are limited according to the installation boundary requirements.

[0042] For each candidate interior module, instead of simply determining whether it belongs to the category of modules that can be replaced, we further calculate its combinatorial fit into the initial combinatorial task set: in, Indicates the suitability of the group; This indicates the degree of regional matching between the candidate interior module and the target installation area; This indicates the interface adaptability of the candidate interior module's set of connection bits, set of adjacent bits, and set of yield bits to the current cabin layout constraints; This indicates the completeness of the coverage of the target installation area by the authentication witness token corresponding to the candidate interior module; , , These represent the weight coefficients for region matching, interface adaptation, and token coverage, respectively.

[0043] When implementing, it is possible to take , , As a set of execution examples, the regional matching degree is obtained by the overlap ratio between the target installation area of ​​the candidate interior module and the corresponding target installation area in the basic cabin data; the interface adaptability is determined by whether the connection position can be hooked, whether the adjacent position can be attached, and whether the clearance position avoids the airbag placement area and the occupant contact area; the token coverage integrity is determined by whether the effective area of ​​the centralized authentication witness token in the module token mapping completely covers the corresponding target installation area.

[0044] when When the value is not less than a preset combination threshold, the candidate interior module is written into the initial combination task set; when If the value is less than the preset combination threshold, it will not be written into the initial combination task set. When multiple candidate interior modules correspond to the same target installation area, they are sorted from high to low according to their suitability, and only the candidate interior module ranked first is retained as the module to be combined for that target installation area.

[0045] The generated initial combination task set includes at least a task identifier, target installation region, module identifier to be combined, module category, priority flag, and token invocation relationship; wherein, the token invocation relationship is used to identify the range of authentication witness tokens that the module to be combined should invoke in the subsequent collaborative generation phase. After completion, the initial combination task set is output for S220 to call.

[0046] S220. The multi-agent system performs collaborative generation on the initial combination task set to obtain a draft of the target interior combination scheme.

[0047] During implementation, the multi-agent system established by S120 continues to participate in the processing. However, in this step, the module generation agent, interface coordination agent, and authentication reasoning agent are actually involved in the collaborative generation. The compensation and reconstruction agent and the verification and adjudication agent remain in a pending state and do not intervene in the current combination generation. The module generation agent reads the task identifier, target installation area, module identifier to be combined, and priority marker from the initial combination task set one by one. Based on the structural parameter set and shape parameter set of the corresponding candidate interior modules, it first generates the candidate arrangement results for each module to be combined. The so-called candidate arrangement result refers to the spatial position, orientation, boundary extension range, and local clearance shape of each module to be combined under the installation reference plane.

[0048] Subsequently, the interface coordination agent reads the candidate arrangement results and the module token mapping set, coordinates the connection relationships, adjacency relationships and yielding relationships between the modules to be combined and between the modules to be combined and the basic cabin data, and generates the interface coordination results.

[0049] During coordination, if the set of connection points of a module to be assembled cannot form a corresponding hooking relationship with the frame connection point or the adjacent module to be assembled, its spatial position and orientation will be adjusted; if the set of adjacent connection points of a module to be assembled does not have a boundary fit relationship with the adjacent module to be assembled, its boundary extension range will be adjusted; if the set of clearance points of a module to be assembled encroaches on the airbag arrangement area, the occupant contact area or the steering mechanism envelope, its local clearance shape will be adjusted.

[0050] Based on the interface coordination results, the authentication reasoning agent invokes the authentication witness token defined by the token invocation relationship to reason and determine the continuity of the scope of action, the integrity of the action boundary, and the cross-module coupling state of each module to be combined.

[0051] If the authentication inference results indicate that a candidate arrangement result satisfies the installation boundary and interface relationship, but its authentication witness token's effective area is broken, obscured, or misaligned across modules, then the corresponding candidate arrangement result will be marked as invalid, and the module will be sent back to generate an agent to regenerate an alternative arrangement result. If the authentication inference results indicate that the candidate arrangement result simultaneously satisfies the installation boundary, interface relationship, and authentication witness token coverage integrity, then it will be written into the target interior combination scheme draft.

[0052] To avoid situations where the combined results are partially feasible but conflicting overall, the interface coordination agent also performs a global coordination on all valid results, ensuring that only one module result is retained in the same target installation area, the same connection position is not repeatedly occupied by multiple modules to be combined, and the same clearance area is not cross-intruded by different modules to be combined.

[0053] After collaborative generation by a multi-agent system, a draft target interior trim combination scheme is obtained. This draft scheme includes at least the spatial location, orientation, connectivity, adjacency, yielding relationships, and authentication witness token invocation status of each module to be combined. Upon completion, the draft target interior trim combination scheme is output for use by the S230.

[0054] S230. Based on the draft of the target interior combination scheme, establish a combination association diagram and generate a combination association data package.

[0055] During implementation, the draft of the target interior assembly scheme is used as the basis for the diagram. First, the nodes and edges in the assembly diagram are defined. Nodes are based on the modules to be assembled, and each node records at least the module identifier, module category, target installation area, spatial location, and authentication witness token invocation status. Edges are used to record the relationships between the modules to be assembled; static boundary records that duplicate the basic cabin data are not created separately to avoid invalid relationships in the assembly diagram. Whether an edge is established between any two modules to be assembled is determined according to the following formula: in, Indicates the module to be combined With the module to be combined The results of establishing the associated edges between them; This indicates whether there is a connection between the two; This indicates whether there is an adjacency relationship between the two; Indicates whether there is a yielding influence relationship between the two; function This tool is used to output whether to build an edge and the edge type based on the connection relationship, adjacency relationship, and yield influence relationship.

[0056] When at least one relationship exists, a corresponding edge is created in the composite association graph; when a connection relationship exists, the edge type is marked as a connection edge; when only an adjacency relationship or a yielding influence relationship exists, the edge type is marked as an adjacency edge. After the graph is constructed, the authentication witness tokens are then written into the composite association graph: the authentication witness tokens corresponding to the local effects of a single module to be combined are recorded as node attribute records, and the authentication witness tokens corresponding to the coupling of cross-module action area continuity, break boundary transmission, or yielding path are recorded as edge attribute records.

[0057] The purpose of this processing is to enable S310 to directly perform authentication witness token propagation and superposition based on node attributes and edge attributes, and to enable S410 to directly determine the reconstruction scope of the hidden compensation structure based on edge type and edge attributes. Finally, the target interior decoration combination scheme draft, combination association graph, node table, edge table, and token attachment table are encapsulated into a combination association data package. The combination association data package includes at least the target interior decoration combination scheme, node identifier set, edge identifier set, edge type record, and token attachment record, and is available for use by S310, S410, and S510.

[0058] S3 specifically includes the following sub-steps: S310. Read the target design input set and the combined association data package, attach the authentication witness tokens in the module token mapping set to the corresponding nodes and edges in the combined association graph, and generate a token action relationship set.

[0059] During implementation, the target design input set output by S130 and the combined association data package output by S230 are used as the sole inputs for this step. The target design input set includes at least a confirmed module token mapping set, and the combined association data package includes at least a target interior combination scheme, a node table, an edge table, and a token attachment table. First, the authentication inference agent reads the associated token identifier, scope of action, action type, and applicable boundary of the module token mapping set one by one, and matches them with the node attributes and edge attributes in the combined association graph.

[0060] During matching, instead of uniformly attaching all authentication witness tokens to nodes or edges, the type of object to which the authentication witness token applies is first determined. When an authentication witness token only represents the local influence of a single module to be combined on the airbag deployment envelope, occupant contact area, structural clearance path, or trim fracture boundary, the authentication witness token is attached as a node attribute of the corresponding node. When an authentication witness token represents the cross-module effect between two modules to be combined, including cross-module clearance transfer, cross-module fracture boundary continuity, and the joint effect of adjacent modules to be combined on the same occupant contact area, the authentication witness token is attached as an edge attribute of the corresponding edge.

[0061] If the same authentication witness token has both local and cross-module effects, it is split into node attachment records and edge attachment records according to the effect type and written separately. A single attachment result does not replace a dual attachment result. During the attachment process, conflict resolution is also required: when the same authentication witness token is repeatedly attached to multiple nodes or multiple edges, and there is no common effect relationship between these attached objects, only the attachment records with the same attachment object type and the same corresponding associated module identifier, effect area, and edge type record are retained, and the remaining attachment records are deleted; when the attached object is a node, the edge type record is not compared.

[0062] When the same authentication witness token is attached to multiple nodes or edges with shared action relationships, all attachment records are retained, and a propagation order field is added to each record to determine the action transmission order during subsequent propagation stacking. After attachment is completed, a token action relationship set is generated. The token action relationship set includes at least the token identifier, attached object identifier, attached object type, action area, action type, applicable boundary, and propagation order, and is available for use by S320.

[0063] S320. Based on the token action relationship set, perform authentication witness token propagation and superposition along the spatial position, connection relationship and adjacency relationship of the composite association graph to generate a composite authentication tensor graph.

[0064] During implementation, the token action relationship set output by S310 is used as direct input, and the propagation path is constructed independently without deviating from the composite association graph. The basic propagation rules are as follows: the authentication witness token corresponding to the node attribute first acts on the node itself, and then propagates to adjacent nodes along the edges directly connected to that node; the authentication witness token corresponding to the edge attribute first acts on the corresponding edge, then acts on the nodes at both ends of that edge, and continues to propagate along edges of the same type. Considering the different transmission strengths of the security effect between connecting edges and adjacent edges, different propagation weights are configured for different edge types during propagation; among them, connecting edges are used to represent direct assembly relationships, and their propagation weight is higher than that of adjacent edges; adjacent edges are used to represent boundary fitting relationships or yielding influence relationships, and their propagation weight is lower than that of connecting edges. For any module to be combined, its comprehensive authentication effect value is calculated according to the following formula: in, Indicates the first The comprehensive authentication value corresponding to each module to be combined; Indicates that it is attached to the The node authentication value on each module node to be combined; Indicates the relationship with the first The first module to be combined is connected to the first The edge authentication effect value is passed from the edge to this node; Indicates the first The propagation weights corresponding to the edges; Indicates the relationship with the first The number of edges directly connected to each module to be combined.

[0065] During the calculation, the authentication reasoning agent first determines the authentication impact value of each node and the authentication impact value of each edge, and then performs the superposition based on the propagation order and propagation weight. If the impact value after propagation is lower than the preset propagation threshold, or if the terminal node with no further connecting edges and adjacent edges has been reached, the propagation along that path is stopped to avoid invalid diffusion.

[0066] After propagation is complete, the authentication effect values ​​of each node and edge in different authentication regions are written into a unified tensor structure according to the region coordinates to generate a combined authentication tensor graph. The combined authentication tensor graph includes at least the authentication region identifier, region coordinates, node source set, edge source set, and region authentication effect value, and is available for use by S330 and S520.

[0067] S330: Based on the combined authentication tensor map, identify regions that are inconsistent with the authentication boundary of the target vehicle model, and generate an authentication distortion diagnosis set.

[0068] During implementation, the target vehicle's certification boundaries are first established based on the target vehicle's basic cabin data and certification requirements. These certification boundaries are not general safety ranges, but rather a set of boundary benchmarks directly involved in the comparison, including at least the airbag deployment boundary, occupant contact sequence boundary, structural clearance lower limit boundary, and trim component fracture boundary. Subsequently, the certification inference agent reads the regional certification action value from the combined certification tensor map for each certification region and compares it with the corresponding boundary benchmark value in the target vehicle's certification boundaries. The distortion deviation is calculated using the following formula: in, Indicates the first Distortion deviation values ​​for each certified area; The combined authentication tensor graph represents the first... Regional certification effect value on each certification area; Indicates the certification boundary of the target vehicle model at the th Boundary reference values ​​on each certified area.

[0069] When making comparisons, when When the distortion exceeds the preset distortion threshold for the corresponding distortion type, it is considered that the first... Each certified area is a certified distortion area; when the certified action value of an area intrudes into or blocks the airbag deployment boundary, it is judged as deployment interference distortion; when the occupant contact sequence corresponding to the certified action value of an area deviates from the boundary reference value, it is judged as contact sequence distortion; when the certified action value of an area is lower than the lower limit boundary of structural clearance, it is judged as insufficient clearance distortion; when the fracture location, fracture extension direction, or fracture continuity corresponding to the certified action value of an area deviates from the fracture boundary of the trim, it is judged as fracture boundary distortion.

[0070] After determining the distortion type, instead of mechanically assigning responsible modules based on the location of the distortion area, the process traces back the set of node sources and edge sources of the distortion area in the combined authentication tensor graph to identify the target interior module that caused the distortion area and the adjacent interior modules that together form the authentication value of that area. This ultimately generates an authentication distortion diagnostic set. The authentication distortion diagnostic set includes at least the distortion area, area coordinates, distortion type, the target interior module that caused the distortion, associated adjacent interior modules, and corresponding edge source identifiers, and is available for use by S410, S520, and S530.

[0071] S4 specifically includes the following sub-steps: S410: Read the target design input set, combine the associated data packets and the authentication distortion diagnosis set, determine the scope of hidden compensation structure reconstruction, and generate the compensation reconstruction task set.

[0072] During implementation, the target design input set output by S130, the combined associated data package output by S230, and the authentication distortion diagnostic set output by S330 are used as the sole inputs for this step, without introducing any new diagnostic sources. First, the authentication distortion region, region coordinates, distortion type, target interior module causing the distortion, associated adjacent interior modules, and corresponding edge source identifiers from the authentication distortion diagnostic set are read, and the target interior module causing the distortion is directly identified as a mandatory reconstruction object.

[0073] Subsequently, based on the edge type record and edge source identifier in the combined associated data packet, it is determined whether there is a cross-module effect. When the edge source identifier indicates that the authentication distortion area is formed by cross-module yielding, cross-module fracture boundary continuity, or joint contact effect, the corresponding adjacent interior module is identified as the extended reconstruction object. When an adjacent interior module is adjacent to the target interior module that caused the distortion, but does not appear in the edge source identifier corresponding to the authentication distortion area and does not assume a cross-module effect, it is not included in the scope of hidden compensation structure reconstruction. To ensure that the processing order of multiple modules to be reconstructed is executable, the reconstruction priority is calculated for each module to be reconstructed. in, Indicates the first The refactoring priority of each module to be refactored; This indicates the distortion deviation value of the authentication distortion area corresponding to the module to be reconstructed; This indicates the degree of cross-module coupling of the module to be refactored in the composite association diagram; This indicates the degree of linkage influence that the module to be reconstructed has on other modules to be combined within the same target installation area; , , These represent the weighting coefficients for the distortion deviation term, the coupling degree term, and the linkage effect term, respectively.

[0074] When implementing, it is possible to take , , As a set of execution examples, modules with larger distortion deviation values ​​are prioritized for processing. After sorting, a set of compensation reconstruction tasks is formed according to reconstruction priority from high to low.

[0075] The compensation and reconstruction task set includes at least a task identifier, distortion type, distortion region coordinates, module identifier to be reconstructed, reconstruction priority, allowed hidden compensation structure types, and corresponding authentication witness token constraints. The allowed hidden compensation structure types are not arbitrarily given but are pre-defined based on the distortion type; for example, expansion interference distortion corresponds to yield cavity structures, guide edge structures, and weakening groove structures, while fracture boundary distortion corresponds to weakening groove structures, back rib structures, and guide edge structures. Upon completion, the compensation and reconstruction task set is output for S420 to call.

[0076] S420: Invoke the compensation and reconstruction agent in the generative engine, and have the module generating agent, interface coordinating agent and authentication reasoning agent work together to perform hidden compensation structure reconstruction on the compensation and reconstruction task set and generate a set of compensation candidate structures.

[0077] During implementation, the compensation and reconstruction agent processes tasks one by one according to the task identifiers in the compensation and reconstruction task set. Each time, it performs hidden compensation structure reconstruction only on one module to be reconstructed and its necessary extended reconstruction objects to avoid overlapping reconstruction actions in different authentication distortion areas. During reconstruction, it first calls the allowed hidden compensation structure types based on the distortion type, and then determines the local adjustment direction by combining the corresponding authentication witness token constraints.

[0078] When the distortion type is deployment interference distortion, a clearance cavity structure or guide edge structure is generated first inside the target interior module that causes the distortion, and a weakening groove structure is added if necessary to control the deployment path; when the distortion type is contact sequence distortion, the buffer layer structure and support rib structure are adjusted first to correct the local support sequence corresponding to the occupant contact area; when the distortion type is insufficient clearance distortion, the internal volume of the clearance cavity structure is increased first or the positional relationship of the back rib structure is adjusted; when the distortion type is fracture boundary distortion, the arrangement position of the weakening groove structure is changed first and the fracture extension direction is corrected in conjunction with the back rib structure.

[0079] The module generating agent (e.g., constructed based on the Diffusion Model or a 3D point cloud generation network) takes the 3D boundary of the target interior module that causes distortion as input and the corresponding authentication witness token constraint (such as the coordinates of the yield space that needs to be avoided) as a conditional mask. It performs sampling and denoising generation in the local feature space to generate the arrangement position, size parameters and local contour of the candidate hidden compensation structure. The interface coordination agent is responsible for verifying whether the candidate hidden compensation structure breaks through the boundary of the target installation area, whether it changes the connection relationship between the set of connection bits and the set of adjacent bits, and whether it intrudes into the existing set of yield bits; the authentication inference agent is responsible for verifying the recovery effect of the candidate hidden compensation structure on the original authentication distortion area and whether it introduces a new authentication distortion area.

[0080] The recovery effect here is not an abstract improvement, but rather refers to whether the distortion deviation value corresponding to the original authentication distortion region decreases after the candidate hidden compensation structure is written into the module to be reconstructed, and whether the decrease reaches a preset recovery threshold. If a candidate hidden compensation structure can reduce the distortion deviation value of the original authentication distortion region, but causes the target installation area to go out of bounds, interface relationship mismatch, or a new authentication distortion region, then the candidate hidden compensation structure is deleted, and the compensation and reconstruction agent regenerates a replacement structure; if a candidate hidden compensation structure simultaneously satisfies the requirements of boundary preservation, interface preservation, and authentication recovery, then it is written into the compensation candidate structure set.

[0081] The compensation candidate structure set includes at least the candidate structure identifier, corresponding task identifier, hidden compensation structure type, placement position, size parameters, associated interface parameters, and recovery effect record. For example, for a certain unfolded interference distortion area, a clearance cavity structure with a depth of 6 mm can be generated on the back side of the screen bezel module, and a guide edge structure can be formed simultaneously on the adjacent trim module. As long as this combination does not change the connection relationship of the original connection set and can reduce the corresponding distortion deviation value, it can be retained as a valid compensation candidate structure. After completion, the compensation candidate structure set is output for S430 to call.

[0082] S430. Based on the compensation candidate structure set and the recovery effect of the corresponding authentication witness token, select the compensation candidate structure that satisfies the authentication constraints and combination constraints, and generate the hidden compensation structure parameter set.

[0083] During implementation, a pre-deletion judgment is first performed on each compensation candidate structure in the compensation candidate structure set: if the recovery effect record shows that the recovery degree of the compensation candidate structure for the original authentication distortion area does not reach the preset recovery threshold, it is directly deleted and will not participate in subsequent screening; when the recovery effect reaches the preset recovery threshold, its ability to maintain interface relationships and overall combination stability is then scored. The comprehensive screening value is calculated according to the following formula: in, Indicates the first The comprehensive screening value of each compensation candidate structure; This indicates the degree to which the compensation candidate structure restores the original authentication distortion area; This indicates the degree to which the compensation candidate structure preserves the connection relationships between the original set of connecting bits, the set of adjacent bits, and the set of yielding bits; This indicates the degree to which the candidate compensation structure maintains the overall stability of the target interior combination scheme; , , These represent the weight coefficients for the recovery rate, interface retention rate, and overall stability retention rate, respectively.

[0084] When implementing, it is possible to take , , As an example of the execution, recovery rate becomes the dominant selection factor. After the calculation is completed, among the remaining compensation candidate structures corresponding to the same task identifier, the compensation candidate structure with the highest comprehensive screening value is selected as the target compensation structure; when multiple compensation candidate structures have the same comprehensive screening value, the compensation candidate structure with smaller adjustment range of layout position and smaller change of associated interface parameters is retained first, so as to reduce the disturbance to subsequent write-back actions.

[0085] Finally, the hidden compensation structure type, placement location, size parameters, and associated interface parameters corresponding to the target compensation structure are written into the hidden compensation structure parameter set. The placement location is used by S510 to determine the parameter write-back location, the size parameters are used by S510 to perform local structure updates, and the associated interface parameters are used by S520 to re-verify interface relationships and authentication consistency. After completion, the hidden compensation structure parameter set is output for use by S510 and S520.

[0086] S5 specifically includes the following sub-steps: S510: Read the combined associated data packet and the hidden compensation structure parameter set, write the hidden compensation structure parameter set back to the target interior module or adjacent interior module in the target interior combination scheme, and generate the reconstructed target interior combination scheme.

[0087] During implementation, the combined association data package output by S230 and the hidden compensation structure parameter set output by S430 are used as the sole inputs for this step. Each parameter record in the hidden compensation structure parameter set includes at least the task identifier, hidden compensation structure type, placement location, size parameters, associated interface parameters, reconstruction priority, and comprehensive screening value; the combined association data package includes at least the target interior decoration combination scheme, node table, edge table, and token attachment table.

[0088] Before writing back, the parameter records in the hidden compensation structure parameter set are mapped to the corresponding target interior module or adjacent interior module according to the task identifier. Then, each parameter record is limited to a local structural area within the module according to its placement position, avoiding expanding the local compensation action into a complete module rewrite. If the same target interior module corresponds to multiple parameter records, they are sorted from high to low according to the reconstruction priority. When the reconstruction priorities are the same, they are sorted from high to low according to the comprehensive screening value, so that the parameter records processed first occupy the local structural area first.

[0089] During the write-back process, if two parameter records apply to the same local structural region and their size parameters or associated interface parameters conflict, the parameter record with the higher reconstruction priority and comprehensive screening value is retained, while the other parameter record is deleted. If two parameter records apply to the same target interior module but correspond to different local structural regions, parallel write-back is allowed. This write-back process does not only write the geometric dimensions but also simultaneously updates the arrangement position, size parameters, and associated interface parameters corresponding to the hidden compensation structure type.

[0090] For example, when the hidden compensation structure type is a clearance cavity structure, at least the positioning coordinates, cavity depth, and cavity boundary of the clearance cavity structure on the back side of the target interior module should be written back; when the hidden compensation structure type is a weakening groove structure, at least the groove starting point, extension direction, and groove depth parameters of the weakening groove structure should be written back.

[0091] After completing the local structure write-back, the node attributes, edge attributes, and token attachment records in the combined associated data packet must also be updated synchronously to ensure that the structural description of the target interior module or adjacent interior module being written back, the connection or adjacency relationship in the node table, and the area of ​​effect record in the token attachment table are consistent with the current write-back result.

[0092] After the update is complete, the reconstructed target interior trim combination scheme is generated. Simultaneously, the updated node table, updated edge table, and updated token attachment table are encapsulated into an updated combination association data package. The reconstructed target interior trim combination scheme includes at least the target interior trim module structure after write-back, the adjacent interior trim module structures, the updated node table, the updated edge table, and the updated token attachment table, and is available for use by S520 and S530.

[0093] S520. Perform a certification consistency check on the reconstructed target interior combination scheme and generate a certification consistency check record.

[0094] During implementation, the reconstructed target interior combination scheme and the updated combination association data package output by S510 are used as direct inputs. The verification and adjudication agent performs re-authentication inference by combining the combination authentication tensor graph output by S320, the authentication distortion diagnosis set output by S330, and the hidden compensation structure parameter set output by S430. Among them, the combination authentication tensor graph output by S320 is updated or recalculated based on the updated node table, the updated edge table, and the updated token attachment table, and the updated combination authentication tensor graph is used as the basis for authentication consistency verification.

[0095] During the verification process, not only the original authentication distortion area is verified, but also adjacent authentication areas that have a direct connection or direct adjacency relationship with the original authentication distortion area in the composite association graph. This is to prevent local elimination or the addition of neighboring areas after the hidden compensation structure is written back. The verification items include at least distortion elimination verification items, new distortion verification items, and interface stability verification items.

[0096] The distortion elimination check item is used to determine whether the distortion deviation value corresponding to the original authentication distortion area has dropped below the preset passing threshold; the new distortion check item is used to determine whether a new authentication distortion area appears after reconstruction and whether the new authentication distortion area exceeds the allowable range. The new authentication distortion area is re-identified based on the updated combined authentication tensor map; the interface stability check item is used to determine whether the target installation area, connection bit set, adjacent bit set, and yield bit set remain compatible. To unify the re-authentication inference results into a decidable result, an authentication consistency check value is calculated for each round of verification: in, This represents the certification consistency verification value; This represents the average residual distortion deviation value of the original distorted area after reconstruction; Indicates the number of newly added authentication distortion areas; This indicates the amount of interface deviation after write-back; , , These represent the weight coefficients for the residual distortion term, the newly added distortion term, and the interface deviation term, respectively.

[0097] When implementing, it is possible to take , , As an example of the implementation, whether the original authentication distortion area has been truly eliminated becomes the primary judgment factor. Specifically, the interface deviation after the write-back is recalculated based on the updated node table, updated edge table, and associated interface parameters, according to the S130 interface deviation calculation rules.

[0098] when When the value is not greater than the preset verification threshold and there are no newly added authentication distortion areas exceeding the allowable range, the reconstructed target interior combination scheme is deemed to have passed the authentication consistency verification; when If the value exceeds the preset verification threshold, or if there are newly added authentication distortion areas exceeding the allowable range, the authentication consistency verification is deemed to have failed. Upon completion, an authentication consistency verification record is generated. This record includes at least the verification round, the verification result for the original authentication distortion areas, the verification result for the newly added authentication distortion areas, the interface stability verification result, the authentication consistency verification value, and a pass flag, and is available for use by the S530.

[0099] S530: Output the modular design results of automotive interior based on the certification consistency verification record.

[0100] During implementation, the system first reads the pass flag, authentication consistency verification value, original authentication distortion area verification result, and newly added authentication distortion area verification result from the authentication consistency verification record. If the pass flag is passed, the reconstruction stops, and the modular design result of the automotive interior is output. The modular design result of the automotive interior includes at least the reconstructed target interior combination scheme, the authentication consistency verification record that passed the authentication consistency verification, the retained hidden compensation structure parameter set, and the updated combination association data package, so that subsequent callers can directly read the final design state and the corresponding verification basis.

[0101] If the target interior module is marked as failed, instead of sending all target interior modules back to S410, the remaining authentication distortion area, the newly added authentication distortion area, the corresponding target interior module that caused the distortion and its associated adjacent interior modules are encapsulated into an updated authentication distortion diagnostic set, and sent back to S410 along with the updated combined associated data package as new input to form a new compensation reconstruction task set.

[0102] To determine whether the iteration has converged, for the th iteration... Calculate the convergence index based on the results of the round reconstruction: in, Indicates the first Convergence metrics after round reconstruction; Indicates the first The total remaining authentication distortion after round reconstruction; Indicates the first The number of newly added authentication distortion areas after the refactoring; , These represent the weight coefficients for the remaining distortion term and the newly added distortion term, respectively.

[0103] When implemented, when the first Convergence index of the wheel When the value is no greater than the preset convergence threshold and the pass mark in the authentication consistency verification record is passed, output the modular design result of the automotive interior; when If the value still exceeds the preset convergence threshold, or if the pass flag still indicates failure, the target interior modules corresponding to the remaining authentication distortion regions and the newly added authentication distortion regions are sent back to S410 to continue the hidden compensation structure reconstruction. After this processing, the write-back result of S510, after being checked by S520, either forms the final output in S530 or forms a local feedback input in S530. The entire method maintains closed-loop consistency between parameter write-back, authentication consistency check, and iterative convergence.

[0104] It should be noted that all weighting coefficients involved in the various formulas of this invention (e.g.) The weights can be manually calibrated and assigned by those skilled in the art based on the expert experience of historical vehicle design, or they can be obtained through machine learning algorithms (such as multilayer perceptron networks or genetic algorithms) using historical qualified and unqualified interior assembly datasets as training samples to adaptively learn the optimal weight distribution. This invention does not limit the specific acquisition method.

[0105] Secondly, the subscript letters and index variables involved in the above embodiments and formulas (including but not limited to) Unless otherwise specified, all are positive integers, used to represent the index number, sequence number, or iteration round of the corresponding object.

[0106] Example 2: Figure 2 As shown, this embodiment provides a modular design system for automotive interiors based on a generative engine, including: The target design input set generation module is used to read the basic cabin data of the target vehicle, the candidate interior module set and the authentication witness token, establish a multi-agent system, and perform unified semantic parsing, role assignment, integrity verification, interface compatibility verification and authentication witness token matching verification on each candidate interior module to generate the target design input set. The combined association data packet generation module is used to extract the modules to be combined based on the target design input set, and the multi-agent system collaboratively generates a draft of the target interior combination scheme, and establishes a combined association data packet containing the spatial location, connection relationship, adjacency relationship and authentication witness token attachment relationship of the modules to be combined; The authentication distortion diagnostic set generation module is used to attach the authentication witness token to the corresponding node and edge of the composite association graph based on the target design input set and the composite association data packet, perform propagation superposition along the composite association graph to generate a composite authentication tensor graph, and identify the authentication distortion area based on the target vehicle model authentication boundary to generate an authentication distortion diagnostic set. The hidden compensation structure parameter set generation module is used to determine the scope of hidden compensation structure reconstruction based on the authentication distortion diagnosis set, the target design input set, and the combined associated data package. It then calls the generative engine to perform hidden compensation structure reconstruction on the target interior module or associated adjacent interior modules that cause distortion, and generates a hidden compensation structure parameter set. The design result output module is used to write back the hidden compensation structure parameter set to the target interior module or adjacent interior module in the target interior combination scheme, generate the reconstructed target interior combination scheme, and perform certification consistency verification on it. When the verification passes, the modular design result of the automotive interior is output. When the verification fails, the target interior module corresponding to the remaining certification distortion area or the newly added certification distortion area is sent back to perform hidden compensation structure reconstruction.

[0107] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.

[0108] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0109] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0110] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A modular design method for automotive interiors based on a generative engine, characterized in that, Includes the following steps: S1. Read the basic cabin data of the target vehicle, the candidate interior module set and the authentication witness token, establish the corresponding multi-agent system, perform unified semantic parsing, role assignment, integrity verification, interface compatibility verification and authentication witness token matching verification on each candidate interior module, and generate the target design input set. S2. Based on the target design input set, extract the modules to be combined, and the multi-agent system collaboratively generates a draft of the target interior combination scheme, and establishes a combination association data package containing the spatial location, connection relationship, adjacency relationship and authentication witness token attachment relationship of the modules to be combined; S3. Based on the target design input set and combined association data package, attach the authentication witness token to the corresponding node and edge of the combined association graph, perform propagation superposition along the combined association graph to generate a combined authentication tensor graph, and identify the authentication distortion area based on the target vehicle model authentication boundary to generate an authentication distortion diagnosis set. S4. Based on the authentication distortion diagnosis set, the target design input set, and the combined associated data package, determine the scope of hidden compensation structure reconstruction, call the generative engine to perform hidden compensation structure reconstruction on the target interior module or associated adjacent interior module that caused the distortion, and generate a hidden compensation structure parameter set.

2. The modular design method for automotive interiors based on a generative engine according to claim 1, characterized in that, Also includes: S5. Write the hidden compensation structure parameter set back to the target interior module or adjacent interior module in the target interior combination scheme, generate the reconstructed target interior combination scheme, and perform certification consistency verification on it. If the verification passes, output the modular design result of the automotive interior. If the verification fails, send back the target interior module corresponding to the remaining certification distortion area or the newly added certification distortion area to perform hidden compensation structure reconstruction.

3. The modular design method for automotive interiors based on a generative engine according to claim 1, characterized in that, S1 specifically includes: Read the basic cabin data, candidate interior module set, and authentication witness token of the target vehicle model. The basic cabin data includes at least the installation reference plane, airbag placement area, occupant contact area, and frame connection position. The candidate interior module set includes at least the module identifier, target installation area, connection position set, adjacent position set, and yield position set. Generate the original design record set. A multi-agent system is established based on the original design record set. A unified semantic parsing and role assignment are performed on each candidate interior module to generate a module token mapping set. Based on the module token mapping set, perform integrity verification, interface compatibility verification, and authentication witness token matching verification on the candidate interior modules. Delete candidate interior modules with incomplete fields, interface conflicts, or incomplete authentication witness token matching, and generate the target design input set.

4. The modular design method for automotive interiors based on a generative engine according to claim 1, characterized in that, S2 specifically includes: Read the target design input set, and extract the modules to be combined from the selected candidate interior modules based on the cabin layout requirements, module replacement requirements and installation boundary requirements of the target vehicle model, and generate the initial combination task set; The initial combination task set is collaboratively generated by a multi-agent system, which coordinates the spatial position, connection relationship, adjacency relationship and yielding relationship of the modules to be combined, and obtains a draft of the target interior combination scheme; Based on the draft of the target interior combination scheme, a combination association diagram is established, and the spatial location, connection relationship, adjacency relationship and authentication witness token attachment relationship of the modules to be combined are encapsulated into a combination association data package.

5. The modular design method for automotive interiors based on a generative engine according to claim 1, characterized in that, S3 specifically includes: Read the target design input set and the composite association data package, and attach the authentication witness tokens in the module token mapping set to the nodes and edges of the composite association graph according to their local or cross-module effects, to generate a token action relationship set; Based on the token action relationship set, the authentication witness token propagation and superposition are performed along the connection and adjacency relationships of the composite association graph. The comprehensive authentication action value of each module to be combined is calculated, and a composite authentication tensor graph is generated. Based on the comparison results between the combined authentication tensor map and the authentication boundary of the target vehicle model, the authentication distortion area is identified, and the distortion is determined to be due to expansion interference, contact sequence, insufficient clearance, or fracture boundary. The target interior module and the associated adjacent interior modules that cause the distortion are identified, and an authentication distortion diagnosis set is generated.

6. The modular design method for automotive interiors based on a generative engine according to claim 1, characterized in that, S4 specifically includes: Read the target design input set, combine the associated data package and the authentication distortion diagnosis set, and determine the scope of hidden compensation structure reconstruction based on the authentication distortion area, distortion type, target interior module that caused the distortion, associated adjacent interior modules and edge source identifier, and generate compensation reconstruction task set; The compensation and reconstruction agent in the generative engine is invoked, and the module-generated agent, interface-coordinating agent, and authentication-inference agent work together to perform hidden compensation structure reconstruction on the compensation and reconstruction task set and generate a set of compensation candidate structures.

7. The modular design method for automotive interiors based on a generative engine according to claim 6, characterized in that, Also includes: Based on the degree of recovery of the original authentication distortion area, the degree of preservation of interface relationships, and the degree of preservation of the overall stability of the target interior combination scheme, the compensation candidate structure that satisfies the authentication constraints and combination constraints is selected, and a hidden compensation structure parameter set is generated.

8. The modular design method for automotive interiors based on a generative engine according to claim 2, characterized in that, S5 specifically includes: Read the combined associated data packet and the hidden compensation structure parameter set, write the hidden compensation structure parameter set back to the target interior module or adjacent interior module in the target interior combination scheme, and synchronously update the node table, edge table and token attachment table to generate the reconstructed target interior combination scheme. A certification consistency check is performed on the reconstructed target interior combination scheme. Based on the updated combination certification tensor diagram, the residual distortion deviation value of the original certification distortion area, the number of newly added certification distortion areas, and the interface deviation after write-back, a certification consistency check record is generated.

9. The modular design method for automotive interiors based on a generative engine according to claim 8, characterized in that, Also includes: Based on the certification consistency verification record, the modular design result of the automotive interior is output when the verification passes. When the verification fails, the target interior module corresponding to the remaining or newly added certification distortion area is sent back for hidden compensation structure reconstruction.

10. A modular design system for automotive interiors based on a generative engine, employing the modular design method for automotive interiors based on a generative engine as described in any one of claims 1 to 9, characterized in that, include: The target design input set generation module is used to read the basic cabin data of the target vehicle model, the candidate interior module set, and the authentication witness token to establish a multi-agent system; The combined associated data packet generation module is used to extract the modules to be combined based on the target design input set, and the multi-agent system collaboratively generates a draft of the target interior combination scheme; The authentication distortion diagnostic set generation module is used to design the input set and composite association data packet according to the target, and attach the authentication witness token to the corresponding node and edge of the composite association graph; The hidden compensation structure parameter set generation module is used to determine the scope of hidden compensation structure reconstruction based on the authentication distortion diagnosis set, the target design input set, and the combined associated data packets. The design result output module is used to write back the hidden compensation structure parameter set to the corresponding target interior module or adjacent interior module in the target interior combination scheme, and generate the reconstructed target interior combination scheme.

Images