Automated customization methods for robot appearance, electronic devices and storage media

Abstract

This invention provides a method, electronic device, and storage medium for automated customization of robot appearance. The method includes: receiving multimodal input; performing hierarchical style semantic extraction using an artificial intelligence model to output a structured style description vector; using at least two artificial intelligence image generation pipelines employing different generation strategies to generate multiple candidate design schemes in parallel based on the structured style description vector; selecting a candidate design scheme from the multiple candidate design schemes; inputting the selected candidate design scheme into a 3D reconstruction model for 3D reconstruction to generate a corresponding 3D geometric model; automatically adapting the generated 3D geometric model to a target carrier to generate a fusion model; and obtaining the robot's appearance scheme based on the fusion model. This significantly reduces labor and time costs while greatly improving design output efficiency and the accuracy of adapting the 3D geometric model to the target carrier.

Description

Technical Field

[0001] This invention belongs to the field of 3D modeling technology, and in particular relates to an automated customization method for robot appearance, electronic equipment, and storage medium. Background Technology

[0002] With the rapid development of the service robot industry, consumers and enterprise clients are increasingly demanding personalized customization of robot appearances. A standardized robot's shell often looks identical, failing to reflect brand differentiation or user individuality. Meanwhile, generative artificial intelligence technology has made breakthroughs in image generation and 3D modeling, making it possible to dynamically model 3D models driven by natural language or reference images. Existing methods for customizing robot shells are as follows: Option A (purely manual option): The model is completed manually by a professional 3D modeler using 3D modeling software.

[0003] Option B (AI generation + manual adaptation): After generating an independent 3D model using an AI-generated model, professionals manually perform geometric adaptation operations and then manually adjust it to an assemblable state.

[0004] Option C (Semi-automatic adaptation scheme): A semi-automatic adaptation method based on surface projection (projecting the shell onto the shell surface) or mesh deformation (deforming the shell to fit the shell surface through control points).

[0005] In the process of implementing this invention, the inventors discovered the following problems with the prior art: Option A: Relies entirely on manual operation, with a single option design cycle ranging from several days to several weeks. The quality of output varies greatly among different modelers, resulting in poor consistency. It fails to provide users with the ability to quickly compare and select from multiple options.

[0006] Option B: The AI ​​generation and geometry adaptation processes are completely disconnected, requiring professional intervention. The geometry adaptation process remains manual, making it difficult to scale up. The adapted model still requires manual inspection for assembly feasibility.

[0007] Option C: In the surface projection method, when projecting the shell model onto the target shell surface, the design details (protrusions, grooves, reliefs, etc.) on the front of the shell are severely flattened or even completely lost, resulting in a significant difference between the projection result and the original design drawing, rendering it unusable. In the mesh deformation method, when deforming the shell using control grids to fit the shell surface, the deformation accuracy is insufficient, leading to penetrating interference in local areas (the shell penetrates into the shell's interior). Furthermore, this method cannot adaptively handle shell surfaces with complex hypercurvature. In the floating-point precision solution method, the geometric solution algorithm based on floating-point approximation fails to solve or generates a large amount of geometric fragmentation when dealing with common non-manifold boundaries (adjacent faces sharing edges of more than two faces) and non-closed surfaces (open surfaces with boundary edges) in the target carrier, making proper trimming impossible. Summary of the Invention

[0008] To address the aforementioned problems in the existing technology, this invention provides a method for automated customization of robot appearance, an electronic device, and a storage medium.

[0009] In a first aspect, embodiments of this disclosure provide a method for automated customization of robot appearance, including: Multimodal intent parsing steps: Receive multimodal input from the user, perform hierarchical style semantic extraction through an artificial intelligence model, and output a structured style description vector; Multi-path parallel design graph generation steps: Use at least two AI image generation pipelines with different generation strategies to generate multiple sets of candidate design schemes in parallel based on structured style description vectors. The candidate design schemes include design images. Image to 3D model generation steps: Select a candidate design scheme from multiple candidate design schemes, input the selected candidate design scheme into the 3D reconstruction model for 3D reconstruction, and generate the corresponding 3D geometric model; Adaptive model fusion steps: The generated 3D geometric model is automatically adapted to the target carrier to generate a fused model, and the robot's appearance scheme is obtained based on the fused model.

[0010] Optionally, automatically adapting the generated 3D geometric model to the target carrier to generate a fused model includes: Load the target carrier, the 3D geometric model, and the predefined fusion boundary mold; Perform coordinate system alignment transformation on the three-dimensional geometric model and scale it proportionally according to the bounding box size of the target carrier so that the geometric center of the three-dimensional geometric model is aligned with the center of the target carrier; A geometric solution algorithm is used, with the fusion boundary mold as the trimming tool, to perform difference operations on the scaled and aligned three-dimensional geometric model to remove the geometric parts that exceed the assembly boundary, and obtain the trimming result; Connectivity component detection is performed on the cropping results to identify all independent geometric regions; the geometric regions are sorted by the number of faces, and fragmented geometric regions with fewer than a preset threshold of faces are removed; Open boundary detection is performed on the clipping results after clearing the fragmented geometric region to identify all boundary edges that are associated with only a single facet. An edge extrusion operation is performed on all detected boundary edges to generate new vertices, and the new vertices are projected to the nearest surface point of the target carrier. Bridging facets are generated based on the original boundary vertices and the projected new vertices to fill the gap between the 3D geometric model and the target carrier, so that the fused model fits seamlessly with the surface of the target carrier at the boundary.

[0011] Optionally, the step of scaling the three-dimensional geometric model proportionally according to the bounding box size of the target carrier to align the geometric center of the three-dimensional geometric model with the center of the target carrier includes: The three-dimensional geometric model is scaled proportionally using a preset scaling factor to ensure that the three-dimensional geometric model completely covers the target carrier; an offset is applied to the three-dimensional geometric model along a set axis to control the coverage density and position of the three-dimensional geometric model completely covering the target carrier.

[0012] Optionally, the adaptive model fusion step may further include: Interactive preview and secondary adjustment steps: The generated fused model is rendered and previewed in real time in 3D, and the fused model is scaled and adjusted based on the user's interaction information. At the same time, interference detection is performed on the fused model and the target carrier in real time.

[0013] Optionally, the scaling and adjustment of the fused model based on user interaction information, and the real-time interference detection of the fused model and the target carrier, include: It allows users to perform multi-axis independent scaling operations on the fused model without changing the geometry of the target carrier, and detects the spatial relationship between the fused model and the target carrier in real time. When scaling of the fused model causes geometric interference, it provides a visual warning.

[0014] Optionally, the method of using at least two AI image generation pipelines employing different generation strategies to generate multiple candidate design schemes in parallel based on structured style description vectors further includes: The known viewpoint rendering of the target carrier is used as the input constraint for image generation, so that each generated candidate design scheme automatically maintains the same contour boundary and proportion relationship as the target carrier.

[0015] Optionally, the method of using at least two AI image generation pipelines with different generation strategies to generate multiple candidate design schemes in parallel based on structured style description vectors also includes: ensuring that all known viewpoint renderings are consistent in style, color, and shape details by semantically associating them with each viewpoint.

[0016] Optionally, when the target carrier includes multiple target components, a unified style semantic is distributed to the generation pipeline of each target component in the multi-path parallel design drawing generation step, thereby ensuring that the appearance style of multiple target components is consistent.

[0017] Secondly, embodiments of this disclosure also provide an electronic device, the electronic device comprising: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the automated customization method for robot appearance as described in any of the first aspects.

[0018] Thirdly, embodiments of this disclosure also provide a computer-readable storage medium storing computer instructions for causing a computer to perform the automated customization method for robot appearance as described in any of the first aspects.

[0019] The automated customization method for robot appearance provided by this invention first parses user input and extracts it into a structured style description vector. Then, it drives multiple AI pipelines employing different generation strategies to generate multiple candidate design schemes in parallel. After the user selects a scheme, the system converts it into a 3D geometric model and automatically and accurately adapts the 3D geometric model to the target carrier through adaptive model fusion. This achieves full-process automation from design concept to directly manufactureable model, thereby significantly reducing labor and time costs while greatly improving design output efficiency and the accuracy of adapting the 3D geometric model to the target carrier. Attached Figure Description

[0020] The above and other objects, features and advantages of this disclosure will become more apparent from the accompanying drawings, in which like reference numerals generally denote like parts.

[0021] Figure 1 A flowchart of a robot appearance automation customization method provided in this embodiment of the disclosure; Figure 2A timing diagram of the automated customization method for robot appearance provided in the embodiments of this disclosure; Figure 3 A flowchart of adaptive model fusion provided in this embodiment of the disclosure; Figure 4 This is a user interaction diagram provided for an embodiment of the present disclosure; Figure 5 A schematic block diagram of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0022] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0023] It should be understood that the following specific examples illustrate the implementation of this disclosure, and those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0024] It should be noted that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice method. Furthermore, this device and / or practice method can be implemented using other structures and / or functionalities besides one or more of the aspects set forth herein.

[0025] It should also be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of this disclosure. The illustrations only show the components related to this disclosure and are not drawn according to the number, shape and size of the components in actual implementation. In actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0026] Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.

[0027] The technical problem solved by this embodiment is as follows: (1) Traditional robot appearance customization relies entirely on manual operation by professional 3D modelers. A single customization solution usually takes several days to several weeks from conceptual design to assembleable 3D model. The labor cost is high and the delivery cycle is long, which cannot meet the growing demand for personalized customization.

[0028] (2) The three-dimensional models produced by existing artificial intelligence generative modeling technology (such as text-generated three-dimensional models) are independent geometric objects. Their outlines and size proportions do not match the target carrier (robot shell). They cannot automatically adapt to the curved outlines and installation boundaries of the target carrier and still require a lot of manual intervention for geometric trimming and alignment adjustment.

[0029] (3) Existing solutions can only generate single-style results. Users lack the ability to compare multiple solutions in parallel and iterate quickly, resulting in low satisfaction and a lengthened design cycle due to repeated modifications.

[0030] (4) There is a semantic gap between two-dimensional conceptual design drawings and three-dimensional assemblable models—the problem of automatically deriving multi-view consistent design drawings from single-view two-dimensional design drawings has not been effectively solved, leading to multi-view Figure 3 Geometric distortion occurs during dimensional reconstruction due to inconsistencies between viewpoints.

[0031] The basic generation pipeline is "text → image → 3D model": the user inputs a text description, the AI ​​image generation model produces a concept design drawing, and then the 3D reconstruction model converts the 2D image into a 3D geometry. However, the end point of this pipeline is an independent 3D model—it has no geometric connection with the actual robot shell and cannot be directly assembled and used.

[0032] This embodiment discloses a method for automated customization of robot appearance, including: Multimodal intent parsing steps: Receive multimodal input from the user, perform hierarchical style semantic extraction through an artificial intelligence model, and output a structured style description vector; Multi-path parallel design graph generation steps: Use at least two AI image generation pipelines with different generation strategies to generate multiple sets of candidate design schemes in parallel based on structured style description vectors. The candidate design schemes include design images. The artificial intelligence model and AI image generation pipeline can be implemented based on the applicant's Tripo 3D generation model. The generation pipeline in this embodiment is an automated process that transforms user-style descriptions into adaptive design drawings based on the AI ​​model and constrained by the target carrier outline.

[0033] The candidate design schemes specifically include the following key information: Multi-view consistent images: The candidate solution is not a single image, but a set of design drawings containing multiple known perspectives (such as front, side, etc.). These views are consistent in style, color, and shape details to solve the problem of geometric distortion caused by inconsistent perspectives when reconstructing from 2D to 3D.

[0034] Image to 3D model generation steps: Select a candidate design scheme from multiple candidate design schemes, input the selected candidate design scheme into the 3D reconstruction model for 3D reconstruction, and generate the corresponding 3D geometric model; Adaptive model fusion steps: The generated 3D geometric model is automatically adapted to the target carrier to generate a fused model, and the robot's appearance scheme is obtained based on the fused model.

[0035] Optionally, automatically adapting the generated 3D geometric model to the target carrier to generate a fused model includes: Load the target carrier, the 3D geometric model, and the predefined fusion boundary mold; Perform coordinate system alignment transformation on the three-dimensional geometric model and scale it proportionally according to the bounding box size of the target carrier so that the geometric center of the three-dimensional geometric model is aligned with the center of the target carrier; A geometric solution algorithm is used, with the fusion boundary mold as the trimming tool, to perform difference operations on the scaled and aligned three-dimensional geometric model to remove the geometric parts that exceed the assembly boundary, and obtain the trimming result; Connectivity component detection is performed on the cropping results to identify all independent geometric regions; the geometric regions are sorted by the number of faces, and fragmented geometric regions with fewer than a preset threshold of faces are removed; Open boundary detection is performed on the clipping results after clearing the fragmented geometric regions to identify all boundary edges that are associated with only a single facet. An edge extrusion operation is performed on all detected boundary edges to generate new vertices, and the new vertices are projected to the nearest surface point of the target carrier. Bridging faces are generated based on the original boundary vertices and the projected new vertices to fill the gap between the 3D geometric model and the target carrier, so that the fused model fits seamlessly with the surface of the target carrier at the boundary, thereby solving the problem of visible gaps between the outer shell and the inner shell after Boolean clipping.

[0036] Edge extrusion refers to the operation of pulling out a new geometric surface along the normal direction or a specified direction of the selected edge in 3D modeling.

[0037] Optionally, the step of scaling the three-dimensional geometric model proportionally according to the bounding box size of the target carrier to align the geometric center of the three-dimensional geometric model with the center of the target carrier includes: The three-dimensional geometric model is scaled proportionally using a preset scaling factor to ensure that the three-dimensional geometric model completely covers the target carrier; an offset is applied to the three-dimensional geometric model along a set axis to control the coverage density and position of the three-dimensional geometric model completely covering the target carrier.

[0038] Optionally, the adaptive model fusion step may further include: Interactive preview and secondary adjustment steps: The generated fused model is rendered and previewed in real time in 3D, and the fused model is scaled and adjusted based on the user's interaction information. At the same time, interference detection is performed on the fused model and the target carrier in real time.

[0039] Optionally, the scaling and adjustment of the fused model based on user interaction information, and the real-time interference detection of the fused model and the target carrier, include: It allows users to perform multi-axis independent scaling operations on the fused model without changing the geometry of the target carrier, and detects the spatial relationship between the fused model and the target carrier in real time. When scaling of the fused model causes geometric interference, it provides a visual warning.

[0040] Optionally, the method of using at least two AI image generation pipelines employing different generation strategies to generate multiple candidate design schemes in parallel based on structured style description vectors further includes: The known viewpoint rendering of the target carrier is used as the input constraint for image generation, so that each generated candidate design scheme automatically maintains the same contour boundary and proportion relationship as the target carrier.

[0041] Each design scheme is generated under a template-constrained image generation mechanism. That is, during generation, a known viewpoint rendering of the target carrier (such as robot chest armor, helmet, etc.) is used as the input constraint. Therefore, the design drawings in the candidate design schemes automatically maintain the outline boundaries and proportions that are completely consistent with the target carrier, but the internal design elements (such as lines, patterns, color distribution) are freely created according to the user's intention and different generation strategies (style transfer, image editing, etc.).

[0042] Users select a satisfactory multi-view design from multiple candidate 2D design schemes. This selected multi-view design is then input into an image-to-3D model generation system. Examples include diffusion-based models, Neural Radiation Field (NeRF)-based reconstruction models, 3D Gaussian splash-based reconstruction models, and multi-view stereo matching-based reconstruction models, thereby generating a 3D geometric model.

[0043] Optionally, the method of using at least two AI image generation pipelines with different generation strategies to generate multiple candidate design schemes in parallel based on structured style description vectors also includes: ensuring that all known viewpoint renderings are consistent in style, color, and shape details by semantically associating them with each viewpoint.

[0044] Optionally, when the target carrier includes multiple target components, a unified style semantic is distributed to the generation pipeline of each target component in the multi-path parallel design drawing generation step, thereby ensuring that the appearance style of multiple target components is consistent.

[0045] In a specific application scenario, such as Figure 1 and Figure 2 As shown, the method in this embodiment includes the following stages: Phase 1: Multimodal Intent Parsing This stage accepts multimodal mixed input from users (natural language text descriptions and / or style reference images), and performs hierarchical style semantic extraction using a large language model. The large language model can be one such as Qianwen or Doubao.

[0046] The hierarchical style semantic extraction includes four levels: First layer: Extract the overall style and layout direction (such as mecha style, rounded cartoon style, cyberpunk style, etc.); Second layer: Extract the shape details of each area (the division method and decorative features of the upper area, middle area, lower area, and side area); The third layer: Extracting the color distribution scheme (primary color, secondary color, accent color and their spatial distribution relationship); Fourth layer: Extract the characteristics of decorative elements (the specific shape and position of lines, relief patterns, inlays, etc.); This stage outputs a structured style description vector, which serves as the semantic driving signal for subsequent image generation stages. It also outputs a user-oriented design summary for human-computer interaction confirmation.

[0047] Phase Two: Generation of Multi-Path Parallel Design Graphs The system simultaneously launches at least two independent AI image generation pipelines, each employing a different generation strategy (such as style transfer route, image editing route, etc.) to generate multiple candidate design schemes in parallel.

[0048] Geometric constraint mechanism: Each generation pipeline uses a known viewpoint rendering of the target carrier as input constraints (i.e., "template-constrained generation"). This mechanism ensures that the AI-generated design drawings automatically maintain the same contour boundaries and scale relationships as the target carrier, meaning that no matter how creative the AI ​​model is, its output design drawings always remain within the geometric contours of the target carrier.

[0049] Multi-perspective consistency guarantee: The system ensures a high degree of consistency in style, color, and shape details among design drawings from different perspectives through a semantic association mechanism, providing high-quality input for subsequent 3D reconstruction.

[0050] The generation process uses a server-side push mechanism to provide real-time feedback on the progress of each pipeline to the user, allowing the user to freely compare and choose from multiple solutions once all are ready.

[0051] Component selection and customization range: The system allows users to select customized target parts, including but not limited to chest armor, helmet, arm guards, leg armor, and shoulder armor. Each part corresponds to an independent standard part model and a fusion boundary mold.

[0052] The system also supports a "full-body integrated customization" mode: users can customize all the robot's exterior parts in a unified style at once, and the system will automatically apply the same style semantics to the generation pipeline of each part to ensure that the style of all parts of the whole body is consistent and coordinated.

[0053] Users can also choose to customize individual components independently, offering a high degree of flexibility.

[0054] Phase 3: Image to 3D Model Generation The system converts multi-view design drawings of user-selected schemes into 3D models to generate large-scale models. These large-scale models take data in various forms, such as images, voxels, or point clouds, as input and generate high-precision 3D models including surface textures (PBR materials) end-to-end, without requiring manual intervention in the modeling process.

[0055] Image-to-3D model generation can also be used to create large models, such as 3D generation models based on diffusion models, reconstruction models based on neural radiation fields, reconstruction models based on 3D Gaussian splashing, and reconstruction models based on multi-view stereo matching.

[0056] Phase 4, Adaptive Model Fusion: like Figure 3 As shown, the adaptive model fusion algorithm automatically adapts the initial 3D model generated by artificial intelligence to the target carrier, outputting a directly manufacturable industrial-grade model. The specific steps of adaptive model fusion are as follows: Step 1: Model Import and Preprocessing Load the target carrier standard parts, the initial 3D model generated by artificial intelligence, and the predefined fusion boundary mold. Perform transformation and normalization processing on each model.

[0057] Step 2, Spatial Coordinate Alignment and Adaptive Scaling: Perform coordinate system alignment transformation on the initial model to align it with the target carrier; scale the initial model proportionally using a preset scaling factor based on the bounding box size of the target carrier to make it slightly larger than the target carrier to ensure complete coverage; align the geometric center of the scaled model with the center of the target carrier; support applying fine-tuning offsets along a specified axis to precisely control coverage density and position.

[0058] Step 3, Precise geometric blending and cropping: A precise geometry solving algorithm (non-floating-point approximation) is employed, using the fusion boundary mold as a trimming tool. It performs a difference operation on the scaled and aligned initial model to automatically remove redundant geometry exceeding the assembly boundary. This precise geometry solving algorithm can directly handle industrial-grade fusion molds containing geometric defects such as non-manifold boundaries and non-closed surfaces, without requiring any preprocessing or repair of the mold.

[0059] Geometric problem-solving algorithms refer to systematic methods used to solve geometric problems (such as length, angle, area, volume, positional relationships, etc.), including coordinate methods and volume formula methods. The coordinate method (spatial vectors) establishes a three-dimensional coordinate system and uses vector operations to solve for line-to-line angles, line-to-plane angles, plane-to-plane angles, point-to-plane distances, etc.

[0060] Step 4: Multi-level topology analysis and debris cleaning: The trimmed results undergo connected component detection to identify all independent geometric regions (facet islands). They are then sorted by the number of faces, and fragmented islands below a preset threshold are automatically removed, retaining only the main geometric regions. This strategy ensures the topological integrity and fabrication feasibility of the output model.

[0061] Boundary Gap Filling: The system performs open boundary detection on the clipping results after fragment cleaning, identifying all boundary edges in the model that are referenced by only a single facet. These boundary edges are typically open contours generated by geometric difference operations, with visible gaps between them and the target carrier surface. The system performs edge extrusion on all boundary edges, generating new vertices (extended vertices). Each new vertex is then projected to its nearest point on the target carrier surface, and bridging faces are automatically generated between the original boundary vertices and the projected new vertices. This ensures a seamless fit between the cut boundaries of the fused model and the target carrier surface, eliminating visual gaps and structural discontinuities caused by geometric clipping.

[0062] Step 5: Exporting the industrial-grade model: The merged shell model (fused model) and the target carrier are exported separately into common 3D formats (such as GLB / STL / OBJ / 3MF, etc.). The output model meets industrial manufacturing standards and can be directly used in manufacturing processes such as CNC machining, injection molding, and 3D printing without the need for secondary mold modification. This end-to-end design-to-manufacturing pathway has been validated as a closed loop on an actual production line.

[0063] Phase 5: Interactive Preview and Secondary Adjustments A 3D interactive scene is built on the browser side based on a real-time rendering engine, and the fused result is presented to the user.

[0064] Secondary customization of the outer shell after fusion: Users can independently scale the merged shell model along multiple axes (X / Y / Z axes controlled separately) while keeping the target carrier unchanged, achieving fine-tuning of the coverage effect. One-click simultaneous scaling along all three axes is also supported.

[0065] Real-time interference detection: The system continuously monitors the spatial relationship between the shell model and the carrier. When scaling of the shell model causes geometric interference (penetrating the interior of the carrier), the system immediately issues a visual warning, guiding the user to adjust to a safe range.

[0066] Multi-modal visualization: It offers multiple visualization methods, such as solid rendering, wireframe mode, semi-transparent shell model mode, and carrier visibility switching, to help users comprehensively evaluate the fusion effect from multiple perspectives.

[0067] like Figure 4 As shown, this embodiment encapsulates complex technical processes within the system, allowing users to complete the entire process with minimal effort: a complete customization requires only 3-5 active steps. Steps 1 and 2: Enter style description and select customization target: Users can enter a natural language style description (such as "cyberpunk style, streamlined armor") in the system interface, optionally upload reference images, and then select the target custom part to start generating the product with one click.

[0068] Step 3: Select a satisfactory solution from multiple candidate solutions: The system automatically generates multiple candidate design schemes (including front and side views) in parallel. Users can directly click to select the satisfactory scheme and adjust the armor wrapping depth parameters.

[0069] Step 4 (optional): Interactively adjust in the 3D previewer: After automatically completing 3D reconstruction and model fusion, the system presents a real-time 3D preview in the browser. Users can perform multi-axis scaling and fine-tuning of the fused shell, and the system displays the interference detection results in real time. Switching the display mode allows you to view the adaptation effect between the fusion model and the target carrier (green), and verify the fit between the outer shell and the housing.

[0070] Step 5: Download the manufacturing-ready model → Put it directly into production: After the user confirms the design, they can download the industrial-grade model file with one click, which can be directly used in manufacturing processes such as 3D printing and CNC machining. Below are actual images of the 3D printed products.

[0071] Throughout the entire process, all technical aspects, including 3D modeling, model fusion, and topology cleaning, are completed automatically by the system, requiring no prior 3D modeling expertise from the user. The end-to-end time from input to obtaining a manufacturable model is controlled within minutes.

[0072] This embodiment has the following effects: (1) Full-process automation: From inputting natural language or images to outputting an assembleable 3D model, the entire process requires no professional 3D modeling skills and can be completed in minutes (existing solutions take several days to weeks).

[0073] (2) Parallel comparison of multiple schemes: The system automatically generates multiple candidate design schemes simultaneously for users to compare and choose from in real time, greatly improving the initial satisfaction rate and reducing the number of iterations for repeated modifications.

[0074] (3) Integrating quality advantages: The adaptive model fusion algorithm has been verified by a large number of engineering experiments (more than 25 rounds of scheme iterations). It is significantly better than existing schemes such as surface projection method, mesh deformation method, and floating point approximation solution method in terms of detail preservation, geometric adaptation accuracy, and industrial compatibility.

[0075] (4) Adjustments are still possible after fusion: Unlike existing "one-time generation and unchangeable" solutions, the fusion results in this embodiment allow users to perform real-time secondary scaling adjustments on the browser side. Interference detection is used to ensure assemblability, achieving a "what you see is what you get" interactive experience.

[0076] (5) Industrial-grade robustness: The fusion algorithm in this embodiment uses precise geometric solution (non-floating point approximation), which can directly process real industrial molds with geometric defects such as non-manifold boundaries and non-closed surfaces without preprocessing repair, thus reducing the threshold for engineering deployment.

[0077] (6) Manufacturing-ready output: The system outputs a fusion model that can be directly applied to industrial manufacturing processes such as 3D printing, CNC machining, and injection molding, achieving a complete end-to-end closed loop from design concept to physical manufacturing (which has been verified on the actual production line).

[0078] (7) Flexible customization granularity: It supports both individual component customization and whole-body integrated customization, balancing precision and efficiency.

[0079] The large language model in this embodiment can be any text understanding model (including but not limited to various large-scale pre-trained language models, open-source language models, etc.). Visual understanding employs any multimodal understanding model (including but not limited to visual-language joint models, image description models, etc.). Input modalities can include voice input, hand-drawn sketch input, 3D scan input, etc.

[0080] Image generation methods can employ any generation method or combination thereof, such as style transfer, image editing, text-to-image generation, conditional control generation, and image adapter.

[0081] The number of parallel pipelines can be expanded to N (N≥2), using N different generation strategies.

[0082] Geometric constraints can be implemented through various mechanisms such as image stitching, conditional coding, and attention guidance.

[0083] The image-to-3D model generation can be any similar 3D model, including but not limited to: 3D generation models based on diffusion models, reconstruction models based on neural radiation fields, reconstruction models based on 3D Gaussian splashing, reconstruction models based on multi-view stereo matching, and 3D generation models based on variational autoencoders.

[0084] The input can be an image, voxel, point cloud, or a combination thereof.

[0085] Difference operations can be performed using voxelized Boolean operations, construct solid geometry operations, implicit surface intersection, mesh trimming, etc. Fragment cleaning can be based on any strategy, such as a threshold for the number of patches, a threshold for volume, the area of ​​connected components, surface integrals, or machine learning classifiers.

[0086] Scaling alignment strategies can be based on feature point matching, ICP registration, semantic alignment, and other methods.

[0087] Browser-based 3D rendering can be used for desktop client rendering, mobile augmented reality previews, and virtual reality immersive previews.

[0088] Interference detection can employ any collision detection method, such as bounding box detection, separating axis theorem, ray projection, GJK algorithm, and mesh cross detection.

[0089] Scaling adjustments can be expanded to support more degrees of freedom, such as rotation, translation, and local deformation.

[0090] The target carrier is not limited to a single component; it can be any assembleable component such as a chest armor, helmet, arm guards, leg armor, or shoulder armor, or it can be a customized integrated appearance of the entire robot.

[0091] The applicable objects are not limited to humanoid robots, but can be extended to various carrier forms such as quadruped robots, wheeled robots, industrial robots, consumer robots, drone shells, and smart wearable devices.

[0092] Input is not limited to text and images, but can be extended to any modality such as voice, sketches, 3D scans, and existing 3D models.

[0093] The inventors conducted more than 25 rounds of iterative experiments on the model fusion algorithm, systematically verifying and eliminating the following six types of technical approaches: namely, the six types of technical approaches shown in Table 1 cannot solve the technical problem of this embodiment: Table 1. Technical Route Defects Table The inventors ultimately determined the technical approach of accurate geometric solution and connected component topology cleaning in this embodiment. This solution simultaneously meets four core requirements: detail preservation, curvature matching, industrial mold compatibility, and automatic debris removal.

[0094] The meanings of the terms are shown in Table 2: Table 2. Glossary of Terms This embodiment can be deployed as an online service, supporting concurrent use by multiple users, with the average time for a single customization process completed within minutes.

[0095] The electronic device disclosed in this embodiment includes a memory and a processor. The memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0096] The processor may be a central processing unit (CPU) or other processing unit with data processing and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of this disclosure, the processor is used to execute computer-readable instructions stored in the memory, causing the electronic device to perform all or part of the steps of the automated customization method for robot appearance described in the foregoing embodiments of this disclosure.

[0097] Those skilled in the art will understand that, in order to solve the technical problem of how to achieve a good user experience, this embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of this disclosure.

[0098] like Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present disclosure. It illustrates a structural schematic diagram suitable for implementing the electronic device in the embodiment of the present disclosure. Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments disclosed herein.

[0099] like Figure 5 As shown, an electronic device may include a processing unit (such as a central processing unit, graphics processing unit, etc.) that can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) or a program loaded from a storage device into random access memory (RAM). The RAM also stores various programs and data required for the operation of the electronic device. The processing unit, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0100] Typically, the following devices can be connected to the I / O interface: input devices, such as sensors or visual information acquisition devices; output devices, such as displays; storage devices, such as magnetic tapes or hard drives; and communication devices. Communication devices allow electronic devices to exchange data wirelessly or via wired communication with other devices, such as edge computing devices. Although Figure 5 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have alternatively.

[0101] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processing device, all or part of the steps of the automated customization method for robot appearance according to embodiments of this disclosure are performed.

[0102] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0103] A computer-readable storage medium according to embodiments of the present disclosure stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the automated robot appearance customization method described in the foregoing embodiments of the present disclosure are performed.

[0104] The aforementioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).

[0105] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0106] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0107] In this disclosure, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The block diagrams of devices, apparatuses, devices, and systems involved in this disclosure are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as "comprising," "including," "having," etc., are open-ended terms meaning "including but not limited to," and are used interchangeably with them. The terms "or" and "and" as used herein refer to the terms "and / or," and are used interchangeably with them unless the context clearly indicates otherwise. The term "such as" as used herein refers to the phrase "such as but not limited to," and is used interchangeably with it.

[0108] Additionally, as used herein, the "or" used in a list of items beginning with "at least one" indicates a separate list, such that a list of, for example, "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not imply that the described example is preferred or better than other examples.

[0109] It should also be noted that in the systems and methods of this disclosure, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0110] Various changes, substitutions, and modifications can be made to the techniques described herein without departing from the teachings defined in this embodiment. Furthermore, the scope of this embodiment is not limited to the specific aspects of the processes, machines, manufacturing processes, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufacturing processes, events, means, methods, or actions that perform substantially the same functions or achieve substantially the same results as the corresponding aspects described herein can be utilized. Therefore, this embodiment includes such processes, machines, manufacturing processes, events, means, methods, or actions within its scope.

[0111] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0112] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A method for automated customization of robot appearance, characterized in that, include: Multimodal intent parsing steps: Receive multimodal input from the user, perform hierarchical style semantic extraction through an artificial intelligence model, and output a structured style description vector; Multi-path parallel design graph generation steps: Use at least two AI image generation pipelines with different generation strategies to generate multiple sets of candidate design schemes in parallel based on structured style description vectors. The candidate design schemes include design images. Image to 3D model generation steps: Select a candidate design scheme from multiple candidate design schemes, input the selected candidate design scheme into the 3D reconstruction model for 3D reconstruction, and generate the corresponding 3D geometric model; Adaptive model fusion steps: The generated 3D geometric model is automatically adapted to the target carrier to generate a fused model, and the robot's appearance scheme is obtained based on the fused model.

2. The automated customization method for robot appearance according to claim 1, characterized in that, The process of automatically adapting the generated 3D geometric model to the target carrier to generate a fused model includes: Load the target carrier, the 3D geometric model, and the predefined fusion boundary mold; Perform coordinate system alignment transformation on the three-dimensional geometric model and scale it proportionally according to the bounding box size of the target carrier so that the geometric center of the three-dimensional geometric model is aligned with the center of the target carrier; A geometric solution algorithm is used, with the fusion boundary mold as the trimming tool, to perform difference operations on the scaled and aligned three-dimensional geometric model to remove the geometric parts that exceed the assembly boundary, and obtain the trimming result; Connectivity component detection is performed on the cropping results to identify all independent geometric regions; the geometric regions are sorted by the number of faces, and fragmented geometric regions with fewer than a preset threshold of faces are removed; Open boundary detection is performed on the clipping results after clearing the fragmented geometric region to identify all boundary edges that are associated with only a single facet. An edge extrusion operation is performed on all detected boundary edges to generate new vertices, and the new vertices are projected to the nearest surface point of the target carrier. Bridging facets are generated based on the original boundary vertices and the projected new vertices to fill the gap between the 3D geometric model and the target carrier, so that the fused model fits seamlessly with the surface of the target carrier at the boundary.

3. The automated customization method for robot appearance according to claim 2, characterized in that, The step of scaling the three-dimensional geometric model proportionally according to the bounding box size of the target carrier to align the geometric center of the three-dimensional geometric model with the center of the target carrier includes: The three-dimensional geometric model is scaled proportionally using a preset scaling factor to ensure that the three-dimensional geometric model completely covers the target carrier; an offset is applied to the three-dimensional geometric model along a set axis to control the coverage density and position of the three-dimensional geometric model completely covering the target carrier.

4. The automated customization method for robot appearance according to claim 1, characterized in that, The adaptive model fusion step is followed by: Interactive preview and secondary adjustment steps: The generated fused model is rendered and previewed in real time in 3D, and the fused model is scaled and adjusted based on the user's interaction information. At the same time, interference detection is performed on the fused model and the target carrier in real time.

5. The automated customization method for robot appearance according to claim 4, characterized in that, The scaling and adjustment of the fused model based on user interaction information, and the real-time interference detection of the fused model and the target carrier, include: It allows users to perform multi-axis independent scaling operations on the fused model without changing the geometry of the target carrier, and detects the spatial relationship between the fused model and the target carrier in real time. When scaling of the fused model causes geometric interference, it provides a visual warning.

6. The automated customization method for robot appearance according to claim 1, characterized in that, The method of using at least two AI image generation pipelines with different generation strategies to generate multiple candidate design schemes in parallel based on structured style description vectors also includes: The known viewpoint rendering of the target carrier is used as the input constraint for image generation, so that each generated candidate design scheme automatically maintains the same contour boundary and proportion relationship as the target carrier.

7. The automated customization method for robot appearance according to claim 6, characterized in that, The method of using at least two AI image generation pipelines with different generation strategies to generate multiple candidate design schemes in parallel based on structured style description vectors also includes: ensuring that the design schemes generated for the same target carrier from different perspectives are consistent in style, color and shape details by semantic association between all known viewpoint renderings.

8. The automated customization method for robot appearance according to claim 1, characterized in that, When the target carrier includes multiple target components, a unified style semantic is distributed to the generation pipeline of each target component in the multi-path parallel design drawing generation step, thereby ensuring that the appearance style of multiple target components is consistent.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the automated customization method for robot appearance as described in any one of claims 1-8.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing a computer to perform the automated customization method for the robot appearance as described in any one of claims 1-8.

Images