Data processing system

Abstract

The present application relates to a system that automatically generates multiple product design options using generative AI based on user inputted use, size, material, color, and other constraints, and automatically outputs a script for generating a three-dimensional model that can be used in three-dimensional modeling software based on the user selected option. Thus, even a user without professional three-dimensional modeling knowledge can easily select the best option from multiple design options in a short time and automatically generate a specific three-dimensional model, thereby greatly improving the efficiency of the initial design and prototype manufacturing engineering of new product development.

Description

Technical Field

[0001] The technology disclosed herein relates to a system. Background Technology

[0002] Patent Document 1 discloses a personalized chatbot control method executed by at least one processor, the method comprising: receiving user speech; adding the user speech to a prompt containing instructions related to a chatbot role; encoding the prompt; and inputting the encoded prompt into a language model to generate chatbot speech in response to the user speech.

[0003] Patent document 1: Japanese Patent Application Publication No. 2022-180282.

[0004] The problem this invention aims to solve is that, during the planning stage of new products, creating concrete 3D models from a large number of product concepts or ideas requires a high level of 3D modeling technology or expertise, which in turn requires a significant amount of time and labor. Specifically, it addresses the difficulty for planning leaders to quickly and accurately concretize their ideas into 3D models, hindering efficient visualization or verification, and ultimately making proposals or creating prototypes for internal and external stakeholders. The invention addresses this by providing an environment that can quickly generate multiple design options reflecting various design requirements and constraints, allowing users to easily select and modify options. Based on these selections, a 3D model generation script is automatically output for use in 3D modeling software. This enables even users without specialized knowledge to create 3D models intuitively and efficiently, significantly improving the efficiency and labor-saving aspects of the initial design and prototyping processes in new product development. Summary of the Invention

[0005] The means to solve the problem of the present invention is to provide a system including a user input unit, a generative AI processing unit, and a 3D model generation unit. The user input unit has the function of receiving design requirements from the user, including the product's purpose, size, material, color, and other constraints. The generative AI processing unit has the following functions: receiving the design requirements received from the user input unit as input data, automatically generating multiple product design schemes based on the design requirements, and assigning each design scheme features that reflect the shape optimized for the purpose, specified materials and colors, and other constraints. The 3D model generation unit automatically outputs a 3D model generation script that can be used in 3D modeling software based on the content of the design scheme selected by the user from the design schemes generated by the generative AI processing unit. This script automatically generates a 3D model that reflects the specified size, color, and material information. By adopting such a structure, even if the user does not have professional 3D modeling skills or knowledge, they can easily select the best option from multiple design options in a short time and automatically generate a specific 3D model, thereby greatly improving the efficiency and labor-saving of the initial design and prototyping process of new product development. Attached Figure Description

[0006] Figure 1 This is a conceptual diagram illustrating an example of the configuration of a data processing system according to the first embodiment.

[0007] Figure 2 This is a conceptual diagram illustrating an example of the main functions of the data processing apparatus and smart device according to the first embodiment.

[0008] Figure 3 This is a conceptual diagram illustrating an example of the data processing system configuration in the second embodiment.

[0009] Figure 4 This is a conceptual diagram illustrating an example of the main functions of the data processing device and smart glasses according to the second embodiment.

[0010] Figure 5 This is a conceptual diagram illustrating an example of the data processing system configuration in the third embodiment.

[0011] Figure 6 This is a conceptual diagram illustrating an example of the main functions of the data processing device and head-mounted terminal according to the third embodiment.

[0012] Figure 7 This is a conceptual diagram illustrating an example of the data processing system configuration in the fourth embodiment.

[0013] Figure 8 This is a conceptual diagram illustrating an example of the functions of the main parts of the data processing device and robot according to the fourth embodiment.

[0014] Figure 9 It represents an emotion graph that maps multiple emotions.

[0015] Figure 10 It represents an emotion graph that maps multiple emotions.

[0016] Explanation of reference numerals in the attached figures Data processing systems 10, 210, 310, and 410 12 Data processing device 14 Smart devices 214 Smart Glasses 314 Head-mounted terminal 414 Robot. Detailed Implementation

[0017] Hereinafter, an example of an implementation of the system involved in this disclosure will be described with reference to the accompanying drawings.

[0018] First, let's explain the terms used in the following description.

[0019] In the following embodiments, the processor (hereinafter referred to as "processor") can be a single computing device or a combination of multiple computing devices. Furthermore, a processor can be a single computing device or a combination of multiple computing devices. Examples of computing devices include CPU (Central Processing Unit), GPU (Graphics Processing Unit), GPGPU (General-Purpose computing on Graphics Processing Units), APU (Accelerated Processing Unit), etc.

[0020] In the following implementation, the labeled RAM (Random Access Memory) is a memory that temporarily stores information and is used by the processor as working memory.

[0021] In the following embodiments, the labeled memory is one or more non-volatile storage devices used to store various programs and various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), disk (e.g., hard disk) or magnetic tape, etc.

[0022] In the following implementation, the labeled Communication I / F (Interface) is an interface that includes a communication processor and an antenna, etc. The Communication I / F is responsible for communication between multiple computers. Examples of communication standards applicable to the Communication I / F include wireless communication standards such as 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), or Bluetooth (registered trademark).

[0023] In the following implementation, "A and / or B" is synonymous with "at least one of A and B". That is, "A and / or B" means that it can be only A, only B, or a combination of A and B. Furthermore, in this specification, when "and / or" connects more than three items, the same approach as "A and / or B" applies.

[0024] [First Implementation] Figure 1 An example of the configuration of the data processing system 10 according to the first embodiment is shown.

[0025] like Figure 1 As shown, the data processing system 10 includes a data processing device 12 and an intelligent device 14. An example of the data processing device 12 is a server.

[0026] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0027] The smart device 14 includes a computer 36, a receiver 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the receiver 38, output device 40, camera 42, and communication I / F 44 are also connected to the bus 52.

[0028] The receiving device 38 includes a touchscreen 38A and a microphone 38B, etc., for receiving user input. The touchscreen 38A receives user input generated by contact with an indicator (e.g., a pen or finger). The microphone 38B receives user input generated by sound by detecting the user's voice. The control unit 46A in the processor 46 sends data representing user input received through the touchscreen 38A and microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data representing the user input.

[0029] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting data in a form perceptible to the user 20 (e.g., sound and / or text). The display 40A displays visual information such as text and images according to instructions from the processor 46. The speaker 40B outputs sound according to instructions from the processor 46. The camera 42 is a small digital camera equipped with an optical system such as a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0030] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for sending and receiving various information between processor 46 and processor 28 via network 54.

[0031] Figure 2 An example of the main functions of the data processing device 12 and the smart device 14 is shown.

[0032] like Figure 2 As shown, in the data processing apparatus 12, specific processing is performed by the processor 28. A specific processing program 56 is stored in the memory 32. The specific processing program 56 is an example of a "program" as understood in this disclosure. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 based on the specific processing program 56 executed on the RAM 30.

[0033] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0034] In the smart device 14, the processor 46 performs the receive / output processing. The memory 50 stores the receive / output program 60. The receive / output program 60 is used in conjunction with the data processing system 10 and the specific processing program 56. The processor 46 reads the receive / output program 60 from the memory 50 and executes the read receive / output program 60 on the RAM 48. The receive / output processing is implemented by the processor 46 acting as a control unit 46A based on the receive / output program 60 executed on the RAM 48. Furthermore, the smart device 14 may also have the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and use these models to perform the same processing as the specific processing unit 290. The receive / output processing is implemented by the processor 46 acting as a control unit 46A based on the receive / output program 60 executed on the RAM 48.

[0035] In addition, other devices besides the data processing device 12 may also have a data generation model 58.

[0036] For example, a server device (e.g., a generation server) may have a data generation model 58.

[0037] In this case, the data processing device 12 obtains the processing results (prediction results, etc.) using the data generation model 58 by communicating with a server device that has the data generation model 58. Furthermore, the data processing device 12 can be a server device or a user-owned terminal device (e.g., a mobile phone, robot, home appliance, etc.). Next, an example of the processing performed by the data processing system 10 of the first embodiment will be described.

[0038] (Example 1) The specific processing flow in Example 1 will be described below. Each part of the system described below is implemented by the data processing device 12 and the smart device 14. Furthermore, the data processing device 12 is referred to as the "server" and the smart device 14 is referred to as the "terminal".

[0039] (Implementation methods for carrying out the present invention) A more specific and detailed description is given regarding one embodiment of the present invention. The system of the present invention comprises a terminal and a server, employing a structure where the terminal provides the user interface, and the server handles generative AI processing and 3D model generation. The terminal can utilize any information processing device capable of network connectivity, such as a personal computer, tablet, smartphone, wearable device, or dedicated business terminal. The server utilizes computing resources with generative AI computing capabilities and 3D model generation capabilities, such as cloud servers, local servers, distributed processing servers, and edge servers. Specifically, the system adopts a client-server architecture. The system uses a secure encryption protocol in communication between the terminal and the server. The server has a parallel computing cluster using GPUs. The server executes each processing unit as a containerized microservice. This structure centralizes computationally demanding generative AI inference processing on the server side. This structure significantly relaxes the hardware requirements on the terminal side. The system suppresses terminal battery consumption. The system dramatically improves processing speed. The server distributes requests to multiple computing nodes via a load balancer. The server maintains stable response times even with surges in traffic. The system keeps large-scale language models and image generation models resident in memory. This system minimizes inference latency. It enables 3D model rendering to be performed on the server side. It sends only lightweight drawing commands to the terminal. It reduces network bandwidth consumption. This system is applicable to the early design phase in manufacturing. It significantly reduces designers' trial-and-error time. It enables non-experts to perform advanced 3D design. It improves the resource management efficiency of computer technology itself.

[0040] First, the user input area on the terminal offers multiple input methods, including GUI, voice input, touch operation, pen input, and importing external files. For example, users can specify a wide variety of uses, such as "small storage box for desktop use," "weather-resistant bench for outdoor use," "safe toys for children," "casing for electronic devices," "portable folding containers," "medical equipment storage boxes," and "assembled exhibition equipment," through natural language, options, or by uploading images and drawings. Regarding dimensions, users can input specific numerical values ​​(e.g., 10 cm high, 5 cm wide, 8 cm deep), relative sizes (e.g., the size to hold A4 paper, the size of a palm), and functional constraints such as "maximum weight capacity of 2 kg." Regarding materials, users can specify various options based on considerations such as intended use, cost, environmental impact, processability, strength, weight, and texture, including plastics, metals, wood, glass, ceramics, composite materials, recycled materials, heat-resistant materials, and antibacterial materials. Regarding color, it can provide abstract instructions such as monochrome, multicolor, gradient, pattern, brand color, color matching code (e.g., PANTONE code), and even "bright colors that children like" and "subdued tones." As other constraints, it can simultaneously specify multiple extremely diverse and complex conditions such as "with drawers," "stackable," "waterproof," "impact resistant," "assembly type," "removable and washable," "complies with safety standards," "cost limit," "environmentally friendly," "incorporation of specific logos or designs," and "interchangeability with specific parts." Specifically: The user input section of this terminal converts various input data into structured data. This input section tokenizes text input based on natural language. This input section converts the token sequence into multidimensional word embedding vectors. This input section inputs image data into a convolutional neural network. This input section extracts visual feature vectors from the image. This input section normalizes numerical and categorical data. This input section encodes the normalized data as numerical vectors. This input section combines these vectors of different modalities. This input section generates a combined high-dimensional tensor. This input section packages the high-dimensional tensor into a payload for transmission to the server. This processing transforms vague human intentions into rigorous mathematical expressions that can be processed by computers. This processing significantly improves the accuracy of subsequent generative AI analysis. This system detects missing and contradictory input data. Upon detecting a contradiction, the system automatically generates a prompt to re-enter the data. This system ensures the quality of input data. This system is applied in the fields of furniture design and industrial product prototyping. This system effectively conveys diverse user requirements to the system. This system achieves technical improvements to the data input interface.

[0041] The terminal sends this input data to the server. A generative AI processing unit is located on the server side, where the generative AI analyzes the input design requirements. Generative AI integrates multiple AI modules, including natural language processing, image recognition, structural optimization, material property reasoning, reference to past design databases, user history analysis, similar case retrieval, trend analysis, cost simulation, and environmental load assessment, to make comprehensive judgments. For example, if the purpose is a "storage box," it automatically designs the number and configuration of drawers, the presence or absence of handles, stacking structure, locking mechanism, presence or absence of dividers, anti-slip shape of the bottom, opening and closing method of the lid (sliding, flip-up, magnetic, etc.), material thickness, load-bearing capacity, cost, and mass production feasibility. For example, if it's an "outdoor bench," it automatically determines the seat height, backrest angle, leg strength, weather-resistant coating, anchoring structure, anti-tipping design, maintainability, disassembly and transportation, material corrosion resistance, UV degradation countermeasures, and size optimization based on user body distribution, generating multiple design variations. For example, if it's a "toy," considerations such as rounded corners for child safety, minimum part size to prevent accidental swallowing, ease of assembly, color variations based on color psychology, added educational elements, durability, washability, part replaceability, and legal compliance are automatically reflected in the design. Specifically, this generative AI processing unit possesses a composite AI model combining a large-scale language model based on the Transformer architecture and a graph neural network. This AI model receives high-dimensional tensors from the terminal as input data. This input data, for example, contains a 768-dimensional text embedding vector and a 512-dimensional image feature vector. This AI model uses a self-attention mechanism to calculate the dependencies between elements within the input data. This AI model dynamically weights the importance of usage and constraints. This AI model outputs graph structure data representing the connectivity relationships between nodes and edges in three-dimensional space. This output data, for example, contains a 1024-row, 3-column floating-point tensor representing the three-dimensional coordinates of each part. This output data contains a probability distribution vector representing the material of each part. This processing unit determines whether the maximum value of the output probability distribution vector exceeds a predetermined threshold. If a threshold is exceeded, this processing unit passes the material as the definitive material to the subsequent structural optimization module. This structural optimization module automatically performs a simulation based on the finite element method. It identifies stress concentration areas and automatically increases the wall thickness at these identified areas. This processing employs unconventional mathematical methods that do not rely on human intuition. It autonomously ensures the physical rationality of the design within the computer. This system prevents unnecessary consumption of computational resources. This system significantly improves the quality and safety of the generated 3D models. This system is applied to the automated design of building components and the customized design of medical equipment.

[0042] Generative AI, while referencing past design data, design patterns, material property databases, user input history, industry standards, laws and regulations, cost data, environmental impact data, user evaluation data, and trend information, generates multiple design solutions optimized for specific uses and constraints. For example, it can prioritize shapes and structures that have received high praise in the past for similar applications, or automatically generate design solutions that also consider material strength, processability, cost, environmental impact, sourcing feasibility, processing methods (injection molding, cutting, additive manufacturing, etc.), and surface treatments (coating, electroplating, printing, laser processing, etc.). Furthermore, when prioritizing "handling ease," it can automatically incorporate folding mechanisms, lightweight design, handle configuration, compactness during storage, durability, balance, user handling methods, and analysis of transport operations into the design. For example, in the case of a "medical device storage box," sterilization, airtightness, labeling space, partition flexibility, impact absorption structure in case of drops, durability after opening and closing cycles, drug resistance, cleanability, and color-based identification are also automatically incorporated into the design. Specifically... This generative AI utilizes retrieval-enhanced generation technology in conjunction with an external database. It uses the input component vector as a query to retrieve data from a vector database. This vector database preserves past design data as points in a high-dimensional vector space. The generative AI calculates the cosine similarity between the query vector and each vector in the database. It extracts the design data with the highest similarity as context. The extracted context is then combined into an input tensor. This combined tensor is input into a decoder network. The decoder network outputs parameters for multiple design schemes in parallel. This output includes, for example, numerical arrays representing the volume, surface area, and estimated manufacturing cost of each design scheme. The generative AI applies an evaluation function to each output design scheme. This evaluation function calculates a weighted linear sum of various indices, including intensity, cost, and environmental impact. The generative AI selects a predetermined number of design schemes with the highest evaluation scores as final candidates. This process rapidly explores optimal solutions from a vast design space. This process fundamentally improves existing design processes that rely on human memory and experience. This process achieves data-driven, objective design scheme generation. This system optimizes computational efficiency. This system reduces memory usage. It is applied to lightweight design of aerospace components and aerodynamic optimization design of automotive components.

[0043] Multiple design schemes generated by the generative AI processing unit on the server are sent to the terminal. On the terminal, a wide variety of information is presented regarding each scheme's features, appearance images, dimensions, material information, functional specifications, cost estimates, environmental impact assessments, mass production assessments, user evaluation predictions, legal and regulatory compliance, assembly steps, disassembly steps, maintainability, customizability, component replaceability, and brand image compatibility. For example, 3D preview images of each scheme, convenience assessments by application, cost estimates, environmental impact assessments, assembly animations, disassembly animations, component lists, material lists, processing method lists, recommended production methods, recommended distribution methods, and recommended packaging schemes are also simultaneously provided. Users can refer to this information while selecting the optimal design scheme. Specifically, the server converts the generated 3D structural data of the design schemes into a lightweight polygonal mesh format. The server serializes the converted mesh data and metadata into a payload in a data description language format. The server sends the serialized data to the terminal using an asynchronous communication protocol. The terminal expands the received data into memory. The terminal uses its built-in graphics processing unit to render the 3D preview images in real time. The terminal recalculates the depiction matrix based on changes in the user's viewpoint. This terminal drives pixel shaders on the screen to dynamically depict shadows and reflections. This terminal displays cost and environmental impact values ​​as overlaid infographics. This terminal depicts the evaluation scores of each design scheme as vertex coordinate vectors of a radar chart. This processing efficiently visualizes the heavy computational results on the server side. This processing minimizes network bandwidth consumption. This processing significantly improves the responsiveness of the user interface. This system captures user selection actions through event listeners. This system immediately feeds back the identifier of the selected design scheme to the server. This system is applied to order-taking systems for customized products in e-commerce and inter-enterprise component procurement platforms. This system eliminates information prompt delays. This system improves computer display control technology.

[0044] When a user selects a design scheme, the terminal sends this selection information to the server. The 3D model generation unit on the server automatically generates a 3D model generation script usable in 3D modeling software based on the selected design scheme. This script is output in the script format of open-source 3D modeling tools (e.g., Python, JSON, STEP, STL, OBJ), automatically generating a 3D model that reflects specified dimensions, materials, colors, structure, and functions. For example, for a storage box, it automatically configures the outer frame, drawers, handles, stacking mechanisms, locking mechanisms, partitions, anti-slip bottom, lid opening and closing mechanisms, label spaces, component segmentation, assembly steps, disassembly steps, etc., and also automatically sets material properties reflecting dimensions and material information. For a chair, it automatically generates the seat, backrest, legs, reinforcing structure, coating, anchor points, anti-tipping features, disassembly and transport joints, maintenance access panels, size optimization based on the user's body distribution, and thickness and strength settings for each material. For toys, the automatic design considers safety features such as rounded corners, component segmentation, assembly structure, educational elements, color variations, component replaceability, washability, legal compliance, brand logo placement, and user name imprinting. Specifically, this 3D model generation unit possesses an autoregressive code generation neural network. This neural network receives a set of parameters from the selected design scheme as an input vector. This input vector, for example, contains floating-point arrays representing the dimensions of each component and one-hot encodings representing categorical variables of materials. Based on these inputs, the neural network constructs an abstract syntax tree in stages. The neural network serializes the constructed abstract syntax tree into a text string according to the syntax rules of a specific scripting language. The output data is, for example, text data of a script containing the generation function for vertex coordinates and calls to Boolean operations. This processing unit tests and executes the generated script in a virtual environment. This processing unit automatically detects runtime errors in the script and self-intersections of the generated mesh. When errors are detected, this processing unit executes a feedback loop to correct the corresponding lines of the script. This processing completely eliminates the existing process of manual operation of the design software by humans. This processing automates code generation and verification within the computer. This system reduces script generation speed to the millisecond level. This system ensures the geometric accuracy of the generated 3D models. It is applied to the construction of digital twins and the automatic generation of assets for virtual spaces. This system significantly enhances the automation of data processing.

[0045] Furthermore, the scripts for generating 3D models can incorporate user-tunable parameter settings (such as dimensions, colors, materials, presence or absence of components, and addition / deletion of functions), automatic generation of multiple variants, strength analysis and cost estimation based on application, environmental impact assessment, assembly / disassembly animation generation, automatic generation of component / material lists, recommended processing methods, mass production assessment, user evaluation prediction, and legal compliance checks. In generating these scripts, generative AI comprehensively considers the user's intended use and constraints, past design data, material properties, cost data, environmental impact data, legal and regulatory data, and user evaluation data to automatically determine the optimal 3D model generation steps. Specifically, this 3D model generation unit dynamically implements the concept of parametric modeling within the script. This processing unit defines the dimensions and positional relationships of each component as a combination of variables and mathematical expressions, rather than fixed values. This processing unit constructs a directed acyclic graph (DAG) representing the dependencies between variables within the script. This processing unit scans the DAG when the user changes specific variables. This processing unit embeds the code for the constraint solver, which recalculates all relevant variables in a chain, into the script. This generative AI adds the finite element mesh generation algorithm for strength analysis as a subroutine to the script. The lookup table used for cost calculation in this generative AI is included in the script in a data description language format. This processing performs advanced simulations and redesigns solely through the generated scripts themselves. This processing reduces dependence on external software. This system optimizes memory consumption in the script execution environment. This system minimizes redraw latency when parameters change. This system reduces the computational load for users fine-tuning designs in their local environments. This system is used as a design simulation teaching material and open-source hardware distribution format in educational institutions. This system expands the program generation capabilities of computers.

[0046] Furthermore, even when users input multiple times under different purposes and conditions, the generative AI can learn from past input history and selection preferences, automatically generating optimized proposals and design solutions for each user. For example, for users who have designed "storage boxes" multiple times in the past, it can learn their preferred shapes, functions, materials, colors, cost ranges, brand images, etc., and automatically generate more personalized proposals. In addition, when users input additional requirements or modification instructions for the selected design solution (such as "add another drawer," "change the handle position," "change the material to metal," "change the color to blue," "add a logo," etc.), the generative AI can immediately analyze the content and automatically regenerate a new design solution and a 3D model generation script reflecting the modifications. Specifically, this system records each user's behavior history in a time-series database. This system inputs the recorded time-series data into a recurrent neural network. This neural network continuously updates the latent feature vector representing the user's design preferences. This system automatically assigns this latent feature vector as contextual clues to the generative AI. This system uses reinforcement learning algorithms, using whether the user ultimately adopts the design solution as a reward signal to fine-tune the model's weights. This system executes a natural language processing pipeline to extract changes when additional requests are submitted via text input. This pipeline outputs a dictionary of key-value pairs containing the parameters of the changed object and their new values. The generative AI then merges this dictionary data into the existing set of design parameters. Based on this merged parameter set, the generative AI only performs difference compilation on the corresponding parts of the script used to generate the 3D model. This process avoids the complete regeneration of the script. This process significantly reduces computational resource consumption. This process reduces the turnaround time from correction indication to result display to milliseconds. This system provides a personalized design experience. This system is used as a tool to support the iterative revision work of professional designers. This system improves the adaptive learning capabilities of computers.

[0047] Furthermore, when users upload external design data, images, drawings, sketches, CAD data, 3D scan data, etc., generative AI can analyze this data and automatically redesign and optimize it according to specified uses and constraints, generating new design schemes and 3D model generation scripts. For example, by uploading 3D scan data of an existing product and providing instructions such as "add drawers based on this shape," "change materials," "enlarge dimensions," and "interchangeable with specific parts," generative AI can automatically optimize the design and output 3D model generation scripts. Specifically, the external data acquisition unit unfolds the uploaded 3D scan data as a point cloud tensor into memory. The processing unit performs preprocessing to convert the point cloud tensor into a voxel mesh. The processing unit uses a 3D convolutional neural network to extract spatial feature maps from the voxel data. The processing unit inputs the extracted feature maps into the encoder of the generative AI. The generative AI uses a cross-attention mechanism to combine the input feature maps and the user's text instruction vector. The generative AI outputs the signed distance function of the new shape from the combined representation vector. This output data is a three-dimensional array of continuous scalar values ​​representing the distances of points in space from the object's surface. This processing unit applies the moving cube method to this signed distance function. This processing unit automatically extracts optimized new polygon meshes. This processing reconstructs geometrically accurate solid models from noisy scan data. This processing fully automates reverse engineering based on manual human intervention. This system improves the accuracy of data conversion. This system dramatically reduces processing time from hours to seconds. This system is applied to the manufacture of replacements for damaged parts and the digital restoration of cultural heritage. This system fundamentally improves the three-dimensional data processing capabilities of computers.

[0048] In this embodiment of the invention, by employing a structure where the user input unit is implemented in the terminal, and the generative AI processing unit and the 3D model generation unit are implemented in the server, the computational load on the terminal side can be reduced, while advanced design scheme generation and 3D model generation processing utilizing high-performance generative AI can be achieved on the server side. The terminal can be composed of a personal computer, tablet terminal, smartphone, wearable terminal, business-specific terminal, etc., and the server can be composed of a cloud server, local server, distributed processing server, edge server, etc. Specifically: This system achieves optimal task partitioning for edge computing and cloud computing. This terminal executes only lightweight front-end applications. This terminal is dedicated to capturing and processing user input events. This server utilizes dedicated hardware equipped with a tensor operation unit. This server executes inference of large-scale models with billions of parameters at high speed. This system serializes data sent and received between the terminal and the server into binary format. This system minimizes the size of the communication payload. This system improves tolerance to network latency and packet loss. This system batch processes multiple inference requests on the server side. This system maximizes the utilization of the memory bandwidth of parallel processing processors. This system improves throughput. This architecture ensures stable operation even in bandwidth-constrained environments such as mobile lines. It maximizes the efficiency of hardware resource utilization. This system is used in globally deployed design collaboration platforms. This system addresses the technical challenges of data synchronization and computing resource management in distributed systems. This system improves the performance of the entire computer network.

[0049] Thus, according to an embodiment of the present invention, by combining advanced automated design and automated 3D model generation processing on the server side, centered on generative AI, with an intuitive user interface on the terminal side, the efficiency of the initial design and prototyping processes in traditional new product development can be significantly improved, enabling even users without specialized knowledge to easily and quickly generate various 3D models automatically. Specifically: This system integrates design space exploration algorithms and machine learning models. This system uses gradient descent to automatically explore the optimal solution for design parameters. This system incorporates user feedback into the loss function. This system continuously improves the output accuracy of the model. This system performs topology optimization of the generated 3D model in the background. This system mathematically derives shapes that ensure the required strength while eliminating material waste. This processing generates non-intuitive optimal shapes that are difficult for human intuition to achieve. This system seamlessly connects the data pipeline from design to manufacturing. This system eliminates data format conversion errors. This system reduces the lead time required for prototype manufacturing from weeks to minutes. This system drastically reduces development costs. This system is applied to system integration with cloud manufacturing services. This system functions as a foundational technology for on-demand manufacturing. This system brings more than just improved business efficiency; it represents a technological innovation based on computer-driven autonomous design and generation. This system expands the application of information processing devices to the core processes of manufacturing.

[0050] (System Composition) The system described in this embodiment includes a user input unit, a data transmission unit, a generative AI processing unit, a design scheme prompting unit, a user selection unit, a 3D model generation unit, a script output unit, a history management unit, and an external data acquisition unit. The user input unit allows users to input the product's purpose, size, material, color, and other constraints on a terminal. For example, it can specify various uses such as "small storage box for desktop use," "weather-resistant bench for outdoor use," "safe toy for children," "casing for electronic devices," "portable folding container," "medical device storage box," and "assembled exhibition equipment" through natural language, options, images, drawings, voice input, touch operation, pen input, and external file import. Furthermore, regarding size, it can provide specific or abstract instructions such as "10 cm high, 5 cm wide, 8 cm deep," "size that can hold A4 paper," and "maximum weight capacity of 2 kg." Regarding materials, it can specify various options such as "plastic," "metal," "wood," "glass," "composite materials," "recycled materials," "heat-resistant materials," and "antibacterial materials." Regarding color, it can provide indications such as "monochrome," "multicolor," "gradient," "brand color," "color matching code," "bright colors favored by children," and "subdued hues." As other constraints, it can simultaneously specify multiple options such as "with drawers," "stackable," "waterproof," "impact-resistant," "assembly-type," "removable and washable," "compliant with safety standards," "cost limit," "environmentally friendly," "incorporating specific logos or designs," and "interchangeable with specific components." The data transmission unit is responsible for sending all information input by the user input unit to the server, and also performs network communication encryption, data format conversion, retransmission control in case of errors, and input data integrity checks. Specifically, each module of this system runs as an independent process in memory space. This system uses inter-process communication to execute high-speed data transfer between modules. This user input unit reduces the multi-modal data input into a single data object. This user input unit passes voice input to the speech recognition engine and converts it into text data. This user input unit samples the touch operation trajectory as an array of two-dimensional coordinates. This data transmission unit serializes the reduced data object into a byte stream. This data transmission unit uses transport layer security protocols to encrypt the byte stream. This data transmission unit uses checksums to verify data integrity. In the event of a network disconnection, the data transmission unit temporarily saves the data to a local cache. Once communication is restored, the data transmission unit automatically performs retransmission processing in the background. This process completely prevents data loss even in unstable network environments. This process improves the system's fault tolerance. This system performs type and range checks on the input data at the terminal side in advance. This system prevents invalid data from being sent to the server. This system suppresses unnecessary consumption of server computing resources.This system is used in smart factories to input requirements from field operators. It provides a robust foundation for data collection.

[0051] The generative AI processing unit runs on a server, analyzing design requirements sent by users and automatically generating multiple product design schemes with optimal shape, structure, materials, colors, and functions based on usage and constraints. For example, for a "storage box," it automatically designs the number and configuration of drawers, the presence or absence of handles, stacking structure, locking mechanism, presence or absence of dividers, anti-slip shape of the bottom, opening and closing method of the lid, material thickness, load-bearing capacity, cost, and mass production feasibility. For an "outdoor bench," it automatically determines the seat height, backrest angle, leg strength, weather-resistant coating, anchoring structure, anti-tipping design, maintainability, disassembly and transportation, material corrosion resistance, UV degradation countermeasures, and size optimization based on user body distribution, generating multiple design variations. For a "toy," considerations such as rounded corners for child safety, minimum component size to prevent accidental swallowing, ease of assembly, color variations based on color psychology, addition of educational elements, durability, cleanability, component replaceability, and legal compliance are also reflected in the automatic design. The generative AI processing unit references past design data, design patterns, material property databases, user input history, industry standards, laws and regulations, cost data, environmental impact data, user evaluation data, trend information, etc., and generates multiple design solutions optimized for the intended use and constraints. For example, when emphasizing "portability", it incorporates folding mechanisms, lightweight design, handle configuration, compactness during storage, durability, balance, user handling methods, and analysis of handling operations into the automated design. Specific examples of prompts that the generative AI can read include: "Purpose: Small storage box for desktop use; dimensions: 10cm high, 5cm wide, 8cm deep; material: plastic; color: white; requirements: with drawers, stackable, cost capped at 1000 yen, emphasis on mass production, environmentally friendly design, brand logo on the side, shape that allows for one-handed carrying, modular, removable and washable, anti-slip bottom," and "Purpose: Weather-resistant bench for outdoor use; dimensions: seatable for two adults; material: composite of metal and wood; color: natural wood color and black; requirements: anchoring structure, anti-tipping design, UV protection, easy disassembly and transport, emphasis on maintenance, cost capped at 50,000 yen, design consistent with brand image." Specifically, the generative AI processing unit inputs the prompts into a natural language processing model. This model segments the prompts into sub-words and converts them into an array of token IDs. The model then uses an embedding layer to obtain a dense vector representation of the token ID array. This model uses a multi-head attention mechanism to calculate the contextual relationships between tokens. For example, the strong correlation between "outdoor" and "weather resistance" is extracted as an attention weight matrix. This processing unit provides this context vector as a conditional input to the diffusion model. The diffusion model starts from a random noise tensor and removes noise in stages.This diffusion model generates 3D voxel data consistent with the conditional vectors. The output data is, for example, a 3D mesh with a resolution of 64×64×64, where each voxel has a continuous value from 0 to 1 representing the occupancy probability. This processing unit performs thresholding on the voxel data to extract the shape of the entity. This processing unit calculates the volume and surface area of ​​the extracted shape and inputs it into the cost simulation module. This simulation module calculates an estimated manufacturing cost by referring to a material unit price database. This processing deterministically derives specific 3D shapes that meet physical constraints from text-based abstract requirements. This processing reduces the cognitive load on human designers to zero. This system outputs multiple variants within seconds through parallelization of the generation process. This system is applied to the appearance design of personal mobility tools and the automatic generation of custom furniture. This system technically demonstrates the ability to perform computer-based creative tasks.

[0052] The design scheme prompting department sends multiple design schemes generated by the generative AI processing department to the terminal. The terminal displays information from various perspectives on each scheme, including features, appearance images, dimensions, material information, functional specifications, cost estimates, environmental impact assessments, mass production assessments, user feedback predictions, legal and regulatory compliance, assembly steps, disassembly steps, maintainability, customizability, component replaceability, and brand image compatibility. For example, 3D preview images, convenience assessments by application, cost estimates, environmental impact assessments, assembly animations, disassembly animations, component lists, material lists, processing method lists, recommended production methods, recommended distribution methods, and recommended packaging schemes are also simultaneously displayed. The user selection department allows users to choose the best scheme from the multiple suggested options. For example, users can rotate and zoom in / out of the 3D preview while comparing and discussing the options, or compare detailed information, cost, environmental impact, mass production, and legal and regulatory compliance of each scheme, while referring to user feedback predictions and brand image compatibility. Furthermore, additional expectations or correction instructions can be input for the user's selected scheme (e.g., "add another drawer," "change the handle position," "change the material to metal," "change the color to blue," "add a marker," etc.). This information is then sent again to the generative AI processing unit, which automatically generates a new design scheme reflecting the corrections. Specifically: The design scheme prompting unit caches the multidimensional evaluation data received from the server in the terminal's memory. The prompting unit invokes a web-based graphical interface to depict a 3D preview on the canvas. The prompting unit hooks into user touch swipes and mouse drag events. The prompting unit updates the virtual camera's quaternion based on the incremental values ​​of the events. The prompting unit re-renders the scene using the updated camera matrix. The user selection unit monitors the state of UI components on the screen. When the user presses the selection button for a specific design scheme, the selection unit obtains the unique ID of that scheme. When additional correction instructions are input, the selection unit analyzes the text using a natural language processing module. The selection unit extracts pairs of parameter names and changed values ​​for the correction objects. The selection unit asynchronously sends this differential data to the server in a data description language format. This process provides a seamless user experience without requiring screen reloading. This process optimizes the terminal's processor utilization. This system reduces network communication overhead. This system is applied to interactive design review systems and customer-participatory product customization tools. This system improves human-computer interface technology.

[0053] The 3D model generation department automatically generates 3D model generation scripts usable in 3D modeling software based on the user-selected design scheme. For example, it outputs scripts in open-source 3D modeling tool formats (Python, JSON, STEP, STL, OBJ, etc.) to automatically generate 3D models reflecting specified dimensions, materials, colors, structures, and functions. For instance, for a storage box, it automatically configures the outer frame, drawers, handles, stacking mechanisms, locking mechanisms, partitions, anti-slip bottom, lid opening and closing mechanisms, label spaces, component segmentation structures, assembly steps, and disassembly steps, and automatically sets material properties reflecting dimensions and material information. For a chair, it automatically generates the seat, backrest, legs, reinforcing structures, coatings, anchors, anti-tipping mechanisms, disassembly and transport joints, maintenance access panels, dimension optimization based on the user's body distribution, and thickness and strength settings for each material. For toys, the automatic design considers safety features such as rounded corners, component segmentation, assembly structure, educational elements, color variations, component replaceability, cleanability, legal compliance, brand logo configuration, and user name engraving. The script output unit sends the 3D model generation script generated by the 3D model generation unit to the terminal, allowing users to download it or import it into external 3D modeling software. For example, users can download the script and edit and utilize it within the company's design department, or send it directly to external 3D printers or manufacturers. Specifically, the 3D model generation unit maps the parameter group of the design scheme to an internal abstract syntax tree. This generation unit generates nodes defining the geometric primitives of each component. This generation unit constructs edges defining Boolean operations between components. This generation unit traverses the abstract syntax tree using a depth-first search. During the traversal, this generation unit successively generates code strings that conform to the syntax rules of the target language. This generation unit, for example, outputs a group of functions calling the 3D processing library as text. This script output unit buffers the generated text data as a character-encoded file in memory. This output unit applies a hash function to the file to generate a checksum for tamper-proof protection. This output unit sends the file and checksum as multipart form data to the terminal. The terminal saves the received data to local storage. This processing represents complex 3D shapes as lightweight text data. This processing drastically reduces data transfer volume from gigabytes to kilobytes. This system saves storage capacity. This system enables the generated scripts to be easily managed through a version control system. This system is applied to the sharing of design data in distributed development environments and the direct input of data into automated manufacturing production lines. This system fundamentally improves the efficiency of data representation and transmission.

[0054] The History Management Department centrally manages user input history, selection history, generated design schemes, 3D model generation scripts, download history, correction history, and user preference data. For example, for users who have designed "storage boxes" multiple times in the past, it learns their preferred shapes and functions, materials, colors, cost ranges, and brand image, enabling it to automatically generate more personalized proposals. The External Data Acquisition Department analyzes external design data, images, drawings, sketches, CAD data, and 3D scan data uploaded by users. It automatically redesigns and optimizes these data according to specified uses and constraints, generating new design schemes and 3D model generation scripts. For example, by uploading 3D scan data of an existing product and providing instructions such as "add drawers based on this shape," "change materials," "enlarge dimensions," and "interchangeability with specific components," generative AI can automatically optimize the design and output a 3D model generation script. Specifically, this History Management Department uses a hybrid storage system combining relational and non-relational databases. This department stores structured parameter data in tables. This department saves scripts and images as unstructured data in document storage. This management department accumulates user operation logs as time-series data. This system inputs this time-series data into a Long Short-Term Memory (LSTM) network. This network probabilistically predicts the user's next design requirements. The external data acquisition department parses the uploaded design data as a boundary representation model. This acquisition department extracts topological and geometric information from the boundary representation model. This acquisition department converts the extracted information into graph-structured data. This generative AI applies a graph convolutional network to this graph-structured data. This generative AI updates the coordinates of nodes while maintaining the characteristics of the existing shape and satisfying specified constraints. This process maximizes the reusability of existing assets. This process eliminates the computational costs associated with designing from scratch. This system optimizes database query response time. This system continuously improves the inference accuracy of machine learning models. This system is applied to reverse engineering and the digital archiving of traditional products. This system improves the technical framework for data management and reuse.

[0055] Thus, the system involved in this embodiment includes a user input unit, a data transmission unit, a generative AI processing unit, a design scheme prompting unit, a user selection unit, a 3D model generation unit, a script output unit, a history management unit, and an external data acquisition unit. Through the coordinated actions of these units, the system enables even users without professional 3D modeling skills and knowledge to easily select the best option from multiple design options in a short time and automatically generate a specific 3D model. The prompt text for the generative AI to read can include, for example, "Purpose: Medical device storage box; Dimensions: Width 30 cm, Height 20 cm, Depth 15 cm; Material: Antibacterial plastic; Color: White; Conditions: Emphasis on sterilization, airtightness, label pasting space, partition flexibility, impact absorption structure in case of drop, durability after opening and closing cycles, drug resistance, cleanability, color-based identification, cost limit of 3000 yen, compliance with laws and regulations," comprehensively including purpose, dimensions, material, color, and conditions. Specifically, this system enables loosely coupled coordination between the processing units via asynchronous message queues. This system adopts an event-driven architecture. This system dynamically scales horizontally to instances of a specific module when the load is concentrated there. It maintains a constant overall system throughput. This generative AI analyzes the sprawling input prompts using a natural language understanding model. The model extracts each condition as a key-value pair. It transforms the extracted condition set into an input format for a constraint satisfaction problem. The system uses a constraint solver to determine the solution region in the parameter space that simultaneously satisfies all conditions. It generates multiple initial seeds from this solution region through random sampling. It performs gradient-based optimization starting from various seeds. This process mathematically resolves trade-offs between opposing constraints. This process automatically derives Pareto optimal solutions. This system achieves optimization in multidimensional spaces that exceed the limits of human cognition. This system maximizes hardware utilization efficiency through dynamic allocation of computational resources. This system is applied to the stringent specification design of medical equipment and the optimization design of aircraft components. This system improves the overall processing capabilities of computer systems.

[0056] (Implementation steps) Step 1: User Input. The user uses the user input section on the terminal to input the product's purpose, size, material, color, and other constraints. For example, the purpose can be specified through various methods such as natural language, options, images, drawings, voice input, touch operation, pen input, and importing external files, such as "small storage box for desktop use," "weather-resistant bench for outdoor use," "safe toy for children," "casing for electronic devices," "portable folding container," "medical equipment storage box," and "assembled display equipment for exhibitions." Size can be input including specific numerical values, relative sizes, and functional constraints. Multiple materials, colors, and other constraints can also be specified simultaneously. Specifically, the terminal binds event listeners to each input field of the user interface. The terminal captures the user's keystrokes and mouse clicks in real time. The terminal immediately performs regular expression validation on the input strings. When image files are dragged and dropped, the terminal analyzes the file header to verify the format's appropriateness. The terminal reads the appropriate image data as a byte array in memory. The terminal maps this input data to a key-value data structure. Upon detecting completion of input, this terminal serializes the data structure into a string. This process ensures the consistency of input data format on the terminal side. This process significantly reduces the error rate in subsequent communication and server processing. This system maintains the response speed of the input interface in milliseconds. This system eliminates user input pressure. This system normalizes data from multiple input devices into a unified format. This system is used in design applications that support multiple platforms. This system improves the technical reliability of data input processing.

[0057] Step 2: The input data sending terminal sends all user-input data to the server via the data sending unit. Simultaneously, network communication encryption, data format conversion, retransmission control in case of transmission errors, and input data integrity checks are also implemented. Specifically: This terminal stores the serialized string in the body of the communication request. This terminal appends an authentication token and content type to the request header. This terminal calls the communication protocol stack to generate data packets. This terminal performs a handshake to establish a secure communication channel with the server. This terminal sends the generated data packets via the network interface card. This terminal waits for a response from the server after sending. If there is no response within the specified timeout period, this terminal uses an exponential backoff algorithm to gradually extend the retransmission interval while attempting retransmission. This server decrypts the payload of the received data packets. This server uses a pattern validator to verify the structural integrity of the data. This process completely prevents eavesdropping and tampering of data along the communication path through cryptographic algorithms. This process improves the system's robustness to packet loss and network latency. This system suppresses unnecessary consumption of communication resources. This system is applied to a secure cloud-based design platform. This system improves data transmission technology between computer networks.

[0058] Step 3: The generative AI processing unit on the design scheme generation server analyzes the design requirements sent by the user and automatically generates multiple product design schemes with optimal shape, structure, material, color, and function based on the purpose and constraints. For example, for a "storage box," it automatically designs the number and configuration of drawers, the presence or absence of handles, stacking structure, locking mechanism, presence or absence of partitions, anti-slip shape of the bottom, opening and closing method of the lid, material thickness, load-bearing capacity, cost, and mass production feasibility. For an "outdoor bench," it automatically determines the seat height, backrest angle, leg strength, weather-resistant coating, anchoring structure, anti-tipping design, maintainability, disassembly and transportation, material corrosion resistance, UV degradation countermeasures, and size optimization based on the user's body distribution, generating multiple design variations. As a specific example of a prompt text for the generative AI to read, one could include: "Purpose: Small storage box for desktop use; Dimensions: 10 cm high, 5 cm wide, 8 cm deep; Material: Plastic; Color: White; Requirements: With drawer; Stackable; Cost limit 1000 yen; Emphasis on mass production; Environmentally friendly design; Brand logo on the side; Shape that allows for one-handed carrying; Assembled; Removable for washing; Anti-slip surface on the bottom." Specifically, the server enqueues the received request data into the input queue of the inference engine. The inference engine dequeues the data and populates it into a tensor for batch processing. The inference engine transfers this batch processing tensor to the memory of the parallel processing processor. The generative AI uses a Transformer-based encoder to map the request text into a 768-dimensional latent space vector. The generative AI uses this latent vector as conditional input to the generator of a generative adversarial network. The generator generates a 3D voxel mesh. The discriminator determines whether the generated voxel mesh satisfies the physical constraints and specified conditions and outputs a score. This generative AI repeatedly performs generation loops while fine-tuning the generator's weights until the score exceeds a predetermined threshold. The output data is, for example, a structured array containing vertex coordinates, normal vectors, and material identifiers for each design scheme. This processing utilizes the learning results of complex probability distributions in high-dimensional space. This processing achieves flexible shape generation that rule-based programs cannot describe. This system pushes parallel computing capabilities to the limit. This system significantly reduces inference time. This system is applied to the automation of industrial design and the procedural generation of game assets. This system dramatically improves the computational efficiency of computers.

[0059] Step 4: Design Scheme Prompts and Selection. Multiple design schemes generated by the generative AI processing unit are sent to the terminal via the design scheme prompting unit. The terminal displays information from various perspectives, including features, appearance images, dimensions, material information, functional specifications, cost estimates, environmental impact assessments, mass production assessments, user evaluation predictions, legal and regulatory compliance, assembly steps, disassembly steps, maintainability, customizability, component replaceability, and brand image compatibility. Users can select the optimal design scheme by referring to 3D preview images, convenience assessments based on intended use, cost estimates, environmental impact assessments, assembly animations, disassembly animations, component lists, material lists, processing method lists, recommended production methods, recommended distribution methods, and recommended packaging schemes. Specifically, the system's design scheme prompting unit receives multiple design scheme data output from the generative AI processing unit. This design scheme data is structured as a multidimensional tensor containing voxel data representing 3D shapes and various attribute values. The design scheme prompting unit temporarily stores the received multidimensional tensor in main memory. The design scheme prompting unit analyzes the stored multidimensional tensor. Based on the analysis results, the design scheme prompting unit generates 3D preview image data. This design scheme prompts the system to send the generated 3D preview image data to the terminal. The terminal receives the 3D preview image data via a network interface. The terminal renders the received 3D preview image data on the display. In this rendering process, the terminal utilizes the parallel computing cluster of the GPU. The terminal calculates the vertex coordinates and normal vectors of the polygon mesh. The terminal performs real-time shading processing based on the calculation results. The terminal calculates the brightness value of each pixel based on the position of the light source and the viewpoint vector. The system receives selection operations from the user via a touch panel or mouse. The system extracts the identifier corresponding to the selected design scheme. The system sends the extracted identifier to the server. Through this process, the system achieves the filtering of design schemes that accurately reflect the user's intent. This system enables the preview display of high-precision 3D models while suppressing communication bandwidth consumption. This technology can be applied to various use cases such as product design processes in manufacturing and prototype evaluation in the construction field.

[0060] Step 5: Inputting Additional Requirements and Modification Instructions and Regenerating the Design. Users can input additional requirements and modification instructions for the selected design scheme (e.g., "Add another drawer," "Change the handle position," "Change the material to metal," "Change the color to blue," "Add a logo," etc.). This information is sent again to the generative AI processing unit, which automatically generates a new design scheme reflecting the modifications. A specific example of a prompt for the generative AI to read could be: "Purpose: Storage box, Dimensions: 10 cm high, 5 cm wide, 8 cm deep, Material: Metal, Color: Blue, Conditions: Add two drawers, change the handle position to the center, add a brand logo," etc. Specifically, the system's terminal receives additional requirements and modification instructions from the user. This terminal inputs the received natural language text into the natural language processing model. This natural language processing model is a large-scale language model based on the Transformer architecture. The input data for this natural language processing model is a sequence of word embedding vectors. This sequence of embedding vectors is, for example, represented as a 512-dimensional floating-point array. As a specific example of the input data, vector data derived from the text "add another drawer" can be given. As another specific example, vector data derived from the text "change the material to metal" can be given. This natural language processing model uses a self-attention mechanism to calculate the contextual dependencies of the input vector sequence. This natural language processing model calculates an attention score matrix representing the degree of association between words. This natural language processing model updates the feature vectors based on the calculated attention score matrix. This natural language processing model inputs the updated feature vectors into a fully connected layer. This natural language processing model outputs a modification intent vector based on the calculation results. This modification intent vector is represented, for example, as coordinate values ​​in a 1024-dimensional latent space. As a specific example of the output data, a vector containing a structural change flag indicating the addition of a drawer can be given. As another specific example, a vector containing a material ID indicating a material change can be given. This generative AI processing unit inputs the output modification intent vector into a subsequent 3D shape generation model. This 3D shape generation model uses a generative adversarial network to generate new 3D voxel data. This 3D shape generation model combines existing shape tensors and modification intent vectors to perform 3D convolution operations. Through this process, the system automatically executes accurate 3D shape changes based on the user's ambiguous natural language instructions. This system significantly reduces the time designers spend on trial and error. This technology can be applied to fields requiring rapid prototyping, such as furniture design and industrial product customization.

[0061] Step 6: Automatic Generation of 3D Model Generation Scripts. Based on the user's final selected design scheme, the 3D model generation unit on the server automatically generates 3D model generation scripts usable in 3D modeling software. For example, it outputs the scripts in open-source 3D modeling tool formats (Python script, JSON format, STEP format, STL format, OBJ format, etc.), automatically generating 3D model content reflecting specified dimensions, materials, colors, structures, functions, etc. Specifically, the 3D model generation unit of this system acquires the 3D voxel data of the selected design scheme. This unit converts the acquired 3D voxel data into polygon mesh data. In this conversion process, the unit executes the moving cube algorithm. It extracts isosurfaces in voxel space. It extracts the vertex coordinates and face connectivity information of the generated polygon mesh. The extracted geometric information is input into the script generation module. This module uses a recurrent neural network to convert the geometric information into a syntax tree. It scans the syntax tree using a depth-first search algorithm. Based on the scan results, it generates text data conforming to a specific programming language. This data contains a set of commands executable on 3D modeling software. The 3D model generation unit saves the generated text data as a 3D model generation script. This 3D model generation unit automatically inserts variable definitions for specified dimensions and materials into the script. Through this processing, the system automatically converts complex 3D shapes into a human-readable and editable script format. This system ensures data compatibility between different CAD software. This system significantly improves the reusability of design data. This technology can be applied to various use cases, such as collaborative design using open-source modeling tools and providing data to automated manufacturing lines.

[0062] Step 7: Script Output and Download The 3D model generation script generated by the 3D model generation unit is sent to the terminal via the script output unit. Users can download the script and edit and utilize it within the company's design department, or send it directly to external 3D printers and processors. Specifically, the script output unit of this system encrypts the generated 3D model generation script. In this encryption process, the script output unit applies an Advanced Encryption Standard (AES) algorithm. The script output unit packages the encrypted script data. The script output unit appends the target address and error detection code to the data packet header. The script output unit sends the packaged data to the terminal via a secure communication protocol. The terminal receives the data packet via a network interface. The terminal combines the received data packet. The terminal decrypts the combined data. The terminal saves the decrypted script data in local storage. The terminal transmits the saved script data to the external 3D printer control device. In this transmission process, the terminal triggers the conversion of geometric information within the script data into G-code. The terminal executes slicer software in the background. The terminal outputs the generated G-code via a serial communication interface. Through this process, this system achieves seamless data linkage from design data to the manufacturing device. This system ensures data security while promoting the automation of the manufacturing process. This technology can be applied to fields requiring rapid data sharing, such as distributed manufacturing and on-demand manufacturing services.

[0063] Step 8: Historical Management and External Data Import. The Historical Management Department centrally manages user input history, selection history, generated design schemes, 3D model generation scripts, download history, modification history, and each user's preference data. When users upload external design data, images, drawings, sketches, CAD data, 3D scan data, etc., the External Data Import Department analyzes this data and can automatically redesign and optimize it according to specified uses and constraints, generating new design schemes and 3D model generation scripts. Specifically, the Historical Management Department records user operation logs in a time-series database. The Historical Management Department converts the recorded operation logs into feature vectors. The Historical Management Department inputs the feature vectors into a clustering algorithm. This clustering algorithm classifies the user's design preferences. The External Data Import Department receives uploaded external data. The External Data Import Department preprocesses the received external data. For example, when 3D scan data is input, the External Data Import Department performs denoising processing on the point cloud data. The External Data Import Department applies a statistical outlier removal algorithm. The External Data Import Department inputs the denoised point cloud data into a 3D convolutional neural network. The input data for this 3D convolutional neural network is a tensor containing spatial coordinates and color information. A specific example of this input data is a point cloud tensor with 1024 vertices. Another specific example is a voxel mesh tensor with a resolution of 256. This 3D convolutional neural network extracts local geometric features from the input tensor. This 3D convolutional neural network propagates the extracted features to fully connected layers. This 3D convolutional neural network outputs the shape category probability and size parameters of the object. A specific example of this output data is a score of 0.95 indicating that the object is a "cylinder". Another specific example is numerical data indicating that the object has a radius of 50 mm. This generative AI processing unit inputs the output shape category probability and size parameters into subsequent optimization algorithms. Through this processing, this system automatically recovers accurate design requirements from incomplete external data. This system makes reverse engineering of existing products more efficient. This technology can be applied to various use cases such as recovering CAD data from old drawings and designing customized parts from physical objects.

[0064] (specific use case) For example, imagine a planning manager at an office supplies manufacturer exploring the development of a new type of storage box for office desks. This manager uses the user input section on a terminal to input the following: purpose: "storage box for office desks"; dimensions: "30 cm wide, 10 cm high, 15 cm deep"; material: "recycled plastic"; color: "white"; and other constraints: "with drawers," "stackable," "cost capped at 1000 yen," "brand logo on the side," "anti-slip design on the bottom," "assembly type," "removable and washable," "emphasis on mass production," and "environmentally friendly design." Input can be performed through various methods, including natural language input, numerical input, option selection, image attachment, voice input, and importing external files. The input data is sent to the server via a data transmission unit. Specifically, the system's user input section receives various input data from the planning manager. This user input section comprehensively processes natural language text, numerical data, and image data. This user input section performs morphological analysis on the input natural language text. This user input section extracts keywords from the analysis results. This user input section maps the extracted keywords to a semantic vector space. The user input unit normalizes the input numerical data. It then stores the normalized numerical data in a constraint matrix. Next, it inputs the image data into a convolutional neural network. The user input unit extracts visual feature vectors from the image data. It combines the semantic vectors, the constraint matrix, and the visual feature vectors. Finally, it generates a comprehensive feature tensor from the combined data. This comprehensive feature tensor is then passed to the data sending unit. The data sending unit packages the comprehensive feature tensor and sends the packaged data to the server. Through this processing, the system integrates input data from different modalities into a single mathematical representation. This system significantly improves the computational efficiency in subsequent AI processing. This technology can be applied to fields requiring a high degree of information integration, such as product planning stages with complex requirement definitions and system design with multiple overlapping constraints.

[0065] The generative AI processing unit on the server analyzes the received design requirements and automatically generates multiple storage box design schemes with optimal shape, structure, material, color, and function based on the intended use and constraints. For example, it automatically designs the number and configuration of drawers, the presence or absence of handles, stacking structure, locking mechanism, presence or absence of dividers, anti-slip shape of the bottom, opening and closing method of the lid, material thickness, load-bearing capacity, cost, mass production capability, environmental impact, brand logo placement, assembly steps, disassembly steps, etc., while referring to past design data, material property database, cost data, environmental impact data, user evaluation data, trend information, etc., to generate multiple design schemes optimized for the intended use and constraints. As a specific example of the prompt text for the generative AI to read, it is: "Purpose: Storage box for use on an office desk, Dimensions: 30 cm wide, 10 cm high, 15 cm deep, Material: Recycled plastic, Color: White, Conditions: With drawers, Stackable, Cost limit 1000 yen, Brand logo on the side, Anti-slip on the bottom, Assembled, Disassembleable and washable, Emphasis on mass production capability, Environmentally friendly design." Specifically, the generative AI processing unit of this system analyzes the received comprehensive feature tensor. This generative AI processing unit inputs the comprehensive feature tensor into a generator of a conditional generative adversarial network. The input data to this generator is a combination of the comprehensive feature tensor and a random noise vector. As a specific example of this input data, a tensor combining a 512-dimensional conditional vector with a specified purpose and size and a 256-dimensional Gaussian noise vector can be given. As another specific example, a 128-dimensional constraint vector encoding cost limits and material information can be given. This generator uses multiple deconvolution layers to expand the input data to spatial dimensions. This generator outputs the material presence probability in each cell of a 3D voxel mesh. As a specific example of this output data, a continuous value from 0 to 1 representing the occupancy probability of each voxel in a 64-resolution 3D mesh can be given. As another specific example, a classification distribution representing the material IDs assigned to each voxel can be given. This generative AI processing unit applies a threshold to the output material presence probability. This generative AI processing unit determines voxels with probabilities above the threshold as entities. This generative AI processing unit performs structural analysis simulations on the identified voxel groups. Based on the simulation results, it evaluates load-bearing capacity and durability. Through this processing, the system rapidly explores optimal shapes that do not rely on human intuition or conventional methods. This system automates the verification of physical constraints in the early stages of design, significantly reducing rework. This technology can be applied to the automated design of office furniture, the housing design of electronic devices, and other product development requiring the fulfillment of multiple constraints.

[0066] Multiple design schemes generated by the generative AI processing unit are sent to the terminal via the design scheme prompting unit. The terminal displays information from various perspectives on each scheme, including its features, appearance images, dimensions, material information, functional specifications, cost estimates, environmental impact assessments, mass production feasibility assessments, user evaluation predictions, legal and regulatory compliance, assembly steps, disassembly steps, maintainability, customizability, component replaceability, and brand image compatibility. For example, 3D preview images, ease-of-use assessments, cost estimates, environmental impact assessments, assembly animations, disassembly animations, component lists, material lists, processing method lists, recommended production methods, recommended distribution methods, and recommended packaging schemes are also simultaneously provided. The responsible person refers to these information while selecting the optimal design scheme. Specifically, the system's design scheme prompting unit receives multiple design scheme data output from the generative AI processing unit. This design scheme data is a composite data structure containing 3D shape data, cost predictions, and environmental impact scores. The design scheme prompting unit parses the received composite data structure. It then passes the 3D shape data to the rendering engine. The rendering engine uses ray tracing algorithms to generate a high-precision 3D preview image. The design scheme prompting unit passes the cost prediction and environmental impact score to the charting module. This charting module converts the numerical values ​​into visual chart data. The design scheme prompting unit generates a user interface screen that integrates the 3D preview image and chart data. The design scheme prompting unit outputs the generated screen data to the terminal's display. The system monitors user selection actions. The system stores the identifier of the selected design scheme in memory. Through this processing, the system provides users with multi-faceted evaluation indicators in an intuitive and understandable format. The system supports the designer's decision-making process based on quantitative data. This technology can be applied to product review meetings involving multiple stakeholders, customer presentations, and other situations requiring rapid consensus.

[0067] For the design scheme selected by the person in charge, additional requests or modification instructions can be input, such as "add another drawer," "change the handle position," "change the material to metal," "change the color to blue," and "add a logo." This information is then sent to the generative AI processing department, which automatically generates a new design scheme reflecting the modifications. A specific example of a prompt for the generative AI to read is: "Purpose: Storage box for use on an office desk; Dimensions: 30 cm wide, 10 cm high, 15 cm deep; Material: Metal; Color: Blue; Conditions: Increase the number of drawers to 2; change the handle position to the center; add a brand logo; add a non-slip surface to the bottom; modular design; removable and washable; emphasize mass production; environmentally friendly design." Specifically, the system's terminal receives additional requests and modification instructions from the person in charge. This terminal inputs the received natural language text into a natural language processing model. This natural language processing model is a large-scale language model based on the Transformer architecture. The input data to this natural language processing model is a sequence of word embedding vectors. As a specific example of the input data, vector data derived from the text "Add drawers to 2 layers" can be cited. As another specific example, vector data derived from the text "Move the handle to the center" can be cited. This natural language processing model uses a self-attention mechanism to calculate the contextual dependencies of the input vector sequence. Based on the calculation results, this natural language processing model outputs a revised intent vector. As a specific example of the output data, a vector containing a structural change indicator indicating the addition of drawers can be cited. As another specific example, a displacement vector indicating the coordinate movement of the handle can be cited. This generative AI processing unit inputs the output revised intent vector into the subsequent 3D shape generation model. This 3D shape generation model combines the existing 3D voxel data and the revised intent vector. This 3D shape generation model inputs the combined data into a 3D convolutional layer to perform local shape updates. Through this processing, this system automatically executes accurate 3D shape changes from the user's ambiguous natural language instructions. This system significantly reduces the time required for designers to experiment. This technology can be applied to customized product design and agile development processes that allow for immediate reflection of customer feedback.

[0068] Based on the final selected design scheme, the 3D model generation unit on the server automatically generates a 3D model generation script usable in 3D modeling software. For example, it outputs the script in the script format of open-source 3D modeling tools (Python script, JSON format, STEP format, STL format, OBJ format, etc.), automatically generating a 3D model that reflects the specified dimensions, materials, colors, structure, and functions. For storage boxes, it automatically configures the outer frame, drawers, handles, stacking concave-convex structures, locking mechanisms, partitions, anti-slip bottom, lid opening and closing mechanisms, label spaces, the segmentation structure of each component, assembly steps, disassembly steps, etc., and also automatically sets the materials reflecting the dimensions and material information. Specifically, the 3D model generation unit of this system acquires the 3D voxel data of the final selected design scheme. This 3D model generation unit converts the acquired 3D voxel data into polygon mesh data. In this conversion process, this 3D model generation unit executes the moving cube algorithm. This 3D model generation unit extracts the vertex coordinates and face connection information of the generated polygon mesh. This 3D model generation unit inputs the extracted geometric information into the script generation module. This script generation module uses a recurrent neural network to convert geometric information into a syntax tree. This module scans the syntax tree. Based on the scan results, this module generates text data conforming to a specific programming language. This text data contains a set of commands executable on 3D modeling software. This 3D model generation unit saves the generated text data as a script for 3D model generation. This 3D model generation unit automatically inserts material setting commands into the script. These material setting commands contain physically based rendering parameters such as the reflectivity and transmittance of the specified materials. Through this process, the system automatically converts complex 3D shapes into a human-readable and editable script format. This system ensures data compatibility between different CAD software. This system significantly improves the reusability of design data. This technology can be applied to various use cases, such as collaborative design using open-source modeling tools and providing data to automated manufacturing pipelines.

[0069] The 3D model generation script generated by the 3D model generation department is sent to the terminal via the script output department. The responsible person can download the script and edit and apply it within the company's design department, or send it directly to external 3D printers or processors. The history management department centrally manages the responsible person's input history, selection history, generated design schemes, 3D model generation scripts, download history, correction history, and the preferred data of each responsible person. When the responsible person uploads external design data, images, drawings, sketches, CAD data, 3D scan data, etc., the external data import department analyzes this data and can automatically redesign and optimize it according to specified uses and constraints, generating new design schemes and 3D model generation scripts. Specifically, the script output department of this system encrypts the generated 3D model generation script. In this encryption process, the script output department applies an advanced encryption standard algorithm. The script output department datagrams the encrypted script data. The script output department sends the datagrammed data to the terminal via a secure communication protocol. The terminal combines the received data packets. The terminal decrypts the combined data. The terminal saves the decrypted script data to local storage. This terminal forwards the saved script data to the external 3D printer control device. During this forwarding process, the terminal triggers the conversion of geometric information within the script data into G-code. The system's historical management department records the responsible person's operation log into a time-series database. This historical management department converts the recorded operation log into feature vectors. This historical management department inputs the feature vectors into a clustering algorithm. This clustering algorithm classifies the responsible person's design preferences. The system's external data import department receives uploaded external data. This external data import department preprocesses the received external data. This external data import department inputs the preprocessed data into a 3D convolutional neural network. This 3D convolutional neural network extracts local geometric features from the input data. This 3D convolutional neural network outputs the object's shape category probability and size parameters. This generative AI processing department inputs the output shape category probability and size parameters into subsequent optimization algorithms. Through this processing, the system automatically reconstructs accurate design requirements from incomplete external data. This system makes reverse engineering of existing products more efficient. This technology can be applied to various use cases, such as reconstructing CAD data from old drawings and designing customized parts from physical objects.

[0070] (Application Example 1) The specific processing flow corresponding to Use Case 1 will be described below. The various parts of the system described below are implemented by the data processing device 12 and the intelligent device 14. Furthermore, the data processing device 12 is referred to as the "server" and the intelligent device 14 is referred to as the "terminal".

[0071] (Implementation methods for carrying out the present invention) A more specific and detailed description is provided regarding one embodiment of the present invention. The system involved in this embodiment comprises a terminal and a server, employing a structure where the terminal provides the user interface, and the server handles generative AI processing and 3D model generation. The terminal can utilize any information processing device capable of network connectivity, such as a personal computer, tablet, smartphone, industrial-grade terminal, wearable terminal, or touchscreen terminal integrated into a control panel within a factory. The server utilizes computing resources with generative AI computing capabilities and 3D model generation capabilities, such as cloud servers, local servers, distributed processing servers, edge servers, and factory-internal servers. Specifically, this system adopts a client-server distributed computing architecture. The terminal of this system functions as an edge device collecting input data from users. The terminal serializes the collected input data into a lightweight data format. The terminal sends the serialized data to the server using an asynchronous communication protocol. The server distributes the received data to multiple computing nodes via a load balancer. The computing nodes of this server constitute a parallel computing cluster using GPUs. The computing nodes unfold the generative AI model into memory. The computing nodes perform inference processing on the input data in parallel. The server aggregates the inference results from each computing node. This server performs 3D model generation processing based on the aggregated results. Through this processing, the system efficiently and in a distributed manner executes generative AI processing that requires massive computing resources. This system is not dependent on terminal-side hardware constraints and provides highly detailed 3D model generation in real time. This technology can be applied to scalable systems such as edge computing environments within factories and cloud-based design platforms deployed globally.

[0072] First, the user input section on the terminal offers multiple input methods, including GUI, voice input, touch operation, pen input, import of external files, and automatic acquisition of on-site information via barcode or RFID readers. Users can simultaneously specify the shape of the object being handled (e.g., cuboid, cylinder, sphere, irregular shape, bag, plate, complex curved shape, soft material, liquid container, powder bag, etc.), dimensions (e.g., 40cm wide × 30cm deep × 20cm high, 7cm diameter × 25cm high, 2mm thick plate, maximum external dimensions within 100mm, etc.), weight (e.g., 5kg, 100g, 20kg, maximum 30kg, minimum 10g, etc.), dimensions and constraints of the workspace (e.g., 1m wide × 1m deep × 1.5m high, ceiling height limit 1.2m, presence or absence of obstacles, movement range restrictions, interference avoidance with adjacent equipment, and collaborative action area with the operator). The requirements include: (domain, height limitations of the conveyor system, etc.); gripping methods (e.g., suction cups, grippers, magnetic force, vacuum, clamps, tweezers, hooks, straps, vacuum adsorption, adhesive pads, air blowing, electrostatic adsorption, soft finger grippers, multi-finger robotic hands, soft grippers for collaborative robots, etc.); desired materials (e.g., aluminum, resin, stainless steel, carbon, rubber, food hygiene materials, heat-resistant materials, chemical-resistant materials, antistatic materials, antibacterial materials, lightweight composite materials, resins for 3D printing, etc.); and required special functions (e.g., heating, cooling, weight sensors, image sensors, force sensors, static elimination, antibacterial coatings, wear detection, self-diagnostic functions, IoT communication functions, QR code reading for workpiece identification). Features include: automatic workpiece identification AI, anomaly detection AI, remote monitoring, self-positioning correction, automatic component replacement, and automated cleaning; cost limits (e.g., 100,000 yen, 300,000 yen, 1,000,000 yen, cost per unit, cost during mass production, component unit price, maintenance costs, etc.); durability (e.g., 100,000 cycles, 1,000,000 cycles, impact resistance, wear resistance, corrosion resistance, chemical resistance, heat resistance, low temperature resistance, long-term lubrication-free operation, maintenance-free operation, component replacement cycle, MTBF, etc.); hygiene requirements (e.g., food factory requirements, medical site requirements, emphasis on cleanability, prevention of foreign matter contamination, aseptic operation, cleaning fluid resistance, autoclave requirements, and detachable components). Features include: cleaning and unloading, antibacterial coating, and foreign object detection sensors; setup space (e.g., maximum configurable dimensions, interference avoidance with surrounding equipment, compatibility with existing equipment, portable design, ease of relocation, etc.); operating speed (e.g., handling time per cycle, maximum acceleration, maximum speed, acceleration / deceleration control, impact buffering mechanism, synchronization control with the conveyor line, etc.); mass production capability (e.g., component standardization, reduced assembly time, reduced component quantity, modular design, 3D printer compatibility, ease of on-site assembly, etc.); and maintainability (e.g., ease of disassembly, component replaceability, lifespan prediction of consumable components, remote diagnostic functions, self-diagnostic functions, maintenance history management, etc.).The availability of replaceable parts (e.g., consumables list, replacement cycle, feasibility of on-site replacement, replacement operation procedures, standardization of replaceable parts, etc.), on-site safety standards (e.g., collaborative robot standards, ISO specification compliance, fail-safe design, emergency stop function, safety cover, photoelectric sensor linkage, operator detection function, automatic stop in case of abnormality, etc.), legal and regulatory compliance (e.g., electrical safety law, food hygiene law, medical device regulations, CE marking, RoHS directive, safety specifications of various countries, etc.), brand image (e.g., corporate logo engraving, brand color adoption, design guideline compliance, appearance design patent response, etc.), and further, inherent special requirements on-site (e.g., explosion-proof, dust-proof, UV-resistant, radiation-resistant, chemical-resistant, oil-resistant, salt-resistant, weather-resistant, low-noise design, static electricity countermeasures, anti-static, environmentally friendly design, recycling response, carbon neutrality design, etc.) are extremely diverse and complex conditions. Specifically, the user input unit of this system comprehensively processes signals from multiple input devices. This user input unit converts the capacitance change from the touch screen into two-dimensional coordinate data. This user input unit samples the analog acoustic signal from the voice input microphone into digital waveform data. This user input unit converts digital waveform data into spectral data using a Fast Fourier Transform. The spectral data is then input into an acoustic model and decoded into text data. Optical reflection signals from a barcode reader are converted into digital codes. This user input unit normalizes these diverse input data. The normalized data is mapped to a multi-dimensional constraint vector. This constraint vector maintains the shape, size, weight, workspace constraints, gripping method, desired material, special functions, cost limits, durability, and hygiene requirements of the transported object as independent dimensions. The user input unit sends the generated constraint vector to the server. Through this processing, the system unifies the complex and diverse requirements of the field into a single, computer-processable mathematical expression. This system absorbs differences in input modalities, improving the versatility of subsequent AI processing. This technology can be applied to engineering tasks requiring the simultaneous consideration of extremely diverse parameters, such as defining the requirements of robotic arms in manufacturing sites and specifying the specifications of automated equipment.

[0073] For example, if the user inputs: "Object: Rectangular cardboard box, dimensions 40cm × 30cm × 20cm, weight 5kg, operating space: 1m wide × 1m deep × 1.5m high, gripping method: suction cup, material: aluminum, special function: built-in weight sensor, maximum cost: 100,000 yen, durability: over 100,000 cycles, hygiene requirements: suitable for food factories, installation space: maximum external dimensions within 1.2 meters, operating speed: within 3 seconds per cycle, mass production capability: standardized parts, maintainability: disassembled and washable." Replaceable parts: Suction cup is replaceable; On-site safety standards: Complies with collaborative robot standards; Laws and regulations: Complies with the Food Sanitation Law; Brand image: Company logo engraving; or enter: Object: Cylindrical plastic bottle, dimensions: diameter 7 cm × height 25 cm, weight 500 g; Working space: width 80 cm × depth 80 cm × height 1.2 m; Holding method: Gripper; Material: Resin; Special functions: Equipped with image sensor; Cost limit: 50,000 yen; Durability: 500,000 cycles; Hygiene requirements: Emphasis on cleanability; The system can be configured to meet various requirements, including: maximum external dimensions within 1 meter; operating speed within 2 seconds per cycle; mass production capability: compatible with 3D printers; maintainability: easy component replacement; replaceable parts: replaceable fingertip pads; on-site safety standards: compliant with ISO specifications; laws and regulations: CE marking; brand image: using brand colors. Alternatively, the system can specify details such as: object: irregularly shaped food packaging, maximum dimensions 30cm x 20cm x 10cm, weight 1kg; operating space: 1.5m wide x 1m deep x 1m high; gripping method: vacuum adsorption; material: stainless steel; special function: antibacterial coating; cost limit: 200,000 yen; durability: 200,000 cycles; hygiene requirements: aseptic operation; setup space: maximum external dimensions within 1.5 meters; operating speed: within 4 seconds per cycle; mass production capability: modular design; maintainability: easy disassembly and cleaning; replaceable parts: replaceable vacuum pads; on-site safety standards: fail-safe design; laws and regulations: RoHS directive; brand image: compatible with design patents. This allows for flexible specification of detailed requirements for each specific site. Specifically, the user input section of this system feeds the detailed request text into the natural language processing (NLP) model. This NLP model is a large-scale language model based on the Transformer architecture. The input data to this NLP model is a sequence of word embedding vectors. As a concrete example of this input data, the vector data derived from the text "Object: A rectangular cardboard box, dimensions 40 cm × 30 cm × 20 cm" can be cited. Another concrete example is the vector data derived from the text "Holding method: Suction cup, Material: Aluminum, Special function: Built-in weight sensor". This NLP model uses a self-attention mechanism to calculate the contextual dependencies of the input vector sequence. Based on the calculation results, this NLP model outputs the required vector.This requirement extracts a vector, for example, expressed as coordinate values ​​in a 2048-dimensional latent space. As a specific example of the output data, a vector containing categorical variables specifying the suction cup method could be given. Another specific example could be a vector containing numerical data representing the spatial constraint of a maximum external dimension of 1.2 meters. The user input unit converts the output requirement extraction vector into a structured data format. The user input unit then sends the structured data to the server. Through this processing, the system accurately extracts the parameters required for the design from unstructured text. The system incorporates human input variations to achieve consistent definition of design requirements. This technology can be applied to systems that automatically convert diverse customer requirements into specifications, as well as equipment adjustment systems based on verbal instructions from field operators.

[0074] The terminal sends this input data to the server. A generative AI processing unit is located on the server side, where the generative AI analyzes the input design requirements. The generative AI integrates multiple AI modules, including natural language processing, image recognition, structural optimization, material property reasoning, reference to past design databases, user history analysis, similar case retrieval, trend analysis, cost simulation, durability assessment, hygiene assessment, legal and regulatory compliance assessment, on-site feedback learning, on-site accident data analysis, equipment operation data analysis, component life prediction AI, anomaly detection AI, on-site operation video analysis AI, and on-site voice analysis AI, to make a comprehensive judgment. For example, when the suction cup method is specified, the system automatically designs the surface condition and weight of the object, handling speed, constraints of the working space, hygiene requirements, on-site temperature and humidity conditions, on-site dust, oil, and moisture environment, suction cup wear, deterioration, and replacement history, power consumption, noise, and maintenance history of the vacuum generating mechanism, chemical resistance, heat resistance, and wear resistance of the pad material, pad replaceability, pad cleanability, pad wear detection function, vacuum leak detection function, optimal layout of vacuum piping, selection of vacuum generating source, vacuum pressure control algorithm, automatic adjustment function of adsorption force, recovery action when adsorption fails, automatic cleaning mechanism of suction cup, automatic replacement mechanism of suction cup, and linkage of suction cup inventory management, etc. In the case of a gripper, the system automatically optimizes the number of fingers, their range of motion, fingertip shape, anti-slip material, presence or absence of force sensors, finger replaceability, gripping force, opening and closing speed, foreign object ingress prevention structure, cleanability, durability, cost, fingertip pad wear detection, automatic fingertip pad replacement mechanism, gripper self-alignment mechanism, gripper self-diagnostic function, gripper remote monitoring function, automatic stop function in case of gripper malfunction, on-site operator detection linkage function, and on-site safety standard compliance. In the case of a magnetic gripper, the system automatically designs the type of magnet, magnetic strength, loading and unloading mechanism, material suitability for the object, magnetic shielding, anti-misoperation measures, cost, durability, hygiene, automatic magnet loading and unloading mechanism, magnet wear detection, magnet temperature rise detection, and automatic stop function in case of magnet malfunction. Specifically, the generative AI processing unit of this system analyzes the received design requirement data. This generative AI processing unit inputs the design requirement data into the multilayer sensor. The input data to this multilayer perceptron is a multidimensional feature vector encoded with design requirements. A specific example of this input data could be a normalized vector containing the object's weight, dimensions, and required durability cycle count. Another specific example could be a vector with uniquely thermally encoded categorical variables for the specified gripping method and material. This multilayer perceptron performs a nonlinear transformation on the input vector. This multilayer perceptron outputs the optimal values ​​for each design parameter. A specific example of this output data could be a two-dimensional array representing the optimal configuration coordinates of the suction cup.As another concrete example, the continuous values ​​representing the necessary attractive forces of a vacuum generating mechanism can be cited. This generative AI processing unit generates an initial model of a 3D shape based on the optimal output values. This generative AI processing unit performs a structural analysis simulation based on the finite element method on the initial model. This generative AI processing unit evaluates the simulation results as a reward function for a reinforcement learning agent. This reinforcement learning agent iteratively updates the shape parameters in a way that maximizes the reward. Through this processing, the system autonomously explores non-habitual optimal designs that do not rely on human rules of experience. This system rapidly derives design schemes that satisfy complex physical constraints. This technology can be applied to fields requiring high levels of engineering knowledge, such as the automated design of robot end effectors and the design of components for industrial machinery operating in special environments.

[0075] Generative AI generates multiple design schemes optimized for specific applications and constraints by referencing past design data, design patterns, material property databases, user input history, industry standards, laws and regulations, cost data, durability data, hygiene data, user evaluation data, trend information, on-site feedback, on-site accident data, equipment operation data, component life prediction data, anomaly detection data, on-site operation video data, and on-site voice data. For example, the system automatically generates solutions with four suction cups, six suction cups, a telescopic robotic arm, resin-based joints for lightweight design, sensors integrated into the gripper front, antibacterial coatings, self-diagnostic capabilities, IoT communication functionality, easy-to-clean disassembly structures, durable reinforcements, cost-effective use of standard components, fail-safe designs compliant with site safety standards, collaborative robot solutions considering operator-assisted actions, dust-proof, chemical-resistant, heat-resistant, cold-resistant, UV-resistant, radiation-resistant, and low-noise designs tailored to the site environment, aesthetic designs that align with brand image, and designs that comply with relevant laws and regulations—variations responding to diverse site requirements. Specifically, the system's generative AI inputs past design data and site feedback data into a graph neural network. The input data to this graph neural network is graph structure data containing adjacency matrices representing connections between components and attribute vectors for each component. As a specific example of the input data, a list of nodes and edges representing the connection between the suction cup and the robotic arm can be given. As another specific example, tensors assigning attribute values ​​such as material, weight, and cost to each node can be given. This graph neural network uses a message-passing algorithm to propagate features between nodes. This graph neural network outputs the latent representation vector of the entire graph. As a specific example of the output data, numerical data representing the durability score of the entire design can be given. As another specific example, a score representing the probability that a specific component meets on-site safety standards can be given. This generative AI inputs the output latent representation vector into the decoder of a conditional variational autoencoder. This decoder generates multiple diverse design variants from the latent representation vector. This generative AI performs constraint satisfaction determination on each generated variant. This generative AI only filters design schemes that satisfy the constraints. Through this process, this system integrates past insights as mathematical models and applies them to new designs. This system not only suggests a single optimal solution but also suggests multiple alternatives with various trade-offs. This technology can be applied to automatically select alternative components based on supply chain changes and automatically generate derived models suitable for different on-site environments.

[0076] Multiple design schemes generated by the generative AI processing unit on the server are sent to the terminal. On the terminal, a wide variety of information is presented, including features, appearance images, dimensions, material information, functional specifications, cost estimates, durability assessments, hygiene assessments, mass production assessments, user feedback predictions, legal and regulatory compliance, assembly and disassembly steps, maintainability, customization, component replaceability, brand image compliance, on-site safety standard compliance, collaboration with on-site operators, compatibility with on-site equipment, suitability for the on-site working environment, suitability for on-site maintenance systems, suitability for on-site training systems, on-site IoT connectivity, on-site data connectivity, and on-site DX implementation. For example, 3D preview images, convenience assessments by application, cost estimates, durability assessments, hygiene assessments, assembly animations, disassembly animations, component lists, material lists, processing method lists, recommended production methods, recommended distribution methods, recommended packaging schemes, on-site operation video simulations, simulations of collaborative actions with on-site operators, interference check simulations with on-site equipment, on-site IoT connectivity simulations, and on-site maintenance operation simulations. Users can refer to this information while selecting the optimal design scheme. Specifically, the design suggestion unit of this system receives multiple design scheme data output from the generative AI processing unit. This design scheme data is a composite data structure containing 3D shape data, cost predictions, durability assessment scores, and on-site operation video simulation data. The design suggestion unit parses the received composite data structure. The design suggestion unit then passes the 3D shape data to the rendering engine. This rendering engine uses ray tracing algorithms to generate a high-precision 3D preview image. The design suggestion unit passes the on-site operation video simulation data to the physics engine. This physics engine uses rigid body mechanics and collision detection algorithms to calculate the robot's motion trajectory. The design suggestion unit generates animation data based on the calculation results. The design suggestion unit generates a user interface screen that integrates the 3D preview image, animation data, and various assessment scores. The design suggestion unit outputs the generated screen data to the terminal's display. The system monitors user selection operations. The system maintains the identifier of the selected design scheme in memory. Through this processing, the system provides users with multi-faceted evaluation indicators and dynamic simulation results in an intuitive and understandable form. The system supports the designer's decision-making process based on quantitative data and visual verification. This technology can be applied to essential engineering tasks such as pre-implementation verification of robot systems and layout discussions of production lines to confirm dynamic behavior.

[0077] When a user selects a design scheme, the terminal sends this selection information to the server. The 3D model generation unit on the server automatically generates a 3D model generation script usable in 3D modeling software based on the selected design scheme. This script is output in formats such as those of open-source 3D modeling tools (Python, STEP, STL, OBJ, 3MF, IGES, VRML, FBX, etc.), automatically generating a 3D model that reflects specified shape, size, material, structure, function, special functions, durability, hygiene requirements, on-site safety standards, on-site IoT connectivity, and on-site DX propulsion. For example, it automatically configures components such as suction cup robotic arms, joints, components with built-in sensors, reinforcing structures, removable and washable structures, antibacterial coatings, IoT communication modules, self-diagnostic functions, on-site personnel detection sensors, on-site IoT interface components, and on-site maintenance support components, and also automatically sets material information reflecting dimensions and materials. Furthermore, the script for generating the 3D model can also incorporate user-tunable parameter settings (such as dimensions, materials, presence or absence of components, function addition / deletion, IoT linkage settings, on-site personnel detection settings, on-site maintenance support settings, on-site DX advancement settings, etc.), automatic generation of multiple variants, durability analysis and cost estimation based on application, hygiene assessment, assembly / disassembly animation generation, automatic generation of component / material lists, processing method recommendation, mass production assessment, user evaluation prediction, legal and regulatory compliance check, on-site IoT linkage assessment, and on-site DX advancement assessment. Specifically, the 3D model generation unit of this system acquires the 3D voxel data of the selected design scheme. This 3D model generation unit converts the acquired 3D voxel data into polygon mesh data. In this conversion process, this 3D model generation unit executes the moving cube algorithm. This 3D model generation unit extracts the vertex coordinates and face connection information of the generated polygon mesh. This 3D model generation unit inputs the extracted geometric information into the script generation module. This script generation module uses a recurrent neural network to convert the geometric information into a syntax tree. This script generation module scans the syntax tree. Based on the scan results, it generates text data conforming to a specific programming language. This text data contains a set of commands executable on the 3D modeling software. The 3D model generation unit saves the generated text data as a script for 3D model generation. This 3D model generation unit automatically inserts parametric variable definitions into the script. These variable definitions provide an interface for users to subsequently change dimensions and materials. Through this process, the system automatically converts complex 3D shapes into a human-readable and editable parametric script format. This system ensures data compatibility between different CAD software. This system significantly improves the reusability and customizability of design data.This technology can be applied to a variety of use cases, such as collaborative design using open-source modeling tools and providing data to automated manufacturing lines.

[0078] In addition, even when users input multiple times for different purposes and under different conditions, generative AI can learn from past input history and selection preferences, on-site feedback, on-site accident data, equipment operation data, component life prediction data, anomaly detection data, on-site operation video data, on-site voice data, etc., and automatically generate proposals and design schemes optimized for each user.

[0079] For example, for users who have designed "suction cup" systems multiple times in the past, the system can learn their preferred suction cup configurations and materials, cost range, brand image, on-site IoT connectivity, and on-site DX propulsion capabilities, automatically generating more personalized proposals. Furthermore, when additional requirements and correction instructions are input for the user's selected design scheme (e.g., "increase the number of suction cups," "extend the length of the robotic arm," "change the material to stainless steel," "add sensors," "add IoT connectivity," "add on-site personnel detection," "add on-site maintenance support," "add on-site DX propulsion," etc.), the generative AI can immediately parse this content and automatically regenerate a new design scheme and 3D model generation script reflecting the corrections. Specifically, the generative AI of this system inputs the user's past input history and selection preferences into a recurrent neural network. The input data to this recurrent neural network is an embedded vector sequence of operation logs arranged in a time series. As a specific example of this input data, the sequence data of material IDs selected in the past 5 design sessions can be cited. As another specific example, the feature vector sequence of natural language text as additional requirement input can be cited. This recurrent neural network uses long short-term memory units to learn long-term dependencies in time-series data. The recurrent neural network outputs a vector of latent user preferences. As a specific example of this output data, a vector of weight coefficients representing a preference for durability over cost can be given. Another specific example is the probability distribution of liking a particular brand color. This generative AI uses the output preference vector as a conditional vector for generating the next design solution. Upon receiving additional requests from the user, the generative AI immediately uses a natural language processing model to extract a revised intent vector. The generative AI combines the revised intent vector with existing design data and inputs it into a 3D shape generation model. This 3D shape generation model performs local shape updates based on the combined data. Through this process, the system continuously learns the user's individual design style, improving the accuracy of personalized proposals. This system makes the user's latent needs explicit through an interactive design process. This technology can be applied to expert systems that learn the proprietary techniques of skilled designers and to recommendation systems that propose products optimized for each customer.

[0080] Furthermore, when users upload external design data, images, drawings, sketches, CAD data, 3D scan data, on-site operation video data, on-site voice data, on-site IoT data, and on-site DX data, the generative AI analyzes this data, comprehensively considering the specified purpose and constraints, on-site feedback, on-site accident data, equipment operation data, component life prediction data, anomaly detection data, on-site operation video data, on-site voice data, on-site IoT data, and on-site DX data, and automatically redesigns and optimizes the system according to the specified purpose and constraints, generating new design schemes and 3D model generation scripts. For example, by uploading the 3D scan data of an existing end effector and providing instructions such as "add a suction cup based on this shape," "change the material," "enlarge the size," "make it interchangeable with specific components," "add IoT linkage function," "add on-site personnel detection function," "add on-site maintenance operation support function," and "add on-site DX propulsion function," the generative AI can automatically optimize the design and output a 3D model generation script. Specifically, the external data acquisition unit of this system receives the uploaded external data. This external data acquisition unit preprocesses the received external data. For example, when 3D scan data is input, the external data acquisition unit performs denoising processing on the point cloud data. The external data acquisition unit inputs the denoised point cloud data into a 3D convolutional neural network. The input data to this 3D convolutional neural network is a tensor containing spatial coordinates and color information. A specific example of this input data is a point cloud tensor with 1024 vertices. Another specific example is a voxel mesh tensor with a resolution of 256. This 3D convolutional neural network extracts local geometric features from the input tensor. This 3D convolutional neural network propagates the extracted features to a fully connected layer. This 3D convolutional neural network outputs the shape category probability and size parameters of the object. A specific example of this output data is a score representing the probability that the object is an "existing end effector" with a probability of 0.95. Another specific example is numerical data representing the flange diameter of the object as 50 mm. This generative AI processing unit inputs the output shape category probability and size parameters into a subsequent optimization algorithm. This generative AI processing unit combines the extracted geometric features with an additional instruction vector from the user. This generative AI processing unit inputs the combined data into the conditional generative model to generate new design solutions. Through this processing, the system automatically reconstructs accurate design requirements from incomplete external data and integrates them with new requirements. This system seamlessly combines reverse engineering of existing products with custom design. This technology can be applied to various use cases, such as the retrofitting of old equipment and the automatic design of interface adapters for other companies' products.

[0081] Thus, in this embodiment of the invention, by adopting a structure where the user input unit is implemented on the terminal, and the generative AI processing unit and the 3D model generation unit are implemented on the server, the computational load on the terminal side can be reduced, while the generation of highly designed schemes utilizing high-performance generative AI and the processing of 3D model generation are achieved on the server side. The terminal can be composed of personal computers, tablet terminals, smartphones, industrial-specific terminals, wearable terminals, or terminals built into control panels within factories, etc., and the server can be composed of cloud servers, local servers, distributed processing servers, edge servers, or local servers within factories, etc. Specifically, this system adopts a client-server distributed computing architecture. This architecture eliminates the constraints of computing resources on the terminal side. This system maintains large-scale language models and 3D convolutional neural network models with large parameters configured on a parallel computing cluster using image processing devices on the server side. Through this configuration, the system significantly reduces the computation time required for inference processing. This system executes a lightweight web browser and dedicated applications on the terminal side. The terminal serializes the input data from the user and sends it to the server. The server converts the received data into tensor format. The server inputs the converted tensor data into the artificial intelligence model. This artificial intelligence model performs optimization calculations in a feature space of hundreds of millions of dimensions. Through these calculations, the system generates high-precision vertex coordinate groups and polygon data for a 3D model. The system encodes the generated data into a lightweight intermediate representation format. The server returns the encoded data to the terminal. The terminal uses graphics processing hardware to render the received data onto the screen. Through this series of processing steps, the system suppresses terminal battery consumption and heat generation. The system also maximizes the utilization of scalable computing resources on the server side. As a result of this technology, the system can perform highly AI-driven design tasks even from low-performance control terminals in mobile environments and factories. Specific application areas of this system include ad-hoc fixture design at production line start-up and maintenance / inspection sites in manufacturing.

[0082] Thus, according to an embodiment of the present invention, by combining highly automated design / automatic 3D model generation processing on the server side, centered on generative AI, with an intuitive user interface on the terminal side, the efficiency of the initial design and prototyping processes in the design / prototyping of robotic arms and end effectors for factory robots can be significantly improved. This enables a system where even users without specialized knowledge can easily and quickly automatically generate diverse 3D models. Specifically, this system has an interface layer that converts intuitive human input into machine-readable high-dimensional vector data. This interface layer uses natural language processing and image recognition models to encode user-input natural language text and 2D sketch images into feature vectors. The system transmits the encoded feature vectors to the generative AI model on the server side. This generative AI model possesses a self-attention mechanism based on the Transformer architecture. This self-attention mechanism calculates the dependencies between the elements of the input feature vectors. Through this calculation, the model maps physical constraints such as the joint angles of the robotic arm and the gripping force of the end effector within the latent space. The model decodes 3D voxel data and point cloud data from the mapped latent representation. In this decoding process, the model functions as a generator within a generative adversarial network (GAN). The system inputs the generated 3D data into a physics simulation engine. This engine automatically performs stress distribution and interference checks on the generated shape. The system feeds the simulation results back to the reinforcement learning algorithm as a reward signal. Through this feedback loop, the system autonomously improves the physical appropriateness of the design. As a result, this system reduces the trial-and-error process that would take human designers days to a few seconds to minutes of computer processing. This system eliminates the human element in the design process, guaranteeing a quality level above a certain standard. Specific application areas of this system include rapid changeover of robotic grippers and on-demand manufacturing of customized parts in smart factories producing a wide variety of products in small quantities.

[0083] (System Composition) The system described in this embodiment includes a user input unit, a data transmission unit, a generative AI processing unit, a design scheme prompting unit, a user selection unit, a 3D model generation unit, a script output unit, a history management unit, and an external data acquisition unit. The user input unit allows users to input information required for the design of the robotic arm and end effector of the factory robot on a terminal. For example, as the shape of the object being handled, it can specify cuboids, cylinders, spheres, irregular shapes, bags, plates, soft materials, liquid containers, powder bags, etc.; as dimensions, it can input plates with dimensions of 40 cm wide × 30 cm deep × 20 cm high, 7 cm diameter × 25 cm high, and 2 mm thick, with a maximum external dimension of 100 mm or less; as weight, it can specify 5 kg, 100 g, 20 kg, a maximum of 30 kg, and a minimum of 10 g, etc. Regarding the dimensions and constraints of the workspace, input options include 1 meter wide × 1 meter deep × 1.5 meters high, ceiling height limit, presence or absence of obstacles, range of motion limitation, interference avoidance with adjacent equipment, collaborative action area with operators, and height limit of the transport assembly line. For gripping methods, options include suction cups, grippers, magnets, vacuum grippers, clamps, tweezers, hooks, straps, vacuuming, adhesive pads, air blowing, electrostatic adsorption, soft finger grippers, multi-finger robot grippers, and soft grippers for collaborative robots. For desired materials, input options include aluminum, resin, stainless steel, carbon fiber, rubber, food hygiene materials, heat-resistant materials, chemical-resistant materials, antistatic materials, antibacterial materials, lightweight composite materials, and 3D printing resins. For required special functions, input options include heating, cooling, weight sensors, image sensors, force sensors, static elimination, antibacterial coatings, wear detection, self-diagnostic functions, IoT communication functions, QR code readers for workpiece identification, automatic workpiece identification AI, anomaly detection AI, remote monitoring functions, self-position correction functions, automatic component replacement functions, and automated cleaning functions. As for cost limits, users can specify 100,000 yen, 300,000 yen, 1,000,000 yen, cost per unit, cost during mass production, unit price of components, maintenance costs, etc. For durability, users can input 100,000 cycles, 1,000,000 cycles, impact resistance, wear resistance, corrosion resistance, chemical resistance, heat resistance, low temperature resistance, long-term lubrication-free operation, maintenance-free operation, component replacement cycle, MTBF, etc. For hygiene requirements, users can specify requirements for food factories, medical settings, emphasis on cleanability, prevention of foreign matter contamination, aseptic operation, cleaning fluid resistance, autoclave compatibility, removable cleaning, antibacterial coating, and inclusion of foreign matter detection sensors, etc. Users can also simultaneously input multiple settings, including operating speed, mass production capability, maintainability, availability of replacement parts, on-site safety standards, legal and regulatory compliance, brand image, and specific on-site requirements. Specifically, the user input unit temporarily stores the multi-dimensional parameter group as a structured data object in memory. This system performs preprocessing to input the various parameters into the embedding layer of the natural language processing model.In this preprocessing, the system normalizes and converts the dimensions and weight of numerical data into continuous value vectors. The system converts the shape and material of categorical data into one-hot encoded or learned word embedding vectors. The system combines these heterogeneous vectors to generate a single high-dimensional input tensor. For example, the system represents a value such as 40 cm wide, 30 cm deep, 20 cm high, and 5 kg weight as a four-dimensional floating-point array. The system represents the category of suction cups as a 512-dimensional dense vector. The system combines these and uses them as input data to the artificial intelligence model. The system uses mean imputation and a prediction model to automatically imput missing values ​​in the input data. The system uses a rule-based inference engine to pre-detect logical contradictions between input constraints. For example, the system generates a warning sign when a heat-resistant material is specified along with a low-cost upper limit. As a result, the system ensures the quality of the input data for the subsequent artificial intelligence model. The system improves the inference accuracy of the artificial intelligence model and prevents the generation of invalid design solutions. As a specific application area of ​​this system, it is used in the component definition stage of automated equipment with complex intertwined components as a support tool to enable business operators and field technicians to input components into the system without omission.

[0084] The data transmission unit is responsible for sending all information input from the user input unit to the server. It also performs network communication encryption, data format conversion, retransmission control in case of transmission errors, integration checks of input data, automatic data collection from field terminals, automatic transmission of field information based on barcodes and RFID, real-time transmission of field IoT data, and uploading and transmission of field operation video and audio data. Specifically, this data transmission unit controls the communication module of the terminal and establishes a secure transport layer protocol. This data transmission unit groups the input structured data and multimedia data. This data transmission unit uses public-key cryptography to encrypt the grouped data. This data transmission unit aggregates time-series temperature and vibration data obtained from field sensors according to a specific time window. This data transmission unit formats the aggregated time-series data into a tensor format that can be input into a one-dimensional convolutional neural network. This data transmission unit samples frames from the field operation video and performs downsampling processing to reduce resolution. This data transmission unit streams the downsampled image tensor to the server. This data transmission unit constantly monitors the network bandwidth. This data transmission unit performs adaptive bit rate control to dynamically increase the compression rate of transmitted data when bandwidth is reduced. It calculates the hash value of transmitted packets, enabling data integrity verification on the server side. As a result of this technology, even in unstable wireless communication environments within factories, this system ensures that large volumes of design requirement data and on-site context data arrive at the server reliably and securely without loss. Specifically, this system is used in edge computing environments where real-time feedback of on-site conditions from manufacturing sites with high levels of radio interference and factories in remote areas is transmitted to artificial intelligence servers in the cloud.

[0085] The generative AI processing unit operates on the server, analyzing the design requirements sent by the user. It integrates multiple AI modules, including natural language processing, image recognition, structural optimization, material property reasoning, reference to past design databases, user history analysis, similar case retrieval, trend analysis, cost simulation, durability assessment, hygiene assessment, legal and regulatory compliance assessment, on-site feedback learning, on-site accident data analysis, equipment operation data analysis, component life prediction AI, anomaly detection AI, on-site operation video analysis AI, and on-site voice analysis AI, to make a comprehensive judgment. For example, when the suction cup method is specified, the system automatically designs the surface condition and weight of the object, handling speed, workspace constraints, hygiene requirements, on-site temperature and humidity conditions, on-site dust / oil / moisture environment, suction cup wear / deterioration / replacement history, power consumption / noise / maintenance history of the vacuum generating mechanism, chemical resistance / heat resistance / wear resistance of the suction cup material, suction cup replaceability, suction cup cleanability, suction cup wear detection function, vacuum leak detection function, optimal layout of vacuum piping, selection of vacuum generating source, vacuum pressure control algorithm, automatic adjustment function of adsorption force, recovery action when adsorption fails, automatic cleaning mechanism of suction cup, automatic replacement mechanism of suction cup, and linkage of suction cup inventory management, etc. In the case of a gripper, the system automatically optimizes the number of fingers, the range of motion of the fingers, the shape of the fingertips, the anti-slip material, the presence or absence of force sensors, finger replaceability, gripping force, opening and closing speed, foreign object ingress prevention structure, cleanability, durability, cost, fingertip pad wear detection, automatic fingertip pad replacement mechanism, automatic centering mechanism of the gripper, self-diagnostic function of the gripper, remote monitoring function of the gripper, automatic stop function in case of gripper malfunction, on-site operator detection linkage function of the gripper, and on-site safety standard compliance of the gripper. In the case of a magnetic gripper, the system automatically designs the type of magnet, magnetic strength, loading and unloading mechanism, material suitability of the object, magnetic shielding, prevention of malfunction, cost, durability, hygiene, automatic magnet loading and unloading mechanism, magnet wear detection, magnet temperature rise detection, and automatic stop function in case of magnet malfunction. As a specific example of a prompt text that the generative AI can read, it could be something like: "Object: A rectangular cardboard box, 40cm x 30cm x 20cm, weighing 5kg; Operating space: 1m wide x 1m deep x 1.5m high; Holding method: Suction cup; Material: Aluminum; Special function: Built-in weight sensor; Cost limit: 100,000 yen; Durability: Over 100,000 cycles; Hygiene requirements: Suitable for food factories; Installation space: Maximum external dimensions within 1.2 meters; Action speed: Within 3 seconds per cycle; Mass production capability: Standardized parts; Maintainability: Disassembleable and washable; Replaceable parts: Replaceable suction cup; On-site safety standards: Complies with collaborative robot standards; Laws and regulations: Complies with the Food Sanitation Law; Brand image: Company logo engraving," etc. Specifically, this generative AI processing unit employs an integrated learning architecture that integrates multiple specialized neural network models.This processing unit uses a large-scale language model's tokenizer to segment the input prompt into a tokenized sequence. This processing unit converts the tokenized sequence into multi-dimensional word embedding vectors. This processing unit inputs the converted vectors into a Transformer-based encoder. This encoder extracts the contextual meaning of the input data and outputs a 512-dimensional context vector. This processing unit inputs this context vector into a graph neural network for structure optimization. This graph neural network generates a graph structure with the components of the robot gripper as nodes and the physical connections between components as edges. This graph neural network assigns material and size parameters to each node. For example, to determine the configuration of the suction cups, this processing unit enables a reinforcement learning agent to move in virtual space. This agent receives the object's center of gravity position and surface roughness data as input. This agent outputs the optimal coordinates of the suction cups as a 3D vector. This processing unit performs a stress analysis simulation based on the finite element method on the output coordinate group in the background. This processing unit incorporates the maximum stress value and deformation obtained from the simulation results into a loss function. This processing unit uses gradient descent to optimize the thickness of the components and the configuration of the reinforcing ribs. This processing unit generates structured data as output data, including the 3D bounding box coordinates, material IDs, degrees of freedom of connecting joints, and estimated cost values ​​for each component. This output data is then passed to the subsequent 3D model generation module. As a result of this technology, the system is not simply a text response, but autonomously synthesizes usable mechanical design data that conforms to physical laws and manufacturing requirements. This system uses mathematical optimization methods to address trade-offs between multiple variables that human engineers cannot fully consider. Specific application areas of this system include the conceptual design of end effectors during the startup of new production lines and the redesign of fixtures to improve the yield of existing equipment.

[0086] The design scheme prompting department sends multiple design schemes generated by the generative AI processing department to the terminal. The terminal displays information from various perspectives on each scheme, including features, appearance images, dimensions, material information, functional specifications, cost estimates, durability assessments, hygiene assessments, mass production assessments, user feedback predictions, legal and regulatory compliance, assembly steps, disassembly steps, maintainability, customization, component replaceability, brand image compliance, on-site safety standard compliance, collaboration with on-site personnel, interchangeability with on-site equipment, suitability for the on-site working environment, suitability for on-site maintenance systems, suitability for on-site education systems, on-site IoT connectivity, on-site data connectivity, and on-site DX implementation. For example, 3D preview images, convenience assessments by application, cost estimates, durability assessments, hygiene assessments, assembly animations, disassembly animations, component lists, material lists, processing method lists, recommended production methods, recommended distribution methods, recommended packaging schemes, on-site operation video simulations, simulations of collaborative actions with on-site personnel, interference inspection simulations with on-site equipment, on-site IoT connectivity simulations, and on-site maintenance operation simulations are also simultaneously displayed. Specifically, this design solution's prompting unit uses the terminal's graphics processing unit to render multi-dimensional design data received from the server in real time. This prompting unit reads the vertex data and normal vectors of the generated 3D model. This prompting unit executes shader programs, using physically based rendering techniques to depict the luster of metal and the texture of resin on the screen. This prompting unit generates visual components such as radar charts and bar graphs for the evaluation indicators of each design scheme. This prompting unit normalizes values ​​such as cost and durability, calculating them as scores from 0 to 100. This prompting unit binds this score to the corresponding widget on the user interface. To generate assembly animations, this prompting unit interpolates the transformation matrix from the local coordinate system to the global coordinate system of each component in a time-series manner. This prompting unit dynamically depicts the component assembly process based on this interpolated data. This prompting unit continuously calculates the distance between the virtual worker model and the generated robot gripper model in the on-site operation video simulation. This prompting unit executes a collision detection algorithm that highlights warning colors when the distance is below a safety threshold. As a result of this technology, this system transforms complex design data into intuitive and understandable visual information. This system enables users, even those without professional drawing reading skills, to accurately assess the appropriateness of design schemes. As a specific application area, this system is used in design review meetings involving management and on-site operators as a presentation tool to facilitate decision-making among stakeholders based on a shared visual framework.

[0087] The user selection unit allows users to choose the best design from multiple suggested options. Users can compare and discuss these options while rotating and zooming in on a 3D preview, or compare detailed information, cost, durability, hygiene, mass production feasibility, legal and regulatory compliance, on-site IoT integration, and on-site DX implementation. Alternatively, users can consider user feedback predictions, brand image compatibility, coordination with on-site personnel, compatibility with on-site equipment, adaptability to the on-site working environment, adaptability to on-site maintenance systems, adaptability to on-site training systems, on-site IoT integration, on-site data integration, and on-site DX implementation. Furthermore, users can input additional requirements or modification instructions for their selected options (e.g., increasing the number of suction cups, extending the length of the robotic arm, changing the material to stainless steel, adding sensors, adding IoT integration functions, adding on-site personnel detection functions, adding on-site maintenance support functions, adding on-site DX implementation functions, etc.). This information is then sent to the generative AI processing unit, which automatically generates a new design reflecting the modifications. Specifically, this user selection unit monitors input events from the terminal's touch panel or mouse. This selection unit calculates the coordinate changes of a sliding or pinching operation. It converts these changes into a rotation matrix or translation vector for the camera in 3D space. The selection unit uses these converted parameters to update the view matrix, changing the display of the 3D model in real time. When the user clicks on a specific part, the selection unit executes a ray casting algorithm. This algorithm emits virtual rays from 2D coordinates on the screen into 3D space, determining intersecting polygons. The selection unit displays the metadata of the determined part. The selection unit receives modification instructions from the user as natural language text or direct editing of parameters. It extracts the modification instructions as differential data and combines them into the original prompt vector. The selection unit then sends this combined vector back to the generative AI model on the server. This generative AI model maintains the latent representation of the previously generated result as the initial state. Based on the input differential vector, the model calculates small movements within the latent space. The model decodes a new design scheme from the latent representation after this movement. As a result of this technology, the system significantly reduces the generation time of modified schemes by avoiding recalculation from scratch. This system implements an iterative process for interactive design that accurately reflects the user's intent. As a specific application area of ​​this system, it is used in face-to-face business or online collaborative design sessions to finalize the specifications of customized machinery while listening to customer requirements on the spot.

[0088] The 3D model generation unit automatically generates 3D model generation scripts usable in 3D modeling software based on the user-selected design scheme. For example, it outputs scripts in open-source 3D modeling tool formats (Python script, STEP format, STL format, OBJ format, 3MF format, IGES format, VRML format, FBX format, etc.), automatically generating 3D model content that reflects specified shape, size, material, structure, function, special functions, durability, hygiene requirements, on-site safety standards, on-site IoT linkage, and on-site DX propulsion information. For example, it automatically configures components such as suction cup robotic arms, joints, components with built-in sensors, reinforced structures, removable and washable structures, antibacterial coatings, IoT communication modules, self-diagnostic functions, on-site personnel detection sensors, on-site IoT linkage interfaces, and on-site maintenance support, and also automatically sets material properties reflecting dimensional and material information. Specifically, this 3D model generation unit performs syntactic analysis on the structured data output by the generative artificial intelligence model. This generation unit maps the analyzed geometric parameters of each component to function calls in a specific programming language. This generation unit constructs an abstract syntax tree, defining the hierarchical structure and parent-child relationships of components. While traversing this abstract syntax tree, the generation unit sequentially generates the corresponding script code as strings. For example, to generate a cylindrical robotic arm, this generation unit describes a cylinder generation function with radius and height as parameters. This generation unit inserts Boolean sum / difference code at the joints between components. In the material settings, this generation unit defines physically based rendering parameters representing the material's reflectivity or roughness as variables. This generation unit generates code that assigns these variables to the attributes of each component. For electronic components such as sensors or communication modules, this generation unit describes command statements for importing the corresponding models from a pre-prepared standard component library. As output data, this generation unit generates executable text files or binary CAD intermediate files that are thousands of lines long. As a result of this technology, this system converts conceptual design data inferred by artificial intelligence into a standard format that can be directly embedded into existing manufacturing workflows. This system ensures a seamless transition from design data to manufacturing data. As a specific application area of ​​this system, it is used in the process of directly reading the generated model into finite element analysis software for detailed strength calculation, or directly inputting it into the slicing software of a 3D printer for rapid prototyping.

[0089] The script output unit sends the 3D model generation script generated by the 3D model generation unit to the terminal, allowing users to download or import it into external 3D modeling software. For example, users can download the script and edit and apply it within the company's design department, or send it directly to external 3D printers or processors. Specifically, the script output unit uses a lossless compression algorithm to compress the generated script file or CAD data. The output unit applies a hash function to the compressed archive file to calculate a checksum. The output unit transmits the archive file to the terminal via a secure communication protocol. Upon completion of the download on the terminal side, the output unit checks the checksum to verify that no data is missing. The output unit has the capability to directly push data to external cloud-based CAD systems or manufacturing execution systems via an application programming interface (API). During this transmission, the output unit uses an authentication protocol to implement secure token-based access control. The output unit includes middleware that dynamically converts data formats according to the specifications of the target system. As a result of this technology, the system securely and efficiently integrates the generated design assets with external systems. This system prevents data silos and constructs a digital pipeline from design to manufacturing. As a specific application area of ​​this system, it is utilized in cloud manufacturing platforms that automatically transmit design data from fabless companies to their partner's processing plants and immediately perform cost estimation or manufacturing arrangements.

[0090] The History Management Department centrally manages user input history, selection history, generated design schemes, 3D model generation scripts, download history, modification history, each user's preference data, on-site feedback, on-site accident data, equipment operation data, component life prediction data, anomaly detection data, on-site operation video data, on-site voice data, on-site IoT data, and on-site DX data. For example, for users who have designed suction cup methods multiple times in the past, the department learns their preferred suction cup configurations or materials, cost range, brand image, on-site IoT connectivity, and on-site DX implementation, enabling more personalized proposals automatically. Specifically, this History Management Department is composed of a distributed relational database and a non-relational database. The department stores structured parameter data in tables within the relational database. It stores unstructured video or voice data in object storage and records their reference addresses in the database. The department accumulates each user's operation logs in a time-series database. It applies collaborative filtering or content-based recommendation algorithms to this accumulated data set. The department models users' past selection preferences as multi-dimensional vectors. The department calculates the cosine similarity between the user's vector and the vector of the newly generated design scheme. Based on the similarity score, this management department dynamically sorts the order of design suggestions. It uses feedback data from the field or component lifespan data as teacher data to fine-tune the weight parameters of the generative artificial intelligence model. As a result of this technology, the more the system is used, the more precise the design solutions it outputs, tailored to the user's individual preferences or specific factory environments. This system accumulates tacit knowledge, relying on individual design know-how, as explicit knowledge within the system. In specific application areas, this system is used as a knowledge management system to learn from the design processes of skilled designers in large enterprises, providing design support or education for young engineers.

[0091] The external data import unit analyzes user-uploaded external design data, images, drawings, sketches, CAD data, 3D scan data, on-site operation video data, on-site voice data, on-site IoT data, and on-site DX data. It comprehensively considers specified purposes and constraints, on-site feedback, on-site accident data, equipment operation data, component life prediction data, anomaly detection data, and other relevant factors. Based on these requirements, it automatically redesigns or optimizes the design, generating new design schemes or 3D model generation scripts. For example, by uploading existing 3D scan data of an end effector and providing instructions such as "add a suction cup based on this shape," "change material," "enlarge dimensions," "make it compatible with specific components," "add IoT linkage function," "add on-site personnel detection function," "add on-site maintenance support function," and "add on-site DX advancement function," the generative AI can automatically optimize the design and output a 3D model generation script. Specifically, this external data import unit has a data pipeline for standardizing uploaded data in various formats. Upon receiving point cloud data as 3D scan data, this import unit reduces the data size using random sampling or voxel mesh filters. This point cloud data is then input into a 3D convolutional neural network. The network extracts local and global geometric features of the point cloud and outputs a fixed-length global feature vector. Upon receiving a 2D drawing image, this import unit extracts dimension lines or annotation text using optical character recognition (OCR). It then uses an edge detection algorithm to digitize the contour vectors. These extracted feature vectors are combined with modification instruction vectors from the user based on natural language. This combined vector is then used as input to a generative artificial intelligence (AI) model. This generative AI model performs deformation processing, maintaining the existing topology of the input shape while reconfiguring the vertex coordinates of the mesh to meet specified constraints. As a result of this technology, the system can reuse existing physical assets or past design data according to the latest requirements without wasting them. This system fully automates the reverse engineering process, dramatically shortening design delivery times. As a specific application area of ​​this system, it is used in maintenance and repair work where replacement parts of old equipment with lost drawings are immediately redesigned from physical scans and output using a 3D printer.

[0092] Thus, the system involved in this embodiment includes a user input unit, a data transmission unit, a generative AI processing unit, a design scheme prompting unit, a user selection unit, a 3D model generation unit, a script output unit, a history management unit, and an external data import unit. Through the coordinated actions of these units, the system enables users, even without professional 3D modeling skills or knowledge, to easily select the best option from multiple design options in a short time and automatically generate a specific 3D model. Specifically, this system containerizes and implements these multiple independent software modules as a microservice architecture. The system uses a message queue protocol to asynchronously handle communication between containers. The system uses orchestration tools to dynamically expand computing resources based on the load of each module. The system uses event-driven control to manage a series of data flows from user input to the final script output. The system caches intermediate data generated in each processing step in an in-memory database. Through this caching mechanism, the system minimizes the recalculation cost when the user returns to the previous step and changes conditions. The system executes the inference processing of the artificial intelligence model and the rendering processing of the 3D model in parallel threads. Through this parallel processing, the system maintains a high level of responsiveness for the user interface. As a result of this technology, the system seamlessly integrates computationally demanding artificial intelligence processing and real-time user operations on a single platform. This improves the overall availability and fault tolerance of the system. In specific application areas, this system is utilized as a global design sharing platform accessible simultaneously from multiple locations worldwide, or as the backend system of a cloud-based integrated development environment.

[0093] (Implementation steps) Step 1: User requests input Users can input detailed information required for the design of the robotic arm or end effector of the factory robot using the user input section on the terminal. For example, as for the shape of the object being handled, users can specify cuboids, cylinders, spheres, irregular shapes, bags, plates, soft materials, liquid containers, powder bags, etc.; as for dimensions, users can input dimensions such as 40 cm wide × 30 cm deep × 20 cm high, 7 cm diameter × 25 cm high, 2 mm thick plate, and a maximum external dimension of 100 mm or less; as for weight, users can specify 5 kg, 100 g, 20 kg, a maximum of 30 kg, a minimum of 10 g, etc. Regarding the dimensions and constraints of the workspace, input the following: width 1m × depth 1m × height 1.5m; ceiling height limit; presence or absence of obstacles; range of motion limit; interference avoidance with adjacent equipment; collaborative action area with operators; and height limit of the transport line. Regarding the gripping method, select suction cups, grippers, magnets, vacuum grippers, clamps, tweezers, hooks, straps, vacuum adsorption, adhesive pads, air blowing, electrostatic adsorption, soft finger grippers, multi-finger robotic hands, and soft grippers for collaborative robots. Regarding the desired materials, specify aluminum, resin, stainless steel, carbon fiber, rubber, food hygiene materials, heat-resistant materials, chemical-resistant materials, antistatic materials, antibacterial materials, lightweight composite materials, and 3D printing resins. Regarding the required special functions, input heating, cooling, weight sensors, image sensors, force sensors, static elimination, antibacterial coatings, wear detection, self-diagnostic functions, IoT communication functions, QR code readers for workpiece identification, automatic workpiece identification AI, anomaly detection AI, remote monitoring functions, self-positioning correction functions, automatic component replacement functions, and automated cleaning functions. It can also simultaneously input multiple cost limits, durability, hygiene requirements, setup space, operating speed, mass production capability, maintainability, availability of replacement parts, on-site safety standards, legal and regulatory compliance, brand image, and other special on-site requirements. Specifically, in this step, the system dynamically expands the input form in the terminal's memory. The system parses the text string entered by the user in real time. The system uses regular expressions to separate numerical values ​​and units and automatically converts them to the internal standard unit system. The system uses a hash function to map the input categorical data to a unique identifier. When the user inputs vague expressions such as "soft" or "amorphous," the system invokes a natural language processing model. This model retrieves similar objects from a past database and suggests specific physical parameters to the user. The system constructs multiple input constraints as a multi-dimensional constraint matrix in memory. The system assigns weighting coefficients to each element of this matrix. As a result of this technology, the system converts the ambiguity or vagueness of human expression into a strict mathematical expression that can be processed by the machine. By immediately detecting input errors at the front end and prompting corrections, the system eliminates the need for subsequent useless calculations in artificial intelligence processing.As a specific application area for this step, it is utilized by field workers who are not professional CAD operators to intuitively input the required information using a tablet terminal.

[0094] Step 2: Input data and send All information entered in the user input section is sent to the server via the data transmission section. Simultaneously, network communication encryption, data format conversion, retransmission control in case of transmission errors, input data consistency checks, automatic data collection from field terminals, automatic transmission of field information based on barcodes or RFID, real-time transmission of field IoT data, and uploading and transmission of field operation video or voice data are also performed. Specifically, in this step, the system serializes the constructed constraint matrix or multimedia data into a single payload. The system encrypts this payload using an encryption protocol. The system sends the encrypted data as an asynchronous network request to the server endpoint. If packet loss is detected during transmission, the system uses an exponential backoff algorithm to gradually increase the retransmission interval while attempting retransmission. The system subscribes to sensor data streaming from the field programmable logic controller. The system divides this time-series data into batches of a certain size and asynchronously merges them into the main payload. The system calculates inter-frame differences for the uploaded video data and compresses the data size by removing frames with less activity. As a result of this technology, the system efficiently utilizes network bandwidth while seamlessly aggregating information from multiple data sources to the server. This system ensures the reliability and security of communications. As a specific application area, this procedure is utilized for the secure transmission of data from overseas factories or underground manufacturing lines with vulnerable communication infrastructure to cloud servers.

[0095] Step 3: Automatic generation of design solutions based on generative AI The generative AI processing unit on the server analyzes the design requirements sent by the user, and links multiple AI modules such as natural language processing, image recognition, structural optimization, material property reasoning, reference to past design databases, user history analysis, similar case retrieval, trend analysis, cost simulation, durability assessment, hygiene assessment, legal and regulatory compliance assessment, on-site feedback learning, on-site accident data analysis, equipment operation data analysis, component life prediction AI, anomaly detection AI, on-site operation video analysis AI, and on-site voice analysis AI to make comprehensive judgments and automatically generate multiple design schemes optimized for the purpose and constraints. For example, when the suction cup method is specified, the system automatically designs the surface condition and weight of the object, handling speed, constraints of the working space, hygiene requirements, on-site temperature and humidity conditions, on-site dust, oil, and moisture environment, suction cup wear, deterioration, and replacement history, power consumption, noise, and maintenance history of the vacuum generating mechanism, chemical resistance, heat resistance, and wear resistance of the suction cup material, suction cup replaceability, suction cup cleanability, suction cup wear detection function, vacuum leak detection function, optimal layout of vacuum piping, selection of vacuum generating source, vacuum pressure control algorithm, automatic adjustment function of adsorption force, recovery action when adsorption fails, automatic cleaning mechanism of suction cup, automatic replacement mechanism of suction cup, and linkage of suction cup inventory management, etc. As a specific example of a prompt text for generative AI to read, it could be something like: "Object: A rectangular cardboard box, dimensions 40cm x 30cm x 20cm, weight 5kg; operating space: 1m wide x 1m deep x 1.5m high; gripping method: suction cup; material: aluminum; special function: built-in weight sensor; maximum cost: 100,000 yen; durability: over 100,000 cycles; hygiene requirements: suitable for food factories; installation space: maximum external dimensions within 1.2 meters; action speed: within 3 seconds per cycle; mass production capability: standardized parts; maintainability: removable and washable; replaceable parts: replaceable suction cup; on-site safety standards: compliant with collaborative robot standards; laws and regulations: compliant with the Food Sanitation Law; brand image: company logo engraving." Specifically, in this step, the system deserializes the received payload and distributes the data to each AI module. The system uses a large-scale language model to infer the design intent or hidden constraints from the prompt text. The system inputs the inference results as conditional vectors of the latent space into the generator network. This generator network employs a variational autoencoder architecture. It combines a noise vector sampled from a normal distribution with a conditional vector. From this combined vector, the network outputs the occupancy probability of each cell in the 3D voxel mesh. The system applies the moving cube method to the output probability distribution to extract a smooth polygonal mesh. A topology optimization algorithm is then applied to the generated mesh. This algorithm iteratively reduces unnecessary volume while satisfying specified weight or stress conditions. As output data, the system generates 3D shape data for each design scheme, a component composition table, and a list of predicted performance scores.This system passes the output data to subsequent prompts. As a result of this technology, based on mathematical principles, the system creates lightweight and high-strength amorphous shapes that are difficult for human intuition to achieve. The system explores a set of Pareto optimal solutions that simultaneously satisfy multiple constraints in a short time. Specific applications of this step include lightweight design of components in the aerospace or automotive industries, and the automated design of innovative, biomimetic robotic hands.

[0096] Step 4: Design scheme presentation and user selection. The multiple design schemes generated by the generative AI processing unit are sent to the terminal via the design scheme presentation unit, and the terminal presents information from multiple perspectives for each scheme, including features, appearance images, dimensions, material information, functional descriptions, cost estimation, durability assessment, sanitation assessment, mass production assessment, user evaluation prediction, legal and regulatory compliance, assembly steps, disassembly steps, maintainability, customizability, part replaceability, brand image compliance, on-site safety standard compliance, coordination with on-site operators, compatibility with on-site equipment, adaptability to on-site operation environments, adaptability to on-site maintenance systems, adaptability to on-site education systems, on-site IoT linkage, on-site data linkage, on-site DX promotion, and the like. The user selects the optimal design scheme while referring to 3D preview images, usability assessment by purpose, cost estimation, durability assessment, sanitation assessment, assembly animations, disassembly animations, part lists, material lists, processing method lists, recommended production methods, recommended distribution methods, recommended packaging solutions, on-site operation video simulation, coordinated action simulation with on-site operators, interference check simulation with on-site equipment, on-site IoT linkage simulation, on-site maintenance operation simulation, and the like. Specifically, in this step, the system caches multiple sets of design scheme data sent from the server into the local memory of the terminal. The system calls a graphics processing interface to render the 3D model on the browser using hardware acceleration. The system analyzes the metadata of each design scheme, graphs various indicators, and arranges them on a dashboard in the user interface. When the user hovers the mouse over a specific design scheme, the system uses a natural language generation model to summarize the features of that scheme and displays the summary as a tooltip. When receiving a playback request for a simulation video, the system reads kinematic data pre-calculated on the server side. Based on this data, the system updates the joint angles of the robotic arm over time to animate the model on the screen. When the user presses the selection button, the system acquires the unique identifier of the design scheme. The system stores this identifier in the state management module and triggers a trigger for transition to subsequent processing. As a technical effect, the system organizes and presents multi-dimensional design data with a huge amount of information in a form that reduces the user's cognitive load. The system supports users in making data-driven objective decisions. As a specific application field of this step, this step is utilized in competitions or design review meetings of procurement departments that select the scheme with the best balance between cost and performance from multiple design schemes.

[0097] Step 5: Input and Regeneration of Additional Requirements and Correction Instructions. Users can input additional requirements or correction instructions for the selected design scheme (such as increasing the number of suction cups, extending the length of the robotic arm, changing the material to stainless steel, adding sensors, adding IoT linkage functions, adding on-site operator detection functions, adding on-site maintenance support functions, adding on-site DX propulsion functions, etc.). This information is sent again to the generative AI processing department, which automatically generates a new design scheme that reflects the corrections. As a specific example of a prompt text that generative AI can read, it could be something like: "Object: A rectangular cardboard box, dimensions 40cm x 30cm x 20cm, weight 5kg; Operating space: 1m wide x 1m deep x 1.5m high; Holding method: Suction cup; Material: Stainless steel; Special functions: Built-in weight sensor, additional IoT linkage function; Maximum cost: 120,000 yen; Durability: Over 150,000 cycles; Hygiene requirements: Suitable for food factories; Installation space: Maximum external dimensions within 1.2 meters; Action speed: Within 2.5 seconds per cycle; Mass production capability: Standardized parts; Maintainability: Disassembleable and washable; Replaceable parts: Replaceable suction cup; On-site safety standards: Complies with collaborative robot standards; Laws and regulations: Complies with the Food Sanitation Law; Brand image: Company logo engraving," etc. Specifically, in this step, the system inputs the corrected text from the user into a natural language processing model to extract intent. The system categorizes the extracted intent into operation categories such as parameter changes or component additions. This system adds the latent vector of the original design scheme to the classified operation vector in vector space. The added vector is then input into a decoder network to generate a new 3D shape. For example, in response to an instruction to increase the number of suction cups, the system re-executes the suction cup configuration algorithm. This algorithm recalculates the surface area and center of gravity of the object, reconfiguring the optimal coordinate group for the suction cups. In response to material change instructions, the system queries the material database to obtain the density and Young's modulus of the new material. Using these physical properties, the system re-executes stress analysis simulations and automatically recalculates the required wall thickness. The system highlights the regenerated design scheme as a difference from the original scheme. As a result, this technology automatically translates the user's abstract requirements into specific changes in physical parameters. This system eliminates the need for manual CAD model corrections, significantly increasing the speed of design iteration. As a specific application area, this step is utilized in agile product development processes where, after receiving test results from prototypes, improvement requests from the field are immediately reflected in the design data.

[0098] Step 6: Automatic Generation of 3D Model Generation Scripts. Based on the user's final selected design scheme, the 3D model generation unit on the server automatically generates 3D model generation scripts usable in 3D modeling software. For example, it outputs in open-source 3D modeling tool script formats (Python script, STEP format, STL format, OBJ format, 3MF format, IGES format, VRML format, FBX format, etc.), automatically generating a 3D model that reflects the specified shape, size, material, structure, function, special functions, durability, hygiene requirements, on-site safety standards, on-site IoT connectivity, and on-site DX propulsion. Specifically, in this step, the system scans the internal data structure of the determined design scheme. The system converts the geometric information of each component into a boundary representation model. Based on this boundary representation model, the system generates text data that follows the syntax rules of the specified output format. When outputting the script language, the system sequentially records the function calls used to generate each component. The system uses loop processing or conditional branching to optimize the generation code of regularly configured components. When outputting standard 3D data formats, the system records entity definitions conforming to international standards. This system rigorously defines the topological relationships of the faces, edges, and vertices of each component. It embeds material density and creator information as metadata into the file header. The system performs a syntax checker on the generated script code to verify its accuracy. As a result of this technology, the system converts the unique data format generated by artificial intelligence into a common industry-standard format. This system enables open data linkage that is not dependent on software from any specific vendor. Specific application areas of this step include transmitting the generated data to downstream manufacturing support systems to automatically generate machine tool toolpaths, or providing a foundational model for building digital twins of products.

[0099] Step 7: Script Output and Download. The 3D model generation script generated by the 3D model generation department is sent to the terminal via the script output department. Users can download the script and edit and utilize it in the company's design department, or send it directly to external 3D printers or processing companies. Specifically, in this step, the system saves the generated script file to the server's temporary storage. The system generates a unique download address for the file. The system sends this address to the terminal. When receiving a data acquisition request from the terminal, the system segments the file into blocks and streams it. The system deletes the file from the temporary storage after transmission to ensure security. When sending directly to an external 3D printer, the system specifies the printer's network address and establishes socket communication. The system calls the slicer software interface to remotely execute the process of converting the generated model data into mechanical control code. When sending to processing companies, the system automatically uploads the file to the company's server using an encrypted communication path. As a result of this technology, the system eliminates manual operations associated with the handover of design data, preventing data confusion or leakage due to human error. The system minimizes the time lag from design completion to manufacturing commencement. As a specific application area of ​​this step, it is utilized in a seamless supply chain management system that links with on-demand manufacturing services and automatically processes orders as soon as the design is completed.

[0100] Step 8: Historical Management and External Data Import. The Historical Management Department manages user input history, selection history, generated design schemes, 3D model generation scripts, download history, correction history, each user's preference data, on-site feedback, on-site accident data, equipment operation data, component life prediction data, anomaly detection data, on-site operation video data, on-site voice data, on-site IoT data, and on-site DX data in a unified manner. When users upload external design data, images, drawings, sketches, CAD data, 3D scan data, on-site operation video data, on-site voice data, on-site IoT data, and on-site DX data, the External Data Import Department analyzes this data, comprehensively considering the specified purpose and constraints, on-site feedback, on-site accident data, equipment operation data, component life prediction data, anomaly detection data, on-site operation video data, on-site voice data, on-site IoT data, and on-site DX data, and automatically redesigns or optimizes the data according to the specified purpose and constraints, generating new design schemes or 3D model generation scripts. Specifically, in this step, the system asynchronously writes all transaction data occurring in a series of sessions into the data lake. The system then performs a machine learning pipeline on this accumulated big data through batch processing. This system uses on-site anomaly detection data or component life prediction data as target variables and design parameters as explanatory variables to perform regression analysis. From the analysis results, the system extracts the correlations between specific shapes or materials causing failures. These extracted rules are added as penalty terms to the loss function of the generative artificial intelligence model. The system applies an iterative nearest-point algorithm to align externally uploaded 3D scan data with existing standard component models. The system calculates the differences between the scan data and the standard model, quantifying the degree of wear or deformation. To compensate for this wear data, the system automatically corrects the offset of the original design data. As a result, this system constructs a closed-loop continuous improvement cycle that feeds back actual data obtained during the operational phase to the design phase. This system elevates from a static design tool to a self-evolving engineering platform. As a specific application area, this step is utilized in a predictive maintenance-linked product lifecycle management system that automatically optimizes the design of next-generation models based on sensor data collected from operating equipment.

[0101] (specific use case) For example, suppose an automated material handling line in a food factory requires a robotic arm and end effector optimized for loading products from cardboard boxes onto pallets. The object being handled is a rectangular cardboard box, 40 cm wide, 30 cm deep, and 20 cm high, weighing 5 kg. The operating space is 1 m wide, 1 m deep, and 1.5 m high, with ceiling height restrictions and to avoid interference with adjacent equipment. The gripping method must be a suction cup that does not damage the surface of the cardboard box and allows for reliable handling. The material is aluminum-based, with food-grade silicone rubber specified for the suction cups. Special functions include suction status detection and box weight verification via a weight sensor, real-time transmission of operational data via IoT communication, suction cup wear detection and automatic replacement, a detachable and washable structure, an antibacterial coating, coordination detection with on-site operators, automatic stop function in case of abnormalities, and self-diagnostic functions for on-site maintenance support. The cost ceiling is 150,000 yen per unit, with a durability of over 200,000 cycles. Hygiene requirements include resistance to food processing plants and cleaning solutions, easy disassembly for cleaning, and prevention of foreign matter contamination. On-site safety standards require compliance with collaborative robot standards. Legal requirements include compliance with the Food Sanitation Law and the CE marking. Brand image requirements include the use of the company logo and brand colors. Specifically, this system acquires these highly multidimensional and complex design requirements not as simple text information, but as structured data that can be processed by a computer. This data acquisition department uses a large-scale language model to analyze the requirement definition text written in natural language. This large-scale language model adopts the Transformer architecture. This large-scale language model tokenizes the input text. This large-scale language model converts each token into a high-dimensional word embedding vector. This large-scale language model extracts the parameters constituting the design requirements. Through this extraction process, dimensional requirements are represented as three-dimensional numerical vectors. Weight requirements are represented as scalar values. Material requirements are represented as index values ​​from a material database. Special function requirements are represented as arrays of binary flags indicating the presence or absence of each function. This system integrates these extracted parameters. The integrated parameters are then expanded into memory as a design requirement tensor. This design requirement tensor serves as input data for subsequent generative AI modules. This processing method fundamentally differs from the existing mental process by which human designers interpret specifications. This method eliminates the ambiguity of natural language. This method transforms design requirements into a rigorous mathematical expression. Therefore, in the subsequent design scheme generation process, this system can achieve high-speed and comprehensive exploration by a computer. As a result of this technology, errors in interpreting design requirements are reduced. As a result of this technology, the accuracy of data processing in the early stages of the design process is significantly improved. As for specific application areas, this method is not limited to the design of robot end effectors for food factories.This method can be applied to the design of automated assembly lines for automotive parts. It can also be applied to the design of wafer handling mechanisms in semiconductor manufacturing equipment.

[0102] The user inputs all the above requirements in detail in the user input section of the terminal. As an example, the input specifies: "Object: A rectangular cardboard box, dimensions 40cm × 30cm × 20cm, weight 5kg; Working space: 1m wide × 1m deep × 1.5m high; Holding method: Suction cup; Material: Aluminum and food-grade silicone rubber; Special functions: Adsorption status detection, weight sensor, IoT communication, suction cup wear detection, automatic replacement, removable cleaning, antibacterial coating, coordinated motion detection, automatic stop in case of abnormality, self-diagnosis; Cost limit: 150,000 yen; Durability: 200,000 cycles or more; Hygiene requirements: Suitable for food factories, resistant to cleaning solutions, removable for cleaning, prevention of foreign matter contamination; On-site safety standards: Compliant with collaborative robot standards; Laws and regulations: Compliant with the Food Sanitation Law, CE marking; Brand image: Corporate logo engraving, brand colors used." Specifically, the user input section of this terminal preprocesses the input text data in real time. This preprocessing section removes unwanted symbols or spaces from the input text. This preprocessing section standardizes the text. This preprocessing unit segments the standardized text into characters or words. The preprocessing unit stores the segmented data as a one-dimensional array in a buffer. This terminal sends this one-dimensional array to the server via a communication interface. During this transmission process, this terminal encrypts the data. This terminal uses a secure communication protocol to ensure the security of data transmission. The receiving unit of this server receives the encrypted data. This receiving unit decrypts the data. This server inputs the decrypted text data into the natural language processing engine. This natural language processing engine has a pre-learned neural network. This neural network extracts design parameters from the input text data. For example, the input to this neural network is a 512-dimensional vector column after the text string of the above input example is tokenized. The output from this neural network is structured data consisting of key-value pairs corresponding to each design item. As a specific example of the output, for the size key, the output is a numerical array with a width of 40, a depth of 30, and a height of 20. For the weight key, the output is a value of 5. For the material key, the output is an identifier representing aluminum and silicone rubber. This server temporarily saves this structured data to a database. This server utilizes this structured data as the initial condition for subsequent 3D model generation. This processing method automatically converts the user's free-form input into a strict data format that can be directly processed by the computer. This processing method does not replace existing manual data input tasks performed by humans. It automates the structuring process of the input data itself, fundamentally improving data processing efficiency. As a result of this technology, the parsing and processing of input data is accelerated. As a result of this technology, errors in the data handover to subsequent design algorithms are eliminated. In specific application areas, this input interface can be used in industrial robot design systems. This input interface can also be used in custom medical device requirement definition systems.

[0103] The input information is sent to the server via the data transmission unit. The generative AI processing unit comprehensively considers factors such as the number and configuration of suction cups, the length and range of motion of the robotic arm, the structure of the joints, the material and thickness of the suction cups, the selection of the vacuum generating mechanism, the configuration of the weight sensor, the assembly of the IoT communication module, the setting of the wear detection sensor, the design of the detachable cleaning structure, the application of the antibacterial coating, the configuration of sensors for coordinated motion detection, the design of the automatic stop mechanism in case of abnormalities, the assembly of the self-diagnostic algorithm, cost and durability, hygiene, on-site safety standards, laws and regulations, and brand image, and automatically generates multiple design schemes. For example, there are schemes with 4 suction cups, schemes with 6 suction cups, schemes with a telescopic robotic arm, schemes with enhanced automatic suction cup changing mechanisms, schemes with expanded IoT communication functions, schemes prioritizing detachable cleaning, schemes with expanded antibacterial coating coverage, and schemes that improve the accuracy of coordinated motion detection with on-site operators, among other variations. Specifically, the generative AI processing unit of this server has an integrated architecture that combines multiple deep learning models. This generative AI processing unit receives structured input data as a design requirement tensor. It uses this design requirement tensor as a conditional vector. The unit then uses a conditional generative adversarial network (GAN) to generate a potential representation of the 3D shape. The input to this generative network is a 384-dimensional tensor, formed by combining a 128-dimensional normal random number vector with a 256-dimensional conditional vector. The network has multiple fully connected layers and 3D deconvolution layers. It generates 3D voxel data or point cloud data from the input tensor. The output is, for example, a 3D voxel mesh with a resolution of 64×64×64. Each voxel has a probability value representing the presence or absence of material or its type. The generated 3D shape is evaluated using a physics simulation engine. This evaluation process uses the finite element method to calculate stress distribution or deformation. It also uses fluid analysis to calculate the suction force of the vacuum suction cup. The generative AI processing unit uses these physical evaluation results as reward signals. This generative AI processing unit uses a reinforcement learning algorithm to update the weight parameters of the generative network. This reinforcement learning algorithm employs a proximal policy optimization technique. The generative AI processing unit uses a cost prediction model to calculate the manufacturing cost of each design scheme. This cost prediction model is based on a gradient boosting decision tree algorithm learned using past manufacturing data. Based on these multiple evaluation metrics, the generative AI processing unit extracts multiple design schemes on the Pareto optimal boundary. Through this extraction process, various variations of suction cup configurations or robotic arm structures are generated. This processing method discovers optimal solutions that are difficult for human intuition to reach through mathematical exploration in a high-dimensional design space. This processing method replaces the existing trial-and-error design process with a parallel and comprehensive optimization process performed by a computer. As a result of this technology, the speed of design scheme generation is dramatically improved.As a result of this technology, high-quality design solutions that simultaneously meet both physical and cost constraints are consistently obtained. In specific application areas, this generative process can be applied to the design of robot end effectors. It can also be applied to the topology optimization design of lightweight aircraft components.

[0104] Users compare various design schemes on the terminal, considering 3D previews, dimensions, materials, functions, cost estimates, durability assessments, hygiene assessments, assembly and disassembly animations, component lists, on-site simulation videos, IoT linkage assessments, availability of on-site maintenance support functions, and brand image compatibility, to select the best design scheme. Furthermore, users can input additional requests or correction instructions such as "increase the number of suction cups," "extend the length of the robotic arm by 10 cm," "enhance IoT communication functions," "further improve detachability and cleanability," and "change the position of the brand logo," triggering automatic generation of a new design scheme by generative AI. Specifically, the terminal's display control unit renders multiple design scheme data received from the server onto a graphical user interface. This display control unit uses web-based 3D graphics rendering technology to rapidly render 3D models on a browser. Based on user input, the display control unit performs real-time rotation, zooming, and scaling of the 3D model. The terminal visualizes the evaluation indicators of each design scheme as radar charts or bar graphs. The terminal's input receiving unit receives additional requests or correction instructions from the user as natural language text or direct parameter changes. The terminal converts the received correction instructions into difference vectors. This terminal sends the difference vector to the server. The generative AI processing unit of this server uses the received difference vector to update the conditional vector. This generative AI processing unit re-inputs the updated conditional vector into the generative network. In this regeneration process, this generative AI processing unit uses the latent performance of the previous generation result as an initial value. By setting this initial value, this generative AI processing unit performs local shape changes according to the correction instructions while maintaining the basic structure of the design scheme selected by the user. For example, when an instruction to increase the number of suction cups is input, this generative AI processing unit recalculates the voxel configuration probability of the bottom surface region of the end effector. This generative AI processing unit reconfigures the optimal suction cup configuration coordinates while avoiding interference. This processing method directly maps the user's intention as a mathematical constraint into the latent space of the generative model. This processing method realizes a dialogic optimization process between humans and computers. As a result of this technology, the convergence speed of the design scheme required by the user is improved. As a result of this technology, the computational cost required for fine-tuning the design is reduced. As a specific application area, this dialogic design interface can be applied to the custom design of industrial equipment. This conversational design interface can also be applied to online ordering systems for personalized products aimed at consumers.

[0105] Regarding the final selected design scheme, the 3D model generation department automatically generates a 3D model generation script (Python script, STEP format, STL format, etc.) that can be used in 3D modeling software. Users can download the script and generate a 3D model immediately. As a specific example of the prompt text that the generative AI can read, it is: "Object: A rectangular cardboard box, dimensions 40 cm × 30 cm × 20 cm, weight 5 kg, working space: 1 m wide × 1 m deep × 1.5 m high, gripping method: suction cup, material: aluminum and food-grade silicone rubber, special functions: adsorption status detection, weight sensor, IoT communication, suction cup wear detection, automatic replacement, removable cleaning, antibacterial coating, coordinated motion detection, automatic stop in case of abnormality, self-diagnosis, cost limit: 150,000 yen, durability: more than 200,000 cycles, hygiene requirements: food factory compliance, cleaning fluid resistance, disassembly and cleaning capability, prevention of foreign matter contamination, on-site safety standards: compliance with collaborative robot standards, laws and regulations: compliance with the Food Sanitation Law and CE marking, brand image: corporate logo engraving and brand colors adopted." Specifically, the 3D model generation unit of this server converts the voxel data or point cloud data of the determined design scheme into a boundary representation model. In this conversion process, the 3D model generation unit uses the moving cube method to extract isosurfaces. The 3D model generation unit smooths the vertex coordinates of the extracted polygonal mesh. The 3D model generation unit fits the smoothed mesh data onto surface patches of analytical geometry. Based on the fitted surface data, the 3D model generation unit constructs a solid model. The 3D model generation unit extracts the features of the constructed solid model. The 3D model generation unit constructs the history of the extracted features as an abstract syntax tree in memory. The 3D model generation unit scans this abstract syntax tree. Based on the scan results, the 3D model generation unit generates program code for calling the application programming interface (API) of specific 3D CAD software. In this program code generation, a Transformer-based language model specialized for code generation is used. The input to this language model is the node information of the abstract syntax tree and the specification data of the CAD software as the object. The output from this language model is executable scripting language text data. As a concrete example of output, this includes a code block containing function calls to generate a cylinder at specific coordinates and apply fillets with a specified radius. The server compresses the generated script file or a file in a standard 3D data exchange format. The server provides the compressed file to the terminal in a downloadable format. This processing method is not limited to simply outputting shape data, but automatically generates parametric CAD models that retain the design history. This processing method facilitates design changes in subsequent processes. As a result of this technology, the conversion process from design data to manufacturing data is fully automated. As a result of this technology, data compatibility issues between different CAD systems are eliminated. As a specific application area, this script generation technology can be applied to the automation of mold design.This script generation technology can also be applied to the automatic generation of building information model data for buildings.

[0106] Thus, by using the system of this invention, even users without specialized knowledge can design and prototype factory robot arms or end effectors that meet the detailed requirements of various sites in a short period of time and with high precision. It can flexibly and efficiently address various on-site needs such as design verification, simulation before on-site implementation, prototype production before mass production, construction of on-site maintenance systems, on-site IoT linkage, and DX advancement. Specifically, this system utilizes a distributed processing architecture built on a cloud computing environment. By using a parallel computing cluster of multiple image processing semiconductors, the system performs real-time exploration and physical simulation of a vast design space. The system tightly integrates the generative AI module and the physical simulation module. Through this tight integration, the system forms a closed-loop control system that iteratively performs shape generation and performance evaluation. This closed-loop control system solves an optimization problem that maximizes the fitness of the design scheme as the objective function. The system accumulates the design data obtained in each iterative step into a time-series database. The system utilizes the accumulated data as a dataset for relearning the next generation of generative models. This relearning process uses a self-supervised learning method. Through this relearning process, the system's design generation capabilities are continuously improved. This approach elevates existing design processes, which rely on human experience and intuition, into a data-driven, computer-dominated, autonomous engineering process. It combines specific rule-based processes with highly information-processing techniques that incorporate unconventional steps. As a result, the lead time from design to manufacturing is drastically reduced. Furthermore, it enables the quantitative prediction and mitigation of manufacturing defects or performance failures in the early stages of design. Specifically, this integrated design system can serve as an indispensable foundational technology for building smart factories that enable high-variety, low-volume production. It can also be used as part of a platform for digitizing the entire supply chain.

[0107] The specific processing unit 290 sends the result of the specific processing obtained in Embodiment 1 or Application Example 1 described above to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires voice representing the user's input on the result of the specific processing. The control unit 46A sends the voice data representing the user's input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0108] Data generation model 58 is what is known as generative AI (Artificial Intelligence).

[0109] One example of data generation model 58 includes ChatGPT (registered trademark) (Internet search).<URL: https: / / openai.com / blog / chatgpt> Generative AI, etc.

[0110] The data generation model 58 is obtained by performing deep learning on a neural network. A prompt containing instructions is input to the data generation model 58, along with inference data such as speech data representing speech, text data representing text, and image data representing images (e.g., data from still images or moving images). The data generation model 58 infers the input inference data based on the instructions shown in the prompt and outputs the inference result in one or more data forms, such as speech data, text data, and image data. The data generation model 58 may include, for example, text generation AI, image generation AI, and multimodal generation AI. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization. The specific processing unit 290 performs the aforementioned specific processing while using the data generation model 58. The data generation model 58 may also be a fine-tuned model to output inference results from prompts without instructions; in this case, the data generation model 58 can output inference results from prompts without instructions. The data processing device 12, etc., includes multiple data generation models 58, including AI other than generative AI. AI other than generative AI includes, but is not limited to, linear regression, logistic regression, decision trees, random forests, support vector machines (SVM), k-means clustering, convolutional neural networks (CNN), recurrent neural networks (RNN), generative adversarial networks (GAN), or Naive Bayes. Furthermore, AI can also act as an AI agent. Additionally, when the processing described above is performed by AI, this processing may be partially or entirely performed by AI, but is not limited to this example. Moreover, processing performed by AI, including generative AI, can be replaced by rule-based processing, and vice versa.

[0111] Furthermore, the processing performed by the aforementioned data processing system 10 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart device 14, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart device 14. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart device 14 or external devices, and the smart device 14 acquires or collects information required for processing from the data processing device 12 or external devices.

[0112] In the above embodiments, examples of specific processing performed by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the smart device 14.

[0113] [Second Implementation] Figure 3 An example of the configuration of the data processing system 210 according to the second embodiment is shown.

[0114] like Figure 3 As shown, the data processing system 210 includes a data processing device 12 and smart glasses 214. One example of the data processing device 12 is a server.

[0115] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0116] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, and communication I / F 44 are also connected to the bus 52.

[0117] Microphone 238 receives instructions from user 20 by receiving the sounds emitted by user 20. Microphone 238 captures the sounds emitted by user 20, converts the captured sounds into speech data, and outputs it to processor 46. Speaker 240 outputs sound according to the instructions from processor 46.

[0118] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, used to capture the area around user 20 (e.g., the shooting range defined by an angle of view equivalent to the field of vision of an average healthy person).

[0119] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F 44 and 26 is performed in a secure state.

[0120] Figure 4 An example of the main functions of the data processing device 12 and the smart glasses 214 is shown. Figure 4 As shown, in the data processing device 12, specific processing is performed by the processor 28. The memory 32 stores a specific processing program 56.

[0121] The specific processing program 56 is an example of a "program" as disclosed herein. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0122] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290. The specific processing unit 290 can use the emotion-specific model 59 to infer the user's emotions and perform specific processing based on the user's emotions. The emotion inference function using the emotion-specific model 59 (emotion-specific function) includes various inferences and predictions related to the user's emotions, but is not limited to this example. Furthermore, emotion inference and prediction also include, for example, emotion analysis (analysis).

[0123] In the smart glasses 214, the processor 46 performs the receive / output processing. The memory 50 stores the receive / output program 60. The processor 46 reads the receive / output program 60 from the memory 50 and executes the read receive / output program 60 on the RAM 48. The receive / output processing is implemented by the processor 46 operating as a control unit 46A based on the receive / output program 60 executed on the RAM 48. Furthermore, the smart glasses 214 may also have the same data generation model and emotion-specific model as the data generation model 58 and the emotion-specific model 59, and use these models to perform the same processing as the specific processing unit 290.

[0124] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. Each part of the system described below is implemented by the data processing device 12 and the smart glasses 214. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal".

[0125] (Example 1) Since the specific processing flow is the same as that in Embodiment 1 described in the first embodiment above, the description is omitted.

[0126] (Application Example 1) Since the specific processing flow is the same as that in Application Example 1 described in the first embodiment above, the description is omitted.

[0127] The specific processing unit 290 sends the result of the specific processing obtained in Embodiment 1 or Application Example 1 described above to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires voice representing the user's input on the specific processing result. The control unit 46A sends the voice data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0128] Furthermore, the processing performed by the aforementioned data processing system 10 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the smart glasses 214, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the smart glasses 214. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the smart glasses 214 or external devices, and the smart glasses 214 acquires or collects information required for processing from the data processing device 12 or external devices.

[0129] In the above embodiments, examples of specific processing performed by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the smart glasses 214.

[0130] [Third Implementation] Figure 5 An example of the configuration of the data processing system 310 according to the third embodiment is shown.

[0131] like Figure 5 As shown, the data processing system 310 includes a data processing device 12 and a head-mounted terminal 314. An example of the data processing device 12 is a server.

[0132] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0133] The head-mounted terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, display 343, and communication I / F 44 are also connected to the bus 52.

[0134] Microphone 238 receives instructions from user 20 by receiving the sounds emitted by user 20. Microphone 238 captures the sounds emitted by user 20, converts the captured sounds into speech data, and outputs it to processor 46. Speaker 240 outputs sound according to the instructions from processor 46.

[0135] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, used to capture the area around user 20 (e.g., the shooting range defined by an angle of view equivalent to the field of vision of an average healthy person).

[0136] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F 44 and 26 is performed in a secure state.

[0137] Figure 6 An example of the main functions of the data processing device 12 and the head-mounted terminal 314 is shown. Figure 6 As shown, in the data processing device 12, specific processing is performed by the processor 28. The memory 32 stores a specific processing program 56.

[0138] The specific processing program 56 is an example of a "program" as disclosed herein. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0139] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the special processing unit 290.

[0140] In the head-mounted terminal 314, the processor 46 performs the receiving and output processing.

[0141] The memory 50 stores a receive / output program 60. The processor 46 reads the receive / output program 60 from the memory 50 and executes the read receive / output program 60 on the RAM 48. The receive / output processing is implemented by the processor 46 operating the control unit 46A based on the receive / output program 60 executed on the RAM 48.

[0142] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. Each part of the system described below is implemented by the data processing device 12 and the head-mounted terminal 314. In the following description, the data processing device 12 will be referred to as the "server" and the head-mounted terminal 314 will be referred to as the "terminal".

[0143] (Example 1) Since the specific processing flow is the same as that in Example 1 described in the first embodiment above, the description is omitted.

[0144] (Application Example 1) Since the specific processing flow is the same as that in Application Example 1 described in the first embodiment above, the description is omitted.

[0145] The specific processing unit 290 sends the result of the specific processing obtained in Embodiment 1 or Application Example 1 described above to the head-mounted terminal 314. In the head-mounted terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires voice representing the user's input on the specific processing result. The control unit 46A sends the voice data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0146] Furthermore, the processing performed by the aforementioned data processing system 10 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the head-mounted terminal 314, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the head-mounted terminal 314. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the head-mounted terminal 314 or external devices, and the head-mounted terminal 314 acquires or collects information required for processing from the data processing device 12 or external devices.

[0147] In the above embodiments, examples of specific processing performed by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the head-mounted terminal 314.

[0148] [Fourth Implementation] Figure 7 An example of the configuration of the data processing system 410 according to the fourth embodiment is shown.

[0149] like Figure 7 As shown, the data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[0150] The data processing apparatus 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" as understood in this disclosure. The computer 22 includes a processor 28, RAM 30, and memory 32. The processor 28, RAM 30, and memory 32 are connected to a bus 34. Furthermore, the database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. An example of the network 54 includes a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0151] Robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control object 443. Computer 36 includes a processor 46, RAM 48, and memory 50. The processor 46, RAM 48, and memory 50 are connected to a bus 52. Furthermore, the microphone 238, speaker 240, camera 42, control object 443, and communication I / F 44 are also connected to the bus 52.

[0152] Microphone 238 receives instructions from user 20 by receiving the sounds emitted by user 20. Microphone 238 captures the sounds emitted by user 20, converts the captured sounds into speech data, and outputs it to processor 46. Speaker 240 outputs sound according to the instructions from processor 46.

[0153] Camera 42 is a small digital camera equipped with an optical system such as a lens, aperture and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, used to capture the area around user 20 (e.g., the shooting range defined by an angle of view equivalent to the field of vision of an average healthy person).

[0154] Communication I / F 44 is connected to network 54. Communication I / F 44 and 26 are responsible for the transmission and reception of various information between processor 46 and processor 28 via network 54. The transmission and reception of various information between processor 46 and processor 28 using communication I / F 44 and 26 is performed in a secure state.

[0155] The controlled object 443 includes a display device, LEDs for the eyes, and motors for driving the arms, hands, and feet. The posture and movements of the robot 414 are controlled by controlling the motors for the arms, hands, and feet. Some of the robot 414's emotions can be expressed by controlling these motors. In addition, facial expressions of the robot 414 can also be expressed by controlling the illumination state of the LEDs for the robot 414's eyes.

[0156] Figure 8 An example of the main functions of the data processing device 12 and the robot 414 is shown. For example... Figure 8 As shown, in the data processing device 12, specific processing is performed by the processor 28. The memory 32 stores a specific processing program 56.

[0157] The specific processing program 56 is an example of a "program" as disclosed herein. The processor 28 reads the specific processing program 56 from the memory 32 and executes the read specific processing program 56 on the RAM 30. Specific processing is implemented by the processor 28 operating as a specific processing unit 290 according to the specific processing program 56 executed on the RAM 30.

[0158] The memory 32 stores a data generation model 58 and an emotion-specific model 59. The data generation model 58 and the emotion-specific model 59 are used by the specific processing unit 290.

[0159] In robot 414, the processor 46 performs the receive / output processing. The memory 50 stores the receive / output program 60. The processor 46 reads the receive / output program 60 from the memory 50 and executes the read receive / output program 60 on RAM 48. The receive / output processing is implemented by the processor 46 operating the control unit 46A based on the receive / output program 60 executed on RAM 48.

[0160] Next, the specific processing performed by the specific processing unit 290 of the data processing device 12 will be described. Each part of the system described below is implemented by the data processing device 12 and the robot 414. In the following description, the data processing device 12 will be referred to as the "server" and the robot 414 will be referred to as the "terminal".

[0161] (Example 1) Since the specific processing flow is the same as that in Example 1 described in the first embodiment above, the description is omitted.

[0162] (Application Example 1) Since the specific processing flow is the same as that in Application Example 1 described in the first embodiment above, the description is omitted.

[0163] The specific processing unit 290 sends the result of the specific processing obtained in Embodiment 1 or Application Example 1 described above to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the controlled object 443 to output the result of the specific processing. The microphone 238 acquires voice representing the user's input on the result of the specific processing. The control unit 46A sends the voice data representing the user's input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[0164] Furthermore, the processing performed by the data processing system 10 is executed by the specific processing unit 290 of the data processing device 12 or the control unit 46A of the robot 414, but it can also be executed jointly by the specific processing unit 290 of the data processing device 12 and the control unit 46A of the robot 414. Additionally, the specific processing unit 290 of the data processing device 12 acquires or collects information required for processing from the robot 414 or external devices, and the robot 414 acquires or collects information required for processing from the data processing device 12 or external devices.

[0165] In the above embodiments, examples of specific processing performed by the data processing device 12 are given, but the technology disclosed herein is not limited to this, and specific processing may also be performed by the robot 414.

[0166] Furthermore, the emotion-specific model 59, serving as an emotion engine, can determine the user's emotion based on a specific mapping. Specifically, the emotion-specific model 59 can determine the user's emotion based on an emotion graph that serves as a specific mapping (see...). Figure 9 The robot's emotions can be determined by the emotion-specific model 59. In addition, the emotion-specific model 59 can also determine the robot's emotions in the same way, and the specific processing unit 290 can also perform specific processing using the robot's emotions.

[0167] Figure 9This is a diagram representing an emotion map 400 that maps various emotions. In the emotion map 400, emotions are arranged radially from the center in concentric circles. The closer to the center of the concentric circles, the more primitive the emotion is. Further out on the concentric circles, emotions are arranged representing states or actions arising from mood. Emotion is a concept that includes both feelings and mental states. To the left of the concentric circles, emotions generated by reactions occurring in the brain are arranged roughly. To the right of the concentric circles, emotions guided by situational judgments are arranged roughly. Above and below the concentric circles, emotions generated by reactions occurring in the brain and guided by situational judgments are arranged roughly. Furthermore, the emotion of "pleasure" is arranged above the concentric circles, and the emotion of "unpleasantness" is arranged below. Thus, in the emotion map 400, various emotions are mapped according to the structure of emotion generation, while easily generated emotions are mapped nearby.

[0168] These emotions are distributed at the 3 o'clock position on the Emotion Chart 400, and usually fluctuate between peace and unease. In the right half of the Emotion Chart 400, because situational awareness is more dominant than internal feelings, it gives a sense of calm.

[0169] The inner side of the emotion diagram 400 represents the mind, and the outer side of the emotion diagram 400 represents actions. Therefore, the further you go to the outer side of the emotion diagram 400, the more the emotion can be seen (manifested in actions).

[0170] Here, human emotions are based on a balance of various factors such as posture and blood sugar levels. When these balances deviate from the ideal, it indicates unhappiness; when they approach the ideal, it indicates pleasure. In robots, cars, and motorcycles, emotions can also be created based on a balance of factors such as posture and remaining battery power. When these balances deviate from the ideal, it indicates unhappiness; when they approach the ideal, it indicates pleasure. Emotion maps can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Speech Emotion Recognition and Brain Physiological Signal Analysis Systems for Emotion, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). In the left half of the emotion map, emotions belonging to the "Reaction" domain, where sensation is dominant, are arranged. Furthermore, in the right half of the emotion map, emotions belonging to the "Situation" domain, where situational cognition is dominant, are arranged.

[0171] The emotion map defines two types of emotions that promote learning. One is a negative emotion located near the middle of "repentance" or "reflection" on the situation side. That is, when the robot experiences negative emotions such as "I never want to feel this way again" or "I never want to be scolded again." The other is a positive emotion located near "desire" on the response side. That is, when the robot experiences positive feelings such as "wanting more" or "wanting to know more."

[0172] The emotion-specific model 59 feeds user input into a pre-learned neural network to obtain emotion values ​​representing each emotion shown in the emotion graph 400, and determines the user's emotion. This neural network is pre-learned based on multiple learning data sets that combine user input with emotion values ​​representing each emotion shown in the emotion graph 400. Furthermore, this neural network is learned to... Figure 10 As shown in sentiment graph 900, nearby sentiment values ​​are similar to each other. Figure 10 Examples show how emotions such as "peace of mind," "stability," and "reassurance" can result in similar emotional values.

[0173] The above description primarily focuses on the functions of the data processing device 12 to illustrate the system disclosed herein. However, the system disclosed herein may not necessarily be implemented on a server. The system disclosed herein can be implemented as a general information processing system. For example, it can be implemented as a software program running on a personal computer, an application running on a smartphone, etc. The methods disclosed herein can be provided to users in the form of SaaS (Software as a Service).

[0174] In the above embodiments, an example of specific processing performed by a single computer 22 is given. However, the technology disclosed herein is not limited to this, and distributed processing of specific processes by multiple computers, including computer 22, is also possible. For example, the data generation model 58 may also be provided in an external device of the data processing apparatus 12, where data is generated based on input data.

[0175] In the above embodiments, an example of storing a specific processing program 56 in memory 32 is illustrated, but the technology disclosed herein is not limited thereto. For example, the specific processing program 56 may also be stored in a portable computer-readable non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 performs specific processing according to the specific processing program 56.

[0176] Alternatively, the specific processing program 56 can be stored in a storage device such as a server connected to the data processing device 12 via a network 54, and the specific processing program 56 can be downloaded and installed into the computer 22 upon request from the data processing device 12.

[0177] Furthermore, it is not necessary to store the entire specific process 56 in a storage device such as a server connected to the data processing device 12 via the network 54, nor is it necessary to store the entire specific process 56 in the memory 32; a portion of the specific process 56 may also be stored.

[0178] As a hardware resource for performing specific processing, various processors can be used. For example, a CPU can be listed as a processor; it is a general-purpose processor that functions as a hardware resource for performing specific processing by executing software, i.e., a program. Furthermore, as a processor, a dedicated circuit can be listed as a processor; it is a processor with a circuit structure specifically designed for performing specific processing, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit).

[0179] Every processor has built-in or connected memory, and every processor performs specific processes by using memory.

[0180] The hardware resources for performing a specific process can consist of one of these various processors, or a combination of two or more processors of the same or different types (e.g., a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Furthermore, the hardware resources for performing a specific process can also be a single processor.

[0181] As an example of a single processor, the first type consists of a combination of one or more CPUs and software, which functions as a hardware resource to perform specific processing. The second type uses a processor, such as a System-on-a-chip (SoC), which implements the entire system functionality, including multiple hardware resources performing specific processing, using a single IC chip. In this case, the specific processing is implemented using one or more of the aforementioned processors that serve as hardware resources.

[0182] Furthermore, as the hardware architecture of these various processors, more specifically, circuits combining semiconductor elements and other circuit components can be used. Moreover, the specific process described above is merely an example. Therefore, it goes without saying that, without departing from the main point, unnecessary steps can be removed, new steps can be added, or the processing order can be changed.

[0183] The foregoing descriptions and illustrations are detailed explanations of the parts covered by this disclosure and are merely one example of this disclosure. For instance, the descriptions of the above-described structure, function, role, and effect are just one example of the structure, function, role, and effect of the parts covered by this disclosure. Therefore, it goes without saying that, without departing from the spirit of this disclosure, unnecessary parts can be deleted, new elements can be added, or replacements can be made to the foregoing descriptions and illustrations. Furthermore, to avoid confusion and facilitate understanding of the parts covered by this disclosure, explanations of technical common sense that does not require special explanation for implementing this disclosure have been omitted from the foregoing descriptions and illustrations.

[0184] All documents, patent applications and technical standards described in this specification are incorporated herein by reference as if they were specifically and individually described as incorporated by reference.

[0185] In addition, the following notes further explain the above explanation.

[0186] (Note 1) A system includes a user input unit, a generative AI processing unit, and a 3D model generation unit. The user input unit receives multiple requirements from the user for the design of a factory robot's robotic arm or end effector, including the shape, size, weight, working space dimensions and constraints, gripping method, desired materials, required special functions, cost limits, durability, and hygiene requirements of the object being handled. The generative AI processing unit receives the requirements from the user input unit as input data and, based on these requirements, comprehensively considers the object being handled, working space, gripping method, materials, special functions, cost, durability, and hygiene requirements, automatically generates multiple optimal robotic arm or end effector design schemes, and assigns each design scheme characteristics reflecting the gripping method, special functions, materials, cost, durability, and hygiene requirements. The 3D model generation unit, based on the design schemes generated by the generative AI processing unit and selected by the user, automatically outputs a 3D model generation script usable in 3D modeling software. This script automatically generates a 3D model reflecting the specified shape, size, materials, structure, function, special functions, durability, and hygiene requirements.

[0187] (Note 2) As described in Appendix 1, the system is characterized in that the generative AI processing unit automatically designs the number and configuration of suction cups, the length and range of motion of the robotic arm, the structure of the joints, the thickness of the materials, the configuration of the sensors, surface treatment, hygiene measures, and design patterns that take into account cost and durability, based on the shape and weight of the object being transported, constraints of the workspace, gripping methods, special functions, cost, durability, and hygiene requirements. The three-dimensional model generation unit outputs a script for generating a three-dimensional model that reflects these detailed designs.

[0188] (Note 3) As described in Appendix 1, the system is characterized in that, after the user selects a design scheme, the three-dimensional model generation unit combines multiple basic shapes and components to construct a three-dimensional model based on the object being transported and the workspace, the handling method, materials, special functions, cost, durability, hygiene requirements, etc., and automatically generates a script for generating a three-dimensional model that reflects the feature assignment processing performed by generative AI.

Claims

1. A system comprising a user input unit, a generative AI processing unit, and a 3D model generation unit, characterized in that, The user input unit has the function of receiving design requirements from the user, including the product's purpose, size, material, color, and other constraints; the generative AI processing unit has the following functions: receiving the design requirements received from the user input unit as input data, automatically generating multiple product design schemes based on the design requirements, and assigning each design scheme features that reflect the shape optimized for the purpose, the specified material and color, and other constraints. The three-dimensional model generation unit automatically outputs a three-dimensional model generation script that can be used in the three-dimensional modeling software based on the design scheme selected by the user in the design scheme generated by the generative AI processing unit. This script automatically generates a three-dimensional model that reflects the specified size, color and material information.

2. The system as described in claim 1, characterized in that, The generative AI processing unit automatically designs additional components such as drawers and handles according to the intended use and constraints, and the 3D model generation unit outputs a script for generating a 3D model containing these additional components.

3. The system as described in claim 1, characterized in that, After the user selects a design scheme, the 3D model generation unit combines multiple basic shapes to construct a 3D model based on its purpose, size, material, color, and other constraints. It also automatically generates a 3D model generation script that reflects the feature assignment processing performed by generative AI.

Images