An AI-generated scene generation medium that can be parsed by a digital twin engine and system

Abstract

The application relates to the technical field of digital twinning and artificial intelligence, and specifically provides a scene generation medium generated by AI and capable of being parsed by a digital twinning engine and a system. The medium is a structured data file automatically generated by an AI system according to a received natural language description or logical requirement. The core content of the medium comprises a scene element identification part, a spatial relationship description part and an attribute and behavior rule part. The application also provides a digital twinning scene generation system comprising an AI system for generating the above medium and a digital twinning engine. The AI system can convert user requirements into a structured scene description. The digital twinning engine is responsible for parsing the medium, automatically retrieving and assembling three-dimensional models from a model library and generating an interactive digital twinning scene according to the rules defined in the medium. The application realizes end-to-end automatic construction from high-level requirements to three-dimensional scenes, significantly reduces the development threshold and improves construction efficiency and flexibility.

Description

Technical Field

[0001] This invention relates to the fields of digital twin and artificial intelligence technology, and more specifically, to an intermediate medium for automatically constructing a three-dimensional digital twin scene, an AI system for generating the medium, a complete scene generation system including the system, and a method thereof. Background Technology

[0002] Digital twin technology, by constructing virtual mappings of physical entities, is widely used in smart cities, industrial manufacturing, equipment maintenance, and other fields. Currently, the construction of digital twin 3D scenes heavily relies on manual labor. Professionals need to manually create models, set positions, configure attributes and interaction logic using specialized 3D modeling software. The entire process is time-consuming, labor-intensive, has high barriers to entry, is inefficient, and struggles to respond to rapidly changing demands.

[0003] In recent years, some technologies have attempted to leverage artificial intelligence to assist in the generation of 3D content, such as generating individual 3D models from text descriptions. However, these technologies primarily focus on generating isolated model assets, rather than a complete, functional digital twin scene containing multiple objects, spatial relationships, and interactive logic. How to seamlessly integrate AI's understanding and generation capabilities with the rendering and driving capabilities of a digital twin engine to achieve end-to-end automatic generation from high-level intent description to an interactive scene remains a pressing technical problem. Existing solutions lack a structured information carrier (i.e., medium) that can effectively connect AI intent understanding with engine execution. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to overcome the shortcomings of the existing digital twin scene construction process, such as low automation, reliance on manual labor, and low efficiency, and to provide an AI generation medium and system that can automatically convert natural language or logical requirements into a workable 3D scene.

[0005] To achieve the above objectives, this invention provides a scene generation medium generated by AI and parsable by a digital twin engine. This medium is a structured data file automatically generated by the AI ​​system based on a natural language description or logical requirements of the target scene. The medium includes at least: Scene element identification section, used to define one or more 3D model elements that constitute a digital twin scene in a machine-readable format; The spatial relationship description section is used to describe the relative position, scale, or topological relationship between the multiple 3D model elements; The Attributes and Behaviors section is used to define the static attribute parameters and / or dynamic behavior logic of at least one 3D model feature.

[0006] The data structure and semantic format of the medium are predefined, enabling the digital twin engine to parse the medium and automatically call and assemble the corresponding 3D model resources based on the parsing results to generate an interactive digital twin scene.

[0007] Preferably, the medium is a text file in JSON, XML, or YAML format.

[0008] Preferably, the AI ​​system is a trained large language model configured to understand the semantics of unstructured text instructions input by the user and output the scene generation medium that conforms to predetermined data structure requirements.

[0009] The present invention also provides an AI system for generating the above-mentioned scene generation medium, comprising: The requirement understanding module is configured to receive and parse the user's input of scenario construction requirements, and extract the elements, relationships and rule information of scenario construction; The media generation module is configured to understand the information output by the requirement understanding module, generate the media data structure according to the predefined scenario, and assemble and generate the corresponding structured data file. The output module is configured to output the generated scene generation medium.

[0010] The present invention further provides a digital twin scene generation system, comprising: The AI ​​system described above is used to generate scene generation media; The digital twin engine, which communicates with the AI ​​system, is configured as follows: Receive and parse the scene generation medium; Based on the analysis results, the corresponding 3D model is retrieved from the associated 3D model library and instantiated. Based on the description in the medium, the instantiated 3D model is laid out and configured in virtual space to generate a runnable digital twin scene.

[0011] Furthermore, this invention provides a method for generating a digital twin scene, comprising the following steps: S1: Receive and understand scenario building requirements through an AI system; S2: Based on the understanding of the requirements, the AI ​​system automatically generates a structured scene generation medium; S3: Provide the medium to the digital twin engine; S4: The digital twin engine parses the medium and automatically calls up the 3D model resources to generate the corresponding digital twin scene.

[0012] The beneficial effects of this invention are as follows: by introducing a structured scene generation medium automatically generated by AI as a "bridge" connecting high-level intents and low-level rendering execution, a revolutionary automation of the digital twin scene construction process is achieved. Users only need to input natural language descriptions, and the system can automatically generate a "blueprint" that can directly drive the engine to generate scenes, greatly reducing the barrier to entry, improving scene construction efficiency and flexibility, and laying the foundation for the large-scale and personalized application of digital twins. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of the architecture of a digital twin scene generation system provided in an embodiment of the present invention.

[0014] Figure 2 A structured data example diagram (in JSON format) of a scene generation medium provided in an embodiment of the present invention.

[0015] Figure 3 This is a flowchart illustrating a digital twin scene generation method provided in an embodiment of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0017] refer to Figure 1 The digital twin scene generation system 100 in this embodiment mainly includes an AI system 110 and a digital twin engine 120. The AI ​​system 110 receives natural language input 101 from the user (e.g., "Place a CNC machine tool in the center of the factory, place a raw material rack on the left, place a finished product rack on the right, and install a monitoring camera above the machine tool"). After internal processing, it outputs a structured scene generation medium 102. The digital twin engine 120 receives the medium 102, parses it, calls the corresponding model from the 3D model library 130 (e.g., "CNC machine tool.glb", "shelf.fbx", "camera.obj", etc.), and instantiates, renders, and logically binds it according to the layout and attributes defined in the medium. Finally, an interactive smart factory digital twin scene 103 is presented on the client 140.

[0018] The AI ​​system 110 includes a requirement understanding module 111, a media generation module 112, and an output module 113. The requirement understanding module 111 may employ a fine-tuned Large Language Model (LLM), which parses the natural language requirement 101 into structured semantic information, such as extracting entities (machine tool, shelf, camera), spatial relationships (center, left, right, top), and attributes (model, color). The media generation module 112, according to predefined grammar rules, fills the aforementioned semantic information into a template to generate a formatted scene generation media 102. The output module 113 sends the media 102 to a designated interface or storage location.

[0019] refer to Figure 2 This example demonstrates a simplified JSON format of scene generation medium 102. The medium contains an array of "scene_elements" (corresponding to the scene element identifier section), listing the required models, their unique IDs, and resource paths. An array of "spatial_relations" (corresponding to the spatial relationship description section) defines the layout between models using relative coordinates or relational terms. A "properties_and_behaviors" object (corresponding to the property and behavior rules section) defines properties such as the initial state of the machine tool and the rotation angle of the camera, and can be associated with data interfaces or scripts to define behaviors.

[0020] The digital twin engine 120 includes a parser 121, a resource manager 122, and a scene assembler 123. The parser 121 reads the medium 102 and extracts all information. The resource manager 122 loads resources such as meshes and textures from a local or cloud-based model library 130 according to the "resource_path". The scene assembler 123 places model instances in 3D space according to spatial relationships, and configures materials, ties animations, or data-driven logic according to attribute and behavior rules, ultimately generating a scene graph and delivering it to the rendering pipeline.

[0021] refer to Figure 3 The method of this invention includes the following steps: Step S301, the user inputs scene construction requirements. Step S302, the AI ​​system understands the requirements and extracts key information. Step S303, the AI ​​system generates a structured scene generation medium based on a template. Step S304, the digital twin engine receives and parses the medium. Step S305, the engine loads the corresponding 3D model resources according to the medium instructions. Step S306, the engine assembles the scene in 3D space according to the defined layout and rules. Step S307, the final interactive digital twin scene is generated.

[0022] The above embodiments are merely illustrative of the technical solutions of the present invention and are not intended to limit it. Those skilled in the art will understand that modifications or equivalent substitutions can be made to the disclosed embodiments without departing from the spirit and scope of the present invention. The scope of protection of the present invention is defined by the appended claims.

Claims

1. A scene generation medium generated by AI and parsable by a digital twin engine, characterized in that, The medium is a structured data file.

2. It is automatically generated by the AI ​​system based on the natural language description or logical requirements of the target scene, and includes at least: Scene element identification section, used to define one or more 3D model elements that constitute a digital twin scene in a machine-readable format; The spatial relationship description section is used to describe the relative position, scale, or topological relationship between the multiple 3D model elements; The Attributes and Behaviors section is used to define the static attribute parameters and / or dynamic behavior logic of at least one 3D model feature; The data structure and semantic format of the medium are predefined, enabling the digital twin engine to parse the medium and automatically call and assemble the corresponding 3D model resources based on the parsing results to generate an interactive digital twin scene.

3. The scene generation medium according to claim 1, characterized in that, The medium is a text file in JSON, XML, or YAML format, or a binary file in a specific encoding format.

4. The scene generation medium according to claim 1, characterized in that, The AI ​​system is a trained large language model, the natural language description is an unstructured text instruction input by the user, the large language model is configured to understand the semantics of the text instruction and output the scene generation medium that conforms to the predetermined data structure requirements.

5. An AI system for generating a scene generation medium as described in any one of claims 1-3, characterized in that, include: The requirement understanding module is configured to receive and parse the user's input of scenario construction requirements, and extract the elements, relationships and rule information of scenario construction; The media generation module is configured to understand the information output by the requirement understanding module, generate the media data structure according to the predefined scenario, and assemble and generate the corresponding structured data file. The output module is configured to output the generated scene generation medium to a storage location or send it directly to the digital twin engine.

6. A digital twin scene generation system, characterized in that, include: The AI ​​system as described in claim 4 is used to generate a scene generation medium; The digital twin engine, which communicates with the AI ​​system, is configured as follows: Receive the scene generation medium; Analyze the scene element identifiers, spatial relationship descriptions, and attribute and behavior rules in the medium; Based on the analysis results, the corresponding 3D model is retrieved from the associated 3D model library and instantiated. Based on the spatial relationship description and attribute and behavior rules, the instantiated 3D model is laid out and configured in the virtual space to generate a runnable digital twin scene.

7. A method for generating a digital twin scene, characterized in that, Includes the following steps: Receive and understand scenario building requirements through an AI system; Based on the understanding of the requirements, the AI ​​system automatically generates a structured scene generation medium, which includes scene element identifiers, spatial relationship descriptions, and attribute and behavior rules. The scene generation medium is provided to the digital twin engine; The digital twin engine analyzes the medium and automatically calls up 3D model resources based on the analysis results to generate the corresponding digital twin scene.

Images