AI-native digital twin scene construction method and related devices

Abstract

The present application discloses a method and related device for constructing a digital twin scene based on AI native, which relates to the field of software technology and is applied to a large language model to obtain input information for a target digital twin scene and perform intent recognition on the input information; if the intent recognition result of the input information contains goal orientation, determine the target business process for achieving goal orientation; perform function calls of API functions according to the target business process to achieve goal orientation in the target digital twin scene. The present application constructs digital twin scenes based on the AI ​​native idea, and supports autonomous completion of business process orchestration and function calls based on goal orientation, which can greatly improve the efficiency of product scene construction.

Description

Technical Field

[0001] The present application relates to the field of software technology, and in particular to an AI-native digital twin scene construction method and related devices. Background Art

[0002] As technology advances, the technology for building digital twin scenarios (such as digital twin cities) will become more sophisticated and intelligent. It will be able to simulate complex systems more accurately, provide decision makers with deeper insights, and play a key role in more areas. Digital twin scenario building technology is a bridge between the physical world and the digital world.

[0003] The traditional technical route is process-oriented software design, with GUI (Graphical User Interface) menu interface as the main interaction method. Often a business process requires the operation of a large number of UI menus and editing panels, which has problems such as high learning threshold and cumbersome operation process. In addition, due to the process orientation based on a limited problem domain, it is impossible to autonomously complete business process orchestration based on goal orientation. Summary of the invention

[0004] In view of the above problems, this application provides a method and related devices for constructing digital twin scenarios based on AI native to achieve the purpose of autonomously completing business process orchestration based on goal orientation. The specific scheme is as follows:

[0005] In a first aspect, the present application provides a method for constructing a digital twin scene based on AI native, wherein the method for constructing a digital twin scene based on AI native is applied to a large language model, and the method for constructing a digital twin scene based on AI native includes:

[0006] Obtain input information for the target digital twin scene, and perform intent recognition on the input information;

[0007] If the intention recognition result of the input information includes a goal orientation, determining a target business process for achieving the goal orientation;

[0008] The function call of the API function is performed according to the target business process to achieve the target orientation in the target digital twin scenario.

[0009] In a possible implementation, the determining a target business process for achieving the goal orientation includes:

[0010] Retrieve the API description document corresponding to the target digital twin scenario, where the API description document is used to record the description information of the atomic API function in the constructed atomic function set. The atomic API function is obtained by encapsulating the API function of the smallest granularity business process in the target digital twin scenario. The description information includes the function call description of the atomic API function and the dependency description of the business process with other business processes.

[0011] Constructing a business process tree corresponding to the target digital twin scenario according to the dependency description, wherein the business process tree can represent the dependency relationships of all business processes in the target digital twin scenario;

[0012] A tree node corresponding to the target orientation in the business process tree is determined, and all business processes under the tree node in the business process tree are taken as the target business processes.

[0013] In a possible implementation, the function call of the API function according to the target business process includes:

[0014] Determine a business process sequence as leaf nodes of the tree node from the target business process;

[0015] The atomic API functions corresponding to the business process sequence are sequentially used as target atomic API functions to be called;

[0016] A target function call description corresponding to the target atomization API function is obtained from the description information, and a function call is performed on the target atomization API function according to the target function call description.

[0017] In a possible implementation, the AI-native digital twin scene construction method further includes:

[0018] If the intention recognition result of the input information contains the target question, calling the knowledge enhancement retrieval engine, wherein the digital twin knowledge base is imported into the knowledge enhancement retrieval engine;

[0019] Retrieving a target answer corresponding to the target question from the digital twin knowledge base through the knowledge enhancement retrieval engine;

[0020] The target answer is output.

[0021] In a possible implementation, the AI-native digital twin scene construction method further includes:

[0022] If the intention recognition result of the input information includes a modeling requirement, calling a service function description document, wherein the service function description document is used to record a modeling function description of the modeling model;

[0023] A target modeling model that can realize the modeling requirement is determined based on the modeling function description, and the target modeling model is run to realize the modeling requirement.

[0024] A second aspect of the present application provides an AI-native digital twin scene construction device, which is applied to a large language model. The AI-native digital twin scene construction device includes:

[0025] An intention recognition module, used to obtain input information for the target digital twin scene and perform intention recognition on the input information;

[0026] A function calling module is used to determine the target business process for realizing the goal orientation if the intention recognition result of the input information contains the goal orientation; and to make a function call of the API function according to the target business process to realize the goal orientation in the target digital twin scenario.

[0027] In a possible implementation, the AI-native digital twin scene construction device further includes:

[0028] The knowledge retrieval module is used to call the knowledge enhancement retrieval engine if the intention recognition result of the input information contains a target question, and the digital twin knowledge base is imported into the knowledge enhancement retrieval engine; through the knowledge enhancement retrieval engine, the target answer corresponding to the target question is retrieved from the digital twin knowledge base; and the target answer is output.

[0029] The third aspect of the present application provides a computer program product, including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements the AI-native digital twin scene construction method of the above-mentioned first aspect or any implementation method of the first aspect.

[0030] A fourth aspect of the present application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0031] The memory is used to store computer programs;

[0032] The processor is used to execute the computer program so that the electronic device can implement the AI-native digital twin scene construction method of the above-mentioned first aspect or any implementation method of the first aspect.

[0033] The fifth aspect of the present application provides a computer storage medium, which carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the AI-native digital twin scene construction method of the above-mentioned first aspect or any implementation method of the first aspect.

[0034] Through the above technical solution, the present application provides a method and related device for constructing a digital twin scene based on AI native, which is applied to a large language model to obtain input information for the target digital twin scene and perform intent recognition on the input information; if the intent recognition result of the input information contains goal orientation, determine the target business process for achieving goal orientation; perform function calls of API functions according to the target business process to achieve goal orientation in the target digital twin scene. This application constructs digital twin scenes based on AI native ideas, and supports autonomous completion of business process orchestration and function calls based on goal orientation, which can greatly improve the efficiency of product scene construction. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the accompanying drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and the originals and elements are not necessarily drawn to scale.

[0036] Figure 1 A flowchart of a method for constructing a digital twin scene based on AI native provided in an embodiment of the present application;

[0037] Figure 2 A partial flow chart of a method for constructing a digital twin scene based on AI native provided in an embodiment of the present application;

[0038] Figure 3 A schematic diagram of a business process decomposition hierarchy and conceptual system provided for an embodiment of the present application;

[0039] Figure 4 A schematic diagram of a business process disassembly of a typical digital twin scenario provided in an embodiment of the present application;

[0040] Figure 5 A schematic diagram of a CUI interactive interface constructed based on an AI-native digital twin scene provided in an embodiment of the present application;

[0041] Figure 6 Another partial flow chart of a method for constructing a digital twin scene based on AI native provided in an embodiment of the present application;

[0042] Figure 7Another flowchart of a method for constructing a digital twin scene based on AI native provided in an embodiment of the present application;

[0043] Figure 8 Another flowchart of a method for constructing a digital twin scene based on AI native provided in an embodiment of the present application;

[0044] Fig. 9 A schematic diagram of the structure of an AI-native digital twin scene construction device provided in an embodiment of the present application;

[0045] Fig.10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] The following describes the embodiments of the present application in conjunction with the drawings in the embodiments of the present application. The terms used in the implementation method section of the present application are only used to explain the specific embodiments of the present application, and are not intended to limit the present application.

[0047] The embodiments of the present application are described below in conjunction with the accompanying drawings. Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0048] The terms "first", "second" etc. in the specification of the application and the above-mentioned drawings are used to distinguish similar objects, and need not be used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable in appropriate circumstances, and this is only to describe the distinction mode adopted by the objects of the same attributes when describing in the embodiments of the application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0049] To facilitate understanding of this application, the following first describes the relevant concepts involved in this application:

[0050] Digital twin city: is an innovative urban management model that realizes real-time monitoring, analysis and simulation of the city's operating status by building digital copies of the city's physical entities. This technology combines multiple technical means such as the Internet of Things (IoT), big data analysis, artificial intelligence (AI), and geographic information systems (GIS). Its advantage is that it can provide more accurate decision-making support. City managers can use this data for urban planning, traffic management, environmental monitoring, etc., thereby improving the efficiency and response speed of urban management. With the development of technology, digital twin cities are becoming an important part of smart city construction. It can not only improve the level of intelligence in cities, but also provide residents with a safer, more convenient and comfortable living environment. Through digital twin technology, future urban management will be more efficient, transparent and humane.

[0051] Digital twin scene construction: It is an advanced information technology that simulates, analyzes and optimizes scenes in the real world by creating digital copies of physical entities. This technology involves multiple fields, including but not limited to 3D modeling, data integration, scene editing and real-time simulation.

[0052] AI native: refers to the use of artificial intelligence (AI) technology as a core component rather than an additional feature during the design, development and operation process. AI-native systems and products consider the integration and application of AI from the beginning, thus achieving smarter and more automated solutions. Its advantages include:

[0053] 1) User experience: Provide a more personalized and intelligent user experience.

[0054] 2) Efficiency improvement: Improve work efficiency through automation and process optimization.

[0055] 3) Decision support: Provide data-based insights to assist decision making.

[0056] 4) Accelerated innovation. AI-native products can quickly adapt to changes and promote innovation.

[0057] CUI (Conversational User Interface) interaction paradigm: is an emerging human-computer interaction method. It uses artificial intelligence technology, especially natural language processing (NLP) technology, to allow users to communicate with the system through natural language. This interaction method is gradually becoming a new trend in digital product design. Compared with the traditional GUI (graphical user interface), it has the advantages of being more intuitive and convenient.

[0058] AI Agent Design Patterns: refers to a set of design principles and methods used when building artificial intelligence agents (AI Agents). AI Agents generally refer to intelligent systems that can perform tasks, make decisions, and autonomously interact with the environment or users.

[0059] 3D AIGC (3D Artificial Intelligence Generated Content) capability: refers to the process of automatically generating 3D content using artificial intelligence technology. This technology can be applied in many fields, including but not limited to game development, film production, virtual reality (VR), augmented reality (AR), architectural visualization and product design.

[0060] As AI technology continues to advance, AI native will become more popular and become a key force in promoting digital transformation in all walks of life. It will help enterprises and organizations achieve more efficient operations, create more value, and bring users a more intelligent service experience.

[0061] Currently, there are two main technical routes for building digital twin scenarios. One is the traditional process-oriented software design architecture, and the other is the AI-injected technology route that embeds AI capabilities as tools.

[0062] The traditional technical route is process-oriented software design, with GUI menu interface as the main interaction method. Often a business process requires the operation of a large number of UI menus and editing panels, which has problems such as high learning threshold and cumbersome operation process. In addition, due to the process orientation based on a limited problem domain, it is impossible to independently complete the business process orchestration based on goal orientation.

[0063] The AI ​​intelligence technology route is still a system architecture oriented to process design, but AI technology will be integrated into some specific business processes or functional modules. This model positions AI as an enabling tool, and AI helps people complete specific tasks. However, most of the work is still done by humans, including setting work goals, decomposing target sub-processes, and completing sub-process tasks one by one based on UI menus. In this model, although the introduction of AI capabilities in specific functions and local modules has achieved performance, efficiency and innovation capabilities, the overall software interaction paradigm is still based on the GUI menu interface, with a high learning threshold and cumbersome operation procedures. At the same time, it is still a process-oriented software design architecture, and it is impossible to autonomously complete business process orchestration based on goal orientation.

[0064] In order to solve the above problems, the embodiment of the present application provides a method for constructing a digital twin scene based on AI native. The following is a detailed introduction to the method for constructing a digital twin scene based on AI native in the embodiment of the present application in conjunction with the accompanying drawings.

[0065] See also Figure 1 , Figure 1 A flowchart of a method for constructing a digital twin scene based on AI native is provided in an embodiment of the present application. Figure 1 As shown, an AI-native digital twin scene construction method provided in an embodiment of the present application is applied to a large language model (LLM, Large Language Model), which may include steps S10 to S30, and these steps are described in detail below.

[0066] S10, obtaining input information for the target digital twin scene, and performing intent recognition on the input information.

[0067] In the embodiments of the present application, the AI ​​Agent (intelligent agent) design pattern is introduced. The assistant AI Agent adopts an anthropomorphic role division method to realize the upgrade of the interactive interface constructed by the digital twin scene from the traditional GUI to the GUI interaction paradigm, and supports the autonomous completion of business process orchestration and function calls based on goal orientation. Specific functions include conversational interaction, business process orchestration, and professional knowledge Q&A. The AI-native CUI provides a more natural and personalized user experience than the traditional GUI through conversational interaction. CUI is not visually restricted and is easy to access. This interaction method shows obvious advantages in improving efficiency and user satisfaction.

[0068] Based on this, the user's input information (including text, voice, etc.) for the specified digital twin scene (i.e., the target digital twin scene) can be obtained through the interactive interface, and the intention of the input information can be recognized to determine the function to be implemented.

[0069] It should be noted that in the embodiment of the present application, the assistant AI Agent follows the principle of Conway's Law and defines the duties and responsibilities of the "digital employee" in the organizational structure with anthropomorphic roles. Among them, Conway's Law states: "The architecture of the system is subject to the communication structure between the organization and the people using the system." This means that the design of the digital system must take into account the communication and collaboration methods within the organization.

[0070] S20: If the intention recognition result of the input information includes goal orientation, determine a target business process for achieving the goal orientation.

[0071] In an embodiment of the present application, if the intention recognition result of the input information includes goal orientation, that is, the function to be implemented is business process orchestration, the business process (that is, the target business process) required to implement the goal orientation in the target digital twin scenario can be determined.

[0072] The current mainstream large-model application development frameworks all have the ability to call functions, such as LangChain, Bisheng, Zhipu AI Open Platform, etc. After the large language model fully learns the corresponding API (Application Programming Interface) description document, it can choose to call the correct API interface and combination based on the context and specific context, and extract API parameters from the dialogue information.

[0073] See also Figure 2 , Figure 2 A partial flow chart of a method for constructing a digital twin scene based on AI native provided in an embodiment of the present application. Figure 2 As shown, an embodiment of the present application provides an AI-native digital twin scene construction method, wherein "determining a target business process for achieving goal orientation" in step S20 may include steps S201 to S203, and these steps are described in detail below.

[0074] S201, call the API description document corresponding to the target digital twin scene. The API description document is used to record the description information of the atomic API function in the constructed atomic function set. The atomic API function is obtained by encapsulating the API function of the smallest granularity business process in the target digital twin scene. The description information includes the function call description of the atomic API function and the dependency description of the business process with other business processes.

[0075] S202, construct a business process tree corresponding to the target digital twin scenario according to the dependency description. The business process tree can represent the dependency relationships of all business processes in the target digital twin scenario.

[0076] S203, determining a tree node corresponding to the target orientation in the business process tree, and taking all business processes under the tree node in the business process tree as target business processes.

[0077] In the embodiment of the present application, guided by the top-level business objectives, the business processes involved in the target digital twin scenario are disassembled and atomically packaged, so as to build a multi-level business process architecture. Specifically:

[0078] 1) Business process decomposition. For the granularity and hierarchical definition of business process decomposition, the present application embodiment introduces the concept system setting of agile development Scrum (Scrum Framework, which means Scrum framework in Chinese. It is an agile project management method for developing, delivering and continuously supporting complex products), and defines three levels: epic, user story and task chain. Figure 3 , Figure 3A schematic diagram of a business process decomposition hierarchy and conceptual system provided in an embodiment of the present application. Figure 3 As shown, the top-level business goal contains multiple epics (including epic 1, epic 2, ..., epic N). Taking epic 1 as an example, epic 1 contains multiple user stories (including user story 1, user story 2, ..., user story M). Taking user story 1 as an example, the user story contains multiple task chains (including task chain 1, task chain, ..., task chain Q). Among them, epics usually represent a larger function or business goal, which may require multiple iterations to complete. It is a large and complex work unit. Usually, an epic will contain multiple user stories; user stories represent a smaller and specific functional requirement, which can usually be completed within one iteration. It is a small and simple work unit; task chains are more specific than user stories, provide more details, and include a sequence of execution tasks to complete a specific single business goal, which is generally a fixed task workflow.

[0079] See also Figure 4 , Figure 4 A schematic diagram of the business process disassembly of a typical digital twin scenario provided in the embodiment of this application. Figure 4 As shown, from the top-level business goals, it can be divided into three epics: loading scene templates, adding scene content, and editing scene properties. Taking adding scene content as an example, it includes three user stories: adding scene baseplate, adding 3D model, and adding special effect components. Taking adding scene baseplate as an example, it includes three task chains: adding GIS data service, adding fine model data, and adding BIM (Building Information Modeling) data. Taking adding fine model data as an example, its task workflow includes model selection, pose setting, and graphics rendering.

[0080] 2) Atomic capability encapsulation. The task chain after the above business process is disassembled is encapsulated into an atomic API function. The embodiment of the present application can adopt the RestfulAPI (Representational State Transfer ApplicationProgramming Interface) design specification to support calling in the form of services. Then, the API description document is written in accordance with the format specification defined by the large model application development framework and provided to the large language model for learning, so as to realize the function call of multiple atomic API functions based on the CUI conversation-driven large language model, and then realize a complete business process orchestration and execution.

[0081] Based on this, in the embodiment of the present application, the API description document corresponding to the target digital twin scene is retrieved, and the API description document records the description information of each atomic API function in the atomic function set, wherein the atomic API function is the smallest granularity business process in the target digital twin scene (i.e. Figure 3 The description information of the atomic API function includes, on the one hand, a function call description (such as the purpose of the atomic API function, the interface call location, etc.), and a description of the dependency relationship between the business process corresponding to the atomic API function and other business processes.

[0082] Taking the task chain of adding fine model data as an example, the API functions of model selection, pose setting and graphics rendering in its task workflow can be encapsulated into an atomic API function. Correspondingly, its description information includes the function call description of the atomic API function on the one hand, and the dependency description between adding fine model data and adding scene baseplate and adding scene content on the other hand. Similarly, the task chain of adding GIS data service can also be encapsulated into another atomic API function. Its description information includes the function call description of the atomic API function on the one hand, and the dependency description between adding GIS data service and adding scene baseplate and adding scene content on the other hand.

[0083] Based on this, we can construct a business process tree for the target digital twin scenario based on the dependency description of each atomic API function in the atomic function set, that is, a tree structure that represents each layer of business processes. This business process tree can represent the dependency of all business processes. Figure 4 As shown in the figure, building a twin scene is the root node, and its child nodes are loading scene templates, adding scene content, and editing scene properties. Taking adding scene content as an example, its child nodes are adding scene base, adding 3D models, and adding special effect components. Taking adding scene bottom as an example, its child nodes are adding GIS data services, adding fine model data, and adding BIM data. Adding GIS data services, adding fine model data, and adding BIM data are all leaf nodes of the business process tree.

[0084] Further, determine the tree node corresponding to the goal orientation in the business process tree. For example, if the goal orientation is to add a scene baseplate, then you can locate the tree node of adding a scene baseplate in the business process tree and set all business processes under the tree node, including adding GIS data services, adding fine model data, and adding BIM data, as target business processes.

[0085] S30, making function calls to the API functions according to the target business process to achieve target orientation in the target digital twin scenario.

[0086] In an embodiment of the present application, after determining the target business process, a function call can be made to the API function involved in the target business process based on Function call to achieve target-oriented construction in the target digital twin scenario.

[0087] See also Figure 5 , Figure 5 A schematic diagram of a CUI interactive interface based on AI-native digital twin scene construction provided in an embodiment of the present application. Continuing with the example of adding a scene baseplate as a goal-oriented approach, adding GIS data services, adding fine model data, and adding BIM data can be displayed as target business processes on the interactive interface, and users can perform editing operations such as modification and deletion. When the user finishes editing, the API function call can be made according to the edited target business process, and finally the digital twin scene construction result is output in the display area.

[0088] See also Figure 6 , Figure 6 Another partial flow chart of a method for constructing a digital twin scene based on AI native provided in an embodiment of the present application. Figure 6 As shown, an AI-native digital twin scene construction method is provided in an embodiment of the present application, wherein "perform a function call of an API function according to a target business process" in step S30 may include steps S301 to S303, and these steps are described in detail below.

[0089] S301, determining a business process sequence that is a leaf node of a tree node from a target business process.

[0090] S302, sequentially taking the atomic API functions corresponding to the business process sequence as target atomic API functions to be called.

[0091] S303, obtaining a target function call description corresponding to the target atomic API function from the description information, and performing a function call on the target atomic API function according to the target function call description.

[0092] In the embodiment of the present application, for the target business process, multiple business processes as leaf nodes of the tree node corresponding to the target orientation can be determined to form a business process sequence. Continuing to take the target orientation as adding a scene base plate as an example, the target business process under the corresponding tree node includes adding GIS data services, adding fine model data, and adding BIM data, among which the business processes as leaf nodes of the tree node include adding GIS data services, adding fine model data, and adding BIM data. According to the execution order of the business processes, the business process sequence can be determined as adding GIS data services → adding fine model data → adding BIM data.

[0093] Furthermore, according to the execution order of the business process sequence, the atomic API functions corresponding to adding GIS data services, adding fine model data, and adding BIM data are sequentially used as atomic API functions to be called (i.e., target atomic API functions). For the target atomic API function, its corresponding function call description (i.e., target function call description) can be obtained from the description information, and then the function call of the target atomic API function is performed according to the target function call description.

[0094] In this regard, the embodiments of the present application can encapsulate the atomic capabilities that have been disassembled based on business goals (i.e., goal-oriented) into an API, and write an API description document in accordance with the format specifications defined by the corresponding large language model, so as to complete multiple function calls based on the CUI conversational-driven large language model, thereby realizing a complete business process orchestration and execution.

[0095] In a possible implementation, the large language model can also be learned in an external knowledge base manner to complete professional knowledge question answering. Figure 7 , Figure 7 Another flowchart of a method for constructing a digital twin scene based on AI native is provided in an embodiment of the present application. Figure 7 As shown, an AI-native digital twin scenario provided in an embodiment of the present application may also include steps S40 to S60, and these steps are described in detail below.

[0096] S40, if the intention recognition result of the input information contains the target question, call the knowledge enhancement retrieval engine, and the digital twin knowledge base is imported into the knowledge enhancement retrieval engine.

[0097] S50, through the knowledge enhancement retrieval engine, retrieve the target answer corresponding to the target question from the digital twin knowledge base.

[0098] S60, outputting the target answer.

[0099] In an embodiment of the present application, if the intention recognition result of the input information includes a question input by the user (i.e., the target question), that is, the function to be implemented is professional knowledge question and answer, a knowledge enhancement retrieval engine (which may be RAG, the full name of which is Retrieval-Augmented Generation) can be called for this purpose, and the digital twin knowledge base has been imported into the knowledge enhancement retrieval engine.

[0100] Drive the knowledge enhancement retrieval engine, retrieve the answer corresponding to the target question (i.e., the target answer) from the digital twin knowledge base through the knowledge enhancement retrieval engine, and output the target answer in the interactive interface.

[0101] In this regard, the embodiment of the present application implements large language model plug-in knowledge enhanced retrieval based on RAG technology, which can greatly improve the ability of the large language model in professional field knowledge understanding, multi-source information integration, etc., so that the assistant AI Agent has the ability to answer professional knowledge related to digital twin scenario construction, can understand complex professional terms, and provide context-sensitive and accurate answers.

[0102] In one possible implementation, expert AI Agent, based on the new productivity brought by 3D AIGC, can penetrate all aspects of the digital twin scene construction business process with atomic AI capabilities, which can greatly improve the construction efficiency and automation and intelligence level of 3D assets. Figure 8 , Figure 8 Another flowchart of a method for constructing a digital twin scene based on AI native is provided in an embodiment of the present application. Figure 8 As shown, an AI-native digital twin scenario provided in an embodiment of the present application may also include steps S70 to S80, and these steps are described in detail below.

[0103] S70, if the intention recognition result of the input information includes modeling requirements, retrieve the service function description document, which is used to record the modeling function description of the modeling model.

[0104] S80, determining a target modeling model that can realize the modeling requirement based on the modeling function description, and running the target modeling model to realize the modeling requirement.

[0105] In the embodiment of the present application, if the intention recognition result of the input information contains the user's modeling requirements, the service function description document can be retrieved to obtain the modeling function description of the modeling model recorded therein. The modeling models based on the core 3D AIGC capability include NeRF (full name Neural Radiance Fields) neural radiation field modeling, single image generation model and 3D Gaussian splash modeling. Among them,

[0106] NeRF neural radiation field modeling is used to create high-resolution three-dimensional environmental models. By analyzing a large amount of two-dimensional image data and using deep learning algorithms to infer the three-dimensional structure of the scene, modeling time can be greatly shortened, costs can be reduced, and an immersive experience can be provided. The NeRF technology ecosystem is developing rapidly, and domestic and foreign companies and research institutions represented by Google (Mip-NeRF-360), NVIDIA (instant-ngp), and Shanghai Artificial Intelligence Laboratory (Shusheng*Tianji) have all opened their own NeRF neural network models to the public.

[0107] The single-image generation model is a technology that generates a 3D model from a single image. It combines the advanced technologies in computer vision and machine learning. The core algorithm steps designed include image analysis, depth estimation, 3D reconstruction, texture mapping, and detail optimization. Currently, the representative open source models include Shap-E, ONE-2-3-45, and Tsinghua CRM.

[0108] 3D Gaussian splat modeling is a mathematical method that simulates Gaussian distribution in three-dimensional space to generate random data points with specific characteristics. This technology is widely used in computer graphics, physics simulation, machine learning and other fields. Compared with NeRF neural radiation field, 3D Gaussian splat has advantages including stronger real-time rendering capabilities, higher computing efficiency and less memory usage.

[0109] In practical applications, the large language model can be used by the expert AIAgent to identify user intent and call the 3D AIGC MaaS (Model as a Service) service. The user intent recognition implementation of the expert AIAgent is similar to that of the assistant AIAgent. By letting the large language model learn the 3D AIGC MaaS service function description document in advance, it helps it to reason about the relevance and accurately match the user intent.

[0110] In practical applications, the data base can support the unified aggregation, fusion, management and sharing of multi-source heterogeneous spatiotemporal data. MaaS can also be used to build a full-process, componentized general AI base to provide lightweight, assemblable MaaS capabilities, including full-lifecycle management services for large language models and multimodal large models, as well as cognitive enhancement services for feature fields and industry applications.

[0111] The technical architecture of the AI ​​base includes application development, cognitive enhancement, and basic model development. Application development includes service orchestration, microservices, and tool environment integration. Cognitive enhancement includes model fine-tuning, prompt engineering, RAG, AI Agent, application development, and model security governance. Basic model development includes data annotation, model development, model training, and model reasoning.

[0112] Basic model development focuses on creating and implementing basic large models, covering the entire process from data preparation and annotation to model design, development, training, and result prediction through model reasoning. In addition, it also includes providing the necessary computing resources and software framework to support the efficient operation and processing capabilities of the model.

[0113] Cognitive enhancement: To make up for the limitations of existing large models in specific industry knowledge and logical reasoning, MaaS cognitive enhancement components are introduced to ensure the effectiveness and applicability of these models in actual industry applications, so that they can be more accurately embedded and serve specific industry environments. This technology covers the following key aspects:

[0114] 1) Model fine-tuning: Targeted adjustments are made to large models to suit the needs and context of specific industries.

[0115] 2) Prompt Engineering: Optimizing the input prompts that the model receives to improve its performance on specific tasks.

[0116] 3) RAG. The retrieval mechanism is combined to enhance the model’s generation capability, enabling it to use external information sources for more accurate information retrieval and content generation.

[0117] 4) AI Agent: Deploy intelligent agents to perform complex tasks and improve the model’s ability to interact and adapt in dynamic environments.

[0118] 5) Application development: Encapsulate MaaS into microservices, provide service orchestration capabilities for different scenarios and business goals, and provide relevant tools and environment integration.

[0119] 6) Model security governance. The AI ​​model security governance service ensures the security, reliability and compliance of the model through a series of measures, reduces the risks of data leakage, manipulation, unfair decision-making and illegal use, and enhances the transparency and explainability of the model.

[0120] Applying the embodiments of this application to the construction of digital twin cities can improve the efficiency of city construction and enhance the user interaction experience and intelligence level of the digital twin scene construction system. The digital twin scene construction method based on AI native provided by the implementation example of this application is designed based on the AI ​​native idea to upgrade the traditional GUI to the CUI interaction paradigm. At the same time, the AI ​​Agent design mode is introduced to support the autonomous completion of business process orchestration and function calls based on goal orientation, and the 3D AIGC capability is introduced to achieve high automation and intelligence of 3D asset modeling.

[0121] In summary, this application can greatly improve product interaction experience, scene construction efficiency and 3D asset modeling intelligence. Compared with other mainstream technical solutions, this application has obvious technical advantages in three dimensions: human-computer interaction, process experience and intelligent production, as follows:

[0122] Human-computer interaction. Based on the CUI interaction paradigm, it fully taps and releases the capabilities of the large language model, can understand the user context, and thus provide a more personalized interaction experience. It supports multimodal conversational interaction, has online learning capabilities, and can continuously learn from new interaction data to improve its intelligence level.

[0123] Process experience. Relying on the powerful reasoning ability of the large language model and combined with the prompt word engineering, it can quickly identify user intentions and pain points, provide personalized work processes, automate many repetitive and time-intensive processes, reduce cumbersome processes, reduce manual intervention, and improve overall efficiency.

[0124] Intelligent production. Relying on 3D AIGC's powerful 3D content creation capabilities, including NeRF neural radiation field, single image generation model and 3D Gaussian splashing technologies, it can quickly generate high-quality 3D models, greatly improving design and modeling efficiency. Automated content generation reduces dependence on professionals, thereby reducing production costs. At the same time, it can generate realistic 3D models and scenes, improving the quality of visual content.

[0125] The above introduces an AI-native digital twin scene construction method provided in an embodiment of the present application. The following will introduce a device for executing the above-mentioned AI-native digital twin scene construction method.

[0126] See also Fig. 9 , Fig. 9 A schematic diagram of the structure of a digital twin scene construction device based on AI native provided in an embodiment of the present application. Fig. 9 As shown, an AI-native digital twin scene construction device provided in an embodiment of the present application includes:

[0127] The intention recognition module 10 is used to obtain input information for the target digital twin scene and perform intention recognition on the input information;

[0128] The function calling module 20 is used to determine the target business process for achieving the goal orientation if the intention recognition result of the input information contains the goal orientation; and perform function calls of the API function according to the target business process to achieve the goal orientation in the target digital twin scenario.

[0129] In a possible implementation, the function calling module 20 for determining a target business process for implementing a goal orientation is specifically used to:

[0130] Retrieve the API description document corresponding to the target digital twin scenario. The API description document is used to record the description information of the atomic API function in the constructed atomic function set. The atomic API function is obtained by encapsulating the API function of the smallest granularity business process in the target digital twin scenario. The description information includes the function call description of the atomic API function and the dependency description of the business process with other business processes. Build a business process tree corresponding to the target digital twin scenario according to the dependency description. The business process tree can characterize the dependency relationship of all business processes in the target digital twin scenario. Determine the tree node corresponding to the target orientation in the business process tree, and take all business processes under the tree node of the business process tree as the target business process.

[0131] In a possible implementation, the function calling module 20 for calling the function of the API function according to the target business process is specifically used to:

[0132] Determine a business process sequence as a leaf node of the tree node from the target business process; sequentially use the atomic API functions corresponding to the business process sequence as the target atomic API functions to be called; obtain the target function call description corresponding to the target atomic API function from the description information, and perform a function call on the target atomic API function according to the target function call description.

[0133] In a possible implementation, the above-mentioned AI-native digital twin scene construction device further includes:

[0134] The knowledge retrieval module is used to call the knowledge enhancement retrieval engine if the intention recognition result of the input information contains the target question, and the digital twin knowledge base is imported into the knowledge enhancement retrieval engine; through the knowledge enhancement retrieval engine, the target answer corresponding to the target question is retrieved from the digital twin knowledge base; and the target answer is output.

[0135] In a possible implementation, the above-mentioned AI-native digital twin scene construction device further includes:

[0136] The modeling module is used to retrieve the service function description document if the intention recognition result of the input information contains modeling requirements. The service function description document is used to record the modeling function description of the modeling model; based on the modeling function description, the target modeling model that can achieve the modeling requirements is determined, and the target modeling model is run to achieve the modeling requirements.

[0137] It should be noted that the specific functions of each module in the AI-native digital twin scene construction device provided in the embodiment of the present application can be found in the corresponding public part of the above-mentioned AI-native digital twin scene construction method embodiment, and will not be repeated here.

[0138] The present application also provides an electronic device in an embodiment. Fig.10 , Fig.10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device in the embodiment of the present application may include but is not limited to fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Fig.10 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0139] like Fig.10 As shown, the electronic device may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 1001, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage device 1008 to a random access memory (RAM) 1003. When the electronic device is powered on, various programs and data required for the operation of the electronic device are also stored in the RAM 1003. The processing device 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0140] Typically, the following devices may be connected to the I / O interface 1005: an input device 1006 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 1007 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 1008 including, for example, a memory card, a hard disk, etc.; and a communication device 1009. The communication device 1009 may allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although Fig.10 An electronic device having various devices is shown, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0141] Also provided in an embodiment of the present application is a computer program product, including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any one of the AI-native digital twin scene construction methods provided in the embodiments of the present application.

[0142] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When one or more computer programs are executed by an electronic device, the electronic device can implement any one of the AI-native digital twin scene construction methods provided in the embodiment of the present application.

[0143] It should also be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed over multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the drawings of the device embodiments provided by the present application, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines.

[0144] Through the description of the above implementation mode, the technicians in the field can clearly understand that the present application can be implemented by means of software plus necessary general hardware, and of course, it can also be implemented by special hardware including special integrated circuits, special CPUs, special memories, special components, etc. In general, all functions completed by computer programs can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be various, such as analog circuits, digital circuits or special circuits. However, for the present application, software program implementation is a better implementation mode in more cases. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a readable storage medium, such as a computer floppy disk, a U disk, a mobile hard disk, a ROM, a RAM, a disk or an optical disk, etc., including a number of instructions to enable a computer device (which can be a personal computer, a training device, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0145] In the above embodiments, all or part of the embodiments may be implemented by software, hardware, firmware or any combination thereof. When implemented by software, all or part of the embodiments may be implemented in the form of a computer program product.

[0146] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions may be transmitted from a website site, a computer, a training device, or a data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, training device, or data center. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device, a data center, etc. that includes one or more available media integrations. The available medium may be a magnetic medium, (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)), etc.

Claims

1. A method for constructing a digital twin scene based on AI native, characterized in that: The AI-native digital twin scenario construction method is applied to a large language model, and the AI-native digital twin scenario construction method includes: Obtain input information for the target digital twin scene, and perform intent recognition on the input information; If the intention recognition result of the input information includes a goal orientation, determining a target business process for achieving the goal orientation; Performing function calls of API functions according to the target business process to achieve the target orientation in the target digital twin scenario; Wherein, the determining of the target business process for achieving the target orientation includes: Retrieve the API description document corresponding to the target digital twin scenario, where the API description document is used to record the description information of the atomic API function in the constructed atomic function set. The atomic API function is obtained by encapsulating the API function of the smallest granularity business process in the target digital twin scenario. The description information includes the function call description of the atomic API function and the dependency description of the business process with other business processes. Constructing a business process tree corresponding to the target digital twin scenario according to the dependency description, wherein the business process tree can represent the dependency relationships of all business processes in the target digital twin scenario; Determine a tree node corresponding to the goal orientation in the business process tree, and use all business processes under the tree node in the business process tree as the target business process; The function call of the API function according to the target business process includes: Determine a business process sequence as leaf nodes of the tree node from the target business process; The atomic API functions corresponding to the business process sequence are sequentially used as target atomic API functions to be called; A target function call description corresponding to the target atomization API function is obtained from the description information, and a function call is performed on the target atomization API function according to the target function call description.

2. The method for constructing a digital twin scene based on AI native according to claim 1, characterized in that: The AI-native digital twin scene construction method further includes: If the intention recognition result of the input information contains the target question, calling the knowledge enhancement retrieval engine, wherein the digital twin knowledge base is imported into the knowledge enhancement retrieval engine; Retrieving a target answer corresponding to the target question from the digital twin knowledge base through the knowledge enhancement retrieval engine; The target answer is output.

3. The method for constructing a digital twin scene based on AI native according to claim 1, characterized in that: The AI-native digital twin scene construction method further includes: If the intention recognition result of the input information includes a modeling requirement, calling a service function description document, wherein the service function description document is used to record a modeling function description of the modeling model; A target modeling model that can realize the modeling requirement is determined based on the modeling function description, and the target modeling model is run to realize the modeling requirement.

4. A digital twin scene construction device based on AI native, characterized in that: The digital twin scene construction device based on AI native is applied to a large language model, and the digital twin scene construction device based on AI native includes: An intention recognition module, used to obtain input information for the target digital twin scene and perform intention recognition on the input information; A function calling module, for determining a target business process for realizing the target orientation if the intention recognition result of the input information includes the target orientation; and performing a function call of an API function according to the target business process to realize the target orientation in the target digital twin scenario; The function calling module for determining the target business process for implementing the target orientation is specifically used for: Retrieve the API description document corresponding to the target digital twin scenario, the API description document is used to record the description information of the atomic API function in the constructed atomic function set, the atomic API function is obtained by encapsulating the API function of the smallest granularity business process in the target digital twin scenario, and the description information includes the function call description of the atomic API function and the dependency description of the business process and other business processes; construct a business process tree corresponding to the target digital twin scenario according to the dependency description, the business process tree can characterize the dependency relationship of all business processes in the target digital twin scenario; determine the tree node corresponding to the target orientation in the business process tree, and take all business processes under the tree node in the business process tree as the target business process; The function calling module for performing a function call of an API function according to the target business process is specifically used for: Determine a business process sequence as a leaf node of the tree node from the target business process; use the atomic API functions corresponding to the business process sequence as the target atomic API functions to be called in turn; obtain the target function call description corresponding to the target atomic API function from the description information, and make a function call to the target atomic API function according to the target function call description.

5. The AI-native digital twin scene construction device according to claim 4 is characterized in that: The AI-native digital twin scene construction device also includes: The knowledge retrieval module is used to call the knowledge enhancement retrieval engine if the intention recognition result of the input information contains a target question, and the digital twin knowledge base is imported into the knowledge enhancement retrieval engine; through the knowledge enhancement retrieval engine, the target answer corresponding to the target question is retrieved from the digital twin knowledge base; and the target answer is output.

6. A computer program product, characterized in that It includes computer-readable instructions, which, when executed on an electronic device, enable the electronic device to implement the AI-native digital twin scene construction method as described in any one of claims 1 to 3.

7. An electronic device, characterized in that: The method comprises at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is used to execute the computer program so that the electronic device can implement the AI-native digital twin scene construction method as described in any one of claims 1 to 3.

8. A computer storage medium, characterized in that: The storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the AI-native digital twin scene construction method as described in any one of claims 1 to 3.

Images