CN-119885874-B - Large model digital twin body system

CN119885874BCN 119885874 BCN119885874 BCN 119885874BCN-119885874-B

Abstract

The invention relates to a large-model digital twin body system, and belongs to the field of artificial intelligence or digital twin. The system comprises a man-machine interaction subsystem, a generation type large model, a generation type content processing subsystem, an instruction router and a real-time controllable rendering engine, wherein the man-machine interaction subsystem receives user input and analyzes intention to realize man-machine interaction, the generation type large model processes user voice or text input to generate instruction and non-instruction type generation content and transmits the instruction and non-instruction type generation content to the generation type content processing subsystem to analyze the instruction type content and then transmit the instruction type content and an execution result to the instruction router, the non-instruction type content and the execution result are transmitted back to the generation type large model to support multi-round question-answer interaction, and the instruction router maintains an execution relation between an instruction queue and a logic directed graph and transmits the execution result back. The method realizes the deep fusion and interaction of physical and digital spaces, and a user can intuitively communicate with the system in a context-supported multi-round dialogue mode, and can obtain instant, accurate and efficient feedback, thereby improving experience and interaction efficiency.

Inventors

QIU YUMING
YONG QIDONG
SUN QILONG

Assignees

中国科学院重庆绿色智能技术研究院

Dates

Publication Date: 20260512
Application Date: 20241230

Claims (7)

1. A large model digital twin body system, comprising: The system comprises a data integration and analysis platform, a digital mapping static three-dimensional model, real-time dynamic data, a data integration and analysis platform, a data analysis platform and a data analysis platform, wherein the digital copy of an actual physical system of a physical space is constructed in a digital space to reflect the state, performance and change of the actual physical system in real time, so that the fusion and interaction of the physical space and the digital space are realized; the real-time controllable rendering engine integrates and renders the static three-dimensional model and the dynamic data in real time to realize the visualization of the running condition of the actual physical system, receives a rendering control instruction analyzed by the instruction router, regulates and controls a rendering area, a rendering mode and rendering parameters, and sends a rendering result to a display terminal of the man-machine interaction subsystem; The instruction router receives and analyzes the instruction of the instruction issuing party and forwards the analyzed instruction to the corresponding instruction execution unit, collects the instruction execution result and transmits the instruction execution result back to the instruction issuing party, and simultaneously transmits the instruction execution result to the generated content processing subsystem for processing and then returns to the generated large model to realize the global context; The system comprises a man-machine interaction subsystem, a generation type large model and a man-machine interaction subsystem, wherein the man-machine interaction subsystem receives man-machine interaction input of a user, recognizes man-machine interaction actions, processes man-machine interaction data, analyzes interaction intention, calls a related system to realize interaction intention, and presents a man-machine interaction result; receiving the content returned by the generated content processing subsystem and transmitting the content to the man-machine interaction subsystem, and simultaneously recording the instruction execution result into a context, wherein the context is used for supporting the subsequent multi-round question-answer interaction dialogue and decision; The system comprises a generation content processing subsystem, a generation type large model, a non-instruction type generation content processing subsystem, a command router, a generation type large model and a command processing subsystem, wherein the generation content processing subsystem receives generation content from the generation type large model, directly transmits back the generation type large model for the non-instruction type generation content, analyzes the instruction type generation content, firstly transmits analysis results to the command router, then collects command execution results transmitted back by the command router, processes and transmits back to the generation type large model, and then collects command execution results transmitted back by the command router; wherein the generating content processing subsystem comprises: the generation content router analyzes the generation content of the generation type large model, and transmits the instruction type generation content to a corresponding instruction analyzer; An instruction parser for converting the instruction type generated content into an instruction set and forming an instruction queue or an instruction logic directed graph to be transmitted to the instruction router; the execution result encapsulator is used for reforming and encapsulating the execution result fed back by the instruction router and transmitting the execution result to the large generation model; And the physical control instruction distributor receives the physical control instruction analyzed by the instruction router and transmits the physical control instruction to an executor in a physical system to control a physical object, and transmits the execution result of the physical control instruction back to the instruction router.
2. The large model digital twin body system of claim 1, wherein the instruction of the instruction issuing party comprises a single instruction, an instruction queue and an instruction logic directed graph, the instruction router maintains the sequence, synchronism, asynchronism and context execution relationship of the instruction queue and the instruction logic directed graph, the instruction execution result is packaged in batches or whole batches according to instruction logic and instruction requirements and is transmitted back to the instruction issuing party, and the execution unit comprises a real-time controllable rendering engine and a physical control instruction distributor.
3. The large model digital twin body system according to any of claims 1-2, wherein the man-machine interaction subsystem is provided with a speech synthesis module for playing the man-machine interaction result.
4. The large model digital twin body system according to any of claims 1-2, further comprising a virtual simulation model for performing virtual simulation calculation according to user input, generated instructions or predefined business scenarios, and applying simulation calculation results to the static three-dimensional model, and visually displaying the results of the virtual simulation calculation to verify business scenarios in advance.
5. The large model digital twin body system as defined in claim 4, wherein the human-machine interaction subsystem supports a sub-window mode to expose results of the virtual simulation calculations.
6. The large model digital twin body system of any of claims 1-2, further comprising a business simulation component that performs business simulation calculations based on user input, generated instructions, or predefined business scenarios, and applies business simulation calculation results to the static three-dimensional model to visually simulate business critical business processes, critical facility equipment, and operational logic or operational methods to support training and maintenance.
7. The large model digital twin body system as defined in claim 6, wherein the human-machine interaction subsystem supports a sub-window mode to present the business simulation calculation results.

Description

Large model digital twin body system Technical Field The invention belongs to the field of artificial intelligence or digital twinning, and relates to a large-model digital twinning body system. Background With the continuous acceleration of digital transformation steps of the economic society, the digital twin technology is becoming an important force and increasingly becoming a hot spot technology for competitive exploration of various industries. By virtue of its unique advantages, digital twin technology has become an important tool and core driving force to drive enterprise digital transformation and promote digital economic development. However, although the digital twin technology has great potential and value, the current interaction mode is still mainly limited to the traditional and relatively single interaction modes such as mouse operation, 3D glasses and the like, the interaction mode is relatively single and has low efficiency, and the application range and the user experience of the digital twin system are limited. In recent years, the generation type artificial intelligence technology based on a large language model is prominent in the military, and the development speed and the excellent performance of the technology are rapid, so that unprecedented innovation is brought. The technology makes smooth and reliable multi-round voice dialogue possible, and greatly enriches the ways and scenes of man-machine interaction. The user can realize seamless communication with the digital twin system through natural language, thereby obtaining the needed information and service more intuitively and conveniently. However, although large language models present great advantages in terms of voice interactions, how to organically fuse together a generative large model and a digital twin system safely and reliably remains an unsolved problem. Disclosure of Invention In view of the above, it is an object of the present invention to provide a large model digital twin system. In order to achieve the above purpose, the present invention provides the following technical solutions: A large model digital twin system comprising: the data integration and analysis platform constructs a digital copy of an actual physical system of a physical space in a digital space to reflect the state, performance and change of the actual physical system in real time, so as to realize the fusion and interaction of the physical space and the digital space; the digital mapping static three-dimensional model of the physical space in the digital space; Real-time dynamic data for driving the visualization virtual operation of the static three-dimensional model, the real-time dynamic data being provided by the data integration and analysis platform; the real-time controllable rendering engine integrates and renders the static three-dimensional model and the dynamic data in real time to realize the visualization of the running condition of the actual physical system, receives a rendering control instruction analyzed by the instruction router, regulates and controls a rendering area, a rendering mode and rendering parameters, and sends a rendering result to a display terminal of the man-machine interaction subsystem; The instruction router receives and analyzes the instruction of the instruction issuing party and forwards the analyzed instruction to the corresponding instruction execution unit, collects the instruction execution result and transmits the instruction execution result back to the instruction issuing party, and simultaneously transmits the instruction execution result to the generated content processing subsystem for processing and then returns to the generated large model to realize the global context; The system comprises a man-machine interaction subsystem, a generation type large model and a man-machine interaction subsystem, wherein the man-machine interaction subsystem receives man-machine interaction input of a user, recognizes man-machine interaction actions, processes man-machine interaction data, analyzes interaction intention, calls a related system to realize interaction intention, and presents a man-machine interaction result; Receiving the content returned by the generated content processing subsystem and transmitting the content to the man-machine interaction subsystem, and simultaneously recording the returned content into a context to support subsequent multi-round question-answer interaction dialogue and decision; The system comprises a generation content processing subsystem, a generation type large model, a command router, a generation content processing subsystem, a command execution result processing subsystem and a command execution result processing subsystem, wherein the generation content processing subsystem receives generation content from the generation type large model, directly transmits back the generation type large model for the non-command generation content, analyzes the command generation con