MUTUALLY GENERATED ARTIFICIAL INTELLIGENCE SYSTEM BASED ON MULTIDIMENSIONAL SPATIO-TEMPORAL INFORMATION VECTOR GRAPHICS

Abstract

The present invention relates to a mutually generative AI system based on multidimensional spatio-temporal information vector graphics. Based on a geographic information system, a computer-aided design platform, and a data source, the present invention constructs a multidimensional vector spatio-temporal large-scale model terminal, an intelligent agent terminal for processing multidimensional spatio-temporal information, and an intelligent information system application terminal. It also pre-trains a large-scale model capable of understanding professional engineering knowledge systems, data processing, multidimensional vector graphics, and special graphical literal documents for multidimensional spatio-temporal data such as two-dimensional and three-dimensional vectors.in order to achieve the expression and mutual generation between the multidimensional vector graph and special spatio-temporal information, and the description of professional documents, and to form an intelligent data processing system application in the field of engineering. (Figure for the abbreviation: Fig.1),

Description

Title of the invention: MUTUALLY GENERATED ARTIFICIAL INTELLIGENCE SYSTEM BASED ON MULTIDIMENSIONAL SPATIO-TEMPORAL INFORMATION VECTOR GRAPHICS Technical field

[0001] The present invention relates to artificial intelligence (AI) techniques and engineering applications thereof, in particular to a mutually generative AI system based on multidimensional spatio-temporal information vector graphics. Technical background

[0002] In recent years, AI technologies, represented by large models, have developed rapidly. Large model technology, trained on substantial datasets, exhibits strong semantic understanding and inference capabilities and shows potential for solving complex problems in pronunciation, translation, medicine, education, finance, and art. Multimodal large models combine the capabilities of large models in natural language processing and understanding and generation with other modal data (such as visual, audio, etc.) and mimic the characteristics of multimodal information processing in human cognition by integrating diverse inputs and outputs such as text, images, and voices to provide a richer and more natural interactive experience.However, at present, the data processed by AI technology, including the large multimodal model, still remain in types of text, pronunciation, image, video and others, which are essentially text and matrix formats, and which mainly serve areas of literary and artistic creation, question-answering system, phonetic translation, data synthesis and analysis, etc., are not yet involved in a human engineering application domain characterized by vector format data.Humankind has been confronted with rich multimedia spatiotemporal data in the real world since birth, data which encompasses various types such as text, audio, video, drawing, and images, and complex characteristics such as numbers, texts, attributes, metadata, interactive processes, and spatiotemporal relationships; therefore, humankind possesses intelligence and even wisdom. Language. AI is merely an organic part of multimedia spatiotemporal data, and humankind primarily represents and interprets data using text, vector, and raster graphics. AI is not limited to current textual or statistical intelligence, or to computer vision-based spatial intelligence, but also includes spatiotemporal AI (STAI). In the future, the large model will evolve into a spatiotemporal large model (STLM), and the large language model (LLM) will be just a subset of the spatiotemporal large model (STLM). In an intelligent spatiotemporal processing system, a vector model will play a significant role and be widely used.

[0003] A geographic information system (GIS) is a computer system used to collect, store, manage, calculate, analyze, display, and describe the location and distribution of geospatial objects and to solve user problems. Computer-aided design (CAD) is a computer system that uses a computer and a graphics device to provide technical designers with functions such as geometric modeling, feature calculation, and computer graphics. Currently, GIS and CAD have been widely used in fields such as spatial data management, engineering design, and computer graphics, and have accumulated a large amount of computer-generated data, including electronic maps, engineering design drawings, production and construction drawings, etc.All of these are data types based on vector spatial data. Vector data is essentially an abstract data format, representing an abstract human representation of spatial objects and engineering processes. Whether in GIS or CAD, it involves vector graphics that are processed through artificial interactive participation. Furthermore, the creation of specific graphical literal documents is artificially performed based on the actual needs of the human knowledge system and engineering processes. GIS can perform automated processing due to its topological spatial relationships, whereas CAD lacks a comprehensive spatiotemporal relationship description and attribute expression, making intelligent processing difficult.In general, vector graphics and the processing of special graphic engineering documents related to GIS and CAD involve intelligent one-way mutual generation, not duplex; therefore, intelligent automated processing is not possible.

[0004] Large model technology allows for the mutual generation of the vector graphic and the graphic literal document, but at present, research or application on the large model still focuses on textual data or Raster graphics are commonly used in applications such as text-to-raster video conversion, but rarely in modeling vector data such as points, lines, surfaces, volumes, and networks. Therefore, given that a large number of multidimensional vector graphics and special graphical literal documents require processing within the field of human engineering, there is an urgent need to develop a system for understanding, analyzing, and processing multidimensional spatiotemporal information vector graphics and for generating the corresponding graphical literal document. Description of the invention

[0005] In view of the above problem, the invention proposes a mutually generative AI system based on multidimensional spatio-temporal information vector graphics.

[0006] The present invention implements a mutually generative AI system based on multidimensional spatio-temporal information vector graphics, which includes a multidimensional vector spatio-temporal large model terminal, an intelligent agent terminal for processing multidimensional spatio-temporal information and an intelligent information system application terminal.

[0007] The multidimensional vector spatiotemporal large model terminal is used to fully model and express spatiotemporal data of various types and to construct a multimodal vector spatiotemporal large model. The multimodal vector spatiotemporal large model is used to mutually express, understand, and analyze multidimensional vector spatiotemporal data and natural language descriptions, including texts and voices, in order to achieve learning reinforcement of vertical domain knowledge within a professional engineering knowledge system, and then to enhance the processing capacity of the engineering domain knowledge and the data processing process. The spatiotemporal data of various types includes texts, voices, images, videos, and multidimensional vector graphics.

[0008] The intelligent agent terminal for processing multidimensional spatio-temporal information is used to automatically analyze operational process tasks involving the mutual processing of multidimensional spatio-temporal data of various types and to break down the tasks into simple subtasks, and to autonomously execute, transmit, and resolve the subtasks based on the learning and inference capabilities of the large vector spatio-temporal model, in combination with the processing and analysis capabilities of a geographic information system (GIS) or a computer-aided design platform. computer (CAD) and by learning and referring to a task execution procedure of a typical business process.

[0009] The intelligent information system application terminal is used to transform multidimensional spatio-temporal data of various types into corresponding natural language descriptions based on the deductive and comprehension capabilities of the large vector spatio-temporal model, in order to achieve AI understanding of multidimensional vector graphics and multidimensional spatio-temporal data of various types, to achieve interaction, analysis, and mutual generation of multidimensional vector graphics and special graphical literal documents in the field of engineering driven by an intelligent agent by means of simple inputs including voice and text, and to accurately obtain parameters in multidimensional spatio-temporal data, automatically renew multidimensional spatio-temporal data linked to parameters,then to automatically generate and renew the multidimensional vector graphic or the special graphic literal document using the deduction of the large vector spatio-temporal model and the processing of the intelligent agent.

[0010] According to a preferred embodiment, industry information from a professional engineering knowledge system is embedded in the terminal of a large multidimensional vector spatio-temporal model, in order to realize learning in reinforcement of the vertical domain knowledge of a professional engineering knowledge system, and then to strengthen the processing and understanding capacity of the engineering domain knowledge and the data processing process.

[0011] A procedure of the large vector space-time model constructed by the multidimensional large vector space-time model terminal includes: the steps of describing in natural language the multidimensional vector space-time data, of pre-training and precisely fitting the large vector space-time model, and of understanding and analyzing multidimensional vector space-time data.

[0012] The fact that industry information from a professional engineering knowledge system is embedded in the terminal of a large, multidimensional, vector-based spatio-temporal model, in order to achieve learning in reinforcement of the vertical domain knowledge of a professional engineering knowledge system, includes: the steps of using engineering domain data as a pre-training dataset to precisely fit a large general model; and of pre-training the large general model by expanding a vertical engineering domain vocabulary and using an engineering domain dataset in order to to embed industry information from a professional engineering knowledge system into the larger overall model.

[0013] The fact that industry information from a professional engineering knowledge system is embedded in the terminal of a large multidimensional vector spatio-temporal model, in order to enhance the processing and understanding capacity of the engineering domain knowledge and the data processing process, includes: the steps of describing a data processing process with clear engineering characteristics; of defining a data processing language of the engineering domain; and of using the engineering domain data processing process which is fully expressed as a precisely fitted dataset to train the large general model, in order to achieve the embedding of knowledge of the data processing process, and to enhance the processing and understanding capacity of the engineering domain knowledge and the data processing process.

[0014] Describing said multidimensional vector spatio-temporal data in natural language includes the step of transforming vector spatio-temporal data into various types of points, lines, surfaces, and volumes with continuous spatial-temporal information (x, y, z, t) and attribute information, including, but not limited to: Geographic Information Systems (GIS) and Computer-Aided Design (CAD) according to the training, understanding, and processing needs of said large vector spatio-temporal model, and making the transformed natural language describe all the information with the original vector spatio-temporal data, including, but not limited to: geometry types, geometric features, display styles, reference point coordinates,Relative geometric data based on the reference point, temporal information, and attribute information of a vector object.

[0015] Pre-training and precisely fitting the large vector spatio-temporal model includes: the steps of establishing a dataset that includes the natural language description of the multidimensional vector spatio-temporal data and features of the vector object, and continuously pre-training and precisely fitting the large general model on the basis of the pre-trained dataset and the precisely fitted dataset, in order to form a large vector spatio-temporal model with the ability to understand vector spatio-temporal data.

[0016] Understanding and analyzing multidimensional vector spatio-temporal data includes: the steps of inputting vector spatio-temporal data described by natural language into the large vector spatio-temporal model; to output features of the vector object in understanding from the large vector space-time model; to input a part of the features of the vector object into the large vector space-time model; to output complete vector space-time data in understanding or analysis from the large vector space-time model; and further to accomplish understanding and processing geometric features and attribute features of vector space-time data and to identify and process spatial and space-time relationships of vector space-time data on the basis of the space-time features or data output from the vector object and using the analysis capability of the large vector space-time model.

[0017] In which, understanding and processing geometric features of vector spatio-temporal data includes the step of processing the geometric coordinates of the vector object itself, including, but not limited to, modifying and compiling a shape, size, and position.

[0018] Understanding and processing attribute features of vector spatio-temporal data includes the step of processing attribute information of the vector object itself, including, but not limited to, modifying, querying, analyzing and counting it.

[0019] Identifying and processing spatio-temporal relationships of vector spatio-temporal data includes the step of identifying and processing topological spatial relationships, sequential spatial relationships, and metric spatial relationships between vector objects, where the topological spatial relationship refers to a relationship of association, proximity, inclusion, intersection, overlap, and separation between spatial objects; the sequential spatial relationship refers to an order in which spatial objects or events are arranged in space, including orientation relationships such as front-back, left-right, top-bottom, and east-west-south-north; the metric spatial relationship refers to a relationship of distance or proximity between spatial objects.

[0020] According to a preferred embodiment, the intelligent agent terminal for processing multidimensional spatio-temporal information comprises a task planning module, a task memory module and a task action module, where the task planning module, the task memory module and the task action module respectively represent a corresponding processor.

[0021] The task planning module is used to solve and plan a task of an interactive operational process to process vector graphics included in multidimensional spatio-temporal data processing of various types based on domain knowledge, the operational process, data understanding and the text generation capability of the large model vector spatio-temporal, and each subtask is a general logical processing performed independently or a geospatial analysis operation, and has explicit inputs and outputs; the subtasks interact to complete a complex operational process processing.

[0022] The task memory module is used to make each task sense information from an environment or obtain information from memory in order to provide data required by each task and to store data of running processes; the intelligent agent to process multidimensional spatio-temporal information accumulates data and experience and progressively completes a self-evolution by means of the task memory module, in order to provide iterative capacity support to the large vector spatio-temporal model.

[0023] The task action module is used to execute each subtask planned by the large vector spatio-temporal model to a concrete result; the task is executed using the capabilities, including but not limited to, inference of the large vector spatio-temporal model, spatial analysis and processing of the geographic information system; and processing of the computer-aided mapping system; the task interacts with interactive objects, including but not limited to sensors, model libraries, controllers and databases.

[0024] According to a preferred embodiment, a procedure by which the application terminal of an intelligent information system performs AI understanding of multidimensional vector graphics and multidimensional spatiotemporal data of various types, and the interaction, analysis, and mutual generation of the multidimensional vector graphic and the special graphic literal document, includes the steps of processing and understanding the multidimensional spatiotemporal data, generating the intelligent agent to process multidimensional spatiotemporal information, and mutually generating the multidimensional vector graphic and the special graphic literal document. More specifically, it includes the steps:

[0025] (SI) To input multidimensional spatio-temporal data into the application terminal of an intelligent information system, and to transform the multidimensional spatio-temporal data into natural language descriptions that can be processed by the large vector spatio-temporal model, in order to perform AI-based analysis and understanding of multidimensional spatio-temporal information vector data and temporal data within the multidimensional spatio-temporal data based on the inference and understanding capabilities of the large vector spatio-temporal model.

[0026] (S2) To enter user instructions by interactive means, including of the phonic input and the text input, in which the phonic input is converted into text by phonic recognition; and to transform user instruction inputs described by natural language into formatted instructions to compile, query, analyze and output data from the information system based on the understanding, analysis and processing capabilities of the large vector spatio-temporal model, in order to generate the intelligent agent to process multidimensional spatio-temporal information.

[0027] (S3) To perform the automatic generation of vector graphics multidimensional and special graphic literal documents by automatic and intelligent means, and bidirectionally renew AI applications in mutual generation based on the intelligent agent to process multidimensional spatio-temporal information and using its multi-round merge and execution call, in order to mutually generate multidimensional vector graphics and multidimensional spatio-temporal data in various types.

[0028] According to a preferred embodiment, establishing a dataset that includes a natural language description of the multidimensional vector spatiotemporal data and features of the vector object, and continuously pre-training and precisely fitting the large general model on the basis of the pre-trained dataset and the precisely fitted dataset includes the steps:

[0029] (Tl) To collect massive vector spatio-temporal data and natural language text descriptions and vector object characteristic to the corresponding vector spatio-temporal data, in order to form a vector spatio-temporal data training dataset, in which the vector spatio-temporal data include predefined geometric structure information of points, lines, surfaces and volumes, and attribute text information that corresponds to a geometric solid.

[0030] (T2) To use the large general model as a basis for pre-training using the training dataset of vector spatiotemporal data, to sample vector text using random walk as training sample; and to input them into the large general model in the form of generative GPT and fine-tune the model to achieve efficient knowledge injection into vector information.

[0031] (T3) To collect data in various question-and-answer formats, including vector data - topological relation, vector data - attribute information, and attribute description - vector data, to form a dataset of precisely fitted vector spatio-temporal model, and of precisely fitting the large general model in pre-training in the T2 stage to achieve efficient alignment between natural languages ​​and vector data.

[0032] According to a preferred embodiment, the task planning module searches for similar tasks in the engineering process library based on a task description. The task planning module includes methods of task planning without information feedback and task planning with information feedback, in which the feedback comes from the environment, the user, or the execution result of the large vector space-time model, and the operational process resolves into several subtasks to execute them, which are linked together in a cascade or tree structure, and each of which determines a subsequent subtask based on an execution result performed at each time.

[0033] According to a preferred embodiment, the task memory module stores information sensed from the environment, task execution records, and task execution results, and uses stored memories to promote future actions; the task memory module includes stored data and operation.

[0034] The stored data includes information stored in short term and entered in a pop-up window and information stored in long term in an external vector that is quickly sought and queried.

[0035] In a preferred embodiment, the task action module uses an instrument to perform and complete each task, and the instrument includes the large vector spatiotemporal model itself and an external instrument that includes an algorithmic model, a program compilation, a database, and an API. The task action module further uses a spatial database to perform partial spatial analysis, uses a model in a model database to perform a specific task, uses an API to obtain real-time or historical data, and the specific task refers to a task that corresponds to each model in the model database.

[0036] According to a preferred embodiment, automatically generating the multidimensional vector graphic or the special graphic literal document includes the steps of:

[0037] Artificially forming a model of the multidimensional vector graphic or special graphic literal document according to a content requirement of the multidimensional vector graphic or special graphic literal document, or making the large vector spatio-temporal model automatically generate a model of the special graphic literal document according to a content requirement, wherein the drawing of the special graphic literal document is in vector graphic format multidimensional, or in matrix format converted to multidimensional vector graphics; and

[0038] Generate parameters of various types in the multidimensional vector graph or the special graphic literal document model, accurately obtain parameters in the multidimensional vector spatio-temporal data, and merge the parameters with the models to generate the multidimensional vector graph or the special graphic literal document, by means of information and analysis by an information system or deduction and understanding by the large vector spatio-temporal model.

[0039] According to a preferred embodiment, the mutual generation of the multidimensional vector graph and the special graphic literal document is based on the analysis and understanding capabilities of the large vector spatio-temporal model; after data is entered into the engineering domain or after the special graphic literal document makes relevant spatio-temporal information change, the large vector spatio-temporal model generates a renewed parameter, description or intelligent agent for the multidimensional vector graph, and automatically renews the data of the multidimensional vector graph.

[0040] After drawing or modifying the multidimensional vector graphic causes the multidimensional vector graphic to change, the updated multidimensional vector graphic is obtained through the large vector spatio-temporal model, and corresponding engineering parameters or document description are then updated; which allows for updating content relating to said special graphic literal document.

[0041] Based on a geographic information system (GIS), a computer-aided design (CAD) platform, and a data source, the present invention constructs a large multidimensional vector spatio-temporal model terminal, an intelligent agent terminal for processing multidimensional spatio-temporal information, and an intelligent information system application terminal. It also pre-trains a large model capable of understanding the professional engineering knowledge system, the data processing process, the multidimensional vector graph, and the special graphic literal document for multidimensional spatio-temporal data such as two-dimensional and three-dimensional vectors, in order to achieve the expression and mutual generation between the multidimensional vector graph and special spatio-temporal information, and the description of professional documents.and form an application for an intelligent data processing system in the field of engineering. The present invention will completely liberate, Engineering and technical people, by establishing an AI system in the field of human social engineering, will provide intelligent system support to automatically and rapidly process voice, text, multidimensional vector graphics, and special graphic literal documents in related fields, and has great practicality. Brief description of the figures

[0042] The present invention will be described in more detail below with the aid of the examples of embodiments shown schematically in the figures, so that those skilled in the art may clearly understand a variety of other advantages and benefits. The drawings are used only to show the preferred embodiment and are not considered a limitation of the present invention. In the figures, the same elements, features, and components, which are functionally identical and have the same effect, are each designated by the same reference numerals, unless otherwise indicated. These illustrate:

[0043] Fig. 1 shows a diagram of the mutually generative AI system of the multidimensional spatio-temporal information vector graph in this embodiment.

[0044] Fig. 2 illustrates a flow diagram of the intelligent information system application terminal in the present embodiment based on the large vector spatio-temporal model.

[0045] Fig. 3 illustrates a flow diagram of the large vector spatio-temporal model in the present embodiment which realizes the interaction, analysis and mutual generation of the multidimensional vector graph and the special graphic literal document.

[0046] Figure 4 illustrates another scheme of the mutually generative AI system based on the multidimensional spatio-temporal information vector graph according to the present invention. Detailed description of the implementation methods

[0047] In order to make the aforementioned objective, features, and advantages of the present invention more evident and easier to understand, we describe the present invention in more detail below in conjunction with the drawings and embodiments. It should be understood that the embodiments described herein are used only to explain the present invention and are only a part of the embodiments of the present invention, nor do they include all of the embodiments or limit the present invention.

[0048] With reference to [Fig. 1], a diagram of the mutually generative AI system based on a multidimensional spatio-temporal information vector graph is illustrated. In this embodiment, the mutually generative AI system includes a terminal for a large, multidimensional, vector-based spatiotemporal model, a terminal for an intelligent agent to process multidimensional spatiotemporal information, and a terminal for an intelligent information system application. The intelligent agent is a small, independent module or application capable of performing a single task. The information-processing agent is a small information-processing module.

[0049] The multidimensional vector spatio-temporal large-scale model terminal is used to fully model and express spatio-temporal data of various types; that is, to fully model and express spatio-temporal data in different types. Spatio-temporal data of various types includes text, voice, images, videos, and multidimensional vector graphics.

[0050] The multidimensional vector spatio-temporal large model terminal is used to construct a multimodal vector spatio-temporal large model, which is used to mutually express, understand and analyze multidimensional vector spatio-temporal data and natural language description, including texts and voices, in order to achieve learning in reinforcement of vertical domain knowledge of a professional engineering knowledge system, and then to strengthen the processing capacity to the knowledge of the engineering domain and the data processing process.

[0051] Preferably, industry information from a professional engineering knowledge system is embedded in the terminal of a large multidimensional vector space-time model, in order to realize learning in reinforcement of the vertical domain knowledge of a professional engineering knowledge system, and then to strengthen the processing and understanding capacity of the engineering domain knowledge and the data processing process.

[0052] A procedure for the large space-time vector model constructed by the multidimensional space-time vector terminal includes: the steps of describing the multidimensional space-time vector data in natural language, pre-training and precisely fitting the large space-time vector model, and understanding and analyzing the multidimensional space-time vector data. To better express the structure of the mutually generative AI system, [Fig. 1] exemplifies elements used from the large space-time vector model constructed by the multidimensional space-time vector terminal, which does not mean that the multidimensional space-time vector terminal is used solely for constructing the large space-time vector model. The elements used from the large space-time vector model constructed by the terminal are shown below. multidimensional vector space-time model includes the natural language description of multidimensional vector space-time data (describe multidimensional vector space-time data in natural language), pre-training and fine fitting to the large vector space-time model (pre-train and fine fitting the large vector space-time model), and understanding and analysis of multidimensional vector space-time data (understand and analyze multidimensional vector space-time data).

[0053] In which, the fact that industry information from a professional engineering knowledge system is embedded in the terminal of a large multidimensional vector spatio-temporal model, in order to achieve learning in reinforcement of the vertical domain knowledge of a professional engineering knowledge system, comprises: the steps of using engineering domain data as a pre-training dataset to precisely fit a large general model; and of pre-training the large general model by expanding a vertical engineering domain vocabulary and using an engineering domain dataset in order to embed industry information from a professional engineering knowledge system into the large general model.The grand general model, also known as the "foundational grand model," refers to a large language model that can be applied across domains and tasks, and it is generally trained and provided by large institutions that prioritize the model's general capability and generalizability. Consequently, the spatiotemporal grand model discussed in this article is a vertical domain grand model that strengthens capability in a particular subject area based on the grand general model.

[0054] The fact that industry information from a professional engineering knowledge system is embedded in the terminal of a large multidimensional vector spatio-temporal model, in order to enhance the processing and understanding capacity of the engineering domain knowledge and the data processing process, includes: the steps of describing a data processing process with clear engineering characteristics; of defining a data processing language of the engineering domain; and of using the engineering domain data processing process which is fully expressed as a precisely fitted dataset to train the large general model, in order to achieve the embedding of knowledge of the data processing process, and to enhance the processing and understanding capacity of the engineering domain knowledge and the data processing process.

[0055] The natural language description of multidimensional vector spatio-temporal data includes, but is not limited to, vector spatio-temporal data in various types of points, lines, surfaces and volumes with continuous temporal spatial information (x, y, z, t) and attribute information, which include geographic information systems (GIS) and computer-aided mapping (CAD);A natural language transforms into descriptive material according to the needs of training, understanding, and processing the large vector spatiotemporal model, and the transformed natural language describes all information with original vector spatiotemporal data, which includes, but is not limited to, geometry types, geometric features, display styles, reference point coordinates, relative geometric data based on the reference point, temporal information, and attribute information of a vector object.

[0056] Pre-training and precisely fitting the large vector spatio-temporal model includes: the steps of establishing a dataset that includes the natural language description of the multidimensional vector spatio-temporal data and features of the vector object, and continuously pre-training and precisely fitting the large general model on the basis of the pre-trained dataset and the precisely fitted dataset, in order to form a large vector spatio-temporal model with the ability to understand vector spatio-temporal data.

[0057] Preferably, establishing a dataset that includes a natural language description of the multidimensional vector spatiotemporal data and features of the vector object, and continuously pre-training and precisely fitting the large general model on the basis of the pre-trained dataset and the precisely fitted dataset includes the following steps:

[0058] (Tl) To collect massive vector spatio-temporal data and natural language text descriptions and vector object feature to the corresponding vector spatio-temporal data, in order to form a vector spatio-temporal data training dataset, wherein the vector spatio-temporal data include predefined geometric structure information of points, lines, surfaces and volumes, and attribute text information that corresponds to a geometric solid.

[0059] (T2) To use the large general model as a basis for pre-training using the training dataset of vector spatiotemporal data, to sample vector text using a random walk as the training sample; and to input them into the large general model in the form of generative GPT and precisely adjust the model in order to achieve efficient knowledge injection into vector information.

[0060] (T3) To collect data in various question-and-answer formats, including vector-relation topological data, vector-information attribute data and attribute-data vector description, to form a precisely fitted vector spatio-temporal model dataset, and to precisely fit the large general model in pre-training in the T2 stage to achieve efficient alignment between natural languages ​​and vector data.

[0061] Thanks to the three-step method above, the terminal of the large multidimensional vector spatio-temporal model realizes how to construct the large vector spatio-temporal model.

[0062] Understanding and analyzing multidimensional vector spatio-temporal data includes: the steps of inputting vector spatio-temporal data described by natural language into the large vector spatio-temporal model; of outputting features of the vector object in understanding from the large vector spatio-temporal model; of inputting a part of the features of the vector object into the large vector spatio-temporal model; of outputting complete vector spatio-temporal data in understanding or analysis from the large vector spatio-temporal model;and further to accomplish understanding and processing geometric and attribute characteristics of vector spatio-temporal data and to identify and process spatial and spatio-temporal relationships of vector spatio-temporal data based on the spatio-temporal characteristics or data output from the vector object and using the analytical capability of the large vector spatio-temporal model. ;

[0063] Wherein, understanding and processing the geometric characteristics of vector spatiotemporal data includes the step of processing the geometric coordinates of the vector object itself, including, but not limited to, modifying and compiling its shape, size, and position. Understanding and processing the attribute characteristics of vector spatiotemporal data includes the step of processing attribute information of the vector object itself, including, but not limited to, modifying, querying, analyzing, and counting it.Identifying and processing spatiotemporal relationships of vector spatiotemporal data includes the step of identifying and processing topological spatial relationships, sequential spatial relationships, and metric spatial relationships between vector objects, where the topological spatial relationship refers to a relationship of association, neighborhood, inclusion, intersection, overlap, and separation between spatial objects; the sequential spatial relationship refers to an order in which spatial objects or events are arranged in space. including orientation relationships such as front-back, left-right, up-down, and east-west-south-north; the metric spatial relationship refers to a relationship of distance or proximity between spatial objects.

[0064] The intelligent agent terminal for processing multidimensional spatio-temporal information is used to automatically analyze operational process tasks, mutually process multidimensional spatio-temporal data of various types, and break down tasks into simple subtasks. It is also used to autonomously execute, transmit, and resolve subtasks based on the learning and inference capabilities of the large vector spatio-temporal model, in combination with the processing and analysis capabilities of a geographic information system (GIS) or a computer-aided design (CAD) platform. This involves learning and referring to a task execution procedure of a typical operational process, which refers to a process or step used to process the spatio-temporal information data mentioned in this article. These are the data processing methods related to GIS and CAD.such as processes for drawing and compiling drawings, adding attributes, and analyzing spaces.

[0065] Preferably, the intelligent agent terminal for processing multidimensional spatio-temporal information includes a task planning module, a task memory module and a task action module.

[0066] The task planning module is used to solve and plan a task of an interactive operational process to process vector graphics included in multidimensional spatio-temporal data processing of various types based on domain knowledge, the operational process, data understanding, and the text generation capability of the large vector spatio-temporal model. Each subtask is a general logical process performed independently or a geospatial analysis operation, and has explicit inputs and outputs. The subtasks interact to complete a complex operational process. As a concrete example, the task planning module searches for similar tasks in the engineering process library based on a task description.The task planning module includes methods of task planning without information feedback and task planning with information feedback, in which the feedback comes from the environment, the user or the execution result of the large vector spatio-temporal model, and the operational process resolves into several subtasks to execute them, which link together in a cascade or tree structure, and each of which determines a subsequent subtask based on an execution result made each time.

[0067] The task memory module is used to make each task sense information from an environment or obtain information from memory in order to provide data required by each task and to store data of running processes; the intelligent agent to process multidimensional spatio-temporal information accumulates data and experience and progressively completes a self-evolution by means of the task memory module, in order to provide iterative capacity support to the large vector spatio-temporal model.As a concrete example, the task memory module stores information sensed from the environment, task execution records, and task execution results, and uses stored memories to promote future actions; the task memory module includes stored data and an operation; in which the stored data includes short-term stored information entered into a pop-up window and long-term stored information in an external vector that is quickly sought and queried.

[0068] The task action module is used to execute each subtask planned by the large vector space-time model to a concrete result; the task is executed using the capabilities of, including but not limited to, the large vector space-time model's inference, spatial analysis, and geographic information system processing capabilities; and the computer-aided mapping system processing capabilities; the task interacts with interactive objects, including but not limited to sensors, model libraries, controllers, and databases. As a concrete embodiment, the task action module uses an instrument to perform and complete each task, and the instrument includes the large vector space-time model itself and an external instrument that includes an algorithmic model, program compilation, a database, and an API.The task action module further uses a spatial database to perform partial spatial analysis, uses a model in a model database to perform a specific task, uses an API to obtain real-time or historical data, and the specific task refers to a task that corresponds to each model in the model database.

[0069] To better express the structure of the mutually generative AI system, [Fig. 1] exemplifies the three modules of the intelligent agent terminal for processing multidimensional spatiotemporal information, which does not mean that the intelligent agent terminal for processing multidimensional spatiotemporal information includes only these three modules. The three modules of the intelligent agent terminal for processing multidimensional spatiotemporal information consist of task scheduling (a task scheduling module), memory task-based (a task-based memory module) and task-based action-based (a task-based action module).

[0070] The intelligent information system application terminal is used to transform multidimensional spatio-temporal data of various types into corresponding natural language descriptions based on the deductive and comprehension capabilities of the large vector spatio-temporal model, in order to achieve AI understanding of multidimensional vector graphics and multidimensional spatio-temporal data of various types, to achieve interaction, analysis, and mutual generation of multidimensional vector graphics and special graphical literal documents in the field of engineering driven by an intelligent agent by means of simple inputs including voice and text, and to precisely obtain parameters in multidimensional spatio-temporal data, automatically renew multidimensional spatio-temporal data linked to parameters,then to automatically generate and renew the multidimensional vector graphic or the special graphic literal document using the deduction of the large vector spatio-temporal model and the processing of the intelligent agent. In the embodiment of the present invention, the multidimensional vector graphic refers to two-dimensional and three-dimensional vector data plus temporal data that are applicable in the field of engineering; the special graphic literal document refers to a report or document with a mixture of graphics and text that has a specific engineering meaning or application.

[0071] Preferably, a procedure by which the intelligent information system application terminal performs AI understanding of multidimensional vector graphics and multidimensional spatiotemporal data of various types, and interaction, analysis, and mutual generation of the multidimensional vector graphics and the special graphical literal document, includes the steps of processing and understanding the multidimensional spatiotemporal data, generating the intelligent agent for processing multidimensional spatiotemporal information, and mutually generating the multidimensional vector graphics and the special graphical literal document. The intelligent agent for processing multidimensional spatiotemporal information refers to an intelligent agent that processes multidimensional spatiotemporal data, for example, reads, modifies, and analyzes it. More specifically, it includes the steps:

[0072] (SI) To input multidimensional spatio-temporal data into the application terminal of an intelligent information system, and to transform the multidimensional spatio-temporal data into natural language descriptions that can be processed by the large vector spatio-temporal model, in order to perform analysis and to understand in an AI-like way multidimensional spatio-temporal information vector data and temporal data within multidimensional spatio-temporal data based on the ability to deduce and understand the large vector spatio-temporal model.

[0073] (S2) To enter user instructions by interactive means, including of the phonic input and the text input, in which the phonic input is converted into text by phonic recognition; and to transform user instruction inputs described by natural language into formatted instructions to compile, query, analyze and output data from the information system based on the understanding, analysis and processing capabilities of the large vector spatio-temporal model, in order to generate the intelligent agent to process multidimensional spatio-temporal information.

[0074] (S3) To perform the automatic generation of vector graphics multidimensional or special graphic literal documents by automatic and intelligent means, and bidirectionally renew AI applications in mutual generation based on the intelligent agent to process multidimensional spatio-temporal information and using its multi-round merge and execution call, in order to mutually generate multidimensional vector graphics and multidimensional spatio-temporal data in various types.

[0075] In order to better express the structure of the mutually generative AI system, [Fig.1] exemplifies the understanding of AI in multidimensional vector graphics and multidimensional spatio-temporal data of various types achieved by the intelligent information system application terminal, and elements of interaction, analysis and mutual generation in multidimensional vector graphics and special graphic literal document, which does not mean that the intelligent information system application terminal only complements these technical contents.The elements consist of processing and understanding multidimensional spatio-temporal data (analyzing and understanding multidimensional spatio-temporal information vector data and temporal data within multidimensional spatio-temporal data using AI), generating the intelligent agent to process multidimensional spatio-temporal information (generating the intelligent agent to process multidimensional spatio-temporal information), and mutual generation between the multidimensional spatio-temporal information vector graph and the special graphical literal document (mutually generating the multidimensional spatio-temporal information vector graph and the special graphical literal document).

[0076] The process of automatically generating the multidimensional vector graphic or the special graphic literal document includes the steps of:

[0077] First, artificially form a model of the multidimensional vector graphic or special graphic literal document according to a content requirement of the multidimensional vector graphic or special graphic literal document, or make the large vector spatio-temporal model automatically generate a model of the special graphic literal document according to a content requirement, in which the graphic of the special graphic literal document is in multidimensional vector graphic format, or in matrix format converted to multidimensional vector graphics;next, generate parameters of various types in the multidimensional vector graph or the special graphic literal document model, accurately obtain parameters in the multidimensional vector spatio-temporal data, and merge the parameters with the models to generate the multidimensional vector graph or the special graphic literal document, by means of information gathering and analysis through an information system or deduction and understanding through the large vector spatio-temporal model.

[0078] The mutual generation of the multidimensional vector graph and the special graphic literal document is based on the analysis and understanding capabilities of the large vector spatio-temporal model; after data is entered into the engineering domain or after the special graphic literal document makes relevant spatio-temporal information change, the large vector spatio-temporal model generates a renewed parameter, description or intelligent agent for the multidimensional vector graph, and automatically renews the data of the multidimensional vector graph.

[0079] After drawing or modifying the multidimensional vector graphic causes the multidimensional vector graphic to change, the updated multidimensional vector graphic is obtained through the large vector spatio-temporal model, and corresponding engineering parameters or document description are then updated; which allows for updating content relating to said special graphic literal document.

[0080] Figure 2 illustrates a flow diagram of the intelligent information system application terminal in the present embodiment based on the large vector space-time model, which performs mutual expression and generation between natural languages ​​such as speech input and text input and data such as the multidimensional vector graph and the special graphic literal document based on the large vector space-time model. "The vector graph and temporal analysis and understanding," "the vector graph and the temporal application instruction," and "the intelligent information-processing agent" in the [Fig.2] can be understood to correspond to the previous steps S1-S3, and will not be repeated one by one.

[0081] Figure 3 illustrates a flow diagram of the large vector space-time model in the present embodiment, which performs the interaction, analysis, and mutual generation of the multidimensional vector graph and the special graphic literal document. Based on the understanding and analysis capabilities of the large vector space-time model, multidimensional space-time data of various types, such as engineering data, graphic literal documents, and other sources of space-time data with vector graphics of various types (two-dimensional drawings, three-dimensional drawings, and other types of vectors) and data (vector graph parameters, vector time parameters, vector graph descriptions, and intelligent vector graph agents), are generated and renewed mutually.The intelligent vector graphics agent refers to a small module for processing (reading, modifying, storing) vector graphics.

[0082] In summary, based on the geographic information system (GIS), the computer-aided design (CAD) platform, and a data source, the present invention constructs the multidimensional vector spatio-temporal large model terminal, the intelligent agent terminal for processing multidimensional spatio-temporal information, and the intelligent information system application terminal, and pre-trains a large model capable of understanding the professional engineering knowledge system, the data processing, the multidimensional vector graph, and the special graphic literal document for multidimensional spatio-temporal data such as two-dimensional and three-dimensional vectors plus temporal, in order to achieve the expression and mutual generation between the multidimensional vector graph and special spatio-temporal information, and the description of professional documents,and to form an application for an intelligent data processing system in the field of engineering. The present invention will completely liberate engineering and technical personnel by establishing an AI system in the field of human social engineering, will provide intelligent system support for automatically and rapidly processing voice, text, multidimensional vector graphics, and special graphic literal documents in related fields, and has high practicality.

[0083] The above examples of the mutually generative AI system based on a multidimensional spatio-temporal information vector graph are only schematic. The units described therein as descriptions of detached parts may or may not be physically separated, and the parts displayed as units These modules may or may not be physical units; that is, they may be located in a single place or distributed across multiple network units. It is possible to select some or all of the modules according to the actual needs for achieving the objective of this embodiment. A person skilled in the art can understand and implement it without creative effort.

[0084] The embodiments of each component of the present invention can be implemented in hardware or in software modules running on one or more processors, or in a combination thereof. Those skilled in the art should understand that a microprocessor or a digital signal processor (DSP) can be used in practice to perform some or all of the functions of some or all of the components of an electronic device according to the embodiments of the present invention. The present invention can also be implemented as a device or apparatus (for example, a computer program and a computer program product) used to perform all or part of the methods described herein. Such a program implemented in the present invention can be stored on a computer-readable medium, or can be in the form of one or more signals.Such a signal can be downloaded from a website, or provided on a carrier signal, or in any other form.

[0085] For example, [Fig. 4] illustrates the mutually generative AI system based on a multidimensional spatiotemporal information vector graph that can be implemented according to the present invention. The mutually generative AI system based on a multidimensional spatiotemporal information vector graph comprises a processor 1010 and a computer program in the form of memory 1020 or a computer-readable medium. The memory 1020 can be electronic memory such as flash memory, an EEPROM (electrically erasable read-only programmable memory), an EPROM, a hard disk drive, or a ROM. The memory 1020 includes a storage space 1030 containing program code 1031 used to execute one of the steps of the method above. For example, the 1030 program code storage space may include 1031 program code of various types which is used to implement the different steps of the above method.These program codes can be read from one or more computer program products, or written to one or more of these computer program products, which include program code storage media such as hard drives, compact discs (CDs), memory cards, or floppy disks. These products are typically portable or fixed storage units, which may have storage segments, storage spaces, etc., similar to this arrangement in memory 1020 in the mutually generative AI system based on a multidimensional spatiotemporal information vector graph in [Fig. 4]. The program code can, for example, be compressed by appropriate means. As a rule, the memory unit includes computer-readable code, that is, code that can be read, for example, by the 1010 processor; the fact that this code is executed by an electronic device causes that electronic device to execute each step of the method described above.

[0086] It should be noted that the content relating to the mutually generative AI system based on the multidimensional spatio-temporal information vector graph may refer to the above embodiments, and will not be repeated here.

[0087] In the preceding detailed description, various features aimed at improving the rigor of the presentation were grouped into one or more examples. It should be clarified, however, that the above description is purely illustrative and is in no way intended to be restrictive. It serves to cover all variants, modifications, and equivalents of the different features and implementation examples. Those skilled in the art, by virtue of their technical knowledge, will clearly see many other examples immediately and directly arising from the above description.

[0088] Finally, it should also be noted that the herein-included relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, but they neither require nor necessarily suggest any such actual relation or sequence between these entities or operations. Furthermore, the terms "comprising" and "possessing," or any variant thereof, are intended to cover non-exclusive inclusion, such that a process, method, object, or terminal device including a series of elements includes not only those elements but also other elements not expressly listed, or a process, method, object, or terminal device further includes the elements that are inherent therein. In the absence of other restrictions, the elements defined by the phrase "including a..."» do not exclude the existence of other identical elements in the process, method, object or terminal equipment which comprises the elements.

[0089] Embodiments of the present invention have been described above in conjunction with the drawings, but the present invention is not limited to the specific embodiments described above, and the specific embodiments are merely indicative and not restrictive. Inspired by the present invention, those skilled in the art may also implement numerous forms without departing from the object of the present invention and the scope of protection of the claims, all of which fall within the protection of the present invention.

Claims

1. Demands A mutually generative AI system based on multidimensional spatiotemporal information vector graphics, comprising: a memory (1020) which stores an instruction, and a processor (1010) which executes said instruction, characterized in that said processor is configured so as: to fully express spatio-temporal data in various types and to construct a large multimodal vector spatio-temporal model which is used to understand and analyze a natural language description of multidimensional vector spatio-temporal data so as to enhance a processing capacity to knowledge of an engineering domain and to a data processing process, where the type of said spatio-temporal data includes texts, voices, images, videos and multidimensional vector graphics, said natural language description of multidimensional vector spatio-temporal data includes texts and voices; to automatically analyze an operational process task, to mutually process said spatio-temporal data of various types and to decompose said operational process task into simple subtasks, and to autonomously execute, transmit and resolve said subtasks based on a learning and inference capability of said large vector spatio-temporal model, in combination with a processing and analysis capability of a geographic information system (GIS) or a computer-aided design (CAD) platform and by learning and referring to a task execution procedure of a typical operational process; and to execute the steps of transforming multidimensional spatiotemporal data of various types into corresponding natural language descriptions based on a capacity for deduction and understanding of said large vector spatiotemporal model, in order to achieve interaction, analysis, and mutual generation to a multidimensional vector graph and a special graphical literal document in the field of engineering driven by an intelligent agent by means of simple inputs including voice and text, and to accurately obtain parameters in said multidimensional spatio-temporal data, automatically renew multidimensional spatio-temporal data linked to the parameters, and then to automatically generate and renew said multidimensional vector graphics or said special graphical literal document using a deductive capability of said large vector spatio-temporal model and processing of said intelligent agent, a procedure for constructing said large vector spatio-temporal model includes: the steps of describing said multidimensional vector spatio-temporal data in natural language, of pre-training and accurately fitting said large vector spatio-temporal model, and of understanding and analyzing said multidimensional vector spatio-temporal data,A multidimensional vector spatio-temporal large model terminal uses engineering domain data as a pre-training dataset to precisely fit a large general model, and pre-trains said large general model by expanding an engineering domain vocabulary and using said pre-training dataset to enhance the execution and understanding capabilities of said large general model with engineering domain knowledge and a data processing method; said processor is further configured to describe a processing method for said engineering domain data, to define a language for the processing method for said engineering domain data, and to use the processing method for said engineering domain data, which is defined as the dataset to be precisely fitted to train said large general model.in order to strengthen the capacity to execute and understand said large general model in the process of processing said data in the engineering domain; The act of describing said multidimensional vector spatio-temporal data in natural language includes the step of transforming, in terms of description using natural language, vector spatio-temporal data into various types of points, lines, surfaces and volumes with continuous spatial-temporal information (x, y, z, t) and attribute information, including the geographic information system (GIS) and computer-aided mapping (CAD) according to the needs of training, understanding and processing said large vector spatio-temporal model, and to make natural language transformed to describe all information with original vector spatio-temporal data, including geometry types, geometric features, display styles, reference point coordinates, relative geometric data based on the reference point, temporal information and attribute information of a vector object;the pre-training and fine-tuning of said large vector space-time model includes: the steps of establishing a dataset that includes said natural language description of multidimensional vector space-time data and features of the vector object, and of continuously pre-training and fine-tuning said large general model on the basis of said pre-training dataset and said dataset to be fine-tuned, in order to form a large vector space-time model with the ability to understand vector space-time data; The process of understanding and analyzing said multidimensional vector spatio-temporal data includes the steps of inputting vector spatio-temporal data described by natural language into said large vector spatio-temporal model; of outputting features of the vector object in understanding from said large vector spatio-temporal model; of inputting a part of the features of the vector object into said large vector spatio-temporal model; of outputting complete vector spatio-temporal data in understanding or analysis from said large vector spatio-temporal model;and further to accomplish understanding and processing geometric features and attribute features of vector spatio-temporal data and to identify and process spatial relationships and spatio-temporal relationships of vector spatio-temporal data on the basis of the spatio-temporal features or data output from the vector object and using an analytical capability of said large vector spatio-temporal model; in which understanding and processing the geometric characteristics of vector spatio-temporal data includes the step of processing the geometric coordinates of the vector object itself, Understanding and processing attribute features of vector spatiotemporal data includes the step of processing attribute information of the vector object itself; identifying and processing spatiotemporal relationships of vector spatiotemporal data includes the step of identifying and processing topological spatial relationships, sequential spatial relationships, and metric spatial relationships between vector objects; The process of establishing a dataset that includes said natural language description of multidimensional vector spatiotemporal data and vector object features, and of continuously pre-training and precisely fitting said large general model on the basis of said pre-training dataset and said dataset to be precisely fitted includes the steps: (Tl) to collect massive vector spatiotemporal data and natural language text descriptions and vector object feature descriptions to the corresponding vector spatiotemporal data, in order to form a vector spatiotemporal data training dataset, where said vector spatiotemporal data include predefined geometric structure information of points, lines, surfaces and volumes, and attribute text information that corresponds to a geometric solid; (T2) to use said large general model as a basis for pretraining using said training dataset of vector spatiotemporal data; to sample vector text by means of a random walk; and to input vector text by sampling as a training sample to the large general model; and (T3) to collect data in various question-answer formats, including vector-topological relation data, vector-attribute information data, and attribute description-vector data, to form a spatial model dataset precisely adjusted vector temporal, and precisely adjust the large general model in pre-training in the T2 stage; said processor is further configured to search for similar tasks in an engineering process library based on task description; where a task scheduling module includes ways of task scheduling without information feedback and of task scheduling with information feedback, said processor is further configured to receive feedback of an execution result from an environment, a user or said large vector space-time model, and to resolve an operational process into several subtasks to execute them; said processor is further configured such that performing interaction, analysis, and mutual generation of a multidimensional vector graph and a special graphical literal document includes the steps of processing and understanding said multidimensional spatiotemporal data, generating said intelligent agent to process multidimensional spatiotemporal information, and mutually generating said multidimensional vector graph and said special graphical literal document; and further includes the steps: (SI) to input said multidimensional spatio-temporal data, and to transform said multidimensional spatio-temporal data into natural language descriptions that can be processed by said large vector spatio-temporal model, in order to perform AI-based analysis and understanding of multidimensional spatio-temporal information vector data and temporal data within said multidimensional spatio-temporal data based on the inference and understanding capability of said large vector spatio-temporal model; (S2) to input user instructions by interactive means, including speech input and text input, and to transform user instruction inputs described by natural language into formatted instructions to compile, query, analyze, and output data from the information system based on the understanding, analysis, and processing capabilities of said large vector spatiotemporal model, in order to generate

2.

3. said intelligent agent capable of processing multidimensional spatio-temporal information; and (S3) to realize the automatic generation of multidimensional vector graphics or special graphic literal documents by automatic and intelligent means, and to renew bidirectional AI applications in mutual generation on the basis of said intelligent agent to process multidimensional spatio-temporal information and using its multi-round merge and execution call, in order to mutually generate said multidimensional vector graphics and multidimensional spatio-temporal data in various types; The automatic generation of multidimensional vector graphics or special graphic literal documents includes the steps of accurately obtaining parameters in said multidimensional vector spatiotemporal data, and merging the parameters with models to generate multidimensional vector graphics or special graphic literal documents.A mutually generative AI system based on multidimensional spatio-temporal information vector graphics according to claim 1, characterized in that processing geometric coordinates of the vector object itself includes the step of modifying and compiling a shape, size and position, processing attribute information of the vector object itself includes the step of modifying, querying, analyzing and counting them; said topological spatial relation refers to a relation of association, neighborhood, inclusion, intersection, overlap and separation between spatial objects; said sequential spatial relation refers to an order in which said spatial objects or events are arranged in space, including orientation relations such as front-back, left-right, top-bottom, and east-west-south-north;said metric spatial relationship refers to a relationship of distance or proximity between said spatial objects. A mutually generative AI system based on multidimensional spatio-temporal information vector graphics according to claim 1, characterized in that said processor is further configured such that: to solve and plan a task of an interactive operational process to process vector graphics included in multidimensional spatio-temporal data processing of various types based on domain knowledge, an operational process, data understanding and text generation capability of said large vector spatio-temporal model, where each subtask is a general logical processing performed independently or a geospatial analysis operation, and has explicit inputs and outputs; the subtasks interact to complete a complex operational process processing;to perform each task by sensing information from an environment or obtaining information from a memory in order to provide data required by each task and to memorize data from running processes, where said intelligent agent for processing multidimensional spatio-temporal information accumulates data and experience and progressively completes a self-evolution in order to provide iterative capacity support to said large vector spatio-temporal model; and to execute each subtask planned by said large vector spatio-temporal model to a concrete result, where the subtask is executed using the inference capabilities of the large vector spatio-temporal model, spatial analysis and processing capabilities of the geographic information system and the processing capabilities of the computer-aided mapping system;The subtask interacts with interactive objects that include sensors, model libraries, controllers, and databases.

4. A mutually generative AI system based on multidimensional spatio-temporal information vector graphics according to claim 1, characterized in that said several subtasks are linked together in a cascade or tree structure, and each subtask determines a subsequent subtask based on a task execution result.

5. A mutually generative AI system based on multidimensional spatio-temporal information vector graphics according to claim 3, characterized in that said processor is further configured to memorize information sensed from a

6.

7. environment, a task execution record, and a task execution result, and to use recorded memories to promote a future action that includes a stored data and operation, where said stored data includes short-term stored information entered into a pop-up window and long-term stored information in an external vector quickly sought and queried. A mutually generative AI system based on multidimensional spatio-temporal information vector graphics according to claim 3, characterized in that said processor is further configured to use an instrument to perform and complete each task, where the instrument includes said large vector spatio-temporal model itself and an external instrument that includes an algorithmic model, program compilation, a database, and an API, and to further use a spatial database to perform partial spatial analysis, use a model in a model database to perform a specific task, use an API to obtain real-time or historical data, where the specific task refers to a task that corresponds to each model in the model database. A mutually generative AI system based on multidimensional spatio-temporal information vector graphics according to claim 1, characterized in that performing the automatic generation of multidimensional vector graphics or special graphic literal documents includes the steps of: artificially forming a model of the multidimensional vector graphic or special graphic literal document according to a content requirement of said multidimensional vector graphic or special graphic literal document, or making said large vector spatio-temporal model automatically generate a model of the special graphic literal document according to a content requirement, where the graphic of the special graphic literal document is in multidimensional vector graphics format, or in matrix format converted to multidimensional vector graphics;and extract and generate parameters of various types in the model of said multidimensional vector graphics or said; special graphic literal document, accurately obtain parameters in said multidimensional vector spatio-temporal data, and merge the parameters with the models to generate said multidimensional vector graphic or said special graphic literal document, by means of making information and analysis by an information system or making deduction and understanding by said large vector spatio-temporal model.

8. A mutually generative AI system based on multidimensional spatio-temporal information vector graphics according to claim 1, characterized in that the mutual generation of said multidimensional vector graphics and said special graphic literal document is based on the analysis and understanding capabilities of said large vector spatio-temporal model; after inputting data into the engineering domain or after the special graphic literal document makes relevant spatio-temporal information change, said large vector spatio-temporal model generates a renewed parameter, description or intelligent agent for the multidimensional vector graphics, and automatically renews the data of the multidimensional vector graphics;after drawing or modifying said multidimensional vector graphic causes said multidimensional vector graphic to change, the updated multidimensional vector graphic is obtained through the large vector spatio-temporal model, and corresponding engineering parameters or document description are then updated; which allows for updating content related to said special graphic literal document.

Images