Method, system and equipment for realizing automatic generation of vehicle user scene function based on AI intelligent agent, and medium

Abstract

The invention discloses a method, system and equipment for realizing automatic generation of a vehicle scene function through an AI intelligent agent and a medium, and the method comprises the steps: decomposing a triggering condition and an execution action of a scene function required by a user through an AI large model according to user demand information; generating a scene function triggering condition and an instruction field for executing an action by using an AI large model, and generating scene function data in a JSON format to realize a vehicle scene function required by a user; further, a large model is used for analyzing the mapping relation between the burying point data and the scene data, and a knowledge base is established; and continuously monitoring the buried point data, and when the system recognizes the buried point combination features meeting the mapping table, automatically generating a scene function, issuing the scene function to a vehicle end, and recommending the scene function to a user for use.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent vehicles, and particularly to a method for automatically generating user-customized vehicle scenario functions by using an AI agent. Background Art

[0002] With the development of artificial intelligence (AI) technology and its wide application in various industries, especially in the Internet of Vehicles, vehicle intelligence has become one of the important development directions of the automotive industry. Although existing vehicles already have a certain degree of automated services, these services often lack sufficient flexibility for user personalization development and are difficult to meet the personalized, special, or variable needs of different users. To improve this situation and provide users with a more intelligent and personalized driving experience, it is necessary to develop new technical solutions to achieve the automatic generation of user vehicle scenario functions. Summary of the Invention

[0003] Referring to Figure 1 , the present invention aims to propose a method for automatically creating vehicle scenario functions required by users through an AI agent. This method can, according to the user's input, parse the trigger conditions and execution action behaviors with the help of an AI large model, generate corresponding instruction attribute parameters in combination with a pre-configured scenario instruction library, and finally form a complete JSON-format instruction file for the vehicle controller to parse and execute the scenario function. This method not only improves the intelligent level of vehicle services but also brings great convenience to users, reduces the difficulty of manual design of scenario functions by vehicle manufacturers or users, and can also meet the variable driving scenario needs of users.

[0004] Step S101: Obtain user requirements Embodiments of the present application are based on the user expressing the scenario function they want to create through voice or other means, and the input content is the text information of the user's requirements for the scenario function.

[0005] Step S102: Identify trigger conditions and execution behaviors In the embodiment of the present application, an AI large model is used to analyze and process the user input text information obtained in step S101, and decompose the triggering conditions (optional) for the execution of the scenario function and at least one execution action. For each triggering condition and execution action, it is necessary to identify its specific operation object, operation vehicle area, and operation action, and output the results in the specified JSON format. In particular, when the user inputs a fuzzy statement, through the AI large model, based on the vectorized instruction mapping table, the user input and the instruction mapping table are subjected to similarity matching in a retrieval-enhanced manner to identify the user's intention and automatically construct all the triggering conditions and execution behavior actions of the scenario function corresponding to the user's needs. There are no specific requirements for the selection of the AI model. A general commercial large model or a large model that has been secondarily fine-tuned and trained for vehicle scenario functions can be used; the use of the large model needs to be completed by importing pre-set prompt words.

[0006] Step S103: Query and match instruction attributes In the embodiment of the present application, according to the information obtained in step S102, the corresponding operation commands and related parameters are searched in the pre-set scenario instruction library. The result containing the complete instruction field information is output, including but not limited to the interface name, vehicle area, data type, etc. When querying and matching instructions in the embodiment of the present application, first, pre-set prompt words are imported into the AI large model, and the large model searches for the relevant parameters of the corresponding operation object from the pre-set scenario instruction library - instruction interface data table according to the operation object; the scenario instruction library needs to be vectorized, and then the input information and the scenario instruction library are subjected to approximate matching, and the found result is given to the large model, and the data of the operation object is output. After the AI large model obtains the data of the operation object, the large model searches for the relevant parameters of the corresponding operation vehicle area from the pre-set scenario instruction library - vehicle area data table according to the operation vehicle area; the vehicle area data table also needs to be vectorized and subjected to approximate matching. After the AI large model obtains the data of the operation object, the large model searches for the relevant parameters of the corresponding operation behavior from the pre-set scenario instruction library - execution data table according to the operation behavior; the execution data table also needs to be vectorized and subjected to approximate matching.

[0007] Step S104: Construct JSON field data In the embodiment of the present application, based on the relevant data of the operation object, operation area, and operation target of the trigger condition obtained in step S103, the AI large model imports the preset prompt words and constructs the corresponding JSON field data structure to ensure that all necessary information is correctly represented. In the embodiment of the present application, based on the relevant data of the operation object, operation area, and operation target of the execution action obtained in step S103, the AI large model imports the preset prompt words and constructs the corresponding JSON field data structure to ensure that all necessary information is correctly represented. The embodiment of the present application pays special attention to processing different types of data (such as enumerated values, numerical values, strings, etc.), and validates the generated data field format through the AI large model to ensure the consistency and accuracy of the output format.

[0008] Step S105: Integrate into a complete scenario function In the embodiment of the present application, all the JSON data fields of the trigger conditions and execution actions generated in the previous step S104 are combined together to form a JSON format data file describing the entire scenario function. The JSON format data file includes the scenario function name, the relationship between events (such as the "AND" and "OR" relationships between multiple trigger conditions), the field list of the trigger condition nodes, and the field list of the execution action nodes, etc.

[0009] Step S106: Send to the vehicle side In the embodiment of the present application, the JSON format scenario function instruction file finally generated in step S105 will be sent from the cloud to the vehicle execution controller side, and the scenario engine software inside the controller will parse it according to the scenario instruction library and execute the corresponding function operations to implement the scenario function required by the user.

[0010] Step S107: Record the user-generated scenario function data and vehicle buried point data In the embodiment of the present application, when generating the user scenario function data, the vehicle buried point data and scenario function data at this time should be recorded simultaneously and stored in the cloud database to provide data for the large model for later learning.

[0011] Step S108: Learn the mapping relationship between the buried point data and scenario data and establish a mapping table In the embodiment of the present application, the large model is used to learn the relationship between the scenario function data recorded in the cloud and the corresponding vehicle buried point data, and establish the corresponding knowledge base, so that when the system monitors that the vehicle buried point combination conditions corresponding to the scenario function are met, the corresponding scenario function can be automatically generated.

[0012] Step S109: Monitor the buried point data, automatically generate the scenario function and push it to the user In the embodiment of the present application, by monitoring the real-time buried point data in the cloud, when the combination of buried point data meets the knowledge base generated in step S108, the large model is called, and starting from step S101, the corresponding scenario function is automatically created and sent to the vehicle terminal for recommendation to users for use.

[0013] The embodiment of the present application discloses an electronic device system, specifically including: a processor for parsing and executing the prompt words of the AI large model; a memory for storing the scenario instruction library; wherein, the system needs to provide a network interface for accessing the AI large model to implement the method for automatically generating vehicle scenario functions. Description of the Drawings Figure 1 It is a main flow schematic diagram of the method for automatically creating vehicle scenario functions provided by the embodiment of the present invention; Figure 2 It is a specific implementation flowchart of the method for automatically creating vehicle scenario functions provided by the embodiment of the present invention; Figure 3 It is a structural schematic diagram of the system for automatically creating vehicle scenario functions provided by the embodiment of the present invention; Figure 4 It is a structural schematic diagram of the electronic device provided by the embodiment of the present invention. Detailed Embodiments

[0014] Exemplary embodiments will be described in detail herein, and examples thereof are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of another identical element in the process, method, article, or device comprising the element. In addition, components, features, and elements with the same name in different embodiments of the present application may have the same meaning or different meanings, and their specific meanings need to be determined based on their explanations in the specific embodiments or further in combination with the context of the specific embodiments. It should be understood that although the terms first, second, third, etc. may be used herein to describe various information, such information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "when" or "while" or "in response to determining". Furthermore, as used herein, the singular forms "a", "an", and "the" are also intended to include the plural forms unless the context indicates otherwise. It should be further understood that the terms "comprising" and "including" indicate the presence of the stated features, steps, operations, components, items, kinds, and / or groups, but do not exclude the presence, occurrence, or addition of one or more other features, steps, operations, components, items, kinds, and / or groups. The terms "or" and "and / or" as used herein are interpreted as inclusive, or meaning any one or any combination. Thus, "A, B, or C" or "A, B, and / or C" means "any one of the following: A; B; C; A and B; A and C; B and C; A, B, and C". An exception to this definition only occurs when the combination of elements, functions, steps, or operations is inherently mutually exclusive in some way.

[0015] It should be understood that although the steps in the flowchart in the embodiments of the present application are sequentially shown according to the indication of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, the execution of these steps has no strict order restriction, and they can be executed in other orders. Moreover, at least a part of the steps in the figure may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed alternately or alternately with at least a part of other steps or sub-steps or stages of other steps. It should be noted that in this article, step codes such as S101 and S102 are used. The purpose is to more clearly and briefly express the corresponding content and does not constitute a substantial limitation in order. Those skilled in the art may execute S102 first and then S101 during specific implementation, but these should all be within the protection scope of the present application. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application. In the subsequent description, the suffixes such as "module", "component" or "assembly" used to represent elements are only for the convenience of the description of the present application, and they have no specific meaning themselves. Therefore, "module", "component" or "assembly" can be used interchangeably.

[0016] Refer to Figure 2 , a method for automatically creating vehicle scenario functions required by users through an AI agent provided by an embodiment of the present application. This method can be executed by a device for automatically creating vehicle scenario functions required by users provided by an embodiment of the present application. This device is implemented in a software manner and can be deployed on electronic devices such as in-vehicle terminals. The method for the AI agent provided by this embodiment to automatically create vehicle scenario functions required by users includes:

[0017] Step S201: Prerequisites 1. The AI agent is default deployed in the cloud (when the computing power of the vehicle controller chip is sufficient to support the operation of the large model, the AI agent can also be deployed on the controller on the vehicle side). The AI agent realizes the automatic generation ability of the scenario function of the embodiment of the present application by calling the AI large model multiple times according to the process; 2. The prompt words for the large model to execute tasks in the AI agent need to be set in the cloud memory in advance. For the prompt words for the large model call in the embodiment of the present application, the method of low-sample thinking chain is mainly used; 3. After the data field format and content of the instruction mapping table in the AI agent are vectorized, they need to be stored in the cloud vector database. The main content of the instruction mapping table is as follows: Based on common user inputs such as "a bit hot", it associates the user's target needs, such as air conditioner adjustment. In this way, the large model can further analyze the operation purpose to be executed, such as turning on the air conditioner, turning on / off the cooling function, increasing / decreasing the air conditioner set temperature. The basic content of the mapping table includes: basic user input requirements, mapped vehicle operation instructions, intensity settings, associated conditions, etc.; 4. The data field format and content of the scenario instruction library in the AI agent need to be vectorized and stored in the cloud vector database. The scenario instruction library is mainly divided into 3 types of data tables, namely: 1. Instruction interface data table; 2. Vehicle area data table; 3. Execution data table; 5. Temperature coefficient setting of the AI large model used in this patent: Since the application of the AI large model in this patent belongs to the QA type of application, the set angle value of the temperature coefficient should be ensured to be as accurate and consistent as possible for the task. The recommended range of common temperature coefficient settings is between 0.5 and 1.

[0018] Step S202: Parsing and identifying trigger conditions and execution actions 1. The user inputs requirements through voice, and after being parsed into text by the cloud voice module, it is transmitted to the AI agent; 2. The AI agent calls the AI large model. According to the text information generated after parsing by the voice module, it analyzes the user's intention by importing the prompt words preset in the cloud memory. The prompt word technology adopts the low-sample thought chain, and the main content is as follows: a. Content included in the AI large model prompt words - Role setting: Describes the role and main tasks played by the large model in this step; b. Content included in the AI large model prompt words - Skill 1: Generating scenario card names and introductions: By means of the thought chain, it prompts the large model how to generate scenario card names and introductions step by step based on user inputs, and gives 2 typical cases to improve the output accuracy; c. Content included in the AI large model prompt words - Skill 2: Identifying trigger conditions, where there can be 0 or more trigger conditions: By means of the thought chain, it prompts the large model how to identify and break down all trigger condition information step by step based on user inputs, and gives 2 typical cases to improve the output accuracy; d. Content included in the I large model prompt words - Skill 3: Identifying execution actions, where there can be 1 or more execution actions: By means of the thought chain, it prompts the large model how to identify and break down all execution action information step by step based on user inputs, and gives 2 typical cases to improve the output accuracy; e. Contents included in the prompts of the AI large model - Skill 4: Output the results generated in the first three steps in a specified format: output_SceneName (scene name), output_SceneInfo (scene brief introduction), output_SceneTrigger (trigger condition), output_SceneActions (execution actions); f. The purpose of providing reference examples is to enable the large model to perform few-shot learning to enhance the accuracy of AI large model recognition; Reference examples: User input: "When the driver's door is opened, the reading light is turned on"; Output of scene card name: Turn on the reading light when the driver's door is opened; Output of trigger condition: Open the driver's door Output of execution action: Turn on the reading light g. For cases where the trigger condition and execution behavior cannot be directly parsed from the user input, such as the user inputting fuzzy requirements like "It's a bit hot", the user input needs to be associated with the instruction mapping table through the AI large model to identify the execution action corresponding to the user's actual intention, such as adjusting the air conditioner, turning on / off the cooling (judging according to the season and temperature conditions), increasing / decreasing the air conditioner set temperature. The prompts of the large model adopt low-shot chain of thought; The main execution process is: 1. Vectorize the user input content; 2. Perform similarity retrieval on the vectorized user input and the vectorized database of the instruction mapping table; 3. Use the preset prompts and give the relevant parts of the user input content and the instruction mapping table retrieved in the previous step to the large model for analysis and processing; 4. The large model outputs the execution action that actually needs to be performed; Among them, the instruction mapping table only lists common mapping instructions, such as mapping the instruction of "It's a bit hot" to the instruction of adjusting the air conditioner temperature; Through the large model, instructions similar to or with related meanings in the user input are mapped to the instruction of adjusting the air conditioner temperature, such as "It's too hot", "The temperature is too low"; Among them, for numerical execution actions, such as temperature setting, setting the execution position of actuators such as windows, the large model is made to parse the intensity of the user's requirements through prompts, and the intensity is divided into 3 levels: strong, medium, and weak; They respectively correspond to different execution intensities in the instruction mapping table (this step is ignored for switch-type execution actions); h. Limitation conditions: Limitations on the output format and the number of output characters, as well as the handling and feedback of abnormal working conditions such as the inability to identify the trigger condition or execution action;

[0019] Step S203: Parse and identify the operation object, operation area, and operation target of the trigger condition and execution action 1. Call the large AI model. According to the output result in step S202, parse the specific objects, location areas, and execution targets of the trigger conditions and execution actions respectively. The prompting technique adopts a low-sample thought chain. The main content is as follows: a. Content included in the prompting words of the large AI model - Role setting: Describe the role and main tasks played by the large model in this step; b. Content included in the prompting words of the large AI model - Skill 1: Parse the trigger conditions output in step S202. Prompt the large model to break down the specific object (such as the car door) of each trigger condition, the location area of the trigger condition (such as the driver's or passenger's area), and the action behavior of the trigger condition (such as opening or closing) step by step through the thought chain; and give 2 typical cases to improve the output accuracy; c. Content included in the prompting words of the large AI model - Skill 2: Parse the execution actions output in step S202. Prompt the large model to break down the specific object (such as the reading lamp) of each execution action, the location area of the trigger condition (such as the driver's or passenger's area), and the action behavior of the trigger condition (such as opening or closing) step by step through the thought chain; and give 2 typical cases to improve the output accuracy; d. Content included in the prompting words of the large AI model - Skill 3: Output the result in the specified format: output_Trigger_Obj (trigger object), output_Trigger_Area (trigger area), output_Trigger_Act (trigger action), output_Action_Obj (execution object), output_Action_Area (execution area), output_Action_Act (execution action); give 1 - 2 typical cases to improve the output accuracy; e. Provide reference examples for the purpose of enabling the large model to perform few-shot learning to enhance the recognition accuracy of the large AI model; Reference examples: Input for the execution action: "Open the passenger window"; Output of the execution action parameters: Execution object: window; Execution area: passenger area; Execution action: open; f. Limiting conditions: Limitations on the output format and the number of output characters, as well as the handling and feedback of abnormal working conditions such as being unable to recognize the objects, area locations, and action behaviors of the trigger conditions or execution actions; 2. Call the scenario instruction library indexing module. According to the output_Trigger_Obj (trigger object) and output_Action_Obj (execution object) in the previous step, use the large model to query the corresponding instruction field information in the data table of the scenario instruction library through the thought chain; a. The scenario instruction library mainly includes three types of data tables: interface data table, vehicle area data table, and data type table; b. The interface data table mainly includes: MethodName (interface name), MethodCode (interface code), MethodID (interface ID), AreaName (vehicle area name), DataName (data name), etc.; among them, MethodName is the main retrieval object; c. The vehicle area data table mainly includes: AreaName (vehicle area name), DataType (area data type), TableValue (enumeration data definition), etc.; among them, AreaName is the main retrieval object; d. The data type table mainly includes: DataName (data name), DataType (data type), Accuracy (precision), TableValue (enumeration data definition), etc.; among them, DataName is the main retrieval object; e. First, use the vectorized output_Trigger_Obj as a keyword to query similarity in the interface data table, then pass the output query result and output_Trigger_Obj itself to the large model for processing, and respectively obtain the corresponding AreaName and DataName fields in the result; then, according to the queried AreaName field, search for the corresponding data in the vehicle area data table to obtain the vehicle area data corresponding to the trigger condition; similarly, according to the queried DataName field, search for the corresponding data in the data type table to obtain the action behavior data corresponding to the trigger condition; f. Similar to the previous step, use the vectorized output_Action_Obj as the retrieval keyword to query similarity in the interface data table, then pass the output query result data to the large model for processing, and respectively obtain the corresponding AreaName and DataName fields in the result; then, according to the queried AreaName field, search for the corresponding data in the vehicle area data table to obtain the vehicle area data corresponding to the execution action; similarly, according to the queried DataName field, search for the corresponding data in the data type table to obtain the action behavior data corresponding to the execution action;

[0020] Step S204: Generate JSON data fields for the trigger condition The generation of the trigger condition and the execution action are respectively performed in Step S204 and Step S205 for the following reasons: a. Since there are certain differences in the data field formats of the trigger conditions and execution actions, in order to ensure the accuracy and consistency of the generated JSON data format, it is necessary to separately call the large model to generate data; b. The trigger conditions and execution actions have no context association and can be executed in parallel to improve processing efficiency; 1. Based on the data obtained in step S203, use the large model to generate JSON field data for each trigger condition. The prompting technique adopts a low-sample chain of thought to guide the large model to generate the content of each JSON data attribute field step by step: a. The tirgAreaParamList list includes the vehicle area field data of the trigger condition, and its specific format example is as follows: "dataInfo": { "dataInfoName" : "XXXXXX", "dataType" : "XXXX", "enumValues" :[key:XX; Value:AA......] ..... } b. The trigActParamList list includes the action target field data of the trigger condition. Among them, the data format of the action target field is mainly divided into enumeration type and numerical type according to different data types, and its specific format example is as follows: "dataInfo": { "dataInfoName" : "XXXXXX", "dataType" : "XXXX", "enumValues" :[key:XX; Value:AA......] ...... } c. Combine the vehicle area field data and action target field data of the trigger condition, and its specific data format example is as follows: { "trigAreaParamList" : [{trigAreaParamList},......], "trigAreafields" : [ { "dataType" : "XXX", "value" : "XX" ...... },...... ], "trigActParamList" : [{trigActParamList},......], "trigActfields" : [ { "dataType" : "XXX", "value" : "XX" ...... },...... ], } d. Check the data format and field content for each JSON field generated by the trigger condition: 1. Check whether the data conforms to the JSON format; 2. Check the integrity of the data fields; 3. Check whether each field value in the data is a valid value and whether the numerical value is within the range specified in the data table; 4. When data anomalies are detected, return to step S204 and re - execute the JSON data generation; 5. If the re - generated data fails the check again, then exit the entire execution process and feedback the failure result to the system.

[0021] Step S205: Generate the JSON data fields for the execution action 1. Based on the data obtained in step S203, use the large - model to generate the JSON field data for each execution action. The prompting technique adopts the low - sample chain of thought to guide the large - model to generate the content of each JSON data attribute field step by step: a. The actParamAreaList list includes the vehicle area field data of the execution action. Its specific format is as follows: { "dataType": "XXXX", "value": "XX", "enumValues" : [key:XX; Value:AA......], ...... } b. The actParamActionList list includes the action target field data of the execution action. Among them, the data format of the action target field is mainly divided into the enumeration type and the numerical type according to the different data types. Its specific format is as follows: { "dataType": "XXXX", "value": "XX", "enumValues" : [key:XX; Value:AA......], ...... } c. Synthesize the vehicle area field data and action target field data for the execution action, and its specific format is as follows: { "fields" : [{actParamAreaIdList},{actParamActionList}], "type" : "action", ...... } Among them, since there can be multiple execution actions, when there are multiple execution actions, the data field format of the execution action is as follows: { "fields" : [{actParamAreaIdList1},{actParamActionList1}], "type" : "action", ...... }, { "fields" : [{actParamAreaIdList2},{actParamActionList2}], "type" : "action", ...... }, ...... d. Check the data format and field content of the JSON fields generated for each execution action: 1. Check whether the data conforms to the JSON format; 2. Check the integrity of the data fields; 3. Check whether each field value in the data is a valid value and whether the numerical value is within the range specified in the data table; 4. When data anomalies are detected, return to step S205 and re-execute JSON data generation; 5. If the re-generated data fails the verification again, then exit the entire execution process and feedback the failure result to the system.

[0022] Step S206: Generate the complete scenario function JSON data ​​1. Integrate the data fields of the trigger conditions and execution actions generated in steps S204 and S205 into a complete scenario function JSON data, and its basic data format is as follows: { "SceneId" : "XXXXXXXXXXXX", "SceneName" : "XXXXXX", "SceneInfo" : "XXXXXXXXXXXXXXXX", "fields" : { "triggerNodes" : [{triggerNode1, triggerNode2,......}] ...... }, "actionNodes" : [{actionNode1, actionNode2,......}], ...... } 2. Conduct data format verification on the complete JSON scenario function data to ensure the correctness of the JSON field filling by the large model; if the verification fails, exit the entire execution process and feedback the failure result to the system.

[0023] Step S207: JSON data distribution and scenario function execution 1. In the embodiments of the present application, it is necessary to deploy scenario engine software on the vehicle - end controller to parse and execute the JSON data file of the scenario function; the scenario engine software needs to be deployed in a vehicle - end controller with an operating system (such as Android or Linux operating system); 2. The vehicle - end scenario engine software needs to access the cloud through a 4G / 5G network; the network communication protocol can be communication protocols such as MQTT, http, etc.; 3. After the cloud generates the scenario JSON data file, send the complete scenario JSON data to the vehicle - end, and the vehicle - end scenario engine software parses the JSON scenario data according to the scenario instruction library; 4. After the vehicle - end scenario engine software completes JSON parsing, by subscribing to the corresponding trigger condition interfaces (such as CarService service interface, SOME / IP service interface, DDS service interface, etc.), monitor in real - time whether the trigger condition is triggered and meets the execution condition; 5. When the vehicle - end scenario engine software recognizes that the trigger condition is met, sequentially call the corresponding execution action interfaces (such as CarService service interface, SOME / IP service interface, DDS service interface, etc.) to execute the corresponding scenario function.

[0024] Step S208: Record the vehicle buried point data and the generated scenario function JSON data when creating the scenario function 1. Optionally, in the embodiment of the present application, it is necessary to deploy the vehicle buried point data recording function on the vehicle controller and periodically upload it to the cloud server; 2. When the cloud large model automatically creates the scenario function, automatically record the data of the scenario function, and at the same time record the vehicle buried point data when the user requests to create the scenario function, and store it in the cloud database.

[0025] Step S209: Learn the user's scenario usage habits Optionally, in the embodiment of the present application, by using the large model, import the vehicle buried point data and scenario function data stored in the cloud, analyze the correlation between the two, establish a mapping relationship table between the buried point data combination conditions and the corresponding scenario functions, and save it in the cloud knowledge base, so that when the vehicle data meets the buried point combination conditions, the corresponding scenario function data can be automatically generated.

[0026] Step S210: Monitor the vehicle buried point status and automatically create the scenario function Optionally, in the embodiment of the present application, by monitoring the real-time buried point data in the cloud, when the buried point data combination meets the mapping table generated in step S209, call the large model, execute the task process of automatically creating the scenario function starting from step S202, and send it to the vehicle end, and prompt and recommend it to the user through voice or large screen pop-up window for use.

[0027] Based on the same inventive concept as the foregoing embodiments, refer to Figure 3 , Figure 3 shows an automatic scenario function generation system applicable to the solution of the embodiment of the present invention, such as Figure 3As shown in the figure, the system includes a cloud server 31 and a vehicle terminal 32. Among them, the cloud server 31 is used, on the one hand, to receive user request instructions, such as voice instructions; on the other hand, to generate JSON data of scene functions and send them to the vehicle. The vehicle terminal 32 is used, on the one hand, to read the user's voice demand instructions and upload them to the cloud server 31; on the other hand, to download the JSON data file of scene functions from the cloud server 31 through the scene engine, parse it and send corresponding instructions to the vehicle controller to execute the scene function. The vehicle terminal 32 includes a scene engine, a body controller, various sensors and loads. Among them, the scene engine is used to parse and execute the automatically generated JSON data file of scene functions sent by the cloud server 31 to realize the functions required by users. The body controller has the function of service-oriented architecture, which is used to execute the service request instructions sent by the scene engine, read the input data of vehicle sensors and execute vehicle load functions. After the vehicle terminal controller collects the real-time data of key sensors and executing loads, it periodically reports the data as buried point data to the cloud server. The cloud server includes an AI agent (AI large model) and data such as prompt words for controlling the large model, a scene instruction library, an instruction mapping table, historical buried points and a scene function database, a mapping table of buried point combination condition scene functions, a knowledge base after buried point learning, etc., which are used to identify the user's scene requirements through the AI large model, automatically generate a JSON file of scene function data, and send the edited JSON data file to the vehicle terminal.

[0028] In summary, in the system for automatically generating scene function data provided by the above embodiments, by setting the vehicle and the cloud server to communicate and control each other, the AI automatically generated scene function data is generated and transmitted to improve the convenience and adaptability of scene function generation and enhance the user experience effect. Based on the same inventive concept as the foregoing embodiments, the embodiments of the present invention provide an electronic device, such as Figure 4 As shown in the figure, the electronic device includes: a processor 41 of the vehicle terminal controller and a memory 42 storing a computer program; among them, Figure 4 The processor 41 shown in the figure does not refer to the number of processors 41 being one, but only refers to the positional relationship of the processor 41 relative to other devices. In actual applications, the number of processors 41 can be one or more; similarly, Figure 4The memory 42 shown in the figure also has the same meaning, that is, it is only used to refer to the positional relationship of the memory 42 relative to other devices. In actual applications, the number of memories 42 can be one or more. When the processor 41 runs the computer program, a method for applying to the above-described scenario function JSON data for parsing and execution is implemented. The electronic device may further include: at least one network interface 43 for communication with the cloud. Each component in the electronic device is coupled together through a bus system 44. It can be understood that the bus system 44 is used to implement the connection and communication between these components. In addition to the data bus, the bus system 44 also includes a power bus, a control bus, and a signal bus. However, for the sake of clear illustration, in Figure 4 all kinds of buses are labeled as the bus system 44.

[0029] Among them, the memory 42 can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM, Read Only Memory), a programmable read-only memory (PROM, Programmable Read-Only Memory), an erasable programmable read-only memory (EPROM, Erasable Programmable Read-Only Memory), an electrically erasable programmable read-only memory (EEPROM, Electrically Erasable Programmable Read-Only Memory), a ferromagnetic random access memory (FRAM, ferromagnetic random access memory), a flash memory (Flash Memory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM, Compact Disc Read-Only Memory); the magnetic surface memory can be a disk memory or a tape memory. The volatile memory can be a random access memory (RAM, Random Access Memory), which is used as an external cache. By way of example but not limitation, many forms of RAM are available, such as a static random access memory (SRAM, Static Random Access Memory), a synchronous static random access memory (SSRAM, Synchronous Static Random Access Memory), a dynamic random access memory (DRAM, Dynamic Random Access Memory), a synchronous dynamic random access memory (SDRAM, Synchronous Dynamic Random Access Memory), a double data rate synchronous dynamic random access memory (DDR SDRAM, Double Data Rate Synchronous Dynamic Random Access Memory), an enhanced synchronous dynamic random access memory (ESDRAM, Enhanced Synchronous Dynamic Random Access Memory), a sync link dynamic random access memory (SLDRAM, SyncLink Dynamic Random Access Memory), a direct rambus random access memory (DRRAM, Direct Rambus Random Access Memory).The memory 42 described in the embodiments of the present invention is intended to include, but not limited to, these and any other suitable types of memory. The memory 42 in the embodiments of the present invention is used to store various types of data to support the operation of the electronic device. Examples of such data include: any computer programs for operating on the electronic device, such as operating systems and application programs; contact data; phone book data; messages; pictures; videos, etc. Among them, the operating system contains various system programs, such as the framework layer, the core library layer, the driver layer, etc., for implementing various basic services and processing hardware-based tasks. The application programs can include various application programs, such as a media player (Media Player), a browser (Browser), etc., for implementing various application services. Here, the program for implementing the method of the embodiments of the present invention can be included in the application programs.

[0030] Based on the same inventive concept as the foregoing embodiments, this embodiment also provides a computer-readable storage medium. A computer program is stored in the computer-readable storage medium. The computer-readable storage medium can be a ferromagnetic random access memory (FRAM), a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), a flash memory (FlashMemory), a magnetic surface memory, an optical disc, or a compact disc read-only memory (CD-ROM), etc.; it can also be various devices including one or any combination of the above memories, such as a mobile phone, a computer, a tablet device, a personal digital assistant, etc. When the computer program stored in the computer-readable storage medium is run by a processor, it implements the method for parsing and executing the automatically generated scenario function data as described above. For the specific step flow implemented when the computer program is executed by the processor, please refer to Figure 1 the description of the shown embodiments, which will not be repeated here.

[0031] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification. In this text, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion. In addition to the elements listed, it may also include other elements not explicitly listed. As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims described.

Claims

1. A method for automatically generating user vehicle scenario functions based on an AI agent, characterized in that, The method includes: identifying and decomposing the relevant triggering conditions and execution action behaviors of the vehicle scenario functions required by the user through the large AI model; then querying the scenario instruction parameters corresponding to the triggering conditions and execution action behaviors according to the pre-configured scenario instruction library; and generating a JSON-format scenario function data file for the triggering conditions and execution actions based on the query-matched scenario instruction parameters.

2. The method according to claim 1, wherein Using the large AI model, through low-sample thinking chain prompt engineering, identify the requirements and intentions of the vehicle functions input by the user, and break down the triggering conditions of the scenario function and the execution actions after triggering, where the triggering conditions and execution actions can be one or more; in particular, when the user inputs ambiguous statements, retrieve the instruction mapping table by using the method of vector similarity matching, and through the large AI model, analyze the actual requirements of the user and automatically construct all relevant triggering conditions and execution action behaviors of the scenario function corresponding to the user's requirements.

3. The method according to claim 1, wherein Construct a vectorized scenario instruction library and provide it for the large AI model to retrieve and query. The scenario instruction library is mainly divided into three types of data tables, namely:

1. Instruction interface data table; 2. Vehicle area data table; 3. Execution behavior data table.

4. The method according to claim 1, wherein For the scenario triggering conditions and execution actions, use the large AI model, through low-sample thinking chain prompt engineering, to decompose the respective operation objects, operation areas, and operation behaviors. And query and match the instruction attribute fields that meet the requirements for the operation object in the instruction interface data table; through the low-sample thinking chain prompt engineering of the large AI model, query and match the instruction operation area attribute fields that meet the requirements for the operation area in the vehicle area data table; through the low-sample thinking chain prompt engineering of the large AI model, query and match the instruction operation behavior attribute fields that meet the requirements for the operation behavior domain in the execution behavior data table.

5. An electronic device, characterized in that, Include: A processor; A memory for storing the AI large model prompts and the scenario instruction library; a network interface for accessing the AI large model; wherein, the processor configures the AI large model prompts and the scenario instruction library and accesses the AI large model to implement the vehicle scenario function automatic generation method described in any one of claims 1-4.

Images