Search

CN-121998048-A - Data processing method and system for intelligent automobile intelligent agent

CN121998048ACN 121998048 ACN121998048 ACN 121998048ACN-121998048-A

Abstract

The invention relates to the technical field of intelligent agents, in particular to a data processing method and a system of an intelligent automobile intelligent agent, wherein the data processing method comprises the steps of collecting an MCP process for a vehicle when receiving user intention and extracting current semantics for original input information; searching the knowledge network to obtain matching memory information, inputting the matching memory information into an inference engine to generate an execution plan, calling corresponding MCP processes to execute the execution plans respectively, and taking the execution process as new user interaction data. Aiming at the problems that in the prior art, the vehicle-mounted intelligent agent judges the intention of a user and a calling tool is inaccurate, an intelligent agent and a corresponding knowledge network are built in a vehicle-mounted system, interactive data generated by the user are processed and stored, in the actual prediction process, the memory entries stored in the knowledge network are combined with the MCP process of an application operated by the vehicle to predict, a corresponding execution plan is generated, and the accuracy of judging the intention of the user and the accuracy of the calling tool are improved.

Inventors

  • PAN JINGPENG

Assignees

  • 上海优咔网络科技有限公司

Dates

Publication Date
20260508
Application Date
20251201

Claims (10)

  1. 1. The data processing method of the intelligent automobile intelligent agent is characterized by comprising a memory construction process and an execution process; the memory construction process comprises the following steps: A1, interacting with a user, acquiring user interaction data, and reasoning to obtain an atomic knowledge unit according to the user interaction data; a2, linking the atomic knowledge units to form a knowledge network; the execution process comprises the following steps: step B1, when original input information containing user intention is received, acquiring an MCP process for a vehicle, and extracting current semantics from the original input information; Step B2, searching the knowledge network based on the current semantics to obtain matching memory information; Step B3, inputting the original input information, the matching memory information and the MCP process into an inference engine to generate an execution plan; And step B4, calling the corresponding MCP processes to execute the execution plans respectively, and taking the execution process at this time as new user interaction data.
  2. 2. The data processing method according to claim 1, wherein the step A1 includes: A11, extracting original vehicle input data when a user interacts with in-vehicle equipment, and performing text conversion according to the original vehicle input data to generate a description text; step A12, dividing the description text to obtain a plurality of unit entities; and step A13, logically deriving the description text based on the unit entity to obtain the atomic knowledge unit.
  3. 3. The data processing method according to claim 1, wherein the step A2 includes: a21, carrying out associated node searching on the atomic knowledge units relative to the knowledge graph so as to establish multidimensional links; And step A22, adding the atomic knowledge units to the knowledge graph according to the multi-dimensional links.
  4. 4. A data processing method according to claim 3, wherein said step a21 comprises: step A211, matching the atomic knowledge units with associated nodes in the knowledge graph based on semantic features; Step A212, deducing the correlation between the atomic knowledge units and the associated nodes to obtain associated categories; and step A213, generating the multidimensional link according to the association category.
  5. 5. The data processing method according to claim 4, wherein the step a22 includes: Step A221, determining the relative relation between the atomic knowledge units and the association nodes according to the association category; and step A222, selecting a corresponding fusion rule based on the relative relation, and then updating the atomic knowledge units to the knowledge graph based on the fusion rule.
  6. 6. The data processing method according to claim 1, wherein the step B1 includes: step B11, when the user intention input by the user is received, collecting state data of the vehicle and taking the state data and the user intention as original input; step B12, extracting the currently running MCP process from a MCP host; and step B13, extracting the semantic information from the original input.
  7. 7. The data processing method according to claim 4, wherein the step B2 includes: step B21, searching the matched associated nodes based on the current semantic input knowledge network; Step B22, extracting knowledge slices according to the associated nodes; and B23, assembling all the knowledge slices to obtain the matching memory information and outputting the matching memory information.
  8. 8. The data processing method according to claim 1, wherein the step B3 includes: step B31, inputting the reasoning engine according to the original input information and the matching memory information to obtain planning nodes; step B32, the reasoning engine matches the MCP process according to the planning node to determine a matching process of the planning node to be executed; and step B33, generating the execution plan according to the matching process and the planning node.
  9. 9. The data processing method according to claim 6, further comprising, before performing step B1: And B01, respectively creating corresponding MCP processes by the MCP host when the corresponding vehicle-mounted application is called.
  10. 10. A data processing system of a smart car agent, for implementing a data processing method as claimed in any one of claims 1-9.

Description

Data processing method and system for intelligent automobile intelligent agent Technical Field The invention relates to the technical field of intelligent agents, in particular to a data processing method and system of an intelligent automobile intelligent agent. Background An intelligent in-vehicle infotainment system (IVI) is an intelligent control system configured in an intelligent car to provide a user with functions such as voice control, cabin parameter adaptive adjustment, running entertainment applications, etc. For example, patent document CN202211058831.5 discloses an intelligent cabin domain controller, a control method thereof and a vehicle, and the cabin domain controller is integrated with an instrument, an information entertainment controller, a networking module, a remote management unit, a low-speed pedestrian alarm, fatigue monitoring, a vehicle recorder, head-up display and the like, so that not only is an automobile electronic and electric architecture reconstructed, but also the development trend of the electronic and electric architecture is met, the problems of low bandwidth, network flattening, one ECU unit for each function and the like of the distributed electronic and electric architecture can be optimized, the electric cost of the whole vehicle can be reduced, the value is created for customers, and the customer experience can be improved. By further centralizing the control logic, the development efficiency is improved, and the software iteration update is easy. Meanwhile, the cabin domain controller can improve customer experience through the SOC with high calculation power, and the bottlenecks of distributed architecture controller quantity multifunctional logic dispersion and the like are solved. For another example, patent document with the application number CN202311098248.1 discloses a vehicle-mounted intelligent device interaction method, an instrument panel, an intelligent cabin and a vehicle thereof, wherein the method comprises the steps of acquiring gesture information or touch operation instructions of passengers in the vehicle, opening the movable turnover instrument panel, detecting the passengers in the vehicle by the movable turnover instrument panel, moving to the position where the passengers in the vehicle can operate, and performing intelligent interaction between the movable turnover instrument panel and the vehicle-mounted device. The movable overturning instrument board can be opened according to gestures or touch operation of passengers in the automobile, the movable overturning instrument board can be moved to the position where the passengers in the automobile can operate, intelligent interaction is carried out between the movable overturning instrument board and the vehicle-mounted equipment, the intellectualization and automation of equipment in the automobile cabin are realized, and the user experience is improved. However, in the actual implementation process, the inventor finds that most of existing vehicle-mounted Agent frameworks adopt hard-coded tool lists, lack of dynamic discovery capability, have imperfect discovery mechanism for tool information, and the tool description does not conform to the understanding capability of LLM (such as lack of natural language description), so that the inference decision of the Agent is inaccurate, and finally the accuracy of user experience is affected. Disclosure of Invention Aiming at the problems in the prior art, a data processing method of an intelligent automobile intelligent body is provided; On the other hand, a corresponding system is also provided. The specific technical scheme is as follows: a data processing method of intelligent automobile intelligent agent includes memory construction process and execution process; the memory construction process comprises the following steps: A1, interacting with a user, acquiring user interaction data, and reasoning to obtain an atomic knowledge unit according to the user interaction data; a2, linking the atomic knowledge units to form a knowledge network; the execution process comprises the following steps: step B1, when original input information containing user intention is received, acquiring an MCP process for a vehicle, and extracting current semantics from the original input information; Step B2, searching the knowledge network based on the current semantics to obtain matching memory information; Step B3, inputting the original input information, the matching memory information and the MCP process into an inference engine to generate an execution plan; And step B4, calling the corresponding MCP processes to execute the execution plans respectively, and taking the execution process at this time as new user interaction data. In another aspect, the step A1 includes: A11, extracting original vehicle input data when a user interacts with in-vehicle equipment, and performing text conversion according to the original vehicle input data to generate a descript