CN-121981154-A - Teaching data conversion method and device oriented to intelligent body and electronic equipment

CN121981154ACN 121981154 ACN121981154 ACN 121981154ACN-121981154-A

Abstract

The invention provides a teaching data conversion method and device oriented to intelligent body and electronic equipment, and relates to the technical field of intelligent body. The self-intelligent teaching data conversion method comprises the steps of obtaining teaching data to be converted in a first data format, reading metadata information of the teaching data to be converted, stated in a configuration file, calling a target interface method for the first data format according to the metadata information by using a conversion engine, obtaining characteristic data of a self-intelligent agent when executing a target task from the teaching data to be converted, wherein the conversion engine is a class which is defined by a father class and is used for converting teaching data which are not in a second format into teaching data in the second format, and obtaining the teaching data to be converted in the second format according to the characteristic data by using an aggregation method provided by the conversion engine. The invention can realize standardized conversion of intelligent teaching data with bodies from different sources with high efficiency and low cost.

Inventors

LIU XUECHENG
XIE SHAOXUAN
YAO GUOCAI
NI ZIQIANG

Assignees

北京智源人工智能研究院

Dates

Publication Date: 20260505
Application Date: 20251224

Claims (10)

1. The intelligent teaching data conversion method for the tool is characterized by comprising the following steps of: acquiring teaching data to be converted in a first data format, wherein the teaching data to be converted is generated by an intelligent agent when executing a target task; Reading metadata information of the teaching data to be converted, which is stated in the configuration file; The method comprises the steps of calling a target interface method for a first data format according to metadata information by using a conversion engine, and acquiring characteristic data of the body-building intelligent agent when executing the target task from the teaching data to be converted, wherein the conversion engine is a class defined by a parent class and used for converting teaching data which are not in a second format into the teaching data in the second format, and the target interface method is a specific method realized in the conversion engine and used for extracting the characteristic data from the first format data; And obtaining teaching data to be converted in a second format according to the characteristic data by utilizing the aggregation method provided by the conversion engine.
2. The self-intelligent teaching data conversion method according to claim 1, wherein the metadata information includes storage path information of the teaching data to be converted, and the target interface method includes a task segment number acquisition method; The method for acquiring the feature data of the body intelligent agent when executing the target task from the teaching data to be converted by using the conversion engine according to the metadata information comprises the following steps: positioning root directory information of the teaching data to be converted according to storage path information of the teaching data to be converted in the metadata information by using the conversion engine; and calling the task segment number acquisition method based on the root directory information, so that the task segment number acquisition method starts from the root directory information, traverses each task group data, and calculates and obtains the number of task segments contained in the teaching data to be converted according to the subgroup names of each task group data.
3. The self-intelligent-oriented teaching data conversion method according to claim 2, wherein the target interface method comprises a task segment number acquisition method; The method for acquiring the feature data of the body intelligent agent when executing the target task from the teaching data to be converted by using the conversion engine according to the metadata information comprises the following steps: Positioning catalog information of each task segment contained in the teaching data to be converted according to the storage path information of the teaching data to be converted in the metadata information by using the conversion engine; And calling the task segment containing frame number acquisition method based on the catalog information of each task segment so that the task segment containing frame number acquisition method determines the frame number contained in each task segment from task segment data corresponding to the catalog information of the task segment.
4. The self-intelligent-oriented teaching data conversion method according to claim 3, wherein the target interface method comprises a method of extracting a single frame image; The method for acquiring the feature data of the body intelligent agent when executing the target task from the teaching data to be converted by using the conversion engine according to the metadata information comprises the following steps: and according to the number of task segments contained in the teaching data to be converted, sequentially selecting one task segment from the teaching data to be converted, and executing the following operations: Generating a frame index range based on the number of frames contained in the task segment; And calling the method for extracting the single-frame image according to the catalog information and the frame index range of the task segment so that the method for extracting the single-frame image extracts the single-frame image data from the data of the task segment.
5. The self-intelligent oriented teaching data conversion method according to claim 4, wherein the target interface method comprises a frame state subfield extraction method; The method for acquiring the feature data of the body intelligent agent when executing the target task from the teaching data to be converted by using the conversion engine according to the metadata information, further comprises: Acquiring identification information of a state field in each task segment contained in the teaching data to be converted according to the metadata information; And for each task segment, calling the frame state subfield extraction method according to the catalog information and the frame index range of the task segment, so that the frame state subfield extraction method extracts each state field from the data of the task segment according to the identification information of the state field.
6. The body-oriented intelligent teaching data conversion method according to claim 5, wherein the target interface method comprises a frame action subfield extraction method; The method for acquiring the feature data of the body intelligent agent when executing the target task from the teaching data to be converted by using the conversion engine according to the metadata information, further comprises: Acquiring identification information of action fields in each task segment contained in the teaching data to be converted according to the metadata information; And for each task segment, calling the frame action subfield extraction method according to the catalog information and the frame index range of the task segment, so that the frame action subfield extraction method extracts each action field from the data of the task segment according to the identification information of the action field.
7. Teaching data conversion equipment towards utensil body intelligence, characterized by, include: The system comprises a first acquisition module, a first conversion module and a second acquisition module, wherein the first acquisition module is used for acquiring teaching data to be converted in a first data format, and the teaching data to be converted is generated by an intelligent body when executing a target task; the configuration module is used for reading metadata information of the teaching data to be converted, which is stated in the configuration file; The second acquisition module is used for calling a target interface method for the first data format according to the metadata information by utilizing a conversion engine, and acquiring characteristic data of the body intelligent agent when executing the target task from the teaching data to be converted, wherein the conversion engine is a class defined by a parent class and used for converting teaching data which are not in a second format into the teaching data in the second format, and the target interface method is a specific method realized in the conversion engine and is used for extracting the characteristic data from the first format data; and the aggregation module is used for obtaining teaching data to be converted in a second format according to the characteristic data by utilizing the aggregation method provided by the conversion engine.
8. An electronic device comprising a memory, a processor and a computer program stored on the memory and running on the processor, characterized in that the processor implements the self-intelligent teaching data conversion method according to any of claims 1 to 6 when executing the computer program.
9. A non-transitory computer readable storage medium having stored thereon a computer program, wherein the computer program when executed by a processor implements the self-intelligence oriented teaching data conversion method according to any of claims 1 to 6.
10. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the self-oriented intelligent teaching data conversion method according to any of claims 1 to 6.

Description

Teaching data conversion method and device oriented to intelligent body and electronic equipment Technical Field The invention relates to the technical field of intelligent personal devices, in particular to a teaching data conversion method and device for intelligent personal devices and electronic equipment. Background In the field of intelligent research, acquisition and processing of high-quality teaching data are key to promoting technical development. However, the current data sources are highly heterogeneous, namely, the robot platforms of different manufacturers store operation data in private or mixed formats such as H5 (HIERARCHICALDATA FORMAT, hierarchical data format 5), MCAP (Machine-CENTRIC ARCHIVE Protocol), ROS (Robot Operating System Bag, robot operating system data package), JSON+MP4 and the like, and meanwhile, the data organization modes of peripheral devices such as cameras, force sensors, IMUs (Inertial Measurement Unit, inertial measurement units) and the like are different. The traditional data conversion scheme needs to independently write complete conversion logic for each data source, so that the repeated development cost is high, more than 90% of codes are used for processing common functions such as image reading, frame alignment, abnormal skipping and the like, the system is difficult to maintain, and the whole converter needs to be rewritten when a device type is newly added. Therefore, how to efficiently and inexpensively realize standardized conversion of intelligent teaching data with bodies from different sources is a technical problem to be solved. Disclosure of Invention The invention provides a body intelligent teaching data conversion method, a body intelligent teaching data conversion device and electronic equipment, which are used for solving the defects in the prior art and realizing standardized conversion of body intelligent teaching data from different sources efficiently and at low cost. The invention provides a teaching data conversion method oriented to body intelligence, which comprises the following steps. The method comprises the steps of obtaining teaching data to be converted in a first data format, obtaining the teaching data to be converted in the first data format, wherein the teaching data to be converted is generated by an agent to be subjected to target task execution, reading metadata information of the teaching data to be converted stated in a configuration file, calling a target interface method for the first data format according to the metadata information by using a conversion engine, obtaining characteristic data of the agent to be subjected to target task execution from the teaching data to be converted, wherein the conversion engine is a class defined by a parent class and used for converting teaching data in a non-second format into teaching data in the second format, the target interface method is a specific method implemented in the conversion engine and used for extracting the characteristic data from the first format data, and obtaining the teaching data to be converted in the second format according to the characteristic data by using an aggregation method provided by the conversion engine. The method for converting teaching data for body intelligence comprises the steps of utilizing a conversion engine to obtain characteristic data of an intelligent body during execution of a target task from the teaching data to be converted, positioning root directory information of the teaching data to be converted according to storage path information of the teaching data to be converted in the metadata information, and calling a task segment number obtaining method based on the root directory information to enable the task segment number obtaining method to traverse task group data from the root directory information, and obtaining the number of task segments contained in the teaching data to be converted according to subgroup names of the task group data. The method for converting teaching data for the body intelligence comprises a task segment number obtaining method, wherein the method for converting the teaching data for the body intelligence comprises the steps of utilizing a conversion engine to call a target interface method for a first data format according to metadata information, obtaining characteristic data of the body intelligence when executing the target task from the teaching data for the body intelligence, positioning catalog information of each task segment contained in the teaching data for the conversion according to storage path information of the teaching data for the conversion in metadata information, and calling a task segment containing frame number obtaining method based on the catalog information of each task segment so that the task segment containing frame number obtaining method can determine the frame number contained in each task segment from task segment data corresponding to the catalog information of the task