CN-121997971-A - Body data acquisition planning method and device, electronic equipment and storage medium

CN121997971ACN 121997971 ACN121997971 ACN 121997971ACN-121997971-A

Abstract

The invention provides a method, a device, electronic equipment and a storage medium for planning body data acquisition, which are characterized in that a body capacity map is constructed, nodes are used for representing executable capacity of a robot and target elements in an environment, the nodes are used for representing association relations among the nodes, body data are acquired in real time, current coverage rate of the body data is calculated, one or more candidate data acquisition tasks are determined based on the current coverage rate, expected gains of the coverage rate of the candidate tasks are estimated and executed, the data acquisition tasks are selected from the candidate tasks according to the expected gains, and a data acquisition action sequence is generated.

Inventors

NI ZIQIANG
YAO GUOCAI
ZHANG LEI
YANG XIANG
LIU YOU
XU RUNTIAN
XIE SHAOXUAN
LIU XUECHENG
Jiao Yance

Assignees

北京智源人工智能研究院

Dates

Publication Date: 20260508
Application Date: 20251226

Claims (10)

1. A method for personal data acquisition planning, comprising: constructing a self-capability map, wherein nodes of the map are used for representing the executable capability of a robot and target elements in the environment, and edges of the map are used for representing the association relation between the nodes; Acquiring body data in real time, and calculating the current coverage rate of the body data; Querying the self-capability map based on the current coverage to determine one or more candidate data acquisition tasks and evaluating expected gains for performing each candidate task on the coverage; And selecting a data acquisition task from the candidate tasks according to the expected gain, and generating a data acquisition action sequence according to the selected data acquisition task.
2. The method of claim 1, wherein the constructing a body-building capability map comprises: creating initial capability nodes, environment nodes and initial edges representing initial association between the capability nodes and the environment nodes based on a preset capability list and environment priori information of the robot so as to generate an initial self-body capability map; And receiving feedback information from the robot in the task execution process, and updating the initial physical capability map according to the feedback information so as to obtain a physical capability map.
3. The method of claim 1, wherein the acquiring the body data in real time, calculating the current coverage of the body data comprises: Building a body data set according to the real-time acquisition body data; Calculating a coverage index according to the body dataset, wherein the coverage index comprises an acquired sample number element, a diversity entropy element and a confidence interval element, the acquired sample number element is determined according to the proportion of the acquired data sample number to the expected total sample number, the diversity entropy element is determined according to the distribution entropy values of the body dataset under different categories, positions or conditions, and the confidence interval element is determined based on the performance confidence interval of a model trained by the body dataset on a verification set; And obtaining the current coverage rate of the body data according to the coverage rate index.
4. The method of claim 1, wherein querying the body capability map to determine one or more candidate data acquisition tasks based on the current coverage and evaluating expected gains in performing each candidate task for the coverage comprises: inquiring task nodes associated with environment element nodes and/or robot capability nodes which are not fully covered currently in the body-building capability map based on the current coverage rate, and determining tasks corresponding to the task nodes as the candidate data acquisition tasks; predicting newly-added body data after executing the candidate data acquisition tasks according to the node information related to each candidate data acquisition task in the body-building capacity map; And based on the newly added body data, evaluating the expected gain of the candidate data acquisition task on the current coverage rate.
5. The method according to claim 4, wherein predicting, for each candidate data collection task, newly adding the body data after performing the candidate data collection task according to node information associated with the candidate data collection task in the body capability map comprises: Analyzing at least one environment element node and at least one capability node directly associated with the candidate data acquisition task in the self-capability map; Matching each environment element node with a preset environment-data mapping rule, and determining a data mode which can be acquired under the current environment element; matching each capability node with a preset capability-operation mapping rule, and determining specific data acquisition operation supported by the current capability; Determining a predicted acquisition data mode according to the acquirable data mode and the specific data acquisition operation; Determining a content range of prediction acquisition according to the spatial range, object category or event type parameters in the attribute information of the environment element node and combining the parameters of the specific data acquisition operation; and predicting newly-added body data after executing the candidate data acquisition task based on the data modality and the content range.
6. The method of claim 4, wherein the evaluating the expected gain of the candidate data acquisition task for the current coverage based on the newly added avatar data comprises: Combining the newly added body data into a currently acquired body data set to form a simulation data set; calculating the simulation coverage rate corresponding to the simulation data set; And taking the difference value between the simulated coverage rate and the current coverage rate as the expected gain of the candidate data acquisition task on the current coverage rate.
7. The method of claim 1, wherein when there are a plurality of robots, the generating a data acquisition action sequence according to the selected data acquisition task further comprises: assigning selected different data acquisition tasks to different robots according to the executable capacity of each robot represented in the body energy spectrum; Planning an execution path of each robot assigned with tasks based on geographic position information represented by the environmental element nodes in the body energy spectrum; and generating a data acquisition action sequence for each robot based on the allocated tasks and the execution paths.
8. A personal data acquisition planning apparatus, comprising: the building module is used for building a self-capability map, nodes of the map are used for representing the executable capability of the robot and target elements in the environment, and edges of the map are used for representing the association relation between the nodes; The acquisition module is used for acquiring the body data in real time and calculating the current coverage rate of the body data; An evaluation module for querying the body-building capability map based on the current coverage rate to determine one or more candidate data acquisition tasks and evaluating expected gains for performing each candidate task on the coverage rate; And the generation module is used for selecting a data acquisition task from the candidate tasks according to the expected gain and generating a data acquisition action sequence according to the selected data acquisition task.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of personal data acquisition planning as claimed in any one of claims 1 to 7 when the program is executed by the processor.
10. A non-transitory readable storage medium having stored thereon a computer program, which when executed by a processor implements the method of the personal data acquisition planning of any one of claims 1 to 7.

Description

Body data acquisition planning method and device, electronic equipment and storage medium Technical Field The present invention relates to the field of artificial intelligence technologies, and in particular, to a method and apparatus for planning body data acquisition, an electronic device, and a storage medium. Background In recent years, with the rapid development of autonomous mobile robots and self-contained intelligent technologies, robots are widely expected to be capable of autonomously performing data collection tasks, such as environmental monitoring, map construction, data set collection, etc., in various complex environments. The core requirement of such tasks is that the data acquired by the robot needs to be highly comprehensive and representative to ensure the generalization capability of the subsequent modeling, analysis or machine learning model. The traditional data driving planning method represented by reinforcement learning has poor adaptability and low efficiency in a dynamic environment, and the traditional data driving planning method has the advantages that the environmental adaptability is improved, but the training cost is high, and the decision process is difficult to understand and intervene like a black box. On the other hand, the method for planning tasks by using a Large Language Model (LLM) is faced with the problems of large calculation force requirement, unstable decision, difficulty in accurate butt joint with the self-contained capacity and the like in real-time deployment of robots, and causes the disjointing of planning and specific task targets and physical constraints. In terms of robot data acquisition, traditional schemes often take the form of exhaustive or random exploration to acquire data. In order to improve the data diversity as much as possible, the existing large-scale data acquisition attempts usually make robots perform various operations under different environments to attempt to cover as many situation combinations as possible, and the exploratory acquisition without explicit optimization targets can easily cause data redundancy and key information omission due to lack of quantitative evaluation of data value and coverage progress. Disclosure of Invention The invention provides a body data acquisition planning method, a body data acquisition planning device, electronic equipment and a storage medium, which are used for solving the defects that the body data acquisition method is difficult to understand and intervene, the calculation power requirement is large, the decision is unstable, and the quantitative evaluation of the data value and the coverage progress is lacking in the real-time deployment of robots. The invention provides a body data acquisition planning method, which comprises the following steps: constructing a self-capability map, wherein nodes of the map are used for representing the executable capability of a robot and target elements in the environment, and edges of the map are used for representing the association relation between the nodes; Acquiring body data in real time, and calculating the current coverage rate of the body data; Querying the self-capability map based on the current coverage to determine one or more candidate data acquisition tasks and evaluating expected gains for performing each candidate task on the coverage; And selecting a data acquisition task from the candidate tasks according to the expected gain, and generating a data acquisition action sequence according to the selected data acquisition task. According to the body-building data acquisition planning method provided by the invention, the body-building capability map is constructed, and the method comprises the following steps: creating initial capability nodes, environment nodes and initial edges representing initial association between the capability nodes and the environment nodes based on a preset capability list and environment priori information of the robot so as to generate an initial self-body capability map; And receiving feedback information from the robot in the task execution process, and updating the initial physical capability map according to the feedback information so as to obtain a physical capability map. According to the method for acquiring and planning the body data provided by the invention, the step of acquiring the body data in real time and the step of calculating the current coverage rate of the body data comprises the following steps: Building a body data set according to the real-time acquisition body data; Calculating a coverage index according to the body dataset, wherein the coverage index comprises an acquired sample number element, a diversity entropy element and a confidence interval element, the acquired sample number element is determined according to the proportion of the acquired data sample number to the expected total sample number, the diversity entropy element is determined according to the distribution entropy valu