CN-121979868-A - Data verification method and device for robot

CN121979868ACN 121979868 ACN121979868 ACN 121979868ACN-121979868-A

Abstract

The invention provides a data verification method and device of a robot, and relates to the technical field of artificial intelligence, wherein the method comprises the steps of obtaining data to be verified; the method comprises the steps of obtaining target configuration data corresponding to data to be verified according to equipment model and/or data version corresponding to the data to be verified, inputting the data to be verified, the target configuration data and control data corresponding to the data to be verified into a simulation quality inspection system to obtain a first simulation result corresponding to the data to be verified and a second simulation result corresponding to the control data respectively, and determining the verification result of the data to be verified according to the first simulation result and the second simulation result. According to the embodiment, the target configuration data corresponding to the data to be verified of different data sources can be obtained based on the equipment model and/or the data version, automatic simulation and verification can be further carried out on the data to be verified according to the target configuration data, the individual verification requirements of different robots are effectively met, and the degree of automation is improved.

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 (11)

1. A data verification method of a robot, comprising: Acquiring data to be verified; acquiring target configuration data corresponding to the data to be verified according to the equipment model and/or the data version corresponding to the data to be verified; Inputting the data to be verified, the target configuration data and the control data corresponding to the data to be verified into a simulation quality inspection system to respectively obtain a first simulation result corresponding to the data to be verified and a second simulation result corresponding to the control data; And determining a verification result of the data to be verified according to the first simulation result and the second simulation result.
2. The method according to claim 1, wherein the data to be authenticated comprises sensor data acquired based on at least one sensor, the target configuration data comprises target data; the obtaining the target configuration data corresponding to the data to be verified according to the equipment model and/or the data version corresponding to the data to be verified comprises the following steps: Acquiring the equipment model corresponding to each sensor and the data version corresponding to the data to be verified; determining target mapping relations corresponding to the sensor data respectively from a plurality of preset mapping relations according to the equipment model and the data version, wherein the mapping relations indicate the corresponding relation between field names in the sensor data and target field names in the target data; and converting field names in the sensor data according to the target mapping relation to obtain the target data.
3. The data verification method according to claim 2, wherein the sensor data includes left jaw sensor data and right jaw sensor data, the target data includes jaw opening and closing degree, the target configuration data corresponding to the data to be verified is obtained according to a device model and/or a data version corresponding to the data to be verified, and the method further includes: Determining a clamping jaw opening and closing degree calculation method according to a first equipment model of a sensor corresponding to the left clamping jaw and a second equipment model of a sensor corresponding to the right clamping jaw; and calculating the left clamping jaw sensor data and the right clamping jaw sensor data by using the clamping jaw opening and closing degree calculation method to obtain the clamping jaw opening and closing degree corresponding to the data to be verified.
4. The method for verifying data according to claim 2, wherein inputting the data to be verified, the target configuration data, and the manipulation data corresponding to the data to be verified into a simulation quality inspection system, respectively obtaining a first simulation result corresponding to the data to be verified and a second simulation result corresponding to the manipulation data, includes: Based on the target configuration data, respectively playing back the data to be verified and the control data in a simulation environment of the simulation quality inspection system, and respectively generating a first motion result corresponding to the data to be verified and a second motion result corresponding to the control data; generating a data curve corresponding to each sensor based on the sensor data corresponding to each sensor; And generating a first simulation result corresponding to the data to be verified according to the first motion result and the data curves corresponding to the sensors respectively, and generating a second simulation result corresponding to the control data according to the second motion result.
5. The method according to claim 1, wherein the data to be verified further comprises video data collected by a robot and annotation data corresponding to the video data; the determining the verification result of the data to be verified according to the first simulation result and the second simulation result includes: Inputting the video data, the annotation data, the first simulation result and the second simulation result into a visual quality inspection system, and displaying the video data, the annotation data, the first simulation result and the second simulation result through a plurality of visual windows respectively; Receiving first verification information input by a user based on the visual window, wherein the first verification information indicates whether the consistency verification of the user on the video data, the annotation data, the first simulation result and the second simulation result is passed or not; and generating the verification result according to the first verification information.
6. The data verification method according to claim 1, further comprising, before the verification result of the data to be verified is determined according to the first simulation result and the second simulation result: And receiving second verification information input by a user, wherein the second verification information indicates whether the consistency verification of the first simulation result and the second simulation result by the user is passed or not.
7. The method for data verification according to claim 1, wherein, The target configuration data further includes at least one of playback speed, environment data, and a data source address; And/or the number of the groups of groups, The data to be verified is data subjected to post-processing; and under the condition that the verification result indicates that the data to be verified is wrong, re-performing post-processing on the data to be verified, and re-performing data verification on the data successfully subjected to the post-processing.
8. A data verification device for a robot, comprising: The acquisition module is used for acquiring data to be verified; the configuration module is used for acquiring target configuration data corresponding to the data to be verified according to the equipment model and/or the data version corresponding to the data to be verified; The simulation module is used for inputting the data to be verified, the target configuration data and the control data corresponding to the data to be verified into a simulation quality inspection system, and respectively obtaining a first simulation result corresponding to the data to be verified and a second simulation result corresponding to the control data; and the verification module is used for determining a verification result of the data to be verified according to the first simulation result and the second simulation result.
9. 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 data verification method according to any one of claims 1 to 7 when executing the computer program.
10. A non-transitory computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when executed by a processor, implements the data verification method according to any one of claims 1 to 7.
11. A computer program product comprising a computer program which, when executed by a processor, implements a data verification method as claimed in any one of claims 1 to 7.

Description

Data verification method and device for robot Technical Field The invention relates to the technical field of artificial intelligence, in particular to a data verification method and device for a robot. Background During robotics learning (especially simulated learning and reinforcement learning), high quality training data sets are key to the success of model training. However, the existing training data set often has the problems that (1) personalized verification logic is lacked, because equipment models of robots are various, data format requirements are different, unified data verification cannot be conducted on data from different sources, and (2) the traditional mode is excessively dependent on manpower, users are required to manually load and play data to be verified one by one, so that the efficiency is low, and omission is easy to cause. Therefore, a data verification method of a robot which is convenient to expand and has higher automation degree is needed. Disclosure of Invention The invention provides a data verification method and device for robots, which can obtain target configuration data corresponding to data to be verified of different data sources based on equipment model and/or data version, further can perform automated simulation and verification on the data to be verified according to the target configuration data, effectively meets the personalized verification requirements of different robots, and improves the automation degree of the whole verification link. The invention provides a data verification method of a robot, which comprises the following steps of obtaining to-be-verified data, obtaining target configuration data corresponding to the to-be-verified data according to a device model and/or a data version corresponding to the to-be-verified data, inputting the to-be-verified data, the target configuration data and control data corresponding to the to-be-verified data into a simulation quality inspection system to respectively obtain a first simulation result corresponding to the to-be-verified data and a second simulation result corresponding to the control data, and determining a verification result of the to-be-verified data according to the first simulation result and the second simulation result. Optionally, the data to be verified comprises sensor data acquired based on at least one sensor, and the target configuration data comprises target data; The method comprises the steps of obtaining device models corresponding to all sensors and data versions corresponding to the data to be verified according to the device models and/or the data versions corresponding to the data to be verified, determining target mapping relations corresponding to all the sensor data according to the device models and the data versions from a plurality of preset mapping relations, wherein the mapping relations indicate corresponding relations between field names in the sensor data and target field names in the target data, and converting the field names in the sensor data according to the target mapping relations to obtain the target data. The sensor data comprises left clamping jaw sensor data and right clamping jaw sensor data, wherein the target data comprises clamping jaw opening and closing degree, the target configuration data corresponding to the data to be verified are obtained according to equipment models and/or data versions corresponding to the data to be verified, the method further comprises the steps of determining a clamping jaw opening and closing degree calculation method according to first equipment models of sensors corresponding to the left clamping jaw and second equipment models of sensors corresponding to the right clamping jaw, and the clamping jaw opening and closing degree corresponding to the data to be verified is obtained by calculating the left clamping jaw sensor data and the right clamping jaw sensor data by the aid of the clamping jaw opening and closing degree calculation method. Optionally, the step of inputting the data to be verified, the target configuration data and the control data corresponding to the data to be verified into a simulation quality inspection system to respectively obtain a first simulation result corresponding to the data to be verified and a second simulation result corresponding to the control data comprises the steps of respectively playing back the data to be verified and the control data in a simulation environment of the simulation quality inspection system based on the target configuration data, respectively generating a first motion result corresponding to the data to be verified and a second motion result corresponding to the control data, respectively generating a data curve corresponding to each sensor based on sensor data corresponding to each sensor, respectively generating a first simulation result corresponding to the data to be verified according to the first motion result and the data curve corresponding to each sensor,