CN-121989234-A - Self-learning control method, device, equipment, storage medium and computer program product for online debugging of mechanical arm

CN121989234ACN 121989234 ACN121989234 ACN 121989234ACN-121989234-A

Abstract

The application relates to the technical field of robot control, in particular to a self-learning control method, a device, equipment, a storage medium and a computer program product for online debugging of a mechanical arm. The method comprises the steps of obtaining running state data of a mechanical arm joint and driving related signals to construct a to-be-identified data set, determining task state information based on the to-be-identified data set, generating identification triggering information when a task is switched or an end tool is changed, performing recursive estimation on a dynamic parameter set based on the to-be-identified data set and the task state information after triggering, generating tracking error assessment information based on the dynamic parameter set and the to-be-identified data set, determining identification update control parameters and generating a control gain parameter set, and finally synchronizing the control gain parameter set and the dynamic parameter set, recording debugging data and generating debugging log data.

Inventors

Ou di
XIE PENG
LI KENAN
Ban Bingbing

Assignees

深圳璇玑动力科技有限公司

Dates

Publication Date: 20260508
Application Date: 20260123

Claims (10)

1. The self-learning control method for online debugging of the mechanical arm is characterized by comprising the following steps of: acquiring running state data and driving related signals of each joint of the mechanical arm, and constructing a data set to be identified based on the running state data and the driving related signals; Determining task state information corresponding to the current operation working condition of the mechanical arm based on the data set to be identified, and generating identification triggering information when the task state information indicates task switching or end tool change; when the identification triggering information indicates triggering, based on the data set to be identified and the task state information, updating a dynamic parameter set by adopting recursive estimation type parameter identification processing; Generating tracking error evaluation information based on the dynamic parameter set and the data set to be identified, determining an identification update control parameter based on the tracking error evaluation information, and generating a control gain parameter set based on the dynamic parameter set and the identification update control parameter; And performing parameter synchronization on the control gain parameter set and the dynamic parameter set, and recording debugging data related to the dynamic parameter set, the tracking error evaluation information and the control gain parameter set to generate debugging log data.
2. The method of claim 1, wherein the step of acquiring the operational status data and the drive related signals for each joint of the robotic arm and constructing the set of data to be identified based on the operational status data and the drive related signals comprises: acquiring angular position and angular speed information of each joint in the movement process of the mechanical arm through a preset data acquisition unit, and taking the angular position and the angular speed information as the running state data; collecting current information and moment information of each joint through the preset data collecting unit, and taking the current information and the moment information as the driving related signals; And under a preset control environment, carrying out corresponding association and combination on the running state data and the driving related signals to obtain a data set to be identified for parameter identification processing.
3. The method of claim 1, wherein the step of determining task state information corresponding to the current operating condition of the robotic arm based on the set of data to be recognized and generating recognition trigger information when the task state information indicates a task switch or an end tool change comprises: Extracting working condition characterization information for characterizing the current operation working condition of the mechanical arm based on the to-be-recognized data set, and matching the working condition characterization information with a preset task state set to determine task state information corresponding to the current operation working condition of the mechanical arm; Comparing the task state information corresponding to the current moment with the task state information corresponding to the historical moment in the continuously acquired data set to be identified to obtain task state change judgment information; And generating identification triggering information based on the task state change judging information when the task state change judging information indicates that the task state is switched or the task state information indicates that the end tool is changed.
4. The method of claim 1, wherein the step of updating the set of kinetic parameters using a recursive estimation class parameter identification process based on the set of data to be identified and the task state information when the identification trigger information indicates a trigger, comprises: determining a parameter identification starting condition based on the identification triggering information, and determining a dynamic parameter item to be updated by combining the task state information so as to generate an identification configuration parameter for online parameter identification; inputting the data set to be identified into a preset model identification process under the constraint of the identification configuration parameters, and carrying out recursive estimation on the dynamic parameter item to be updated based on a recursive least square algorithm or an extended Kalman filtering algorithm in the model identification process to obtain candidate dynamic parameters; and carrying out consistency updating on the dynamic parameter records based on the candidate dynamic parameters and the task state information, and determining the updated parameter records as a dynamic parameter set, so that when the identification triggering information indicates triggering, the dynamic parameter set is updated by adopting recursive estimation type parameter identification processing based on the data set to be identified and the task state information.
5. The method of claim 1, wherein the steps of generating tracking error assessment information based on the set of kinetic parameters and the set of data to be identified, and determining an identification update control parameter based on the tracking error assessment information, while generating a set of control gain parameters based on the set of kinetic parameters and the identification update control parameter, comprise: Determining the model consistency of the data set to be identified based on the dynamic parameter set, obtaining model output information corresponding to the data set to be identified, and generating tracking error evaluation information based on deviation between the model output information and running state data in the data set to be identified; determining update strategy information for constraint parameter identification processing based on the tracking error evaluation information, and taking the update strategy information as identification update control parameters, wherein the update strategy information is used for representing learning rate and identification update frequency; and based on the dynamic parameter set and the identification updating control parameter, carrying out self-adaptive updating on the gain parameter of the preset controller to obtain a control gain parameter set.
6. The method of claim 1, wherein the step of parameter synchronizing the set of control gain parameters with the set of dynamics parameters and recording debug data related to the set of dynamics parameters, the tracking error assessment information, and the set of control gain parameters to generate debug log data comprises: Generating parameter release information based on the control gain parameter set and the dynamic parameter set, and sending the parameter release information to a preset node related to mechanical arm control through a preset communication interface so as to trigger a parameter synchronization flow; In the parameter synchronization process, parameter updating processing is performed among the preset nodes based on the parameter release information so as to write the control gain parameter set and the dynamic parameter set into a parameter storage area of a preset control environment, and parameter synchronization of the control gain parameter set and the dynamic parameter set is realized; After the parameter synchronization is completed, recording parameter change information of the dynamic parameter set, error convergence information of the tracking error evaluation information and gain change information of the control gain parameter set, and generating debugging log data based on the recording result.
7. The utility model provides a self-learning controlling means of arm on-line debugging which characterized in that, the device includes: The data acquisition module is used for acquiring the running state data and the driving related signals of each joint of the mechanical arm and constructing a data set to be identified based on the running state data and the driving related signals; The identification triggering module is used for determining task state information corresponding to the current operation working condition of the mechanical arm based on the data set to be identified, and generating identification triggering information when the task state information indicates task switching or end tool change; The parameter updating module is used for updating the dynamic parameter set by adopting recursive estimation type parameter identification processing based on the data set to be identified and the task state information when the identification triggering information indicates triggering; a control gain module for generating tracking error evaluation information based on the dynamic parameter set and the data set to be identified, determining an identification update control parameter based on the tracking error evaluation information, and generating a control gain parameter set based on the dynamic parameter set and the identification update control parameter; and the target module is used for carrying out parameter synchronization on the control gain parameter set and the dynamic parameter set, and recording debugging data related to the dynamic parameter set, the tracking error evaluation information and the control gain parameter set so as to generate debugging log data.
8. A computer device, characterized in that it comprises a memory, a processor and a self-learning control program stored on the memory and capable of on-line debugging of a robot running on the processor, the self-learning control program of on-line debugging of a robot being configured to implement the steps of the self-learning control method of on-line debugging of a robot according to any one of claims 1 to 6.
9. A storage medium, wherein a self-learning control program for on-line debugging of a mechanical arm is stored on the storage medium, and the steps of the self-learning control method for on-line debugging of a mechanical arm according to any one of claims 1 to 6 are realized when the self-learning control program for on-line debugging of a mechanical arm is executed by a processor.
10. A computer program product, characterized in that it comprises a computer program which, when executed by a processor, implements the steps of the self-learning control method for on-line debugging of a robotic arm as claimed in any one of claims 1 to 6.

Description

Self-learning control method, device, equipment, storage medium and computer program product for online debugging of mechanical arm Technical Field The application relates to the technical field of robot control, in particular to a self-learning control method, a device, equipment, a storage medium and a computer program product for online debugging of a mechanical arm. Background Precise control of the robotic arm typically depends on the accuracy of the kinetic model. The existing system generally performs off-line calibration on dynamic parameters such as inertia, friction, gravity center and the like in a factory stage, and accordingly sets relatively fixed control gain parameters for the controller. In practical application, the end effector of the mechanical arm often needs to be replaced by different tools, and the load and the operation condition are changed, so that the dynamic parameters obtained through off-line calibration deviate from the actual operation state, and further the original control parameters are difficult to continuously adapt. Meanwhile, the traditional debugging mode generally depends on manual repeated adjustment of control parameters and track test, the debugging process is complicated, and after task switching or end tool change, resampling and offline modeling are often needed, so that the capability of automatically updating the model and the control parameters in the control process is lacking. In addition, in the control environment, the dynamic parameters and the control gain parameters may be used by different control modules or nodes, and if a unified parameter synchronization and debugging data recording mechanism is absent, consistency and traceability of parameter update are difficult to ensure, so that the stability of the mechanical arm debugging is poor. Therefore, how to improve the stability of the mechanical arm debugging is a technical problem to be solved. Disclosure of Invention The application mainly aims to provide a self-learning control method, device, equipment, storage medium and computer program product for on-line debugging of a mechanical arm, which aim to solve the technical problem of how to improve the stability of the mechanical arm debugging. In order to achieve the above purpose, the application provides a self-learning control method for online debugging of a mechanical arm, which comprises the following steps: acquiring running state data and driving related signals of each joint of the mechanical arm, and constructing a data set to be identified based on the running state data and the driving related signals; Determining task state information corresponding to the current operation working condition of the mechanical arm based on the data set to be identified, and generating identification triggering information when the task state information indicates task switching or end tool change; when the identification triggering information indicates triggering, based on the data set to be identified and the task state information, updating a dynamic parameter set by adopting recursive estimation type parameter identification processing; Generating tracking error evaluation information based on the dynamic parameter set and the data set to be identified, determining an identification update control parameter based on the tracking error evaluation information, and generating a control gain parameter set based on the dynamic parameter set and the identification update control parameter; And performing parameter synchronization on the control gain parameter set and the dynamic parameter set, and recording debugging data related to the dynamic parameter set, the tracking error evaluation information and the control gain parameter set to generate debugging log data. In an embodiment, the step of obtaining the operation state data and the driving related signals of each joint of the mechanical arm and constructing the to-be-identified data set based on the operation state data and the driving related signals includes: acquiring angular position and angular speed information of each joint in the movement process of the mechanical arm through a preset data acquisition unit, and taking the angular position and the angular speed information as the running state data; collecting current information and moment information of each joint through the preset data collecting unit, and taking the current information and the moment information as the driving related signals; And under a preset control environment, carrying out corresponding association and combination on the running state data and the driving related signals to obtain a data set to be identified for parameter identification processing. In an embodiment, the step of determining task state information corresponding to the current operation condition of the mechanical arm based on the to-be-identified data set and generating identification triggering information when the task state information indicat