Industrial robot position correction control method

Abstract

The application discloses an industrial robot position correction control method and relates to the technical field of robot control. A feedforward correction control instruction set Adj with feedforward compensation characteristics is formed, and then is applied to a controller execution path, so that error correction of a motion instruction is realized in advance before the motion instruction is sent, and problems such as unstable dynamic response, static deviation compensation lag, inertia mutation error amplification and the like commonly existing in the starting and stopping stages of an existing industrial robot are effectively solved. After actual motion is completed, the robot motion process is re-collected and calculated to generate an angle time sequence set Ang2 and an actual angle rate set Act2, and an angle error rate set Dif2 used for evaluation and consistent with an initial cycle structure is further constructed. Not only is dynamic feedback evaluation of feedforward control precision realized, but also a data basis is provided for incremental optimization of an error rate response model set Mod, and a complete closed-loop control mechanism is formed.

Description

Technical Field

[0001] This invention relates to the field of robot control technology, specifically to a method for position correction control of an industrial robot. Background Technology

[0002] With the continuous improvement of the intelligence and automation level of manufacturing, industrial robot technology, as a core component of automated production systems, has been widely applied in various manufacturing industries, belonging to the broad technical field of intelligent manufacturing equipment technology. Furthermore, within the industrial robot technology system, control accuracy and repeatability are key performance indicators for measuring operational reliability and production quality, belonging to the technical branch of industrial robot control system optimization. In this branch, the accuracy of robot joint motion control is particularly important as it affects the consistency of the end effector trajectory.

[0003] While most current industrial robot control systems integrate basic control strategies such as closed-loop position control, PID control, and end-effector position feedback, these traditional methods are primarily based on static deviation modeling and hysteresis compensation mechanisms. Specifically, at the robot joint control layer, the common practice is to compensate via a feedback loop after detecting a certain angular deviation in the robot. This "hysteresis control" leads to problems such as untimely short-term angular response and delayed correction signals during high-speed, non-uniform motion changes, especially at the start and end points of the path, resulting in the inability to suppress position errors in a timely manner.

[0004] This control mode is particularly evident when facing dynamic, complex paths or high-speed path segment transitions. Lacking the ability to predict future motion trends, the control system can only passively correct existing errors, lacking an active defense mechanism. Therefore, in applications requiring high dynamic response and high control precision, such as intelligent assembly lines, electronic product welding processes, or high-precision spraying systems, existing control methods cannot fully meet the "fast and accurate" control demands in complex scenarios. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides an industrial robot position correction control method, which solves the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an industrial robot position correction control method, comprising the following steps:

[0007] S1. Real-time acquisition of continuous time-series angle data of each joint of the industrial robot during the movement process, forming an angle time sequence set Ang. At the same time, acquisition of the target angle change rate command received by the industrial robot, generating a target angle rate set Cmd.

[0008] S2. Based on the angle time series set Ang, calculate the rate change during the actual angle change process of the industrial robot to obtain the actual angle rate set Act. Perform difference analysis between the target angle rate set Cmd and the actual angle rate set Act to generate the angle error rate set Dif.

[0009] S3. Using the evaluation angle error rate set Dif2 as feedback input, iteratively generate the updated error rate response model set Mod;

[0010] S4. Based on the error rate response model set Mod, the error of the target action to be executed in the next control cycle of the industrial robot is estimated, and the expected dynamic error value set Pre is generated and fused with the original control command data to form the feedforward correction control command set Adj.

[0011] S5. The industrial robot uses the feedforward correction control instruction set Adj to execute the target action; after execution, the angle timing set Ang is collected again for evaluation and optimization.

[0012] Preferably, S1 includes S11;

[0013] S11. By setting angle detection sensors on each joint of the industrial robot, the joint angle signal is collected at a fixed time sampling period, the angle change value of each joint is continuously recorded during the entire execution cycle, and the data is transmitted and communicated through the Internet of Things and local wireless network.

[0014] The collected angle change values ​​are combined with the collection timestamp of each data point, and then time-aligned and structured for storage to generate an angle time series set Ang with a time sequence structure. The time sampling period range covers the complete motion cycle of the industrial robot.

[0015] Preferably, S1 further includes S12;

[0016] S12. The industrial robot extracts the angle change rate that should be achieved in each control cycle based on the target angle value sequence loaded in the preset task trajectory file and the set execution cycle parameter set, and generates the target angle rate set Cmd.

[0017] The target angle rate set Cmd generates a target angle rate data sequence by performing differential processing on the target angle values ​​in adjacent time periods. Then, the target angle rate data sequence is classified and organized according to joint number and timestamp to form a structured target instruction set, and finally the target angle rate set Cmd is constructed.

[0018] Preferably, S2 includes S21;

[0019] S21. Based on the angle time series set Ang, the angle change of each joint between adjacent time points is differentially processed, and combined with a fixed time sampling period, the actual angular velocity set Vel in each time period is calculated.

[0020] The actual angular velocity set Vel uses the joint number and timestamp of the industrial robot as the index dimension to form a sequence of the actual angular velocity distribution of each joint of the industrial robot during the entire control cycle.

[0021] Based on the actual angular velocity set Vel, a trend analysis of the actual angular velocity distribution sequence over continuous time periods is performed. The actual angular velocity change rate between adjacent time segments is extracted and categorized and integrated in chronological order to form the actual angular velocity set Act.

[0022] Preferably, S2 further includes S22;

[0023] S22. Based on the obtained target angular velocity set Cmd and the actual angular velocity set Act as a comparison basis, perform unified timestamp alignment processing on the two sets;

[0024] After completing the timestamp alignment process, the difference between the target angular velocity and the actual angular velocity of each joint number at each timestamp is calculated to obtain the rate deviation set Tmp representing the response deviation;

[0025] Next, the absolute value of each difference item in the rate deviation set Tmp is processed to remove the directional influence, resulting in the quantization error set Val after quantization.

[0026] The quantization error set Val is sorted and organized in a structured manner according to the joint number and timestamp to generate the angle error rate set Dif, which reflects the dynamic response offset trend within the integrated control cycle of the industrial robot.

[0027] Preferably, S3 includes S31;

[0028] S31. Using the angle error rate set Dif as training samples, input the error trend prediction model constructed using the multivariate linear regression method for training, and obtain the error rate response model set Mod in the subsequent control cycle of the industrial robot.

[0029] The error rate response model set Mod is an error prediction module in the industrial robot control cycle. The input variable is the target angle rate set Cmd that the industrial robot is about to execute, and the output prediction error result is the predicted angle error rate set Pre. The predicted angle error rate set Pre is used to predict the trend of angle error rate change in the next control cycle of the industrial robot.

[0030] Preferably, S3 further includes S32;

[0031] S32. In the subsequent control cycle of the industrial robot, the angle signals of each joint are continuously collected at a fixed time sampling period, and the angular velocity change value between adjacent time points is calculated accordingly. The error is deduced by combining the original control command of the industrial robot, and an evaluation angle error rate set Dif2 is generated to represent the actual response offset of the robot in the current control cycle.

[0032] The predicted angle error rate set Pre is compared with the evaluation angle error rate set Dif2 item by item to obtain the residual data set Res. The residual data set Res is used as the objective function input to incrementally correct the coefficient parameters of the error trend prediction model, generating an updated error rate response model set Mod, which is used to predict the angle error rate change trend in the next period.

[0033] Preferably, S4 includes S41;

[0034] S41. Based on the acquired predicted angle error rate set Pre as the basis for feedforward compensation, the control instructions under each timestamp node and joint number in the target angle rate set Cmd of the next control cycle of the industrial robot are fused.

[0035] The fusion process extracts each timestamp and joint number as a basis, and numerically synthesizes the control command item in the target angle rate set Cmd with the prediction error value in the prediction angle error rate set Pre at the corresponding position to generate the fused control command.

[0036] The fused control commands are structured and summarized according to timestamps and joint numbers to generate a set of feedforward correction control commands Adj, which serves as the control execution input injected into the control system in the next control cycle of the industrial robot, and is used to achieve early compensation of errors before the start of the action.

[0037] Preferably, S5 includes S51;

[0038] S51. In the next control cycle, the industrial robot receives and applies the generated set of feedforward correction control instructions Adj, and executes the predetermined action sequence in chronological order. During the entire execution of the predetermined action sequence, the control system collects the angle signals of each joint in real time at a fixed time sampling period, and records the continuous change process of the angle of each joint over time during the actual work of the industrial robot.

[0039] The control system structures and organizes the collected angle signal timing data to generate an angle timing set Ang2. The angle timing set Ang2 is used to describe the complete motion trajectory of each joint within the execution cycle of the industrial robot. Based on the angle timing set Ang2, the rate is calculated according to the angle difference between adjacent time points and the sampling period to obtain the actual angle rate set Act2. The actual angle rate set Act2 reflects the actual angular velocity response characteristics of each joint of the robot after executing the command.

[0040] Preferably, S5 further includes S52;

[0041] S52. Using the target angular rate set Cmd as a reference, compare the actual response values ​​of each timestamp node and joint number in the actual angular rate set Act2 with the actual response values. Calculate the error by using the item-by-item difference analysis method to obtain the error results for each item. At the same time, to eliminate interference in the positive and negative directions, take the absolute value of each error result to form a preliminary error value set for quantitatively describing the control error amplitude.

[0042] The error results are then categorized in a structured manner according to the timestamp node and joint number dimension to generate the evaluation angle error rate set Dif2. The evaluation angle error rate set Dif2 is used to measure the execution effect of the feedforward compensation control strategy and is a key input set for subsequent feedback training and control optimization.

[0043] The evaluation angle error rate set Dif2 serves as both a feedback sample for the Mod optimization of the error rate response model set in step S3 and a set of execution results used to evaluate the feedforward control effect in step S5.

[0044] This invention provides a position correction control method for industrial robots, which has the following beneficial effects:

[0045] (1) By forming a set of feedforward correction control instructions Adj with feedforward compensation characteristics, and then applying it to the controller execution path, the error of the action instructions is corrected in advance before they are sent. This effectively solves the problems of unstable dynamic response, lag in static deviation compensation and amplification of inertial mutation error that are common in the start-up and stop phases of existing industrial robots. By re-collecting and calculating the robot motion process after the actual action is completed, the angle timing set Ang2 and the actual angle rate set Act2 are generated. Furthermore, the evaluation angle error rate set Dif2, which is consistent with the initial periodic structure, is constructed. This not only realizes the dynamic feedback evaluation of the feedforward control accuracy, but also provides a data basis for the incremental optimization of the error rate response model set Mod, forming a complete closed-loop control mechanism. This not only enables the industrial robot to have the ability to predict the dynamic error trend in the multi-joint coordinated control process, but also significantly reduces the impact of the angular velocity mutation at the start and stop points of the action. It realizes the early response and rapid correction of the dynamic characteristics of the error, significantly improves the accuracy of position correction and execution stability, and makes up for the shortcomings of the traditional static deviation value adjustment method which is difficult to adapt to dynamic control scenarios.

[0046] (2) By using the angle error rate set Dif constructed based on the historical control cycle as the training sample, combined with the time structure input of the target angle rate set Cmd, a multivariate linear regression method is introduced to construct an error rate response model set Mod with dynamic error prediction capability. In subsequent control cycles, the actual execution results are obtained to form an evaluation angle error rate set Dif2, which is compared item by item with the predicted angle error rate set Pre output by the error rate response model set Mod. The residual data set Res is calculated and generated. Based on the current execution feedback results, the coefficient parameters of the error prediction model can be optimized in reverse, and the error rate response model set Mod can be automatically updated in an online incremental manner, thus constructing an error modeling mechanism with cycle-by-cycle self-learning capability.

[0047] (3) By fusing the acquired predicted angle error rate set Pre and the target angle rate set Cmd in a one-to-one correspondence under the dimensions of timestamp and joint number, a feedforward correction control instruction set Adj is constructed and injected into the control system before the start of the actual control cycle, so that the control signal has the ability to correct errors in advance, thereby avoiding the accumulation of errors and transmission to the execution stage, effectively alleviating the initial offset problem caused by response lag or structural inertia, and enhancing the trajectory stability and rate tracking accuracy of the industrial robot in the execution stage. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of the steps of an industrial robot position correction and control method according to the present invention;

[0049] Figure 2This is a graph showing the speed trend of industrial robots. Detailed Implementation

[0050] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0051] Example 1

[0052] This invention provides a position correction control method for an industrial robot. Please refer to [link / reference]. Figure 1 This includes the following steps:

[0053] S1. Real-time acquisition of continuous time-series angle data of each joint of the industrial robot during the movement process, forming an angle time sequence set Ang. At the same time, acquisition of the target angle change rate command received by the industrial robot, generating a target angle rate set Cmd.

[0054] S2. Based on the angle time series set Ang, calculate the rate change during the actual angle change process of the industrial robot to obtain the actual angle rate set Act. Perform difference analysis between the target angle rate set Cmd and the actual angle rate set Act to generate the angle error rate set Dif.

[0055] S3. Using the evaluation angle error rate set Dif2 as feedback input, iteratively generate the updated error rate response model set Mod;

[0056] S4. Based on the error rate response model set Mod, the error of the target action to be executed in the next control cycle of the industrial robot is estimated, and the expected dynamic error value set Pre is generated and fused with the original control command data to form the feedforward correction control command set Adj.

[0057] S5. The industrial robot uses the feedforward correction control instruction set Adj to execute the target action; after execution, the angle timing set Ang is collected again for evaluation and optimization.

[0058] In this embodiment, an error rate response model set Mod is constructed based on the angle error rate set Dif and generated using a linear fitting method. In subsequent control cycles, the predicted angle error rate set Pre and the target angle rate set Cmd output by this model are fused to form a feedforward correction control command set Adj with feedforward compensation characteristics. This Adj is then applied to the controller execution path to achieve error correction before the action command is sent. This effectively solves the problems of unstable dynamic response, lag in static deviation compensation, and amplification of inertial mutation errors commonly found in existing industrial robots during the start-up and stop phases. Furthermore, by re-collecting and calculating the robot's motion process after the actual action is completed, an angle time series set Ang2 is generated, along with the actual... The set of angular rates Act2 is used to further construct the set of angular error rates Dif2 for evaluation, which is consistent with the initial periodic structure. This not only realizes the dynamic feedback evaluation of feedforward control accuracy, but also provides a data foundation for the incremental optimization of the error rate response model set Mod. It forms a complete "prediction-control-feedback-correction" closed-loop control mechanism, which not only enables industrial robots to have the ability to predict dynamic error trends in multi-joint coordinated control, but also significantly reduces the impact of sudden changes in angular velocity at the start and stop points of the action in the control system. It realizes early response and rapid correction of error dynamic characteristics, significantly improves the accuracy of position correction and execution stability, and makes up for the shortcomings of traditional static deviation value adjustment methods that are difficult to adapt to dynamic control scenarios.

[0059] Example 2

[0060] Specifically: S1 includes S11;

[0061] S11. By setting angle detection sensors on each joint of the industrial robot, the joint angle signal is collected at a fixed time sampling period, the angle change value of each joint is continuously recorded during the entire execution cycle, and the data is transmitted and communicated through the Internet of Things and local wireless network.

[0062] The collected angle change values ​​are combined with the collection timestamp of each data point, and then time-aligned and structured for storage to generate an angle time series set Ang with a time sequence structure. The time sampling period range covers the complete motion cycle of the industrial robot, and the sampling interval is set by the user, thereby obtaining a high-resolution continuous data sequence.

[0063] S1 further includes S12;

[0064] S12. The industrial robot extracts the angle change rate that should be achieved in each control cycle based on the target angle value sequence loaded in the preset task trajectory file and the set execution cycle parameter set, and generates the target angle rate set Cmd.

[0065] The target angle rate set Cmd generates a target angle rate data sequence by performing differential processing on the target angle values ​​in adjacent time periods. Then, the target angle rate data sequence is classified and organized according to joint number and timestamp to form a structured target instruction set, and finally the target angle rate set Cmd is constructed.

[0066] The target angle rate set Cmd is used to represent the target rate requirements that each joint of the industrial robot should achieve at different time points. It is a structured expression of the motion intention of the execution path and provides a benchmark for error modeling.

[0067] In this embodiment, by continuously acquiring data on joint angle changes within a complete control cycle, an angle timing set Ang with a timestamp-aligned structure is formed. This set not only ensures the continuity and timing accuracy of the motion process, but also provides high-resolution and highly consistent basic data support for dynamic rate calculation and response modeling in the subsequent control process, significantly improving the ability of the sampled data to express real working conditions.

[0068] Simultaneously, based on the target angle sequence in the preset task trajectory file and combined with the execution cycle parameters, a target angle rate set Cmd is generated in a structured manner. By classifying and managing it according to timestamps and joint numbers, the motion intention of the industrial robot in each control cycle can be accurately expressed in the form of a rate vector. This provides a standardized control benchmark for subsequent error modeling, response comparison, and feedforward correction, significantly improving the control system's ability to identify the difference between the actual response and the ideal command. This enhances the sensitivity and accuracy of error capture, provides a high-confidence data source for subsequent correction mechanisms, and makes up for the unstable error comparison basis caused by insufficient sampling granularity and fuzzy target rate expression in traditional methods.

[0069] Example 3

[0070] Specifically: S2 includes S21;

[0071] S21. Based on the angle time series set Ang, the angle change of each joint between adjacent time points is differentially processed, and combined with a fixed time sampling period, the actual angular velocity set Vel in each time period is calculated.

[0072] The actual angular velocity set Vel uses the joint number and timestamp of the industrial robot as the index dimension to form a sequence of the actual angular velocity distribution of each joint of the industrial robot during the entire control cycle.

[0073] Based on the actual angular velocity set Vel, the distribution sequence trend analysis of actual angular velocity over continuous time periods is performed. The actual angular velocity change rate between adjacent time segments is extracted and categorized and integrated in chronological order to form the actual angular velocity set Act. The actual angular velocity set Act, as an expression of the rate of change of angular velocity, can reflect the robot's response sensitivity in different time periods and is a key input set for error modeling and trend prediction.

[0074] S2 further includes S22;

[0075] S22. Based on the obtained target angular velocity set Cmd and the actual angular velocity set Act as a comparison basis, perform unified timestamp alignment processing on the two sets;

[0076] After completing the timestamp alignment process, the difference between the target angular velocity and the actual angular velocity of each joint number at each timestamp is calculated to obtain the rate deviation set Tmp representing the response deviation;

[0077] Next, the absolute value of each difference item in the rate deviation set Tmp is processed to remove the directional influence, resulting in the quantization error set Val after quantization.

[0078] The quantization error set Val is sorted and organized in a structured manner according to the joint number and timestamp to generate the angle error rate set Dif, which reflects the dynamic response offset trend within the integrated control cycle of the industrial robot.

[0079] The angle error rate set Dif is used to directly quantify the degree of rate response deviation of the industrial robot relative to the control command target during task execution, and serves as the key input set for subsequently constructing the error rate response model set Mod and generating the feedforward correction control command set Adj.

[0080] In this embodiment, by performing differential calculation on the angle time series set Ang and combining it with a fixed time sampling period, a complete actual angular velocity set Vel is constructed. Based on this, the velocity change trend within a continuous time period is analyzed in depth, and the actual angle rate set Act with response dynamic characteristics is extracted. This enables the control system to not only obtain the velocity response information under each control node, but also to grasp its continuous change pattern in the time dimension, thereby effectively revealing the dynamic sensitivity and time series offset characteristics of the actual response of the industrial robot.

[0081] Building upon this foundation, a multi-dimensional difference structure between the actual angle rate set Act and the target command is constructed using the target angle rate set Cmd as a benchmark. By directional normalization and structured archiving of each difference result, an angle error rate set Dif is ultimately generated for fine-grained quantification of dynamic response errors. This set, representing the rate offset between each joint and the control target throughout the entire control cycle under real-world industrial robot operation, significantly improves the accuracy and coverage of error modeling. It not only enables the identification of rate hierarchy differences in the dynamic response process but also solves the problem in traditional solutions where only static angle offsets are collected, failing to capture continuous response anomalies during execution. This lays a highly stable and high-resolution error input foundation for subsequently constructing an error rate response model set Mod with trend prediction capabilities and generating a feedforward correction control command set Adj, ensuring stronger dynamic adaptability of error compensation.

[0082] Example 4

[0083] Please see Figure 1 and Figure 2 Specifically: S3 includes S31;

[0084] S31. Using the angle error rate set Dif as training samples, input the error trend prediction model constructed using the multivariate linear regression method for training, and obtain the error rate response model set Mod in the subsequent control cycle of the industrial robot.

[0085] The error rate response model set Mod is an error prediction module in the industrial robot control cycle. The input variable is the target angle rate set Cmd that the industrial robot is about to execute, and the output prediction error result is the predicted angle error rate set Pre. The predicted angle error rate set Pre is used to predict the trend of angle error rate change in the next control cycle of the industrial robot.

[0086] The training process for the error trend prediction model is as follows:

[0087] Based on the angle error rate set Dif constructed in the previous control cycle as the supervised learning target output variable, and the target angle rate set Cmd, joint torque data set Tor and load state set Lod under the corresponding timestamp in this control cycle as input feature variables, a structured training sample set is constructed.

[0088] Among them, the target angle rate set Cmd provides the ideal rate control target of each joint at a specified time node, the joint torque data set Tor describes the actual force state applied by each joint drive system during the movement, and the load state set Lod characterizes the load distribution and center of gravity offset of the robot end effector at different time points. These multi-dimensional features together reflect the dynamic response basis of the robot under actual working conditions.

[0089] The mapping relationship between the above input feature variables and supervised output variables is modeled and trained using multivariate linear regression. The objective function is to minimize the sum of squared residuals between the predicted output and the actual error. Finally, an error trend prediction model with dynamic error trend prediction capability is fitted. After training, the predicted value is output based on the target angle rate set Cmd to be executed as input, that is, the predicted angle error rate set Pre is constructed to represent the expected error change trend of each joint in the time dimension in the next control cycle.

[0090] The predicted angle error rate set Pre is a key intermediate result used to generate the feedforward correction control instruction set Adj, which supports the control system in injecting error compensation amount in advance before the action is executed, thereby improving the response foresight and control accuracy of industrial robots under dynamic working conditions.

[0091] S3 further includes S32;

[0092] S32. In the subsequent control cycle of the industrial robot, the angle signals of each joint are continuously collected at a fixed time sampling period, and the angular velocity change value between adjacent time points is calculated accordingly. The error is deduced by combining the original control command of the industrial robot, and an evaluation angle error rate set Dif2 is generated to represent the actual response offset of the robot in the current control cycle.

[0093] The predicted angle error rate set Pre and the evaluation angle error rate set Dif2 are compared item by item to obtain the residual data set Res. The residual data set Res is used as the input of the objective function to incrementally correct the coefficient parameters of the error trend prediction model, and generate the updated error rate response model set Mod, which is used to predict the angle error rate change trend in the next period, as shown in Table 1.

[0094] Meanwhile, it should be noted that: the error rate response model set Mod is the error trend prediction model updated in the previous cycle; the target angle rate set Cmd is the target control command set to be executed in the next cycle. When using the multivariate linear regression method to construct the error trend prediction model, the joint torque data set Tor and the load state set Lod are introduced as input feature parameters to enhance the model's adaptability under different motion states and load conditions.

[0095] The joint torque data set Tor is a collection of torque response data collected by the torque monitoring sensor embedded in the industrial robot from each joint execution unit in each control cycle. The torque monitoring sensor analyzes the drive current and the response parameters of the joint motor to calculate the actual force state of each joint at a specified time point in real time. The joint torque data set Tor and the angle error rate set Dif have a unified timestamp correspondence to ensure that the dynamic correlation between error and torque response can be accurately modeled.

[0096] The load state set Lod is periodically collected by the load sensing sensor integrated into the end effector of the industrial robot. The load sensing sensor determines the end load distribution in the current operation cycle based on end gripper feedback, torque sensor and joint acceleration, including tool weight, changes in the weight of the held workpiece and eccentric inertia caused by the adjustment of the gripping method. Each record item in the load state set Lod is accompanied by a timestamp identifier and the input correlation between it and the angle error rate set Dif and the joint torque data set Tor.

[0097] Table 1: Model Prediction and Error Evaluation Table

[0098] Timestamp Joint number Predicted angle error rate set Pre (° / s) Evaluation of the angular error rate set Dif2 (° / s) Residual dataset Res (° / s) T1 J1 0.6 0.5 0.1 T1 J2 0.4 0.7 0.3 T2 J1 0.3 0.5 0.2

[0099] In this embodiment, by using the angle error rate set Dif constructed based on historical control cycles as training samples, combined with the time structure input of the target angle rate set Cmd, a multivariate linear regression method is introduced to construct an error rate response model set Mod with dynamic error prediction capabilities. By introducing the joint torque data set Tor and the load state set Lod as supplementary feature parameters, a multi-input error prediction model that can adapt to changes in operating conditions is constructed, significantly improving the generalization ability and structural adaptability of the error prediction model in complex scenarios. In subsequent control cycles, the evaluation angle error rate set Dif2 formed by the actual execution results is obtained and compared item by item with the predicted angle error rate set Pre output by the error rate response model set Mod, and a residual data set Res is calculated to generate the residual data set Res. The coefficient parameters of the error prediction model can be optimized in reverse based on the current execution feedback results, and the error rate response model set Mod is automatically updated in an online incremental manner, thus constructing an error modeling mechanism with cycle-by-cycle self-learning capabilities.

[0100] While possessing the ability to predict the future trend of angular error rate changes, it can also dynamically adjust the model structure parameters based on the mechanical state data and execution feedback errors collected in real time during the control cycle. This enables the error prediction mechanism to not only reflect historical statistical characteristics but also accurately respond to the real response offset characteristics of the industrial robot under the current task state. As a result, it achieves a higher precision and more robust control compensation basic modeling process under multiple tasks, multiple loads, and multiple working conditions, solving the structural rigidity problem of prediction models relying on static training and lacking adaptability in existing technologies.

[0101] Example 5

[0102] Specifically: S4 includes S41;

[0103] S41. Based on the acquired predicted angle error rate set Pre as the basis for feedforward compensation, the control instructions under each timestamp node and joint number in the target angle rate set Cmd of the next control cycle of the industrial robot are fused.

[0104] The fusion process extracts each timestamp and joint number as a basis, and numerically synthesizes the control command item in the target angle rate set Cmd with the prediction error value in the prediction angle error rate set Pre at the corresponding position to generate a fused control command, which is used to represent the angle rate control amount that should actually be sent after compensation.

[0105] The fused control commands are structured and summarized according to timestamps and joint numbers to generate a set of feedforward correction control commands Adj. This set serves as the control execution input injected into the control system in the next control cycle of the industrial robot, enabling error cancellation before the start of the action and improving execution stability and accuracy.

[0106] S5 includes S51;

[0107] S51. In the next control cycle, the industrial robot receives and applies the generated set of feedforward correction control instructions Adj, and executes the predetermined action sequence in chronological order. During the entire execution of the predetermined action sequence, the control system collects the angle signals of each joint in real time at a fixed time sampling period, and records the continuous change process of the angle of each joint over time during the actual work of the industrial robot.

[0108] The control system structures and organizes the collected angle signal timing data to generate an angle timing set Ang2. The angle timing set Ang2 is used to describe the complete motion trajectory of each joint within the execution cycle of the industrial robot. Based on the angle timing set Ang2, the rate is calculated according to the angle difference between adjacent time points and the sampling period to obtain the actual angle rate set Act2. The actual angle rate set Act2 reflects the actual angular velocity response characteristics of each joint of the robot after executing the command.

[0109] S5 also includes S52;

[0110] S52. Using the target angular rate set Cmd as a reference, compare the actual response values ​​of each timestamp node and joint number in the actual angular rate set Act2 with the actual response values. Calculate the error by using the item-by-item difference analysis method to obtain the error results for each item. At the same time, to eliminate interference in the positive and negative directions, take the absolute value of each error result to form a preliminary error value set for quantitatively describing the control error amplitude.

[0111] The error results are then categorized in a structured manner according to the timestamp node and joint number dimension to generate the evaluation angle error rate set Dif2. The evaluation angle error rate set Dif2 is used to measure the execution effect of the feedforward compensation control strategy and is a key input set for subsequent feedback training and control optimization.

[0112] The evaluation angle error rate set Dif2 serves as both a feedback sample for the Mod error rate response model set optimization in step S3 and an execution result set for evaluating the feedforward control effect in step S5, forming a unified error evaluation basis throughout the entire process of prediction modeling and execution optimization.

[0113] Meanwhile, the change in error between the angle error rate set Dif and the evaluation angle error rate set Dif2 is used as the criterion. If the error reduction is within the preset threshold range, it indicates that the industrial robot's execution effect has been significantly improved after applying the feedforward correction control instruction set Adj, and the error compensation strategy is effective. Otherwise, if the execution effect is not improved, the initialization and reconstruction process of the error trend prediction model is automatically triggered. The angle error rate set Dif is used as the training sample again to execute the model construction and parameter learning process and generate an updated error rate response model set Mod.

[0114] The initial reconstruction process includes using the residual data set Res as the objective function input and employing the recursive least squares method to perform online incremental correction of the regression coefficients in the error trend prediction model to obtain the updated error rate response model set Mod.

[0115] In this embodiment, the predicted angle error rate set Pre and the target angle error rate set Cmd are fused one-to-one in the dimensions of timestamp and joint number to form a feedforward correction control instruction set Adj. This set is then injected into the control system before the start of the actual control cycle, enabling the control signal to have the ability to correct errors in advance. This avoids the accumulation of errors and their transmission to the execution stage, effectively alleviating the initial offset problem caused by response lag or structural inertia, and enhancing the trajectory stability and rate tracking accuracy of the industrial robot during the execution stage.

[0116] During the execution of the feedforward correction control command set Adj, joint angle signals are continuously acquired, and an angle timing set Ang2 and an actual angle rate set Act2 are constructed. Based on these, and combined with the initial target angle rate set Cmd, a new evaluation angle error rate set Dif2 is dynamically generated. This not only accurately characterizes the robot's response deviation in actual control, but also serves as a unified evaluation index throughout the control effect verification and error modeling update stages. This achieves a high degree of coupling between the predictive model and the execution feedback, enabling the error modeling strategy and control compensation mechanism to form a data closed loop.

[0117] Specifically, a comparison strategy based on the change in error amplitude between the angle error rate set Dif and the evaluation angle error rate set Dif2 is introduced as the basis for judging whether the compensation effect is valid. Once the compensation is determined to be invalid, the system will automatically trigger the reconstruction process of the error rate response model set Mod, re-model and train based on the angle error rate set Dif, and incrementally update the model coefficients using the recursive least squares method. This ensures that the model maintains high prediction accuracy and stability during continuous execution, and ensures that the entire error control system has adaptive learning capability and fault-tolerant recovery capability. This solves the problems of existing control methods being unable to dynamically determine the compensation effect and lacking a closed-loop correction mechanism.

[0118] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for position correction and control of an industrial robot, characterized in that: Includes the following steps: S1. Real-time acquisition of continuous time-series angle data of each joint of the industrial robot during the movement process, forming an angle time sequence set Ang. At the same time, acquisition of the target angle change rate command received by the industrial robot, generating a target angle rate set Cmd. S2. Based on the angle time series set Ang, calculate the rate change during the actual angle change process of the industrial robot to obtain the actual angle rate set Act. Perform difference analysis between the target angle rate set Cmd and the actual angle rate set Act to generate the angle error rate set Dif. S3. Using the evaluation angle error rate set Dif2 as feedback input, iteratively generate the updated error rate response model set Mod; S4. Based on the error rate response model set Mod, the error of the target action to be executed in the next control cycle of the industrial robot is estimated, and the expected dynamic error value set Pre is generated and fused with the original control command data to form the feedforward correction control command set Adj. S5. The industrial robot uses the feedforward correction control instruction set Adj to execute the target action; after execution, the angle timing set Ang is collected again for evaluation and optimization.

2. The industrial robot position correction control method according to claim 1, characterized in that: S1 includes S11; S11. By setting angle detection sensors on each joint of the industrial robot, the joint angle signal is collected at a fixed time sampling period, the angle change value of each joint is continuously recorded during the entire execution cycle, and the data is transmitted and communicated through the Internet of Things and local wireless network. The collected angle change values ​​are combined with the collection timestamp of each data point, and then time-aligned and structured for storage to generate an angle time series set Ang with a time sequence structure. The time sampling period range covers the complete motion cycle of the industrial robot.

3. The industrial robot position correction control method according to claim 2, characterized in that: S1 further includes S12; S12. The industrial robot extracts the angle change rate that should be achieved in each control cycle based on the target angle value sequence loaded in the preset task trajectory file and the set execution cycle parameter set, and generates the target angle rate set Cmd. The target angle rate set Cmd generates a target angle rate data sequence by performing differential processing on the target angle values ​​in adjacent time periods. Then, the target angle rate data sequence is classified and organized according to joint number and timestamp to form a structured target instruction set, and finally the target angle rate set Cmd is constructed.

4. The industrial robot position correction control method according to claim 3, characterized in that: S2 includes S21; S21. Based on the angle time series set Ang, the angle change of each joint between adjacent time points is differentially processed, and combined with a fixed time sampling period, the actual angular velocity set Vel in each time period is calculated. The actual angular velocity set Vel uses the joint number and timestamp of the industrial robot as the index dimension to form a sequence of the actual angular velocity distribution of each joint of the industrial robot during the entire control cycle. Based on the actual angular velocity set Vel, a trend analysis of the actual angular velocity distribution sequence over continuous time periods is performed. The actual angular velocity change rate between adjacent time segments is extracted and categorized and integrated in chronological order to form the actual angular velocity set Act.

5. The industrial robot position correction control method according to claim 4, characterized in that: S2 further includes S22; S22. Based on the obtained target angular velocity set Cmd and the actual angular velocity set Act as a comparison basis, perform unified timestamp alignment processing on the two sets; After completing the timestamp alignment process, the difference between the target angular velocity and the actual angular velocity of each joint number at each timestamp is calculated to obtain the rate deviation set Tmp representing the response deviation; Next, the absolute value of each difference item in the rate deviation set Tmp is processed to remove the directional influence, resulting in the quantization error set Val after quantization. The quantization error set Val is sorted and organized in a structured manner according to the joint number and timestamp to generate the angle error rate set Dif, which reflects the dynamic response offset trend within the integrated control cycle of the industrial robot.

6. The industrial robot position correction control method according to claim 5, characterized in that: S3 includes S31; S31. Using the angle error rate set Dif as training samples, input the error trend prediction model constructed using the multivariate linear regression method for training, and obtain the error rate response model set Mod in the subsequent control cycle of the industrial robot. The error rate response model set Mod is an error prediction module in the industrial robot control cycle. The input variable is the target angle rate set Cmd that the industrial robot is about to execute, and the output prediction error result is the predicted angle error rate set Pre. The predicted angle error rate set Pre is used to predict the trend of angle error rate change in the next control cycle of the industrial robot.

7. The industrial robot position correction control method according to claim 6, characterized in that: S3 further includes S32; S32. In the subsequent control cycle of the industrial robot, the angle signals of each joint are continuously collected at a fixed time sampling period, and the angular velocity change value between adjacent time points is calculated accordingly. The error is deduced by combining the original control command of the industrial robot, and an evaluation angle error rate set Dif2 is generated to represent the actual response offset of the robot in the current control cycle. The predicted angle error rate set Pre is compared with the evaluation angle error rate set Dif2 item by item to obtain the residual data set Res. The residual data set Res is used as the objective function input to incrementally correct the coefficient parameters of the error trend prediction model, generating an updated error rate response model set Mod, which is used to predict the angle error rate change trend in the next period.

8. The industrial robot position correction control method according to claim 7, characterized in that: S4 includes S41; S41. Based on the acquired predicted angle error rate set Pre as the basis for feedforward compensation, the control instructions under each timestamp node and joint number in the target angle rate set Cmd of the next control cycle of the industrial robot are fused. The fusion process extracts each timestamp and joint number as a basis, and numerically synthesizes the control command item in the target angle rate set Cmd with the prediction error value in the prediction angle error rate set Pre at the corresponding position to generate the fused control command. The fused control commands are structured and summarized according to timestamps and joint numbers to generate a set of feedforward correction control commands Adj, which serves as the control execution input injected into the control system in the next control cycle of the industrial robot, and is used to achieve early compensation of errors before the start of the action.

9. The industrial robot position correction control method according to claim 8, characterized in that: S5 includes S51; S51. In the next control cycle, the industrial robot receives and applies the generated set of feedforward correction control instructions Adj, and executes the predetermined action sequence in chronological order. During the entire execution of the predetermined action sequence, the control system collects the angle signals of each joint in real time at a fixed time sampling period, and records the continuous change process of the angle of each joint over time during the actual work of the industrial robot. The control system structures and organizes the collected angle signal timing data to generate an angle timing set Ang2. The angle timing set Ang2 is used to describe the complete motion trajectory of each joint within the execution cycle of the industrial robot. Based on the angle timing set Ang2, the actual angle rate set Act2 is obtained by combining the angle difference results between adjacent time points with the sampling period.

10. The industrial robot position correction control method according to claim 9, characterized in that: S5 also includes S52; S52. Using the target angular rate set Cmd as a reference, compare the actual response values ​​of each timestamp node and joint number in the actual angular rate set Act2 with the actual response values. Calculate the error by using the item-by-item difference analysis method to obtain the error results for each item. At the same time, to eliminate interference in the positive and negative directions, take the absolute value of each error result to form a preliminary error value set for quantitatively describing the control error amplitude. The error results are then categorized in a structured manner according to the timestamp node and joint number dimension to generate the evaluation angle error rate set Dif2. The evaluation angle error rate set Dif2 is used to measure the execution effect of the feedforward compensation control strategy and is a key input set for subsequent feedback training and control optimization. The evaluation angle error rate set Dif2 serves as both a feedback sample for the Mod optimization of the error rate response model set in step S3 and a set of execution results used to evaluate the feedforward control effect in step S5.

Images