A method and system for coordinated optimization control of robot arm motion

Abstract

The application discloses a kind of mechanical arm motion coordination optimization control method and system, it includes: based on the initial and desired target pose of mechanical arm determines the difference between them and obtains pose error vector;Determine the vector error of mechanical arm based on pose error vector, and determine the control matrix of mechanical arm based on it;Based on control matrix, the operating parameters of mechanical arm are simulated control, and determine the pose error characteristics of mechanical arm after finishing;Based on pose error characteristics, the pose error state of mechanical arm is analyzed and evaluated, and obtains pose error state evaluation value;Based on pose error state evaluation value, optimization coefficient matrix is determined to optimize control matrix and obtains optimized motion control matrix.The application can pre-insight potential motion performance bottleneck by establishing control matrix and carrying out front simulation and quantitative evaluation, and dynamically optimize through data-driven adaptive optimization coefficient matrix, improve the flexibility and accuracy of mechanical arm, ensure that motion process is stable.

Description

A method and system for optimizing motion coordination of a robotic arm Technical Field

[0001] This invention relates to the field of human robotic arm optimization control technology, and in particular to a method and system for optimizing the motion coordination of a robotic arm. Background Technology

[0002] In modern industrial scenarios such as intelligent manufacturing, precision assembly, and flexible production lines, the robotic arm, as the core execution unit, directly determines the operation quality and efficiency of the entire system through the motion accuracy, coordination, and adaptability of its end effector.

[0003] However, traditional robotic arm control methods typically rely on preset, static commands or fixed trajectories. These methods have significant inherent limitations; inherent deviations such as modeling errors and transmission backlashes in the robotic arm's actuator cannot be compensated for in real time during control. This causes initial small errors to accumulate and amplify during movement, ultimately leading to a severe decrease in end-effector positioning accuracy and failure of precision operations such as grasping and assembly. These problems severely restrict the reliable application of robotic arms in environments with high precision requirements and high dynamic uncertainty. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention provides a method and system for optimizing the motion coordination of a robotic arm, comprising:

[0005] Acquire real-time visual information images of the robotic arm and determine the initial pose of the robotic arm based on the real-time visual information images;

[0006] Determine the desired target pose of the robotic arm, analyze the difference between the desired target pose and the initial pose, and generate a pose error vector of the robotic arm containing six degrees of freedom.

[0007] The vector error of the robotic arm in each degree of freedom is determined based on the pose error vector, and the control matrix of the robotic arm is determined based on the vector error in each degree of freedom.

[0008] The operating parameters of the robotic arm are simulated and controlled based on the control matrix, and the pose error characteristics of the robotic arm in each degree of freedom are determined after the simulation and control.

[0009] Based on the pose error characteristics, the pose error state of the robotic arm in each degree of freedom is analyzed and evaluated to obtain the pose error state evaluation value of the robotic arm in each degree of freedom.

[0010] The optimal coefficient matrix is ​​determined based on the pose error state evaluation value, and the control matrix is ​​optimized based on the optimal coefficient matrix to obtain the optimal motion control matrix of the robotic arm.

[0011] Furthermore, the step of acquiring real-time visual information images of the robotic arm and determining the initial pose of the robotic arm based on the real-time visual information images includes:

[0012] The robot arm's real-time visual information images are acquired using a 3D camera, and the 6D pose of the robot arm's end effector in the camera coordinate system is calculated from the real-time visual information images.

[0013] By using hand-eye calibration, the 6D pose of the end effector in the camera coordinate system is transformed to the coordinate system of the robotic arm, thus obtaining the initial pose of the robotic arm.

[0014] Furthermore, the process of determining the desired target pose of the robotic arm and analyzing the difference between the desired target pose and the initial pose to generate a pose error vector for the robotic arm containing six degrees of freedom includes:

[0015] Determine the desired target pose of the robotic arm, and represent the desired target pose and the initial pose in vector form containing six degrees of freedom, respectively, to obtain the desired target pose vector and the initial pose vector;

[0016] The vector errors between the desired target pose vector and the initial pose vector at each degree of freedom are calculated separately, and the vector errors at each degree of freedom are synthesized to obtain the pose error vector of the robotic arm.

[0017] Furthermore, the step of determining the vector error of the robotic arm in each degree of freedom based on the pose error vector, and determining the control matrix of the robotic arm based on the vector error in each degree of freedom, includes:

[0018] The pose error vector is decomposed to obtain the pose error of the robotic arm in each degree of freedom, and the pose error in each degree of freedom is analyzed to determine the control coefficients in each degree of freedom.

[0019] The control coefficients of each degree of freedom are coupled and synthesized to obtain the control matrix of the robotic arm.

[0020] Furthermore, the analysis of pose errors in each degree of freedom and the determination of control coefficients in each degree of freedom include:

[0021] The pose error on each degree of freedom is divided according to linear position and orientation angle to obtain linear position error and orientation angle error, and the preset correspondence between linear position error and linear position control coefficient and the preset correspondence between orientation angle error and orientation angle control coefficient are determined.

[0022] Each linear position error is mapped to a preset linear position error-linear position control coefficient correspondence to obtain each linear position control coefficient corresponding to each linear position error.

[0023] Each orientation angle error is mapped to a preset orientation angle error-orientation angle control coefficient correspondence to obtain the orientation angle control coefficient corresponding to each orientation angle error.

[0024] The control coefficients for each degree of freedom are constructed based on the linear position control coefficients and the orientation angle control coefficients.

[0025] Furthermore, the simulation control of the robotic arm's operating parameters based on the control matrix, and the determination of the robotic arm's pose error characteristics in each degree of freedom after simulation control, include:

[0026] A pre-defined simulation model of the robotic arm's motion is determined, and the control matrix is ​​input into the simulation model to obtain the simulation results.

[0027] The settling time, overshoot, steady-state error, and number of oscillations of the robotic arm in each degree of freedom are extracted from the simulation results, and the settling time, overshoot, steady-state error, and number of oscillations are determined as the pose error characteristics of the robotic arm in each degree of freedom.

[0028] Furthermore, the analysis and evaluation of the pose error state of the robotic arm in each degree of freedom based on pose error characteristics yields an evaluation value of the pose error state of the robotic arm in each degree of freedom, including:

[0029] Determine the ideal and acceptable worst values ​​for settling time, steady-state error, and number of oscillations respectively, and calculate the settling time score, steady-state error score, and number of oscillations score based on the settling time, steady-state error, and number of oscillations and their corresponding ideal and acceptable worst values ​​respectively;

[0030] Determine the ideal value and upper tolerance limit of the overshoot, and calculate the overshoot score based on the overshoot, ideal value, and upper tolerance limit.

[0031] The reference scores for settling time, steady-state error, oscillation frequency, and overshoot are determined respectively. The differences between the settling time score, steady-state error score, oscillation frequency score, and overshoot score and their corresponding reference scores are calculated to obtain the score difference values ​​for each pose error feature.

[0032] Determine the working task scenario of the robotic arm, and determine the weights of each pose error feature based on the working task scenario;

[0033] The difference in scores for each pose error feature is evaluated to obtain the sub-error state evaluation value of each pose error feature. The sub-error state evaluation value of each pose error feature is then weighted and added with the corresponding weight to obtain the pose error state evaluation value of the robotic arm in each degree of freedom.

[0034] Furthermore, the step of determining the optimization coefficient matrix based on the pose error state evaluation value, and optimizing the control matrix based on the optimization coefficient matrix to obtain the optimized motion control matrix of the robotic arm, includes:

[0035] Determine the historical control coefficients and corresponding historical pose error state evaluation values ​​for each degree of freedom of the robotic arm, and construct an optimization coefficient prediction model for each degree of freedom based on the historical control coefficients and corresponding historical pose error state evaluation values ​​for each degree of freedom and a preset neural network model.

[0036] The pose error state evaluation values ​​of the robotic arm in each degree of freedom are input into the corresponding optimization coefficient prediction model to obtain the optimization coefficients of the robotic arm in each degree of freedom. The optimization coefficients in each degree of freedom are coupled and synthesized to obtain the optimization coefficient matrix of the robotic arm.

[0037] The optimized motion control matrix of the robotic arm is obtained by multiplying the optimization coefficient matrix with the control matrix.

[0038] Furthermore, the step of constructing an optimization coefficient prediction model for each degree of freedom based on the historical control coefficients and corresponding historical pose error state evaluation values ​​for each degree of freedom and a preset neural network model includes:

[0039] Based on the historical pose error state evaluation values ​​and historical control coefficients, corresponding datasets are constructed, and each dataset is input into the corresponding preset neural network model to construct the initial model for predicting the optimization coefficients on each degree of freedom.

[0040] Each dataset is divided into a training set and a test set according to a preset ratio, and the training set and test set are input into the corresponding initial model for predicting optimization coefficients;

[0041] The initial models for predicting each optimization coefficient are trained and tested until they meet the preset convergence conditions, thus obtaining the optimization coefficient prediction models for each degree of freedom.

[0042] The present invention also provides a robotic arm motion coordination optimization control system, comprising:

[0043] The acquisition module is used to acquire real-time visual information images of the robotic arm and determine the initial pose of the robotic arm based on the real-time visual information images.

[0044] The generation module is used to determine the desired target pose of the robotic arm, analyze the difference between the desired target pose and the initial pose, and generate a pose error vector of the robotic arm containing six degrees of freedom.

[0045] The determination module is used to determine the vector error of the robotic arm in each degree of freedom based on the pose error vector, and to determine the control matrix of the robotic arm based on the vector error in each degree of freedom.

[0046] The simulation module is used to simulate and control the operating parameters of the robotic arm based on the control matrix, and to determine the pose error characteristics of the robotic arm in each degree of freedom after the simulation and control.

[0047] The evaluation module is used to analyze and evaluate the pose error state of the robotic arm in each degree of freedom based on the pose error characteristics, and obtain the pose error state evaluation value of the robotic arm in each degree of freedom.

[0048] The optimization module is used to determine the optimization coefficient matrix based on the pose error state evaluation value, and optimize the control matrix based on the optimization coefficient matrix to obtain the optimized motion control matrix of the robotic arm.

[0049] Compared with existing technologies, the beneficial effects of the robotic arm motion coordination optimization control method and system of this invention are as follows:

[0050] This invention achieves real-time perception of the environment and its own state by acquiring real-time visual information images of the robotic arm and calculating the initial pose, enabling the robotic arm to have autonomous positioning capabilities and laying a perceptual foundation for subsequent precise motion control. Combined with the six-degree-of-freedom pose error vector generated by the desired target pose, it can quantify the multi-dimensional spatial deviation between the current state and the target state, thereby transforming the complex motion control problem into a specific mathematical control target.

[0051] This invention uses simulation control to verify and rehearse the control strategy before actual execution, and can detect potential motion jitter, overshoot or instability in advance, which can significantly reduce the risk of collision and debugging costs in actual operation.

[0052] This invention analyzes and evaluates the pose error characteristics of each degree of freedom after simulation, which can identify specific problems in the control process such as response lag, insufficient convergence speed or large steady-state error. It reflects the control performance of each degree of freedom with quantitative evaluation values. Based on these evaluation values, the optimization coefficient matrix can be used to make targeted corrections and adaptive adjustments to the original control matrix, thereby forming an optimized motion control matrix.

[0053] In summary, this invention not only calculates pose error in real time using visual information, but also introduces an online optimization mechanism based on multi-degree-of-freedom error feature analysis. By establishing a control matrix and performing pre-simulation and quantitative evaluation, potential motion performance bottlenecks can be identified in advance. Then, through a data-driven adaptive optimization coefficient matrix, the control strategy is dynamically adjusted, enabling the robotic arm to achieve not only fast and accurate pose alignment when dealing with dynamic tasks, but also ensuring the smoothness, coordination, and energy consumption optimization of the entire motion process. This significantly improves its autonomous operation capability in complex and unstructured environments and the overall intelligence level of the system. Attached Figure Description

[0054] Figure 1 is a schematic diagram of the flow structure of the robotic arm motion coordination optimization control method in an embodiment of the present invention;

[0055] Figure 2 is a schematic diagram of the composition of the robotic arm motion coordination optimization control system in an embodiment of the present invention. Detailed Implementation

[0056] The specific embodiments of this application will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and are not intended to limit the scope of the invention.

[0057] As shown in Figure 1, in an embodiment of this application, a method for optimizing and controlling the motion coordination of a robotic arm is provided, comprising: S100: acquiring real-time visual information images of the robotic arm and determining the initial pose of the robotic arm based on the real-time visual information images; S200: determining the desired target pose of the robotic arm and analyzing the difference between the desired target pose and the initial pose to generate a pose error vector of the robotic arm containing six degrees of freedom; S300: determining the vector error of the robotic arm in each degree of freedom based on the pose error vector, and determining the control matrix of the robotic arm based on the vector error in each degree of freedom; S400: performing simulation control on the operating parameters of the robotic arm based on the control matrix, and determining the pose error characteristics of the robotic arm in each degree of freedom after simulation control; S500: analyzing and evaluating the pose error state of the robotic arm in each degree of freedom based on the pose error characteristics to obtain the pose error state evaluation value of the robotic arm in each degree of freedom; S600: determining the optimization coefficient matrix based on the pose error state evaluation value, and optimizing the control matrix based on the optimization coefficient matrix to obtain the optimized motion control matrix of the robotic arm.

[0058] Furthermore, this invention achieves real-time perception of the environment and its own state by acquiring real-time visual information images of the robotic arm and calculating its initial pose, enabling the robotic arm to have autonomous positioning capabilities. This lays the perceptual foundation for subsequent precise motion control. Combined with the six-degree-of-freedom pose error vector generated by the desired target pose, it can quantify the multi-dimensional spatial deviation between the current state and the target state, thereby transforming the complex motion control problem into a specific mathematical control objective. Through simulation control, this invention can verify and rehearse the control strategy before actual execution, identifying potential motion jitter, overshoot, or instability in advance, significantly reducing the risk of collisions and debugging costs during actual operation. By analyzing and evaluating the pose error characteristics of each degree of freedom after simulation, this invention can identify response lag, insufficient convergence speed, or steady-state error deviations in the control process. This invention addresses specific problems such as large-scale motion control and uses quantitative evaluation values ​​to reflect the control performance of each degree of freedom. Based on these evaluation values, an optimized coefficient matrix is ​​derived, which can be used to specifically modify and adaptively adjust the original control matrix, thereby forming an optimized motion control matrix. In summary, this invention not only calculates pose error in real time using visual information but also introduces an online optimization mechanism based on multi-degree-of-freedom error feature analysis. By establishing a control matrix and performing pre-simulation and quantitative evaluation, potential motion performance bottlenecks can be identified in advance. Then, through a data-driven adaptive optimization coefficient matrix, the control strategy is dynamically adjusted, enabling the robotic arm to achieve not only fast and accurate pose alignment when dealing with dynamic tasks but also ensuring the smoothness, coordination, and energy optimization of the entire motion process. This significantly improves its autonomous operation capability in complex and unstructured environments and the overall system intelligence level.

[0059] In an embodiment of this application, a method for optimizing and controlling the motion coordination of a robotic arm is provided. The method for acquiring real-time visual information images of the robotic arm and determining the initial pose of the robotic arm based on the real-time visual information images includes: acquiring real-time visual information images of the robotic arm through a 3D camera and calculating the 6D pose of the robotic arm's end effector in the camera coordinate system from the real-time visual information images; and converting the 6D pose of the end effector in the camera coordinate system to the coordinate system to which the robotic arm belongs through hand-eye calibration to obtain the initial pose of the robotic arm.

[0060] Specifically, by deploying 3D cameras, high-precision depth point cloud and texture images containing the robotic arm's end effector and its surrounding environment can be captured in real time. Based on this real-time visual information, algorithms such as feature matching, model fitting, or deep learning can be used to directly identify and calculate the 6D pose of the robotic arm's end effector in the camera coordinate system, namely three translational degrees of freedom (X, Y, Z positions) and three rotational degrees of freedom (roll, pitch, and yaw angles around the X, Y, and Z axes). With the help of pre-precise hand-eye calibration, the known fixed spatial transformation relationship between the camera and the robotic arm base is used as a conversion bridge to accurately transform the 6D pose of the end effector in the camera coordinate system to the base coordinate system or tool coordinate system of the robotic arm body, thereby obtaining the real-time and accurate initial pose of the robotic arm in global space. This step provides independent and objective feedback on the end position and posture through direct external visual observation. It provides more reliable positioning information, especially in the presence of modeling errors, joint backlash, or flexible deformation. Since the pose information comes directly from the observation of the real physical world, the robotic arm can know its exact state in the actual workspace, rather than its theoretical state. It can naturally adapt to the reference changes caused by table micro-movements, fixture changes, or thermal deformation, providing a fundamental guarantee for performing tasks in dynamic and unstructured environments.

[0061] In an embodiment of this application, a method for optimizing and controlling the motion coordination of a robotic arm is provided. The method involves determining the desired target pose of the robotic arm, analyzing the difference between the desired target pose and the initial pose, and generating a pose error vector for the robotic arm containing six degrees of freedom. The method includes: determining the desired target pose of the robotic arm, and representing the desired target pose and the initial pose in a vector form containing six degrees of freedom to obtain the desired target pose vector and the initial pose vector; calculating the vector error between the desired target pose vector and the initial pose vector at each degree of freedom, and synthesizing the vector errors at each degree of freedom to obtain the pose error vector of the robotic arm.

[0062] Specifically, after obtaining the current initial pose vector of the robotic arm and the target pose vector preset by the task, a pose error vector is generated. This vector is used to perform degree-of-freedom difference on the two pose vectors. For the translational degree of freedom, the coordinate difference (Δx, Δy, Δz) is directly calculated. For the rotational degree of freedom, the angle deviation of the shortest path (Δα, Δβ, Δγ) needs to be calculated according to the attitude representation method. The errors of these six independently calculated degree-of-freedom vectors are combined into a unified six-dimensional pose error vector, which completely and quantitatively describes all deviations in spatial position and direction from the current state to the target state. This step reduces the high-level spatial task of moving to a certain position and assuming a posture to a specific numerical target that the controller can directly process. The pose error vector becomes the only and clear input command for the entire servo control loop, driving all subsequent calculations to minimize this vector. The perceived information (initial pose) obtained by sensors such as vision is compared with the desired state to generate a continuous and quantifiable feedback signal. This signal not only indicates the direction of motion (positive or negative of the error) but also clarifies the amplitude of motion (magnitude of the error), enabling algorithms such as PID, impedance control, or model predictive control to perform accurate correction calculations based on this, thus achieving true closed-loop feedback control.

[0063] In an embodiment of this application, a method for optimizing the motion coordination of a robotic arm is provided. The method for determining the vector error of the robotic arm in each degree of freedom based on the pose error vector and determining the control matrix of the robotic arm based on the vector error in each degree of freedom includes: decomposing the pose error vector to obtain the pose error of the robotic arm in each degree of freedom, analyzing the pose error in each degree of freedom to determine the control coefficient in each degree of freedom, and coupling and synthesizing the control coefficients in each degree of freedom to obtain the control matrix of the robotic arm.

[0064] Specifically, the six-dimensional pose error vector is mathematically decomposed to directly obtain its independent error components in three translational (Δx, Δy, Δz) and three rotational (Δα, Δβ, Δγ) degrees of freedom. Each error component is analyzed in real time to dynamically determine a preliminary control coefficient for each degree of freedom. As a strongly coupled nonlinear system, the robotic arm's movements in each degree of freedom are not independent. Based on the current configuration and dynamic characteristics of the robotic arm, these preliminary coefficients, along with coupling coefficients describing the mutual influence between degrees of freedom, are comprehensively calculated and organized into a complete 6x6 control matrix. This matrix includes both diagonal elements (main gain) and off-diagonal elements (coupling gain). This step intelligently maps the error vector representing the problem into control commands representing the solution, serving as a bridge between planning and execution. Through the control coefficients determined by the preliminary analysis, preliminary differentiated control is achieved. The off-diagonal control matrix K generated by the coupling synthesis can actively manage the dynamic coupling between degrees of freedom, thus avoiding clumsy movements that affect the entire system and achieving true coordinated and decoupled control.

[0065] In embodiments of this application, a method for optimizing and controlling the motion coordination of a robotic arm is provided. The step of analyzing the pose errors in each degree of freedom and determining the control coefficients for each degree of freedom includes: dividing the pose errors in each degree of freedom according to linear position and orientation angle to obtain linear position error and orientation angle error, and determining a preset correspondence between linear position error and linear position control coefficient and a preset correspondence between orientation angle error and orientation angle control coefficient; mapping each linear position error to the preset correspondence between linear position error and linear position control coefficient to obtain the corresponding linear position control coefficient; mapping each orientation angle error to the preset correspondence between orientation angle error and orientation angle control coefficient to obtain the corresponding orientation angle control coefficient; and constructing the control coefficients for each degree of freedom based on the linear position control coefficients and the orientation angle control coefficients.

[0066] Specifically, the six-degree-of-freedom pose error vector is clearly divided into two categories: linear position error consisting of Δx, Δy, and Δz, and orientation angle error consisting of Δα, Δβ, and Δγ. For these two types of errors, two different mapping relationships are predefined: a preset linear position error-linear position control coefficient correspondence and a preset orientation angle error-orientation angle control coefficient correspondence. In actual operation, each measured linear position error value is input into its corresponding mapping relationship in real time, thereby dynamically querying or calculating an optimal linear position control coefficient. Similarly, each orientation angle error value is also mapped to a corresponding orientation angle control coefficient. These six control coefficients, respectively from the position and attitude channels, are combined to form the initial, independent control coefficient vectors for each degree of freedom. This step, through independent mapping relationships, allows for the configuration of distinctly different control law response characteristics for position and angle errors. For example, a mapping with fast response and high gain can be set for position errors to ensure rapid positioning; while a smoother mapping with less overshoot can be set for angle errors to achieve precise orientation alignment. This effectively avoids the contradiction of balancing the two types of performance when using a uniform gain. The complex multi-degree-of-freedom control parameter tuning problem is decomposed into two more manageable and optimizable sub-problems: position and attitude. This lays a clear and flexible parameter foundation for building a high-performance, highly robust motion controller.

[0067] In the embodiments of this application, a method for coordinated optimization control of robotic arm motion is provided. The method involves simulating and controlling the operating parameters of the robotic arm based on a control matrix, and determining the pose error characteristics of the robotic arm in each degree of freedom after the simulation and control. The method includes: determining a pre-set robotic arm motion simulation model, inputting the control matrix into the robotic arm motion simulation model, and obtaining simulation results; extracting the adjustment time, overshoot, steady-state error, and number of oscillations of the robotic arm in each degree of freedom from the simulation results, and determining the adjustment time, overshoot, steady-state error, and number of oscillations as the pose error characteristics of the robotic arm in each degree of freedom.

[0068] Specifically, a pre-established high-fidelity motion simulation model that accurately reflects the dynamics and kinematics of the robotic arm is invoked. The control matrix generated in the previous step is used as the core control law and input into the simulation model to drive the robotic arm digital model from its initial pose to the desired target pose, thereby obtaining a complete set of simulation results. From these results, four core time-domain indicators that characterize the dynamic response process are precisely extracted and defined as pose error features: settling time (the time required for the error to enter and remain within the allowable range, reflecting the convergence speed), overshoot (the maximum overshoot ratio when the error first crosses the steady-state value, reflecting the system damping and stability), steady-state error (the error remaining after final stabilization, reflecting absolute accuracy), and oscillation count (the number of times the error crosses the steady-state value, reflecting the smoothness of the stabilization process). This set of feature values ​​together constitutes a quantitative, multi-dimensional "health check report" of control performance. This step shifts the debugging and verification process of the controller from a high-risk, high-cost physical entity to a virtual environment. Potential problems can be proactively exposed in the simulation before the control matrix is ​​deployed to the real robotic arm. The four characteristics of settling time, overshoot, steady-state error, and number of oscillations are classic performance indicators in the field of control systems. Extracting them from the simulation results provides an objective, standardized, and comparable quantitative description of the performance of the control matrix.

[0069] In embodiments of this application, a method for optimizing the motion coordination of a robotic arm is provided. The method involves analyzing and evaluating the pose error state of the robotic arm in each degree of freedom based on pose error characteristics to obtain an evaluation value for the pose error state of the robotic arm in each degree of freedom. This includes: determining the ideal and acceptable worst values ​​for the settling time, steady-state error, and number of oscillations, and calculating the settling time score, steady-state error score, and oscillation score based on the settling time, steady-state error, and number of oscillations and their corresponding ideal and acceptable worst values; determining the ideal and tolerance limits for the overshoot, and calculating the overshoot score based on the overshoot and its ideal and tolerance limits; and determining the settling time baseline score, steady-state error baseline score, and steady-state error baseline score. The baseline scores for quasi-scores, oscillation frequency benchmark scores, and overshoot benchmark scores are calculated. The differences between these scores and their corresponding baseline scores are then calculated to obtain the score differences for each pose error feature. The working task scenario of the robotic arm is determined, and the weights of each pose error feature are determined based on this scenario. The score differences for each pose error feature are evaluated to obtain the sub-error state evaluation values ​​for each feature. These sub-error state evaluation values ​​are then weighted and summed with their corresponding weights to obtain the pose error state evaluation values ​​for each degree of freedom of the robotic arm.

[0070] Specifically, ideal values ​​and acceptable worst values ​​are set for settling time, steady-state error, and number of oscillations, respectively, and ideal values ​​and tolerance limits are set for overshoot. The actual measured values ​​of each indicator are converted to a unified scoring scale using a calculation model to obtain scores for settling time, steady-state error, number of oscillations, and overshoot. The calculation model is as follows:

[0071] S = max(0, 100 (Tworst - Tactual) / (Tworst - Tideal)),

[0072] Where S is the score of each pose error feature, Tworst is the upper tolerance limit of each pose error feature, Tactual is the actual value of each pose error feature, and Tideal is the ideal value of each pose error feature; the difference between the current score of each indicator and the corresponding benchmark score is calculated to obtain the score difference value, which directly quantifies the degree to which the current control scheme deviates from the ideal level in each performance dimension. A positive value indicates that it is better than the benchmark, and a negative value indicates that it needs to be improved; the current work task scenario is identified. Different scenarios have different focuses on performance. Precision assembly may give a very high weight to steady-state error, while high-speed grasping pays more attention to adjustment time; the weight coefficients of the four pose error features are dynamically determined accordingly to ensure that the evaluation is highly aligned with the current task objective; the score difference value of each indicator is evaluated and the sub-error state evaluation value is obtained. These four sub-evaluation values ​​are weighted and summed according to the weights determined in the previous step to finally calculate the pose error state evaluation value of the robotic arm in that degree of freedom. This step, through standardization, benchmark comparison, and weighted fusion, condenses complex multi-dimensional performance into a single, objective, and mathematically quantifiable comprehensive score, providing clear and unambiguous input for optimization decisions. By introducing scenario-based weight allocation, the evaluation system is no longer rigid; it understands which performance aspects are most important for the current task, thus achieving contextualized and adaptive evaluation standards. This ensures that subsequent optimization directions remain consistent with the upper-level task objectives, improving the system's practicality and intelligence. The final generated pose error state evaluation value and its composition analysis not only provide an overall score but also form a detailed performance diagnostic report, pinpointing which degree of freedom and which performance indicator is the main weakness, providing extremely precise guidance for generating the next optimization coefficient matrix.

[0073] In an embodiment of this application, a method for optimizing the motion coordination of a robotic arm is provided. The method involves determining an optimization coefficient matrix based on pose error state evaluation values ​​and optimizing the control matrix based on the optimization coefficient matrix to obtain an optimized motion control matrix for the robotic arm. This includes: determining historical control coefficients and corresponding historical pose error state evaluation values ​​for each degree of freedom of the robotic arm; constructing an optimization coefficient prediction model for each degree of freedom based on the historical control coefficients, corresponding historical pose error state evaluation values, and a preset neural network model; inputting the pose error state evaluation values ​​of the robotic arm in each degree of freedom into the corresponding optimization coefficient prediction model to obtain the optimization coefficients for each degree of freedom; coupling and synthesizing the optimization coefficients for each degree of freedom to obtain an optimization coefficient matrix for the robotic arm; and multiplying the optimization coefficient matrix with the control matrix to obtain the optimized motion control matrix for the robotic arm.

[0074] Specifically, the historical control coefficients and corresponding historical pose error state evaluation values ​​for each degree of freedom of the robotic arm are determined. Based on these historical control coefficients, corresponding historical pose error state evaluation values, and a preset neural network model, an optimization coefficient prediction model for each degree of freedom is constructed. This model, by learning the complex nonlinear mapping relationships in historical data, can predict the optimal control coefficients to be adopted based on the input current performance evaluation value. The current pose error state evaluation value calculated in real time is input into the prediction model corresponding to each degree of freedom. The model output is the optimized coefficients for each degree of freedom optimized for the current performance condition. Through a coupled synthesis algorithm, these are organized into a structurally complete optimization coefficient matrix that considers the mutual influence between degrees of freedom. This optimization coefficient matrix is ​​then multiplied with the original basic control matrix to obtain a new optimized motion control matrix. This new matrix integrates the basic control law with the optimization adjustment amount based on the current performance feedback. This step uses a neural network model to uncover hidden patterns in historical data, automatically adopting the best-performing combination of control coefficients based on the current performance. This enables the control system to accumulate experience and improve itself. The more iterations, the more accurate the prediction and optimization. Compared to simple feedback adjustments based on the current error, it can more proactively and smoothly guide the system towards a better performance region, effectively avoiding oscillations caused by feedback lag. The coupling synthesis step ensures that the final output optimization coefficients are not an isolated set of coefficients, but a matrix that reflects the overall coordination relationship of the system. Matrix multiplication embeds this optimization relationship into the original control law in a structured way, which can better maintain the stability of the closed-loop system and ensure that the optimized motion is coordinated and smooth, rather than a rigid combination of individual joints acting independently.

[0075] In an embodiment of this application, a method for optimizing the motion coordination of a robotic arm is provided. The method for constructing an optimization coefficient prediction model for each degree of freedom based on historical control coefficients and corresponding historical pose error state evaluation values ​​and a preset neural network model includes: constructing corresponding datasets based on historical pose error state evaluation values ​​and historical control coefficients, and inputting each dataset into the corresponding preset neural network model to construct an initial model for predicting optimization coefficients for each degree of freedom; dividing each dataset into a training set and a test set according to a preset ratio, and inputting the training set and the test set into the corresponding initial model for predicting optimization coefficients; training and testing each initial model for predicting optimization coefficients until each initial model for predicting optimization coefficients meets a preset convergence condition to obtain the initial model for predicting optimization coefficients for each degree of freedom.

[0076] Specifically, for each degree of freedom, its historical pose error state evaluation value (as input feature) and the historical control coefficients used at the corresponding time moment (as output label) are collected to construct a one-to-one state-parameter dataset. These datasets are divided into a training set for model learning and a test set for verifying generalization ability according to a preset ratio. These data are input into a preset neural network structure to form an initial model for optimizing coefficient prediction for each degree of freedom. The model parameters are updated iteratively using the training set data, and its generalization performance is monitored simultaneously using the test set data. The training process continues until the model's prediction accuracy on the test set reaches a preset standard and the error no longer decreases significantly. At this point, a mature optimized coefficient prediction model that can be deployed online is obtained. In this step, there is often a nonlinear, high-order relationship between the control coefficients and the comprehensive performance evaluation value of the robotic arm, which is difficult to describe with explicit mathematical formulas. The neural network, by learning from historical data, can automatically capture and internalize this complex mapping relationship and has good generalization ability. Even when faced with different performance evaluation inputs, the model can reliably output reasonable optimization coefficient suggestions, breaking through the limitations of traditional analytical methods in modeling. Since the dataset and model are built independently for each degree of freedom, the differences in dynamic characteristics and performance requirements of different degrees of freedom (such as translation and rotation) can be fully considered, thereby customizing optimization strategies for each degree of freedom. This makes the final synthesized control matrix more refined and coordinated to adapt to the overall dynamic characteristics of the robotic arm. This step is a typical offline training process. It uses historical data for preparation, which is fully prepared for subsequent online real-time prediction. Once the model training is completed and the convergence condition is met, its lightweight forward computation process can quickly provide optimization coefficients based on real-time evaluation values, meeting the real-time requirements of the control system.

[0077] As shown in Figure 2, in an embodiment of this application, a robotic arm motion coordination optimization control system is provided, comprising: an acquisition module for acquiring real-time visual information images of the robotic arm and determining the initial pose of the robotic arm based on the real-time visual information images; a generation module for determining the desired target pose of the robotic arm and analyzing the difference between the desired target pose and the initial pose to generate a pose error vector of the robotic arm containing six degrees of freedom; a determination module for determining the vector error of the robotic arm in each degree of freedom based on the pose error vector and determining the control matrix of the robotic arm based on the vector error in each degree of freedom; a simulation module for simulating and controlling the operating parameters of the robotic arm based on the control matrix and determining the pose error characteristics of the robotic arm in each degree of freedom after simulation and control; an evaluation module for analyzing and evaluating the pose error state of the robotic arm in each degree of freedom based on the pose error characteristics to obtain the pose error state evaluation value of the robotic arm in each degree of freedom; and an optimization module for determining an optimization coefficient matrix based on the pose error state evaluation value and optimizing the control matrix based on the optimization coefficient matrix to obtain the optimized motion control matrix of the robotic arm.

[0078] In summary, this invention provides a method and system for optimizing the motion coordination of a robotic arm, comprising: determining the initial and desired target poses of the robotic arm and analyzing the differences between them to obtain a pose error vector; determining the vector error of the robotic arm based on the pose error vector, and determining the control matrix of the robotic arm based on the vector error vector; performing simulation control on the operating parameters of the robotic arm based on the control matrix, and determining the pose error characteristics of the robotic arm after the simulation; analyzing and evaluating the pose error state of the robotic arm based on the pose error characteristics to obtain a pose error state evaluation value; and optimizing the control matrix based on the pose error state evaluation value to obtain an optimized motion control matrix. This invention, by establishing a control matrix and performing pre-simulation and quantitative evaluation, can anticipate potential motion performance bottlenecks, and then dynamically optimize the robotic arm through a data-driven adaptive optimization coefficient matrix, thereby improving the flexibility and accuracy of the robotic arm and ensuring smooth motion.

[0079] Finally, it should be noted that those skilled in the art can obviously make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for optimizing the motion coordination of a robotic arm, characterized in that, include: Acquire real-time visual information images of the robotic arm and determine the initial pose of the robotic arm based on the real-time visual information images; The desired target pose of the robotic arm is determined, and the difference between the desired target pose and the initial pose is analyzed to generate a pose error vector of the robotic arm with six degrees of freedom. Based on the pose error vector, the vector error of the robotic arm in each degree of freedom is determined, and the control matrix of the robotic arm is determined based on the vector error in each degree of freedom. The operating parameters of the robotic arm are simulated and controlled based on the control matrix, and the pose error characteristics of the robotic arm in each degree of freedom are determined after the simulation and control. Based on the pose error characteristics, the pose error state of the robotic arm in each degree of freedom is analyzed and evaluated to obtain the pose error state evaluation value of the robotic arm in each degree of freedom. The process involves determining an optimization coefficient matrix based on the pose error state evaluation value, and then optimizing the control matrix to obtain the optimized motion control matrix for the robotic arm. The process includes simulating the operation parameters of the robotic arm based on the control matrix, and determining the pose error characteristics of the robotic arm in each degree of freedom after simulation control. This includes: determining a pre-set robotic arm motion simulation model and inputting the control matrix into the model to obtain simulation results; extracting the adjustment time, overshoot, steady-state error, and number of oscillations of the robotic arm in each degree of freedom from the simulation results, and defining these as the pose error characteristics of the robotic arm in each degree of freedom. The process further involves analyzing and evaluating the pose error state of the robotic arm in each degree of freedom based on these characteristics, obtaining the pose error state evaluation value for each degree of freedom. This includes: determining the ideal and worst-case values ​​for the adjustment time, steady-state error, and number of oscillations, and then evaluating the values ​​based on these values. The following steps are performed: First, calculate the settling time score, steady-state error score, and oscillation frequency score based on the number of occurrences and their corresponding ideal and worst-case values. Second, determine the ideal and upper tolerance values ​​for overshoot, and calculate the overshoot score based on these values. Third, determine the baseline scores for settling time, steady-state error, oscillation frequency, and overshoot, and calculate the differences between these scores and their corresponding baseline scores to obtain the score differences for each pose error feature. Fourth, determine the robot arm's task scenario and assign weights to each pose error feature based on this scenario. Fifth, evaluate the score differences for each pose error feature to obtain sub-error state evaluation values ​​for each feature, and then weight and sum these sub-error state evaluation values ​​with their corresponding weights to obtain the pose error state evaluation values ​​for each degree of freedom of the robot arm.

2. The method for optimizing and controlling the motion coordination of a robotic arm according to claim 1, characterized in that, The process of acquiring real-time visual information images of the robotic arm and determining the initial pose of the robotic arm based on the real-time visual information images includes: acquiring real-time visual information images of the robotic arm through a 3D camera and calculating the 6D pose of the robotic arm end effector in the camera coordinate system from the real-time visual information images; and converting the 6D pose of the end effector in the camera coordinate system to the coordinate system to which the robotic arm belongs through hand-eye calibration to obtain the initial pose of the robotic arm.

3. The method for optimizing and controlling the motion coordination of a robotic arm according to claim 2, characterized in that, The process of determining the desired target pose of the robotic arm and analyzing the difference between the desired target pose and the initial pose to generate a pose error vector of the robotic arm containing six degrees of freedom includes: determining the desired target pose of the robotic arm, and representing the desired target pose and the initial pose in a vector form containing six degrees of freedom to obtain the desired target pose vector and the initial pose vector; calculating the vector error between the desired target pose vector and the initial pose vector in each degree of freedom, and synthesizing the vector errors in each degree of freedom to obtain the pose error vector of the robotic arm.

4. The method for optimizing and controlling the motion coordination of a robotic arm according to claim 3, characterized in that, The method of determining the vector error of the robotic arm in each degree of freedom based on the pose error vector, and determining the control matrix of the robotic arm based on the vector error in each degree of freedom, includes: decomposing the pose error vector to obtain the pose error of the robotic arm in each degree of freedom, analyzing the pose error in each degree of freedom to determine the control coefficient in each degree of freedom; and coupling and synthesizing the control coefficients in each degree of freedom to obtain the control matrix of the robotic arm.

5. The method for optimizing and controlling the motion coordination of a robotic arm according to claim 4, characterized in that, The analysis of pose errors in each degree of freedom and the determination of control coefficients for each degree of freedom include: dividing pose errors in each degree of freedom according to linear position and orientation angle to obtain linear position error and orientation angle error, and determining preset correspondences between linear position error and linear position control coefficient and between preset correspondences between orientation angle error and orientation angle control coefficient; mapping each linear position error to the preset correspondence between linear position error and linear position control coefficient to obtain the corresponding linear position control coefficient; mapping each orientation angle error to the preset correspondence between orientation angle error and orientation angle control coefficient to obtain the corresponding orientation angle control coefficient; and constructing control coefficients for each degree of freedom based on the linear position control coefficient and the orientation angle control coefficient.

6. The method for optimizing and controlling the motion coordination of a robotic arm according to claim 1, characterized in that, The process of determining the optimization coefficient matrix based on the pose error state evaluation value and optimizing the control matrix based on the optimization coefficient matrix to obtain the optimized motion control matrix of the robotic arm includes: determining the historical control coefficients and corresponding historical pose error state evaluation values ​​of each degree of freedom of the robotic arm; constructing optimization coefficient prediction models for each degree of freedom based on the historical control coefficients and corresponding historical pose error state evaluation values ​​of each degree of freedom and a preset neural network model; inputting the pose error state evaluation values ​​of the robotic arm in each degree of freedom into the corresponding optimization coefficient prediction models to obtain the optimization coefficients of the robotic arm in each degree of freedom; coupling and synthesizing the optimization coefficients in each degree of freedom to obtain the optimization coefficient matrix of the robotic arm; and multiplying the optimization coefficient matrix with the control matrix to obtain the optimized motion control matrix of the robotic arm.

7. The method for optimizing and controlling the motion coordination of a robotic arm according to claim 6, characterized in that, The step of constructing an optimization coefficient prediction model for each degree of freedom based on historical control coefficients and corresponding historical pose error state evaluation values ​​and a preset neural network model includes: constructing corresponding datasets based on historical pose error state evaluation values ​​and historical control coefficients, and inputting each dataset into the corresponding preset neural network model to construct an initial model for predicting optimization coefficients for each degree of freedom; dividing each dataset into training and testing sets according to a preset ratio, and inputting the training and testing sets into the corresponding initial models for predicting optimization coefficients; training and testing each initial model for predicting optimization coefficients until each initial model for predicting optimization coefficients meets the preset convergence conditions to obtain the optimization coefficient prediction model for each degree of freedom.

Images