Industrial robot automatic control system and method

Abstract

The invention discloses an industrial robot automatic control system and method, and relates to the technical field of industrial robots, and the system comprises a data management module, a data analysis module, a motion prediction module and a coordination control module. By accurately analyzing key factors, the system can identify core parameters influencing the motion performance of the robot, the adaptability and working efficiency of the robot in a complex environment are optimized, the motion precision and stability are improved, the system predicts the future motion state and track of the robot in a prospective mode by means of a motion prediction model, and the robot motion performance is improved. Potential fault risks are recognized in advance, production interruption is avoided, the service life of the robot is prolonged, motion parameters are dynamically adjusted according to a prediction result, a control strategy is optimized, and the flexibility and execution efficiency of the robot in a multi-task scene are improved, so that the production efficiency and the product quality are remarkably improved, and the labor cost and operation loss are reduced. And greater economic benefits are created for enterprises.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial robots, and more specifically, to an industrial robot automatic control system and method. Background Art

[0002] With the continuous development of industrial robot technology, the improvement of the factory robotization level has brought profound impacts on the manufacturing industry. The application of robot automatic control systems not only improves production efficiency but also greatly reduces the dependence on the fluctuations of the traditional labor market. In the past, many manufacturing jobs were outsourced to regions with lower labor costs. However, with the maturity of robot technology, more and more enterprises are starting to bring back manufacturing to their home countries. Industrial robots can complete high-precision and highly repetitive tasks, reducing human operation errors, lowering labor costs, and being flexibly applicable in different production environments. Modern industrial robots are easier to deploy and maintain. Thanks to advanced automatic control systems, the operation of robots has become more intelligent and user-friendly. These control systems can monitor the production process in real time and adjust production parameters in a timely manner to cope with changes in market demand.

[0003] In the prior art, it is not convenient to monitor the running state of industrial robots in real time, cannot capture the working state of robots in a timely and accurate manner, is not convenient for detecting abnormal situations, and identifying key factors affecting the motion performance of industrial robots, reducing the adaptability and working efficiency of robots in complex environments. At the same time, it is not convenient to predict the future motion state, motion trajectory, and attitude change trend of industrial robots, not convenient for helping robots adapt to dynamic environments, and not convenient for identifying potential fault risks in advance to avoid production interruptions, thereby reducing the service life and maintenance efficiency of robots.

[0004] In view of the problems in the related art, no effective solution has been proposed yet. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides an industrial robot automatic control system and method, which solve the problems in the above background art that it is not convenient to monitor the running state of industrial robots in real time, cannot capture the working state of robots in a timely and accurate manner, is not convenient for detecting abnormal situations, and identifying key factors affecting the motion performance of industrial robots, reducing the adaptability and working efficiency of robots in complex environments. At the same time, it is not convenient to predict the future motion state, motion trajectory, and attitude change trend of industrial robots, not convenient for helping robots adapt to dynamic environments, and not convenient for identifying potential fault risks in advance to avoid production interruptions, thereby reducing the service life and maintenance efficiency of robots.

[0006] To achieve the above objectives, the present invention is realized through the following technical solutions: According to one aspect of the present invention, an industrial robot automation control system is provided, and the system includes: A data management module, configured to obtain real-time operation data of the industrial robot, process the real-time operation data, and extract operation feature data; A data analysis module, configured to perform feature analysis on the operation feature data by using a feature analysis algorithm to identify key factors affecting the motion performance of the industrial robot; A motion prediction module, configured to establish a motion prediction model based on the key factors to predict the operation state of the industrial robot at a future moment, and identify the motion trajectory and posture change trend; A coordination control module, configured to dynamically adjust the motion parameters of the industrial robot and optimize the motion control strategy based on the motion trajectory and posture change trend, in combination with the real-time operation state of the industrial robot.

[0007] Furthermore, the data management module includes: An acquisition data module, configured to obtain real-time operation data of the industrial robot, and perform denoising, filtering, and smoothing processing on duplicate data, missing values, and outliers in the real-time operation data to obtain accurate real-time operation data; A feature extraction module, configured to perform feature extraction on the obtained accurate real-time operation data by using a clustering algorithm to obtain operation feature data.

[0008] Furthermore, performing feature analysis on the operation feature data by using a feature analysis algorithm to identify key factors affecting the motion performance of the industrial robot includes: Based on the operation feature data, initialize each feature parameter and construct an initial feature matrix; Preset the maximum number of iterations of the feature analysis algorithm. In each iteration, randomly select a feature subset, and calculate the correlation score between each feature and the motion performance by using a score formula to screen out candidate features with the greatest influence on the motion performance; Perform importance evaluation on the screened candidate features, and use a special weight sorting search algorithm to quantify the contribution degree of each feature to the motion performance to determine its importance ranking; According to the importance ranking, identify the key factors with the greatest influence on the motion performance of the industrial robot.

[0009] Furthermore, the score formula is: ; In the formula, M iRepresents the correlation score between the \(i\)-th feature and the motion performance; \(A(i, j)\) represents the weight value of the \(i\)-th feature under the \(j\)-th operating condition; \(D(i, j)\) represents the matching score of the \(i\)-th feature's motion performance under the \(j\)-th operating condition; \(\Delta h\) represents the distance between the \(i\)-th feature and the starting analysis position of the current feature; \(\alpha\) represents the adjustment parameter of the feature weight; \(\beta\) represents the adjustment parameter of the feature matching score; \(\delta\) represents the adjustment parameter of the position distance; \(L\) i Represents the starting analysis position of the \(i\)-th feature.

[0010] Furthermore, conduct an importance assessment on the selected candidate features, and use the feature weight sorting search algorithm to quantify the contribution degree of each feature to the motion performance, and determine its importance ranking, including: Initialize the parameters of the feature weight sorting algorithm, set the maximum search loop times of the feature weight sorting search algorithm, and generate an initial feature weight combination; Update the search loop times of the feature weight sorting algorithm, and determine whether the maximum search loop times are reached. If so, terminate the feature weight sorting algorithm and output the feature weight sorting result; otherwise, continue to perform the sorting search; Use the current feature weight combination searched in the upper layer as the initial feature weight combination for the lower layer search, set the lower layer search loop times, and enter the lower layer search loop; Perform a block structure feature weight combination operation in the lower layer search, and generate a new feature weight combination in combination with the feature optimization algorithm. Arrange each feature weight combination in ascending order of the fitness value of the combination, and select the current feature weight combination; Determine whether the current feature weight combination violates the taboo rule. If it violates the taboo rule, skip the current feature weight combination and select the next feature weight combination to continue the evaluation. Otherwise, determine whether the fitness value of the current feature weight combination has been optimized. If there is optimization, use the current feature weight combination as the current optimal feature weight combination; otherwise, continue the evaluation; Update the optimal feature weight combination and continue the loop until the maximum search loop times are met, return to the upper layer search, and continue to execute the search loop of the upper layer search until the maximum search loop times are reached; finally, output the feature weight sorting result, determine the contribution degree of each feature to the motion performance and its importance ranking.

[0011] Furthermore, perform a block structure feature weight combination operation in the lower layer search, and generate a new feature weight combination in combination with the feature optimization algorithm. Arrange each feature weight combination in ascending order of the fitness value of the combination, and select the current feature weight combination, including: Set the initial parameters of the feature optimization algorithm, and initialize the preliminary feature weight combination and the objective function; Randomly initialize multiple weight combinations in the feature weight space, and calculate their corresponding objective function values; Evaluate each initial feature weight combination using a known feature weight dataset, calculate and record the fitness value of each combination; Arrange all feature weight combinations in ascending order of their fitness values, preferentially select the feature combination with the best performance, and based on the sorting of the fitness values, select the feature weight combination with the most optimization potential; Generate new feature weight combinations according to the fitness value of the current feature weight combination and the optimization goal, and perform ascending sorting again according to the fitness value; select the combination with the optimal fitness value as the current optimal feature weight combination.

[0012] Furthermore, based on key factors, establish a motion prediction model to predict the operating state of an industrial robot at a future moment, and identify the motion trajectory and posture change trends, including: Quantify the key factors, divide them into a training set and a test set according to the data characteristics, and set the maximum number of iterations of the motion prediction model; Generate an initial parameter combination of the motion prediction model through a random initialization strategy and combined with the distribution characteristics of the motion data, and configure key parameters for the motion prediction model; According to the laws of motion trajectory and posture change, dynamically adjust the weights in the objective function of the motion prediction model, combined with the priority setting of key motion characteristics, and use an optimization algorithm to optimize the motion prediction model to identify the performance of the motion prediction model under different motion states; Use the objective function to evaluate the prediction accuracy of the motion prediction model on the test set under each parameter combination, calculate the fitness value, and sort the fitness values of all parameter combinations; screen out the current optimal motion prediction model parameters; According to the non-linear deviation distribution of the motion prediction model, judge whether it is less than the set threshold. If so, adopt a local optimization strategy to adjust some key parameters; otherwise, adopt a global optimization strategy to adjust all key parameters, generate the optimal motion prediction model, and output the motion trajectory and posture change trends.

[0013] Furthermore, according to the laws of motion trajectory and posture change, dynamically adjust the weights in the objective function of the motion prediction model, combined with the priority setting of key motion characteristics, and use an optimization algorithm to optimize the motion prediction model to identify the performance of the motion prediction model under different motion states, including: Randomly sample several motion state data points within the feasible parameter space of the motion prediction model; Use the sampled motion state data points to construct a motion prediction model based on radial basis functions, and calculate the weight coefficients of the radial basis functions; Based on the current motion prediction model, use the Monte Carlo probability sampling algorithm for optimization sampling; Construct a quadratic response surface model, calculate the fitting accuracy of the quadratic response surface model, and determine whether its fitting effect meets the predetermined accuracy requirements. If it meets, continue with new sampling; otherwise, adjust the acceleration factor and re-optimize the parameters of the quadratic response surface model; Evaluate the optimized quadratic response surface model to determine whether the fitting accuracy meets the standard. If it meets the standard, use the quadratic response surface model for motion state prediction; otherwise, add new motion state data points to the training set and continue to optimize the model; Based on the optimized quadratic response surface model, search within the feasible parameter space to identify the optimal motion state prediction parameters and determine the motion state with the minimum prediction error. If the optimization algorithm reaches the maximum number of iterations, output the optimal motion prediction model and identify the performance of the motion prediction model under different motion states.

[0014] Furthermore, the formula for calculating the fitting accuracy of the quadratic response surface model is: ; In the formula, R 2 represents the fitting accuracy of the quadratic response surface model; y a represents the actual response value of the a-th motion state data point; represents the model predicted response value of the a-th motion state data point; n represents the total number of motion state data points; represents the mean value of the actual response values of all motion state data points.

[0015] According to another aspect of the present invention, there is also provided an industrial robot automatic control method, which includes the following steps: S1. Obtain the real-time operation data of the industrial robot, process the real-time operation data, and extract the operation characteristic data; S2. Use the feature analysis algorithm to perform feature analysis on the operation characteristic data to identify the key factors affecting the motion performance of the industrial robot; S3. Based on the key factors, establish a motion prediction model to predict the operation state of the industrial robot at a future moment and identify the motion trajectory and posture change trend; S4. Based on the motion trajectory and posture change trend, combined with the real-time operation state of the industrial robot, dynamically adjust the motion parameters of the industrial robot and optimize the motion control strategy.

[0016] The beneficial effects of the present invention are: 1. The present invention captures abnormal states in a timely manner by obtaining and processing robot operation data in real time, reduces manual intervention, ensures the efficiency and reliability of operation. By precisely analyzing key factors, the system can identify the core parameters affecting the motion performance of the robot, optimize its adaptability and working efficiency in complex environments, improve motion accuracy and stability. With the help of a motion prediction model, the system prospectively predicts the future motion state and trajectory of the robot, identifies potential fault risks in advance, avoids production interruptions, and extends the service life of the robot. According to the prediction results, the motion parameters are dynamically adjusted, the control strategy is optimized, and the flexibility and execution efficiency of the robot in multi-task scenarios are improved. As a result, through automated data analysis and dynamic control, the system significantly improves production efficiency and product quality, reduces labor costs and operation losses, and creates greater economic benefits for enterprises.

[0017] 2. Through the feature analysis algorithm and the feature weight ranking search algorithm, the present invention can accurately identify the key factors affecting the motion performance of the robot, quantify the contribution degree of each feature, determine its importance ranking. The system analyzes the operation feature data in real time, dynamically optimizes the feature weight combination, and improves the motion accuracy and stability of the robot.

[0018] 3. By establishing a motion prediction model, the present invention can accurately predict the future motion trajectory and attitude change trend of the robot, significantly improve its foresight and adaptability. The system uses key factors to optimize the model parameters, dynamically adjusts the weight of the objective function, and ensures prediction accuracy and stability. Through the quadratic response surface model and the Monte Carlo sampling algorithm, the system continuously optimizes the model performance, reduces prediction errors, and improves the execution efficiency of the robot in complex environments. At the same time, the system can identify potential fault risks in advance, avoid production interruptions, and extend the service life of the equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0020] Figure 1 is a schematic block diagram of an industrial robot automation control system according to an embodiment of the present invention; Figure 2 is a flowchart of an industrial robot automation control method according to an embodiment of the present invention.

[0021] In the figure: 1. Data management module; 2. Data analysis module; 3. Motion prediction module; 4. Coordination control module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments.

[0023] In the description of the present invention, unless otherwise specified, the meaning of "a plurality" is two or more. In addition, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0024] According to an embodiment of the present invention, an industrial robot automatic control system and method are provided.

[0025] Now, in combination with the accompanying drawings and specific implementation manners, the present invention will be further described. As Figure 1 shown, an industrial robot automatic control system according to an embodiment of the present invention includes: A data management module 1, configured to obtain real-time operation data of an industrial robot, process the real-time operation data, and extract operation feature data; Specifically, the motion state data of the robot can be collected in real time through sensors (such as encoders, accelerometers, gyroscopes, force sensors, etc.) installed at each joint, end effector, and key parts of the robot, and the relevant data of the robot working environment can also be obtained through external devices (such as vision systems, lidar, environmental sensors, etc.), such as the position, shape, distance, etc. of the target object.

[0026] Specifically, the real-time operation data includes: 1) Motion state data: Joint angles, speeds, and accelerations.

[0027] The position, attitude, speed, and acceleration of the end effector.

[0028] The motion trajectory and path of the robot.

[0029] 2) Force and torque data: The force conditions of the joints and end effectors.

[0030] Torque data collected by torque sensors.

[0031] 3) Control signal data: The output signals of the controller, such as motor drive signals, position control signals, etc.

[0032] Feedback signals, such as the deviation between the actual position, speed and the target value.

[0033] 4) Environmental data: Information such as the position, shape, and color of the target object collected by the vision system.

[0034] Environmental distance and obstacle information collected by lidar or ultrasonic sensors.

[0035] Environmental sensor data such as temperature and humidity.

[0036] 5) Fault and alarm data: Fault codes and alarm information during the operation of the robot.

[0037] Abnormal state data of sensors or actuators.

[0038] Specifically, the operation characteristic data includes: 1) Motion characteristics: Motion smoothness: Calculated by the change rates of speed and acceleration.

[0039] Trajectory accuracy: The deviation between the actual trajectory and the target trajectory.

[0040] Posture stability: The posture change of the end effector.

[0041] 2) Force and torque characteristics: Force uniformity: The distribution of the force on the end effector.

[0042] Torque fluctuation: The change of joint torque.

[0043] 3) Control characteristics: Control error: The deviation between the actual value and the target value.

[0044] Response time: The time from the control signal being issued to the execution being completed.

[0045] 4) Environmental characteristics: Target object recognition accuracy: The deviation between the target position recognized by the vision system and the actual position.

[0046] Environmental adaptability: The execution effect of the robot under different environmental conditions.

[0047] 5) Fault characteristics: Fault frequency: The number of faults occurring per unit time.

[0048] Alarm type: The distribution and severity of different alarms.

[0049] The data analysis module 2 is used to perform feature analysis on the operation characteristic data by using feature analysis algorithms to identify the key factors affecting the motion performance of the industrial robot; Specifically, the key factors affecting the motion performance of the industrial robot include: 1) Motion control parameters: Joint angles, velocities, and accelerations: These parameters directly affect the motion smoothness and accuracy of the robot.

[0050] Position and orientation of the end effector: The motion accuracy and stability of the end effector are important indicators for measuring the performance of the robot.

[0051] Trajectory planning and tracking accuracy: The deviation between the actual motion trajectory and the planned trajectory affects the execution effect of the robot.

[0052] 2) Force and torque characteristics: Joint torque: Fluctuations or uneven distributions of joint torque may cause unsmooth motion or mechanical wear.

[0053] Force on the end effector: The force on the end effector affects its operation accuracy and stability, especially in precision operations or high-load tasks.

[0054] 3) Mechanical structure characteristics: Manipulator rigidity: Insufficient rigidity of the manipulator may cause vibrations or deformations during motion, affecting motion accuracy.

[0055] Joint clearance and wear: Joint clearance and wear can lead to motion errors and performance degradation.

[0056] Manipulator length and weight: The length and weight of the manipulator affect its motion inertia and dynamic response.

[0057] 4) Environmental factors: Working environment temperature and humidity: Changes in environmental temperature and humidity may affect the performance of the robot's mechanical structure and electronic components.

[0058] Target object characteristics: Changes in the shape, weight, and position of the target object affect the grasping and operation accuracy of the robot.

[0059] External disturbances: Such as vibrations, electromagnetic interference, etc. may affect the motion stability and control accuracy of the robot.

[0060] 6) Electrical and drive characteristics: Motor performance: The response speed, torque output, and efficiency of the motor directly affect the motion performance of the robot.

[0061] Power supply stability: Power fluctuations may cause unstable control signals, affecting motion accuracy.

[0062] Driver performance: The control accuracy and response speed of the driver affect the execution effect of the motor.

[0063] The motion prediction module 3 is used to establish a motion prediction model based on key factors to predict the operating state of the industrial robot at a future moment and identify the trends of motion trajectory and posture changes; Specifically, the motion trajectory refers to the motion path of an industrial robot in three-dimensional space, that is, the continuous position change of the end effector or key part of the robot from the starting point to the target point.

[0064] Specifically, the attitude change trend refers to the change of the direction (attitude) of the end effector or key part of the industrial robot over time during the motion process.

[0065] The coordination control module 4 is used to dynamically adjust the motion parameters of the industrial robot and optimize the motion control strategy based on the motion trajectory and attitude change trend, combined with the real-time operating state of the industrial robot.

[0066] It should be explained that through the sensors and data acquisition system on the industrial robot, the motion trajectory, attitude information and operating state data of the robot are obtained in real time. These data include position, speed, acceleration, attitude angle, etc., which can reflect the current working state of the robot; through the feature analysis algorithm, the feature analysis of the operation feature data is carried out to identify the key factors affecting the motion performance of the industrial robot, and a motion prediction model is established based on the key factors to predict the motion state in the future period of time and identify the motion trajectory and attitude change trend. According to the real-time monitoring data, the current state of the robot is compared with the expected state of the motion prediction model. Identify the deviation between the actual motion state of the robot and the predicted value, and analyze the reasons for the deviation, such as motion error, external disturbance, etc.; according to the prediction result and the current deviation, dynamically adjust the motion parameters (such as speed, acceleration, joint angle, etc.). For example, when the robot deviates from the predetermined trajectory, the motion speed or path can be adjusted to ensure that the robot returns to the predetermined motion trajectory again. At the same time, according to the attitude change trend of the robot, the attitude control strategy is adjusted; combined with the real-time operating state of the robot and the motion prediction model, the motion control strategy is adjusted. The purpose of optimizing the strategy is to make the motion of the robot smoother and more efficient, while improving the motion accuracy, reducing energy consumption and working errors; after the adjustment of the motion parameters is completed, the system continues to perform closed-loop control, real-time monitor the adjusted motion effect, and continuously obtain feedback data to correct the control strategy. With the change of the working environment and task requirements, dynamically optimize the motion parameters and control strategy to ensure that the robot can always complete tasks efficiently and accurately in different working scenarios.

[0067] In this alternative embodiment, the data management module 1 includes: The data acquisition module is used to acquire the real-time operation data of the industrial robot, and perform denoising, filtering and smoothing processing on the duplicate data, missing values and outliers of the real-time operation data to obtain accurate real-time operation data; The feature extraction module is used to extract features from the obtained accurate real-time operation data by using the clustering algorithm to obtain operation feature data.

[0068] It should be noted that since the sensor may be interfered by environmental noise, the system will perform denoising processing on the collected raw data. The denoising method can remove high-frequency noise through filtering algorithms (such as low-pass filtering, Kalman filtering, etc.), so as to make the data smoother and more accurate. In practical applications, the sensor may generate missing values due to certain reasons (such as signal loss). The system will use interpolation methods (such as linear interpolation, spline interpolation, etc.) to fill in the missing data points to ensure the integrity of the data. It will also use outlier detection algorithms to identify outliers (such as extreme values or unreasonable jumps) in the data and correct them through correction strategies (such as median substitution, mean correction, etc.), so as to reduce the impact of abnormal data on subsequent analysis. To further improve the accuracy of the data, the system performs smoothing processing on the processed data. For example, it uses a moving average algorithm or an exponentially weighted moving average algorithm to make the data more stable and avoid the interference of short-term fluctuations on the results. After the above series of data processing, the system will obtain a real-time running data set that is accurate, noise-free, missing-free, and outlier-free. Then, the system uses clustering algorithms (such as K-means clustering, DBSCAN clustering, etc.) to perform clustering analysis on the accurate real-time running data. The purpose of clustering is to group data points with similar motion characteristics into one category and identify different motion patterns or behavior patterns. For example, different working stages of the robot (such as start, acceleration, deceleration, stop, etc.) can be separated by the clustering algorithm for further analysis. After clustering, the system will extract key motion characteristics from each cluster. These characteristics may include speed, acceleration, joint angle change, motion smoothness, trajectory deviation, etc., representing different aspects of the robot's motion performance. Feature extraction can be completed by calculating the statistical characteristics (such as mean, variance, etc.) of each cluster center, or using advanced algorithms (such as principal component analysis PCA) to reduce the dimension and extract the main features.

[0069] In this alternative embodiment, a feature analysis algorithm is used to analyze the running feature data, and the key factors affecting the motion performance of the industrial robot are identified as follows: Based on the running feature data, each feature parameter is initialized and an initial feature matrix is constructed; A maximum number of iterations of the preset feature analysis algorithm is set. In each iteration, a random feature subset is selected, and the correlation score between each feature and the motion performance is calculated using a scoring formula to screen out the candidate features with the greatest impact on the motion performance; An importance evaluation is performed on the selected candidate features, and a special weight sorting search algorithm is used to quantify the contribution degree of each feature to the motion performance and determine its importance ranking; According to the importance ranking, the key factors with the greatest impact on the motion performance of the industrial robot are identified.

[0070] Specifically, the feature analysis algorithm is a multiple sequence alignment method based on an improved ant colony algorithm, which changes the pheromone update method, the character selection method, the back-and-forth search of ants between the ant nest and food, and the random assignment of the starting sequence of ants, etc.

[0071] To facilitate the understanding of the above technical solution of the present invention, the following will detail the use of the feature analysis algorithm to analyze the operation feature data in the actual process of the present invention to identify the key factors affecting the motion performance of the industrial robot.

[0072] Step 1: Initialize the feature parameters and construct the initial feature matrix: 1) Data preparation: Collect the operation feature data of the industrial robot, including joint angles, speeds, accelerations, torques, end effector positions, etc.

[0073] Set the feature parameters, for example: Feature 1: Joint angle deviation (unit: degree).

[0074] Feature 2: Speed fluctuation (unit: mm / s).

[0075] Feature 3: Acceleration change rate (unit: mm / s²).

[0076] Construct the initial feature matrix, as shown in Table 1: Table 1: Construct the initial feature matrix 2) Parameter setting: Preset the maximum number of iterations of the feature analysis algorithm: 10 times.

[0077] Adjust the parameters: α = 0.5, β = 0.3, δ = 0.2.

[0078] Position distance: Δh1 = 2, Δh2 = 3, Δh3 = 1.

[0079] Starting analysis position: L1 = 1, L2 = 2, L3 = 3.

[0080] Step 2: Calculate the correlation score between the feature and the motion performance: 1) Select the feature subset: For example, select Feature 1 (joint angle deviation) for analysis.

[0081] 2) Calculate the correlation score: The score formula used is: ; In the formula, M iRepresents the correlation score between the i-th feature and the motion performance; A(i, j) represents the weight value of the i-th feature under the j-th operating condition; D(i, j) represents the matching score of the i-th feature for the motion performance under the j-th operating condition; Δh represents the distance between the i-th feature and the starting analysis position of the current feature; α represents the adjustment parameter for the feature weight; β represents the adjustment parameter for the feature matching score; δ represents the adjustment parameter for the position distance; L i Represents the starting analysis position of the i-th feature.

[0082] Hypothesis: The weight value of feature 1 under condition 1 is A(1, 1) = 0.5, the matching score is D(1, 1) = 0.8, and the distance is Δh = 2.

[0083] The weight value of feature 1 under condition 2 is A(1, 2) = 0.3, the matching score is D(1, 2) = 0.6, and the distance is Δh = 3.

[0084] The weight value of feature 1 under condition 3 is A(1, 3) = 0.2, the matching score is D(1, 3) = 0.4, and the distance is Δh = 1.

[0085] Calculation: For condition 1: A(1, 1) α ×D(1, 1) β ×(Δh) δ = 0.5 0.5 ×0.8 0.3 ×2 0.2 ≈0.752.

[0086] For condition 2: A(1, 2) α ×D(1, 2) β ×(Δh) δ = 0.3 0.5 ×0.6 0.3 ×3 0.2 ≈0.547.

[0087] For condition 3: A(1, 3) α ×D(1, 3) β ×(Δh) δ = 0.2 0.5 ×0.4 0.3 ×1 0.2 ≈0.752.

[0088] Sum of the denominators: 0.752 + 0.547 + 0.329 = 1.628.

[0089] Total score of feature 1: M1 = 0.752 / 1.628 ≈ 0.462.

[0090] Step 3. Repeatedly calculate the correlation scores of other features: 1) Feature 2 (speed fluctuation): Hypothesis: The weight value of Feature 2 under Condition 1, A(2, 1) = 0.4, the matching score D(2, 1) = 0.7, and the distance Δh = 2.

[0091] The weight value of Feature 2 under Condition 2, A(2, 2) = 0.5, the matching score D(2, 2) = 0.9, and the distance Δh = 3.

[0092] The weight value of Feature 2 under Condition 3, A(2, 3) = 0.1, the matching score D(2, 3) = 0.5, and the distance Δh = 1.

[0093] Calculation: For Condition 1: A(2, 1) α × D(2, 1) β × (Δh) δ = 0.4 0.5 × 0.7 0.3 × 2 0.2 ≈ 0.645.

[0094] For Condition 2: A(2, 2) α × D(2, 2) β × (Δh) δ = 0.5 0.5 × 0.9 0.3 × 3 0.2 ≈ 0.828.

[0095] For Condition 3: A(2, 3) α × D(2, 3) β × (Δh) δ = 0.5 0.5 × 0.5 0.3 × 1 0.2 ≈ 0.256.

[0096] Sum of denominators: 0.645 + 0.828 + 0.256 = 1.729.

[0097] Total score of Feature 2: M2 = 0.828 / 1.729 ≈ 0.479.

[0098] 2) Feature 3 (acceleration change rate): Hypothesis: The weight value of Feature 3 under Condition 1, A(3, 1) = 0.3, the matching score D(3, 1) = 0.6, and the distance Δh = 2.

[0099] The weight value of Feature 3 under Condition 2, A(3, 2) = 0.2, the matching score D(3, 2) = 0.4, and the distance Δh = 3.

[0100] The weight value A(3, 3) of Feature 3 under Condition 3 is 0.5, the matching score D(3, 3) is 0.8, and the distance Δh is 1.

[0101] Calculate: For Condition 1: A(3, 1) α × D(3, 1) β × (Δh) δ = 0.3 0.5 × 0.6 0.3 × 2 0.2 ≈ 0.511.

[0102] For Condition 2: A(3, 2) α × D(3, 2) β × (Δh) δ = 0.2 0.5 × 0.4 0.3 × 3 0.2 ≈ 0.406.

[0103] For Condition 3: A(3, 3) α × D(3, 3) β × (Δh) δ = 0.5 0.5 × 0.8 0.3 × 1 0.2 ≈ 0.656.

[0104] The sum of the denominators: 0.511 + 0.406 + 0.656 = 1.573.

[0105] The total score of Feature 3: M3 = 0.656 / 1.573 ≈ 0.417.

[0106] Step 4. Evaluate the importance of the candidate features: 1) Record the candidate features: After the iteration ends, record all candidate features and their scores: Feature 1: 0.462.

[0107] Feature 2: 0.479.

[0108] Feature 3: 0.417.

[0109] 2) Sort: Sort the candidate features according to the scores: Feature 2: 0.479.

[0110] Feature 1: 0.462.

[0111] Feature 3: 0.417.

[0112] 3) Determine the importance ranking: Feature 2 > Feature 1 > Feature 3.

[0113] Step Five: Identify key factors: 1) Analysis result: According to the importance ranking, it is identified that the key factor having the greatest impact on the motion performance of the industrial robot is Feature 2 (speed fluctuation).

[0114] 2) Optimization suggestions: Optimize for speed fluctuation, for example, adjust the speed loop parameters in the control algorithm to improve the motion smoothness.

[0115] In this alternative embodiment, importance evaluation is performed on the selected candidate features, and the contribution degree of each feature to the motion performance is quantified by using the feature weight ranking search algorithm, and its importance ranking includes: Initialize the parameters of the feature weight ranking algorithm, set the maximum search loop times of the feature weight ranking search algorithm, and generate the initial feature weight combination; Update the search loop times of the feature weight ranking algorithm, and determine whether the maximum search loop times is reached. If it is reached, terminate the feature weight ranking algorithm and output the feature weight ranking result; otherwise, continue to perform the ranking search; Use the current feature weight combination of the upper layer search as the initial feature weight combination of the lower layer search, set the lower layer search loop times, and enter the lower layer search loop; Perform the block structure feature weight combination operation in the lower layer search, generate a new feature weight combination in combination with the feature optimization algorithm, arrange each feature weight combination in ascending order of the fitness value of the combination, and select the current feature weight combination; Determine whether the current feature weight combination violates the taboo rule. If it violates the taboo rule, skip the current feature weight combination and select the next feature weight combination to continue the evaluation. Otherwise, determine whether the fitness value of the current feature weight combination has been optimized. If it has been optimized, use the current feature weight combination as the current optimal feature weight combination; otherwise, continue the evaluation; Update the optimal feature weight combination and continue to loop until the maximum search loop times is satisfied, return to the upper layer search, and continue to execute the search loop of the upper layer search until the maximum search loop times is reached; finally, output the feature weight ranking result, determine the contribution degree of each feature to the motion performance and its importance ranking.

[0116] Specifically, the feature weight ranking search algorithm is a hybrid taboo search algorithm. The algorithm partitions the feasible domain, selects sub-regions through the taboo search algorithm based on two neighborhood operations of insertion and exchange, and searches for excellent solutions in the sub-regions by using the taboo search algorithm based on the block structure neighborhood operation.

[0117] It should be noted that the initial parameters of the algorithm are first set, including the maximum number of search loops and the initial feature weight combination; the initial feature weight combination is generated by random initialization of candidate features or a certain heuristic method, representing the preliminary weights of each feature; enter the first-layer search loop, in which the importance of features is evaluated through the feature weight sorting algorithm; in each loop, the algorithm evaluates the impact of each feature on the motion performance according to the current feature weight combination. The algorithm will adjust the feature weights through a certain mechanism in order to achieve the optimal feature combination; after each upper-layer search, the number of search loops is updated, and it is checked whether the maximum number of loops has been reached; if so, the search ends and the current feature weight sorting result is output; in the upper-layer search loop, if it is found that the current feature weight combination has not reached the optimum, the feature weight combination of the upper-layer search result will be used as the initial combination for the lower-layer search; set the number of lower-layer search loops and enter the lower-layer search loop; in the lower-layer search, first perform the "block structure feature weight combination operation". Here, the weight combination of features is updated according to the block structure of features (such as feature groups, feature subsets, etc.). Then, combined with feature optimization algorithms (such as techniques based on genetic algorithms, simulated annealing, lightning search algorithms, etc.), new feature weight combinations are generated and sorted in ascending order according to the fitness values of the combinations; the fitness value reflects the degree of influence of the current feature combination on the motion performance, and the higher the fitness, the greater the contribution of the combination to the motion performance; after each new feature weight combination is generated, the algorithm determines whether the current feature weight combination violates the preset taboo rules. The taboo rules usually include dependencies between features, excessive weights, etc.; if the current combination violates the taboo rules, then this combination is skipped, and the next combination that conforms to the rules is selected for continued evaluation. The purpose of the taboo rules is to ensure that meaningless or unreasonable feature combinations will not be obtained during the search process; if the maximum number of search loops is not reached, the sorting search continues to optimize the feature combination; if the current feature weight combination does not violate the taboo rules, the algorithm further checks whether the fitness value of this combination has been optimized. If the fitness value of the current combination is improved, then the current combination is updated as the optimal feature weight combination. Otherwise, continue to evaluate the next combination; after the lower-layer search is completed and the optimal feature weight combination is determined, return to the upper-layer search and bring the optimal combination back to the upper-layer search. At this time, the search process will continue according to the optimized combination in order to find the global optimum; finally, when all levels of search loops are completed and the maximum number of search loops is satisfied, the entire search process ends. The contribution degree of each feature to the motion performance is output, and all features are ranked according to the contribution degree. Features with high contribution degrees are considered the most important, and conversely, features with low contribution degrees are ranked at the back.

[0118] In this alternative embodiment, a block structure feature weight combination operation is performed in the lower-layer search, and a new feature weight combination is generated in combination with a feature optimization algorithm. The feature weight combinations are arranged in ascending order of the combined fitness value, and the selection of the current feature weight combination includes: Set the initial parameters of the feature optimization algorithm, and initialize the preliminary feature weight combination and the objective function; Randomly initialize multiple weight combinations in the feature weight space and calculate their corresponding objective function values; Use the known feature weight data set to evaluate each initial feature weight combination, and calculate and record the fitness value of each combination; Arrange all the feature weight combinations in ascending order of their fitness values, preferentially select the feature combination with the best performance, and based on the sorting of the fitness values, select the feature weight combination with the most optimization potential; Generate a new feature weight combination according to the fitness value and optimization objective of the current feature weight combination, and perform an ascending order arrangement again according to the fitness value; select the combination with the best fitness value as the current optimal feature weight combination.

[0119] Specifically, the feature optimization algorithm is the lightning search algorithm, and the inspiration for the lightning search algorithm comes from the process of lightning propagation in nature. During a thunderstorm, the propagation path of lightning is often random and can cover a long distance in an extremely short time. Similarly, the lightning search algorithm quickly traverses the search space by simulating the lightning propagation process to find the optimal solution.

[0120] In this alternative embodiment, based on key factors, a motion prediction model is established to predict the operating state of an industrial robot at a future moment, and the motion trajectory and posture change trend are identified, including: Quantify the key factors, divide them into a training set and a test set according to the data characteristics, and set the maximum number of iterations of the motion prediction model; Generate an initial parameter combination of the motion prediction model through a random initialization strategy combined with the distribution characteristics of the motion data, and configure key parameters for the motion prediction model; According to the laws of motion trajectory and posture change, dynamically adjust the weights in the objective function of the motion prediction model, combine the priority settings of key motion features, and use an optimization algorithm to optimize the motion prediction model to identify the performance of the motion prediction model under different motion states; Use the objective function to evaluate the prediction accuracy of the motion prediction model on the test set for each parameter combination, calculate the fitness value, and sort the fitness values of all parameter combinations; screen out the current optimal motion prediction model parameters; According to the non-linear deviation distribution of the motion prediction model, determine whether it is less than the set threshold. If so, adopt a local optimization strategy to adjust some key parameters; otherwise, adopt a global optimization strategy to adjust all key parameters, generate an optimal motion prediction model, and output the motion trajectory and the trend of attitude change.

[0121] It should be noted that, first of all, the key factors affecting the robot's movement are identified, such as speed, acceleration, load, joint angle, external interference, etc. These factors are used to collect data through sensors or computational models and are quantified. For example, discrete sensor signals are converted into numerical features, or certain factors are standardized to ensure data consistency. According to the collected movement data, common partitioning strategies (such as 70% training set and 30% test set) are used to divide the data into a training set and a test set to ensure that the model can be effectively verified on the training data and unseen data. A random initialization strategy is used to generate an initial parameter combination for the movement prediction model. These parameters may include hyperparameters such as the learning rate of the model, the number of neurons in the hidden layer, and the training batch size. The initial value range is appropriately set according to the distribution characteristics of the movement data, such as whether it shows periodic or linear relationships. For the characteristics of the robot's movement, the key parameters of the model are configured. For example, the prediction time step, the frequency of movement state changes, and the number of features required by the movement model can be set. By configuring these key parameters, it is ensured that the movement prediction model can be optimized according to different movement patterns and control requirements. According to the laws of the robot's movement trajectory and attitude changes, combined with historical data, the weights of the objective function in the movement prediction model are dynamically adjusted. For example, speed and acceleration may be more important than attitude changes, so higher weights need to be given in the objective function. Common optimization algorithms (such as genetic algorithms, particle swarm optimization, gradient descent, mode tracking sampling MPS algorithm, etc.) are used to optimize the objective function of the movement prediction model according to the priority settings in different movement states. By continuously adjusting the weights in the objective function, the model can make accurate predictions in multiple movement states. For each set of parameter combinations, the objective function is used to evaluate the prediction accuracy of the model on the test set. The evaluation criteria may include prediction error, mean square error (MSE), or other metrics related to prediction quality. The fitness value of each set of parameter combinations is calculated, and the fitness values of all parameter combinations are sorted. According to the sorting of the fitness values, the parameter combination with the optimal fitness is selected. In this way, the movement prediction model that can most accurately reflect the movement trajectory and attitude change trend is screened out. By analyzing the prediction results of the model, the non-linear deviation of the movement prediction model is evaluated. If the prediction error of the model (such as the residual) is less than the set threshold (for example, a permitted error range is set), it indicates that the performance of the model has reached the expectation. If the non-linear deviation is less than the threshold, a local optimization strategy is adopted to only adjust some key parameters (such as the control parameters of a specific joint or the parameters of a specific movement stage) to further improve the accuracy. If the non-linear deviation is greater than the set threshold, it means that the model still has a large error and needs to be comprehensively optimized.At this time, a global optimization strategy is adopted to readjust all key parameters to ensure that the model can minimize errors as much as possible and improve prediction accuracy. Combining the foregoing local or global optimization results, an optimal motion prediction model is generated. This model can relatively accurately reflect the operating state of the industrial robot at future moments, predicting information such as the motion trajectory, speed, acceleration, and posture of the robot at different time points. Finally, through the optimized motion prediction model, the motion trajectory and posture change trend of the robot at future moments are output. These prediction results can provide important reference bases for the control and scheduling of industrial robots, helping to optimize the motion path of the robot, improve production efficiency, and reduce potential failure risks.

[0122] In this alternative embodiment, according to the laws of motion trajectory and posture change, the weights in the objective function of the motion prediction model are dynamically adjusted. Combining the priority setting of key motion features, the motion prediction model is optimized using an optimization algorithm. The performance of the motion prediction model in different motion states is identified, including: Randomly sample several motion state data points within the feasible parameter space of the motion prediction model; Using the sampled motion state data points, construct a motion prediction model based on radial basis functions and calculate the weight coefficients of the radial basis functions; Based on the current motion prediction model, use the Monte Carlo probability sampling algorithm for optimized sampling; Construct a quadratic response surface model, calculate the fitting accuracy of the quadratic response surface model, and determine whether its fitting effect meets the predetermined accuracy requirements. If it meets, continue with new sampling; otherwise, adjust the acceleration factor and re-optimize the parameters of the quadratic response surface model; Evaluate the optimized quadratic response surface model to determine whether the fitting accuracy meets the standard. If it meets the standard, use this quadratic response surface model for motion state prediction; otherwise, add the new motion state data points to the training set and continue to optimize the model; Based on the optimized quadratic response surface model, search within the feasible parameter space to identify the optimal motion state prediction parameters and determine the motion state with the minimum prediction error. If the optimization algorithm reaches the maximum number of iterations, output the optimal motion prediction model and identify the performance of the motion prediction model in different motion states.

[0123] It should be noted that before starting the optimization, it is first necessary to determine the feasible parameter space of the motion prediction model. This space includes all possible combinations of motion states, such as different speeds, accelerations, loads, etc. By randomly sampling a number of motion state data points within this feasible parameter space, the motion behavior of the robot under different states is simulated. These data points provide training data for subsequent model construction; Based on the sampled motion state data points, a motion prediction model based on the Radial Basis Function (RBF) is constructed. The RBF network is a non-linear model commonly used for function fitting and pattern recognition, and can effectively handle non-linear relationships in high-dimensional spaces; For each motion state data point, the weight coefficients of the RBF network are calculated, and these weights reflect the influence degree of each motion feature on the motion state prediction; After constructing the preliminary RBF motion prediction model, the Monte Carlo probability sampling algorithm is used for optimized sampling. Monte Carlo sampling explores the most likely motion states by randomly selecting parameter combinations and evaluating the model performance under different motion states. This process is explored through a large number of random samplings to ensure that the motion prediction model can cover a wide range of possible motion states; Using the data obtained by Monte Carlo sampling, a quadratic response surface model is constructed. The response surface model establishes the relationship between the input parameters and the prediction results by fitting the training data; Calculate the fitting accuracy of this response surface model and evaluate the accuracy of the current model. If the error of the model fitting is small and meets the predetermined accuracy requirements, continue with the new sampling process; If the fitting effect fails to reach the predetermined accuracy, adjust the acceleration factor of the response surface model to more effectively improve the fitting accuracy; If the fitting effect of the quadratic response surface model is not ideal, adjust the acceleration factor to improve the optimization process. For example, the sampling frequency can be increased or the sampling range can be expanded to obtain more effective motion state data points; According to the adjusted sampling results, re-optimize the parameters of the quadratic response surface model. Through continuous iterative optimization, gradually improve the fitting accuracy of the model until the model performance reaches the expected standard; Evaluate the optimized quadratic response surface model. By comparing the prediction results of the model with the actual motion state data, calculate the fitting error and determine whether it meets the accuracy requirements; If the fitting accuracy of the model has reached the standard, use this quadratic response surface model to predict future motion states; If the fitting accuracy still does not meet the standard, add the new motion state data points to the training set and continue to optimize the model to obtain more accurate predictions; Conduct a final search on the optimized quadratic response surface model. By conducting a systematic search within the feasible parameter space, identify the optimal motion state prediction parameters. These parameters represent the best prediction conditions for the robot under different motion states; Through multiple iterations and optimizations, determine the motion state with the smallest prediction error and use this state as the best future motion state prediction; When the optimization algorithm reaches the maximum number of iterations, output the optimal motion prediction model.This model can make accurate predictions under different motion states, providing support for the control and scheduling of industrial robots; by identifying the performance of the motion prediction model under different motion states, the adaptability and accuracy of the model are evaluated. Through this method, the robot can adjust its motion strategy according to the prediction results, thereby optimizing performance and efficiency.

[0124] Specifically, the optimization algorithm is the mode tracking sampling MPS algorithm. Through mode tracking sampling, the MPS algorithm generates more design sample points in the region where the function model is more likely to have minimum value points, uses the Radial Basis Function (RBF) to determine the region of the global optimal solution, and then constructs a quadratic polynomial response surface model RSM (Response Surface Model) for global convergence determination.

[0125] In this alternative embodiment, the formula for calculating the fitting accuracy of the quadratic response surface model is: ; In the formula, R 2 represents the fitting accuracy of the quadratic response surface model; y a represents the actual response value of the a-th motion state data point; represents the model predicted response value of the a-th motion state data point; n represents the total number of motion state data points; represents the mean value of the actual response values of all motion state data points.

[0126] According to another embodiment of the present invention, as Figure 2 shown, an industrial robot automatic control method is also provided. This method includes the following steps: S1. Obtain the real-time operation data of the industrial robot, process the real-time operation data, and extract operation feature data; S2. Use the feature analysis algorithm to perform feature analysis on the operation feature data to identify the key factors affecting the motion performance of the industrial robot; S3. Based on the key factors, establish a motion prediction model to predict the operation state of the industrial robot at a future moment, and identify the motion trajectory and posture change trend; S4. Based on the motion trajectory and posture change trend, combined with the real-time operation state of the industrial robot, dynamically adjust the motion parameters of the industrial robot and optimize the motion control strategy.

[0127] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An industrial robot automatic control system, characterized in that, The system includes: A data management module, which is used to obtain the real-time operation data of the industrial robot, process the real-time operation data, and extract the operation feature data; A data analysis module, which is used to perform feature analysis on the operation feature data by using a feature analysis algorithm to identify the key factors affecting the motion performance of the industrial robot; A motion prediction module, which is used to establish a motion prediction model based on the key factors to predict the operation state of the industrial robot at a future moment and identify the motion trajectory and posture change trend; A coordination control module, which is used to dynamically adjust the motion parameters of the industrial robot and optimize the motion control strategy based on the motion trajectory and posture change trend in combination with the real-time operation state of the industrial robot.

2. The automated control system of an industrial robot according to claim 1, wherein The data management module includes: A data acquisition module, which is used to obtain the real-time operation data of the industrial robot, and perform denoising, filtering, and smoothing processing on the duplicate data, missing values, and outliers of the real-time operation data to obtain accurate real-time operation data; A feature extraction module, which is used to perform feature extraction on the obtained accurate real-time operation data by using a clustering algorithm to obtain the operation feature data.

3. An industrial robot automatic control system according to claim 1, characterized in that, The performing feature analysis on the operation feature data by using a feature analysis algorithm to identify the key factors affecting the motion performance of the industrial robot includes: Based on the operation feature data, initialize each feature parameter and construct an initial feature matrix; Preset the maximum number of iterations of the feature analysis algorithm. In each iteration, randomly select a feature subset, and calculate the correlation score between each feature and the motion performance by using a score formula to screen out the candidate features with the greatest impact on the motion performance; Perform importance evaluation on the screened candidate features, and use a feature weight sorting search algorithm to quantify the contribution degree of each feature to the motion performance and determine its importance ranking; According to the importance ranking, identify the key factors with the greatest impact on the motion performance of the industrial robot.

4. An industrial robot automatic control system according to claim 3, characterized in that, The score formula is: ； Wherein, M i represents the correlation score between the i-th feature and the motion performance; A(i, j) represents the weight value of the i-th feature under the j-th operating condition; D(i, j) represents the matching score of the motion performance of the i-th feature under the j-th operating condition; Δh represents the distance between the i-th feature and the starting analysis position of the current feature; α represents the adjustment parameter of the feature weight; β represents the adjustment parameter of the feature matching score; δ represents the adjustment parameter of the position distance; L i represents the starting analysis position of the i-th feature.

5. An industrial robot automatic control system according to claim 3, characterized in that, The performing importance evaluation on the screened candidate features and using a feature weight sorting search algorithm to quantify the contribution degree of each feature to the motion performance and determine its importance ranking includes: Initialize the parameters of the feature weight sorting algorithm, set the maximum search loop number of the feature weight sorting search algorithm, and generate an initial feature weight combination; Update the search loop number of the feature weight sorting algorithm, and determine whether the maximum search loop number is reached. If it is reached, terminate the feature weight sorting algorithm and output the feature weight sorting result; otherwise, continue to perform sorting search; Use the current feature weight combination searched in the upper layer as the initial feature weight combination for the lower layer search, set the lower layer search loop number, and enter the lower layer search loop; Perform a block structure feature weight combination operation in the lower layer search, and generate a new feature weight combination in combination with a feature optimization algorithm. Arrange the feature weight combinations in ascending order of the fitness value of the combination, and select the current feature weight combination; Determine whether the current feature weight combination violates the tabu rule. If it violates the tabu rule, skip the current feature weight combination and select the next feature weight combination to continue the evaluation. Otherwise, determine whether the fitness value of the current feature weight combination has been optimized. If it has been optimized, use the current feature weight combination as the current optimal feature weight combination; otherwise, continue the evaluation; Update the optimal feature weight combination and continue the loop until the maximum search loop count is reached. Return to the upper-level search and continue to execute the search loop of the upper-level search until the maximum search loop count is reached; finally, output the feature weight sorting result to determine the contribution degree of each feature to the motion performance and its importance ranking.

6. An industrial robot automatic control system according to claim 5, characterized in that, The operation of the block structure feature weight combination is performed in the lower-level search, and a new feature weight combination is generated in combination with the feature optimization algorithm. Each feature weight combination is arranged in ascending order of the fitness value of the combination, and the selection of the current feature weight combination includes: Set the initial parameters of the feature optimization algorithm, and initialize the preliminary feature weight combination and the objective function; Randomly initialize multiple weight combinations in the feature weight space and calculate their corresponding objective function values; Evaluate each initial feature weight combination using the known feature weight data set, and calculate and record the fitness value of each combination; Arrange all feature weight combinations in ascending order of their fitness values, preferentially select the feature combination with the best performance, and select the feature weight combination with the most optimization potential based on the sorting of the fitness values; Generate a new feature weight combination according to the fitness value and optimization objective of the current feature weight combination, and perform an ascending order arrangement again according to the fitness value; select the combination with the optimal fitness value as the current optimal feature weight combination.

7. An industrial robot automatic control system according to claim 1, characterized in that, Based on the key factors, establish a motion prediction model to predict the operating state of the industrial robot at a future moment, and identify the motion trajectory and posture change trend, including: Quantify the key factors, divide them into a training set and a test set according to the data characteristics, and set the maximum number of iterations of the motion prediction model; Generate an initial parameter combination of the motion prediction model through a random initialization strategy in combination with the distribution characteristics of the motion data, and configure the key parameters for the motion prediction model; Dynamically adjust the weights in the objective function of the motion prediction model according to the laws of the motion trajectory and posture change, combine the priority setting of the key motion features, and use the optimization algorithm to optimize the motion prediction model to identify the performance of the motion prediction model in different motion states; Use the objective function to evaluate the prediction accuracy of the motion prediction model on the test set for each parameter combination, calculate the fitness value, and sort the fitness values of all parameter combinations; screen out the current optimal motion prediction model parameters; Judge whether the non-linear deviation distribution of the motion prediction model is less than the set threshold. If so, adopt a local optimization strategy to adjust some key parameters; otherwise, adopt a global optimization strategy to adjust all key parameters to generate the optimal motion prediction model and output the motion trajectory and posture change trend.

8. An industrial robot automatic control system according to claim 7, characterized in that, According to the laws of motion trajectory and attitude change, dynamically adjust the weights in the objective function of the motion prediction model. Combining with the priority setting of key motion features, use an optimization algorithm to optimize the motion prediction model. The performance of the motion prediction model in different motion states is identified as follows: Randomly sample a number of motion state data points within the feasible parameter space of the motion prediction model; Use the sampled motion state data points to construct a motion prediction model based on radial basis functions and calculate the weight coefficients of the radial basis functions; Based on the current motion prediction model, use the Monte Carlo probability sampling algorithm for optimized sampling; Construct a quadratic response surface model, calculate the fitting accuracy of the quadratic response surface model, and judge whether its fitting effect meets the predetermined accuracy requirements. If it meets, continue with new sampling; otherwise, adjust the acceleration factor and re-optimize the parameters of the quadratic response surface model; Evaluate the optimized quadratic response surface model to judge whether the fitting accuracy meets the standard. If it meets, use this quadratic response surface model for motion state prediction; otherwise, add the new motion state data points to the training set and continue to optimize the model; Based on the optimized quadratic response surface model, search within the feasible parameter space to identify the optimal motion state prediction parameters and determine the motion state with the minimum prediction error. If the optimization algorithm reaches the maximum number of iterations, output the optimal motion prediction model and identify the performance of the motion prediction model in different motion states.

9. An industrial robot automatic control system according to claim 8, characterized in that, The formula for calculating the fitting accuracy of the quadratic response surface model is: ； where R 2 represents the fitting accuracy of the quadratic response surface model; y a represents the actual response value of the a-th motion state data point; represents the model predicted response value of the a-th motion state data point; n represents the total number of motion state data points; represents the mean value of the actual response values of all motion state data points.

10. An industrial robot automatic control method, which adopts the industrial robot automatic control system described in any one of claims 1-9, is characterized in that, The method includes the following steps: S1. Obtain the real-time operation data of the industrial robot, process the real-time operation data, and extract the operation feature data; S2. Use the feature analysis algorithm to analyze the operation feature data and identify the key factors affecting the motion performance of the industrial robot; S3. Based on the key factors, establish a motion prediction model to predict the operation state of the industrial robot at a future moment and identify the motion trajectory and attitude change trend; S4. Based on the motion trajectory and attitude change trend, combined with the real-time operation state of the industrial robot, dynamically adjust the motion parameters of the industrial robot and optimize the motion control strategy.

Images