Industrial robot trajectory optimization control method based on intelligent algorithm

Abstract

The invention relates to the technical field of industrial robot control, and discloses an industrial robot trajectory optimization control method based on an intelligent algorithm. The method comprises the steps that joint position information, tool center point coordinates and a motion time sequence when a robot executes multiple tasks are collected and stored in a track database; after cleaning and screening data, extracting a feature set containing a path point sequence, speed distribution and an acceleration contour; constructing an intelligent optimization algorithm model of a neural network structure, and training by using the feature set to learn a trajectory optimization strategy; analyzing a target position coordinate and a motion constraint condition of the current task to obtain an initial track parameter; and inputting the initial parameters into the trained model, outputting an optimized track sequence containing a path point list and a speed curve, and generating a control instruction to drive the robot to move. The method can adapt to different tasks, improves track rationality and motion stability, and fits industrial production practice.

Description

Technical Field

[0001] This invention relates to the field of industrial robot control technology, specifically to a control method for optimizing the trajectory of industrial robots based on intelligent algorithms. Background Technology

[0002] In intelligent manufacturing systems, industrial robots, with their multi-degree-of-freedom motion capabilities and continuous operational stability, have been deeply integrated into core production processes such as automotive welding, 3C electronics assembly, mechanical parts handling, and precision instrument processing. In these application scenarios, the motion trajectory of industrial robots is directly related to the smoothness of the workflow and the quality of the final product. For example, during automotive body welding, trajectory deviations can lead to misaligned weld points, affecting the structural strength of the car body; during electronic component assembly, insufficient trajectory accuracy can cause component damage or improper installation, increasing production losses.

[0003] Currently, industrial robot trajectory planning largely relies on traditional algorithms. These algorithms typically construct trajectories based on pre-set mathematical models, requiring technicians to input a large number of fixed parameters in advance. However, in actual industrial production, the target position, load weight, and speed limits for different tasks vary significantly, making it difficult for traditional algorithms to flexibly adapt to diverse working conditions. For example, on the same production line, when a robot switches from "heavy parts handling" to "light component assembly," traditional algorithms need to readjust the entire set of parameters, which not only consumes a lot of time but may also cause trajectory transition stutters due to insufficient parameter matching.

[0004] The inherent mechanical structure of industrial robots also poses challenges to trajectory accuracy. Robotic linkages and drive motors undergo slight flexible deformation during long-term operation, especially during high-speed motion or load changes. This deformation can cause uncontrolled offsets in the end effector. Traditional algorithms cannot detect and compensate for these offsets in real time, leading to deviations between the trajectory and the preset path. Furthermore, traditional trajectory planning often neglects the correlation between motion time series and velocity and acceleration, focusing only on the accuracy of path point positions. This can easily result in sudden increases and decreases in speed during movement, which not only accelerates equipment wear but may also cause mechanical vibrations, affecting the robot's service life and operational stability.

[0005] Existing trajectory planning schemes lack effective utilization of historical motion data. Industrial robots generate a large amount of trajectory data during long-term task execution, but this data is mostly stored as records and not used to optimize subsequent trajectory planning. This results in trajectory planning for each new task starting "from scratch," making it impossible to learn from the experience of past successful operations. This reduces trajectory planning efficiency and makes it difficult to continuously improve trajectory accuracy, failing to meet the ever-increasing demands of modern industry for production efficiency and product quality. Summary of the Invention

[0006] The purpose of this invention is to provide a control method for optimizing the trajectory of industrial robots based on intelligent algorithms, so as to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a control method for industrial robot trajectory optimization based on intelligent algorithms, the method comprising: Collect motion trajectory data generated by industrial robots when performing various work tasks. The motion trajectory data includes joint position information, tool center point coordinates and motion time series, and store the motion trajectory data in the trajectory database. The motion trajectory data in the trajectory database is cleaned and filtered to extract the feature set of the motion trajectory data, which includes path point sequence, velocity distribution and acceleration profile. An intelligent optimization algorithm model is constructed. The intelligent optimization algorithm model adopts a neural network structure and is trained using a feature set, so that the intelligent optimization algorithm model learns trajectory optimization strategies. Obtain the task description information of the current industrial robot, which includes the target position coordinates and motion constraints. Parse the task description information to obtain the initial trajectory parameters. The initial trajectory parameters are input into the trained intelligent optimization algorithm model, and the intelligent optimization algorithm model outputs the optimized motion trajectory sequence, which includes a list of path points and a velocity curve. Control commands are generated based on the optimized motion trajectory sequence, and the controller of the industrial robot drives the motors of each joint, so that the industrial robot moves according to the optimized motion trajectory sequence.

[0008] Preferably, the step of collecting motion trajectory data generated by the industrial robot when performing multiple work tasks includes: Joint angle data is collected by encoder sensors installed at the joints of industrial robots; Capture the spatial coordinate data of the tool's center point using a vision system; Record the joint angle data and tool center point coordinate data corresponding to each motion timestamp; Associate the joint angle data, tool center point coordinate data and motion timestamp data with the task identifier, which includes the task type and workpiece number. The associated motion trajectory data is stored in the data table of the trajectory database.

[0009] Preferably, the step of cleaning and filtering the motion trajectory data in the trajectory database includes: Read motion trajectory data from the data table in the trajectory database; Check the integrity and consistency of motion trajectory data, and remove records containing missing or outlier values; Motion trajectory data is filtered according to a preset accuracy threshold, and records with path errors less than the threshold are retained; Extract path point sequences from the filtered records. The path point sequences are composed of the coordinates of the tool's center point in chronological order. Calculate the velocity and acceleration values ​​between the path point sequences to form the velocity distribution and acceleration profile.

[0010] Preferably, the step of constructing the intelligent optimization algorithm model includes: Initialize the layer structure and parameters of the neural network model, which includes an input layer, hidden layers, and an output layer; The path point sequence, velocity distribution, and acceleration profile in the feature set are used as training input data. Set the optimization objective function, which includes minimizing the path and maximizing the smoothness; Use the gradient descent algorithm to adjust the parameters of the neural network model and minimize the optimization objective function; Verify the performance of the trained intelligent optimization algorithm model on the test dataset and save the model parameters.

[0011] Preferably, the step of obtaining the task description information of the current industrial robot includes: Receive task instructions from user input or the host system; Parse the target location coordinates in the task instructions; the target location coordinates include three-dimensional spatial coordinates. Extract the motion constraints from the task instructions. The motion constraints include the maximum speed limit, acceleration limit, and obstacle avoidance area. The target position coordinates and motion constraints are combined to form the initial trajectory parameters.

[0012] Preferably, the step of generating the optimized motion trajectory sequence includes: Input the initial trajectory parameters into the trained intelligent optimization algorithm model; The intelligent optimization algorithm model calculates the list of path points through forward propagation; Generate a continuous velocity curve by interpolation based on the list of path points; Adjust the velocity curve to meet the motion constraints; Output the optimized motion trajectory sequence, including a list of path points and velocity curves.

[0013] Preferably, the step of generating control instructions includes: The optimized motion trajectory sequence is converted into joint space trajectory, and the target angle of each joint is calculated by inverse kinematics. Generate speed control signals for each joint based on the speed curve; Send the target angle and speed control signals to the joint actuators of the industrial robot; Joint actuators control the rotation of motors, enabling the movement of industrial robots.

[0014] Preferably, the method further includes the following steps: Real-time monitoring of the actual movement trajectory of industrial robots, and acquisition of actual joint angles and tool center point coordinates through sensors; Compare the actual motion trajectory with the optimized motion trajectory sequence, and calculate the trajectory tracking error; If the trajectory tracking error exceeds the allowable range, adjust the speed control signal in the control command; Repeat the monitoring and adjustment process until the task is completed.

[0015] Preferably, the step of adjusting the speed control signal in the control command includes: Calculate the speed correction amount based on the magnitude and direction of the trajectory tracking error; The speed correction amount is superimposed on the original speed control signal; The adjusted speed control signal is limited to not exceeding the maximum speed limit in the motion constraints. The adjusted speed control signal is sent to the joint actuator.

[0016] Preferably, the method further includes the following steps: After the task is completed, record the actual motion trajectory data and its performance indicators; Add the actual motion trajectory data to the trajectory database; Update the training set of the intelligent optimization algorithm model with new data and retrain the model; Save the updated model parameters for subsequent tasks.

[0017] Compared with the prior art, the beneficial effects of the present invention are: By collecting motion trajectory data of industrial robots performing various tasks, including joint position information, tool center point coordinates, and motion time series, and storing this data in a trajectory database, we can comprehensively accumulate trajectory information under different working conditions, providing rich practical data support for subsequent trajectory optimization. This data is not isolated fragments of information, but rather encompasses the robot's actual motion state under different task types, load conditions, and motion speeds. This allows subsequent optimization processes to better align with real-world production scenarios, avoiding theoretical planning divorced from reality.

[0018] After cleaning and filtering the data in the trajectory database, feature sets are extracted, including pathpoint sequences, velocity distributions, and acceleration profiles. This process removes outliers and redundant information, allowing the extracted features to more accurately reflect the core characteristics of the trajectory. For example, by filtering out erroneous joint position data caused by temporary sensor malfunctions and by cleaning and organizing chaotic motion time series, the consistency and reliability of the feature data used for subsequent model training are ensured. This makes the trajectory patterns learned by the model more consistent with the robot's actual motion logic and reduces the interference of invalid data on the optimization results.

[0019] A smart optimization algorithm model employing a neural network structure was constructed and trained using a feature set. This enabled the model to learn trajectory optimization strategies, fully leveraging the advantages of neural networks in handling nonlinear problems. Compared to the limitations of traditional algorithms that rely on fixed mathematical models, neural networks can autonomously uncover potential correlations between trajectory parameters from a large amount of feature data. For example, they can identify the matching relationship between path point sequences and velocity distributions, and the influence of acceleration profiles on motion stability, thereby forming more adaptive optimization logic. This data-driven optimization approach eliminates the need for technicians to manually adjust complex parameters, allowing it to autonomously adapt to the needs of different tasks and reducing reliance on human experience.

[0020] Obtaining the description information of the current task, including the target position coordinates and motion constraints, and parsing it to obtain the initial trajectory parameters ensures that the initial trajectory planning is closely aligned with the actual requirements of the current task. In actual production, the target positions of different tasks vary significantly, and the motion constraints also differ. For example, some tasks require rapid movement to improve efficiency, while others require low-speed, stable movement to ensure accuracy. The initial trajectory parameters obtained by parsing the task description information can accurately match the specific requirements of the current task, laying a realistic foundation for subsequent optimization and avoiding deviations in the optimization direction caused by a disconnect between the initial parameters and task requirements.

[0021] Inputting initial trajectory parameters into the trained model outputs an optimized motion trajectory sequence containing a list of path points and velocity curves. This allows the trajectory to meet task requirements while optimizing the motion state. The optimized path point list ensures coordinated movement of all joints during robot motion, reducing unnecessary path detours; the optimized velocity curves prevent sudden speed changes, resulting in smoother robot movement, reduced mechanical vibration, and lower equipment wear. This optimization is not a single-dimensional adjustment but a comprehensive optimization that balances path rationality and motion smoothness, enabling the robot to perform tasks efficiently and stably.

[0022] Control commands are generated based on the optimized motion trajectory sequence. These commands, driven by a controller, drive the joint motors, enabling the robot to move along the optimized trajectory. This effectively translates the optimization results into actual operational behavior. The generated control commands are precisely matched to the robot's hardware structure, ensuring that each joint motor operates collaboratively according to the optimized speed and position requirements, avoiding motion deviations caused by incompatibility between commands and hardware. In actual operation, this process allows the robot's movement to more accurately align with the task objectives, reducing operational errors, improving the smoothness of the production process, and simultaneously lowering the probability of equipment failure due to unreasonable trajectories, reducing maintenance downtime, and ensuring production continuity. Attached Figure Description

[0023] Figure 1 This is a schematic diagram illustrating the working principle of the control method for industrial robot trajectory optimization based on intelligent algorithms described in this invention. Figure 2 A flowchart for collecting motion trajectory data; Figure 3 A flowchart for generating the optimized motion trajectory sequence. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Please see Figure 1This invention provides a control method for industrial robot trajectory optimization based on intelligent algorithms. The method includes: collecting motion trajectory data generated by the industrial robot while performing various tasks; storing this data in a trajectory database, including joint position information, tool center point coordinates, and motion time sequences; cleaning and filtering the motion trajectory data in the database to extract a feature set, including path point sequences, velocity distributions, and acceleration profiles; constructing an intelligent optimization algorithm model using a neural network structure and training it with the feature set to learn trajectory optimization strategies; acquiring the task description information of the industrial robot, including target position coordinates and motion constraints, and parsing it to obtain initial trajectory parameters; inputting these initial trajectory parameters into the trained intelligent optimization algorithm model, which outputs an optimized motion trajectory sequence, including a path point list and velocity curves; and generating control commands based on the optimized motion trajectory sequence, driving the joint motors through the industrial robot's controller to move the industrial robot according to the optimized motion trajectory sequence.

[0026] Example 1: See Figure 2The acquisition of industrial robot motion trajectory data is achieved through encoder sensors integrated into the robot's joints. These encoder sensors are absolute optical encoders with a resolution of millions of pulses per revolution, capable of capturing minute angular changes in the joints in real time. Joint angle data is output as digital signals and transmitted to a data acquisition card via an industrial fieldbus. The data acquisition card filters and digitizes the signals to eliminate electrical noise interference. A vision system is deployed above the robot's working unit. The vision system consists of a high-frame-rate industrial camera and machine vision software. The industrial camera is connected to the processing unit via gigabit Ethernet. The machine vision software calibrates and transforms the captured images, calculating the three-dimensional coordinates of the tool's center point in the base coordinate system. Each motion timestamp is generated by a high-precision clock chip. The timestamp data, joint angle data, and tool center point coordinate data are strictly time-aligned via a hardware synchronization signal, which is uniformly distributed by the robot controller. The process of associating motion trajectory data with task identifiers is completed in the data preprocessing module. The task identifier includes a task type classification code and a workpiece serial number. The task type classification code is defined according to international standard process codes, and the workpiece serial number corresponds to the material tracking number in the production management system. The associated operations employ a database transaction processing mechanism to ensure the atomicity and consistency of data records. The trajectory database adopts a time-series database architecture, with data tables including time index fields, joint angle array fields, tool center point coordinate fields, and task identifier fields. The database engine optimizes the storage efficiency of large-scale time-series data through partitioning table technology. The data acquisition frequency is dynamically adjusted according to the robot's movement speed, using a kilohertz sampling rate for high-speed movement and reducing it to the 100hertz level for low-speed movement, balancing data accuracy and storage overhead.

[0027] The motion trajectory data cleaning process in the trajectory database begins with a data quality check. The data quality check module scans each record in the data table, checking whether joint angle data exceeds mechanical limits and whether the tool center point coordinates are within the robot's workspace envelope. Records containing missing values ​​are identified using database query statements. Outlier detection employs statistical process control methods, calculating the standard deviation of data within the moving window and removing data points deviating more than three times the standard deviation from the mean. Preset accuracy thresholds are derived from the robot manufacturer's technical specifications. Path error calculation is performed by comparing the deviation between the actual tool center point trajectory and the theoretical trajectory, which is generated by a computer-aided manufacturing system. After filtering, the records proceed to the feature extraction stage. Path point sequence reconstruction uses a time alignment algorithm, resampling the tool center point coordinates at millisecond intervals to generate a spatial path point array with a uniform time step. Velocity distribution calculation is based on the time derivative of the path point sequence; the instantaneous velocity vector is obtained by dividing the displacement vector between adjacent path points by the time interval. The magnitude of the velocity vector in three-dimensional space constitutes the velocity distribution curve. Acceleration profiles are derived from the time derivative of the velocity distribution curve, and acceleration values ​​are calculated using the central difference method. Smoothing is achieved using a moving average filter to reduce numerical noise. The feature extraction module outputs a standardized feature set, converts the path point sequence into a homogeneous transformation matrix representation relative to the base coordinate system, and normalizes the velocity distribution and acceleration profile to eliminate the impact of dimensional differences on model training. The entire data cleaning and screening process is embedded in the real-time operating system of an industrial computer, improving computational efficiency through multi-threaded parallel processing, and recording processing logs to the system audit database for quality traceability.

[0028] The hardware configuration of the data acquisition system includes a signal conditioning circuit, which amplifies and isolates the raw encoder signal to prevent ground loop interference. The vision system calibration process employs the Zhang Zhengyou calibration method, with the camera intrinsic and extrinsic parameter matrices updated periodically to compensate for optical system drift caused by mechanical vibration. Motion trajectory data storage utilizes a redundant disk array architecture, and the data backup strategy follows industrial data management standards. Hot backup data is used for real-time querying, while cold backup data is used for long-term archiving. The trajectory database access interface provides both structured query language and application programming interface (API) methods, supporting integration with third-party analysis tools. Outlier handling in trajectory data cleaning incorporates machine learning algorithms; the isolated forest model identifies abnormal patterns in high-dimensional data, assisting traditional statistical methods to improve detection sensitivity. Path error calculation fuses multi-sensor data, with the spatial position measured by the laser tracker used as a benchmark to verify the vision system's accuracy. Feature set construction considers kinematic constraints; the curvature continuity of the path point sequence is ensured through spline interpolation, and the derivative continuity of the velocity distribution and acceleration profile meets the robot's motion smoothness requirements. The data preprocessing pipeline achieves automated scheduling, and the monitoring interface displays data quality indicators and processing progress. Operators can manually intervene to filter threshold parameters.

[0029] The accuracy of industrial robot motion trajectory data acquisition is affected by mechanical transmission errors. A transmission error compensation model is embedded in the data acquisition process, built upon robot calibration data, to correct joint angle readings in real time. Tool center point coordinate calculation involves a complex coordinate system transformation chain, including the robot base coordinate system, flange coordinate system, tool coordinate system, and vision sensor coordinate system. Each coordinate system transformation is achieved through a homogeneous transformation matrix. The time synchronization system adopts the IEEE 1588 precision clock protocol, with clock synchronization errors at each node in the network controlled within microseconds. Task identifier management is integrated with the production order system. Task type classification codes are mapped to the process route library of the manufacturing execution system, and workpiece serial numbers are associated with the bill of materials in the product lifecycle management system. The motion trajectory data storage scheme is designed to support a distributed architecture, with edge computing nodes responsible for local data caching and the cloud platform responsible for historical data archiving. The data cleaning rule engine supports visual configuration, allowing users to customize integrity check rules and anomaly detection algorithms. Accuracy threshold parameters are dynamically adjusted according to processing requirements; strict thresholds are used for precision assembly tasks, while lenient thresholds are used for roughing tasks. Feature extraction of path point sequences includes geometric characteristic analysis, calculating the curvature tensor and torsion distribution of path points to describe the spatial distortion of the trajectory. Velocity distribution analysis takes kinematic constraints into account, compares the velocity extreme points with the robot's dynamic limits, and uses acceleration profile analysis to identify abrupt changes in agility.

[0030] Example 2: The intelligent optimization algorithm model is constructed based on a deep neural network architecture, which consists of a fully connected structure including an input layer, three hidden layers, and an output layer. The number of neurons in the input layer matches the dimension of the feature set, which includes coordinate data of the path point sequence, scalar values ​​of velocity distribution, and vector components of acceleration contours. The hidden layer uses the rectified linear unit activation function, which introduces nonlinear transformation capability. The number of neurons in the hidden layer follows the geometric pyramid decreasing rule. The output layer is designed as a multidimensional regression output, and the number of neurons in the output layer corresponds to the dimension of the optimized trajectory parameters. The weights of the neural network model are initialized using the Xavier initialization method, and the bias term is initialized to zero. The optimization objective function is defined as a multi-objective weighted sum, which includes a path length term and a smoothness term. The path length term is calculated by accumulating the Euclidean distance between trajectory points, and the smoothness term is measured by the integral of acceleration change. The training process uses an adaptive moment estimation algorithm, which dynamically adjusts the learning rate. The training data is divided into batch inputs, with a batch size of 256 samples. The loss function is calculated using mean squared error (MSE), which measures the deviation between the predicted and ideal trajectories. Backpropagation calculates the gradient of the loss function with respect to the weights, and gradient clipping prevents gradient explosion. Training iterations are set to 10,000 epochs, and early stopping is used to monitor changes in the validation set loss value. The validation set randomly selects 20% of the data from the feature set, while the test set uses independently collected trajectory data. The training platform is equipped with a graphics processing unit (GPU) accelerator, which performs parallel matrix operations using the CUDA architecture. Model parameters are saved as binary files, containing the weight matrix and bias vector.

[0031] The task description information acquisition module integrates multiple input interfaces, including a user interface and a manufacturing execution system (MES) interface. The user interface provides a graphical parameter input form, allowing users to input the X, Y, and Z component values ​​of the target position coordinates. The MES interface uses the OPCUA communication protocol and receives production order data in real time. The task instruction parser analyzes the grammatical structure of the input data, ensuring it conforms to a predefined job description language specification. Target position coordinates are extracted through regular expression matching, converting coordinate values ​​to floating-point format. Motion constraint parsing identifies keywords; maximum speed limits are mapped from the robot's technical specifications, and acceleration limits are determined based on mechanical structure strength. Obstacle avoidance areas are defined using bounding box representation, consisting of minimum and maximum coordinate points. The initial trajectory parameter combination process performs data integrity checks, verifying the existence of necessary parameters. Target position coordinates and motion constraints are packaged into structured data objects using the JSON lightweight data exchange format. The coordinate system transformation module unifies the target position coordinates to the robot's base coordinate system, involving homogeneous transformation matrix multiplication. Motion constraints are encoded as mathematical constraint expressions, which include inequality and equality constraints. The initial trajectory parameter generation module outputs a standardized data stream, which is transmitted to the intelligent optimization algorithm model via a network socket. The parameter verification stage checks whether the coordinate values ​​are within the robot's workspace; this workspace check is achieved through inverse kinematics feasibility calculation. The motion constraint rationality verification compares the user-input values ​​with the robot's physical limits; values ​​exceeding the physical limits are automatically adjusted to the allowable range.

[0032] The hidden layer design of the neural network model considers gradient flow characteristics, and residual connections are added between hidden layers to alleviate the gradient vanishing problem in deep networks. Batch normalization layers are inserted before each hidden layer to stabilize the distribution changes during training. The learning rate scheduler uses a cosine annealing strategy, periodically adjusting the learning rate during training. A weight decay term is added to the loss function to control model complexity and prevent overfitting. Training data augmentation techniques introduce random noise to improve model robustness. Model evaluation metrics include average trajectory error and maximum deviation; the average trajectory error is calculated as the average distance between the predicted and actual trajectories. The task description information acquisition system implements an authentication mechanism to ensure secure operation permissions. Task instruction version management records modification history and supports operation rollback. Target position coordinate input supports multiple format conversions, including Cartesian coordinates and joint angle coordinates. The graphical editing tool for motion constraints provides a visual interface, allowing users to drag sliders to adjust parameter values. The initial trajectory parameter data structure includes metadata fields, which record the creation timestamp and operator number. A parameter parsing error handling mechanism records exception logs and provides an error recovery process. The network communication layer implements encrypted data transmission using the TLS security protocol.

[0033] The deployment environment for the intelligent optimization algorithm model is configured with a model server, which provides a REST application programming interface. Inference requests include session identifiers, which are associated with the context of multiple calls. The input data preprocessing pipeline performs feature scaling, normalizing input values ​​to zero mean and unit variance. The model output post-processing program applies a constraint satisfaction algorithm, which corrects trajectory points that violate motion constraints. An inference result caching mechanism stores frequently used trajectory patterns, reducing the computational load on the model. The model performance monitoring dashboard displays response time metrics and alerts on abnormal states. The task description information acquisition process includes a manual review step, verifying the parameter rationality of high-risk tasks. The task instruction syntax definition adopts Backus paradigm, which describes the formal syntax rules of the instructions. The coordinate transformation chain includes the transformation from the tool coordinate system to the world coordinate system, accumulating the product of each transformation matrix. The motion constraint compiler generates intermediate representation code and optimizes the constraint evaluation order. The initial trajectory parameter verification suite runs unit tests, ensuring code quality. The data transmission protocol defines a binary message format, which improves transmission efficiency. The error code system categorizes various exception types, guiding fault diagnosis.

[0034] The hyperparameter tuning of the neural network model uses Bayesian optimization, which constructs a Gaussian process surrogate model. Activation function selection considers the phenomenon of dead neurons, comparing rectified linear units (CLUs) and exponential linear units (EXPs). Loss function weights are determined through Pareto optimization, balancing multiple optimization objectives. Training data sampling strategies address class imbalance by resampling minority class samples. Model ensemble techniques combine multiple neural network models, with voting determining the final output. Feature importance analysis identifies key input variables and guides feature engineering improvements. The task description information acquisition system implements a load balancing mechanism, distributing requests across multiple processing nodes. The instruction parser supports plug-in architecture extensions, allowing for the addition of new instruction syntax. A motion constraint database stores historical constraint templates, supporting fast retrieval. The initial trajectory parameter generation algorithm includes fault tolerance, skipping corrupted input data records. System configuration management maintains parameter configuration files, supporting environment-specific settings. An audit trail records every parameter modification operation, meeting quality management system requirements. Performance analysis tools dissect algorithm time consumption and identify computational bottlenecks.

[0035] Example 3: See Figure 3 The generation of the optimized motion trajectory sequence begins with the input of initial trajectory parameters. These parameters are transmitted to the intelligent optimization algorithm model via a data interface, which employs a high-speed serial communication protocol to ensure real-time data transmission. The intelligent optimization algorithm model loads pre-trained neural network parameters, which are read from non-volatile memory into the memory of the computing unit. Forward propagation computation proceeds layer by layer according to the network hierarchy. The initial trajectory parameters received at the input layer undergo linear transformation and activation function processing. The linear transformation is implemented using weight matrix multiplication, and the activation function is a sigmoid function to introduce non-linear characteristics. The computation of the hidden layers involves large-scale matrix operations, which are accelerated by the parallel computing core of the graphics processing unit. The output layer generates a list of path points, containing a sequence of three-dimensional spatial coordinates, with coordinate values ​​represented in floating-point format for precision.

[0036] The interpolation of the pathpoint list employs a cubic spline interpolation algorithm, which ensures the continuity of the second derivative of the curves between pathpoints. The spacing between interpolation nodes is adaptively determined based on motion accuracy requirements; node density is increased in high-curvature regions and decreased in flat regions. Velocity curve generation is based on time parameterization of the pathpoint list. This time parameterization is initialized using the assumption of uniform motion and then adjusted according to motion constraints. An optimization algorithm is used to ensure the velocity curve satisfies kinematic constraints. This algorithm iteratively adjusts the time allocation to ensure the acceleration profile does not exceed the limits of the mechanical system. The output format of the motion trajectory sequence includes the correspondence between timestamps and spatial coordinates, with timestamps marking the arrival time of each pathpoint with millisecond precision.

[0037] Path smoothness optimization is achieved through the following formula:

[0038] in: This represents the curvature value at the path point. Represents the velocity vector. Represents the acceleration vector. Curvature value. Used to evaluate the smoothness of the path; a smaller curvature value corresponds to a smoother motion trajectory. Velocity vector. The acceleration vector is calculated from the difference between path points. The derivation is based on the second difference of the velocity vector.

[0039] The control command generation phase begins with the conversion of motion trajectory sequences into joint space trajectories, which are obtained through inverse kinematics calculations. Inverse kinematics calculations combine analytical and numerical methods. Analytical methods address the unique geometric configuration of the robot, while numerical methods handle computational difficulties near singular positions. Target angle calculations for each joint consider the physical limitations of the mechanical structure, and joint angle limit checks prevent damage from overtravel. Velocity control signals are generated based on the pseudo-inverse calculation of the Jacobian matrix, which maps the operand space velocity to the joint space velocity. The control signal transmission period is synchronized with the robot controller's sampling period, typically set to the millisecond level. The joint actuator control algorithm employs a dual closed-loop structure of position and velocity loops. The position loop adjusts the joint angles to the target value, while the velocity loop stabilizes the joint motion. Motor drive signals are amplified by a power amplifier, which uses pulse width modulation (PWM) to control the motor current. Feedback data during actual motion is acquired in real-time by an encoder; the encoder data is compared with the target value to generate an error signal. Track tracking accuracy is ensured through a feedforward compensation algorithm, which predicts the system's dynamic characteristics and corrects the control input in advance. Vibration suppression during motion is achieved using input shaping technology, which filters out the frequency components that excite mechanical resonance.

[0040] The forward propagation computation of the intelligent optimization algorithm model includes a normalization step, which scales the input parameters to the numerical range used during model training. Dropout regularization in the hidden layers is disabled during inference to prevent overfitting. Post-processing of the pathpoint list applies curvature continuity constraints to ensure smooth, abrupt robot motion. Velocity curve optimization considers energy consumption metrics, with the integral acceleration squared term evaluating motion efficiency. The time reparameterization of the motion trajectory sequence uses a monotonic function, preserving the pathpoint order while optimizing time allocation. Singularity handling in inverse kinematics calculations employs damped least squares, which adds a damping factor to ensure numerical stability when the Jacobian matrix is ​​singular. Joint angle trajectory generation considers the motion coordination of each axis, achieved through a time synchronization mechanism. Speed ​​control signal limiting prevents exceeding driver capabilities; the limiting value is dynamically adjusted based on motor torque characteristics. Control command transmission uses a deterministic Ethernet protocol, ensuring deterministic communication latency. The self-tuning function of the drive parameters adapts to load changes, and the self-tuning algorithm identifies the inertia of the mechanical system online.

[0041] The optimization process for the path point list includes obstacle avoidance constraints, implemented using an artificial potential field method. Velocity curve generation considers jerk limitations to reduce impact on the mechanical system. Motion trajectory sequence verification is performed in a virtual controller environment that simulates the actual robot motion. Control command verification includes a logic check step, detecting the continuity of joint angle changes. Data recording during actual motion saves operating parameters for subsequent analysis and optimization. Post-processing of the intelligent optimization algorithm model's output includes a trajectory smoothing algorithm, which uses convolutional filtering to remove high-frequency noise. Time optimization of the path point list uses dynamic programming to find the optimal time allocation. Velocity curve constraints are solved using quadratic programming, transforming constraints into optimization objectives. Compressed storage of the motion trajectory sequence uses differential encoding to reduce storage space requirements. Control command generation efficiency is improved through code optimization, which utilizes compilation techniques to increase execution speed.

[0042] Real-time inverse kinematics calculations are ensured through a lookup table method, which pre-calculates joint angles at common locations. Fifth-order polynomials are used for joint spatial trajectory interpolation, ensuring acceleration continuity. Torque limits are considered in the generation of speed control signals, determined based on a motor heating model. Reliability of driver communication is guaranteed through redundancy checks, which detect transmission errors. A system safety monitoring module detects abnormal motion states and triggers an emergency stop signal. The optimization process for the motion trajectory sequence involves multi-objective trade-offs, simultaneously considering both shortest time and minimum energy. The quality assessment of the path point list uses multiple metrics, including path length and curvature extrema. The speed curve optimization algorithm employs gradient descent, iteratively improving the speed distribution. Control command verification is performed on a hardware-in-the-loop simulation platform connected to the actual controller. Motion performance analysis includes vibration spectrum analysis, which identifies resonant frequency points.

[0043] The inference process of the intelligent optimization algorithm model utilizes batch processing technology, simultaneously optimizing multiple trajectory segments. The generation of the path point list considers the workspace boundary, which is constrained using a penalty function method. Velocity curve smoothing employs the phase plane method, optimizing the velocity-acceleration relationship. Visualization of the motion trajectory sequence aids in debugging, utilizing a 3D graphical interface. Control command issuance employs a double-buffering mechanism to avoid command update conflicts. System real-time performance is guaranteed through interrupt service routines, which respond to timer interrupts. The accuracy of inverse kinematics calculations is improved through iterative methods, gradually approximating the exact solution. Joint trajectory generation includes transition segment design, smoothly connecting different trajectory segments. Velocity control signal filtering eliminates high-frequency noise using digital filters. Online adjustment of actuator parameters adapts to mechanical wear, based on model reference adaptive control.

[0044] Example 4: Monitoring the actual motion trajectory of the industrial robot is achieved through a multi-sensor data fusion system, which integrates joint encoders, laser trackers, and a vision measurement unit. The joint encoders are installed at the rear of the servo motors of the robot's six joints. These encoders are multi-turn absolute encoders with a resolution of 24 bits. The laser tracker is positioned at the corner of the robot's working unit, measuring the spatial position of the tool's center point using the interferometer principle. The vision measurement unit consists of a high-speed camera and marker points, which are attached to the surface of the robot's end effector. The sensor data acquisition frequency is uniformly set to 1000 Hz, and the acquisition card uses a synchronization signal to ensure time alignment. Actual joint angle data is read from the encoder's digital interface, and the tool's center point coordinates are calculated using a coordinate transformation chain. The comparison between the actual motion trajectory and the optimized motion trajectory sequence is performed in the trajectory tracking error calculation module, which calculates the position and orientation errors in real time. The position error is defined as the Euclidean distance between the actual tool center point coordinates and the target coordinates, and the orientation error is calculated using quaternion angles. The allowable range of trajectory tracking error is set according to process requirements: 0.1 mm for precision assembly tasks and 1 mm for welding tasks. The error comparison algorithm runs on a real-time operating system, with the calculation cycle synchronized with the control cycle. Error data is stored in a circular buffer, with the buffer size accommodating motion data from the most recent 10 seconds. Error exceeding limits is detected using a sliding window statistical method, with the window size set to 100 sampling points.

[0045] The speed control signal adjustment is based on a proportional-integral-derivative (PID) control algorithm, whose parameters are tuned according to the robot's dynamics model. The proportional coefficient handles the current error, the integral coefficient accumulates historical errors, and the derivative coefficient predicts the error trend. The speed correction is calculated using an incremental formula to avoid integral saturation. A correction limiter prevents excessive adjustment, with the limit threshold set based on the motor's torque capacity. The adjusted speed control signal is transmitted via a fieldbus using the EtherCAT protocol to ensure real-time performance. The signal update cycle is consistent with the controller cycle, typically 1 millisecond. The monitoring data recording system saves the complete motion process, including timestamps, actual coordinates, target coordinates, and error values. Recorded files are stored in binary format to reduce storage space usage. The data playback function supports motion process reproduction and is used for offline analysis and debugging. The system status monitoring interface displays a real-time error curve, with color coding distinguishing between normal and over-limit states. An alarm function triggers when limits are continuously exceeded, alerting the operator with audible and visual signals.

[0046] The calculation of trajectory tracking error considers coordinate system transformation error, which is compensated for through hand-eye calibration. Timestamp alignment of sensor data uses a hardware trigger signal, distributed by the main controller. Periodic error statistics reports are generated, including maximum error, average error, and standard deviation. The error compensation algorithm adapts to different motion stages; the differential coefficient is increased at high speeds, and the integral action is enhanced at low speeds (see Table 1).

[0047] Table 1: Classification and Handling Strategies of Tracking Errors

[0048] The speed control signal adjustment process includes a feedforward compensation stage, which calculates the ideal speed based on a kinematic model. A friction compensation module counteracts Coulomb friction and viscous friction; friction parameters are obtained experimentally. Inertial compensation considers load variations, and inertial compensation parameters are adjusted based on mass properties. A resonance suppression filter eliminates mechanical vibration; this filter employs a notch filter design. Temperature compensation corrects thermal deformation errors; a temperature sensor monitors motor temperature rise. Backlash compensation eliminates transmission backlash; backlash compensation values ​​are calibrated using laser measurement. The actual motion trajectory monitoring system includes data validity checks to identify sensor malfunctions. Abnormal data filtering uses a median filtering algorithm; the median filtering window size is configurable. The data fusion algorithm uses Kalman filtering to estimate the optimal trajectory state. The coordinate system is aligned to the robot's base coordinate system; coordinate transformation is achieved using a homogeneous matrix. Time synchronization accuracy reaches the microsecond level; time synchronization uses the PTP precision clock protocol.

[0049] The analysis of trajectory tracking errors distinguishes between systematic and random errors. Systematic errors are eliminated through parameter calibration, while random errors are suppressed through filtering algorithms. Error trend prediction uses time series analysis to identify periodic errors. Error compensation calculation considers the robot configuration and the differences in error characteristics under different configurations. An error mapping model establishes the relationship between error and joint angles; this model is trained using a neural network. An error compensation table stores compensation values ​​for common positions; this table is established through laser tracker measurements. The speed control signal adjustment algorithm includes an adaptive mechanism that adjusts parameters based on error changes. A fuzzy logic controller handles nonlinear characteristics; its rule base includes expert experience. A neural network controller learns the optimal adjustment strategy and adjusts control parameters online. A genetic algorithm optimizes controller parameters by searching for the optimal parameter combination. A reinforcement learning algorithm autonomously improves the control strategy, guided by a reward function.

[0050] The real-time performance of the monitoring system is verified through benchmark testing, which measures the latency from data acquisition to control output. System resource monitoring tracks CPU utilization, ensuring real-time performance. Data integrity checks verify the transmission process using cyclic redundancy check (CR). System logs record the operating status and are used for fault diagnosis. Performance counters track error processing time and evaluate algorithm efficiency. Actual motion trajectory data is used for online calibration, updating robot kinematic parameters. The tool center point calibration program is executed periodically using the three-point method. Workpiece coordinate system calibration compensates for positioning errors using feature matching. Kinematic parameter identification uses the least squares method, improving model accuracy. Dynamic parameter identification improves control performance through excitation trajectories. The trajectory tracking error handling process includes multiple safety levels: the first level detects errors, the second level implements adjustment strategies, and the third level triggers safety protection. Error thresholds are set in stages to avoid false alarms. Adjustment strategies are progressively strengthened to ensure stability. The safety protection mechanism takes effect immediately to prevent equipment damage. The system uses a state machine to manage the operating modes, which include normal, adjustment, and protection modes.

[0051] Example 5: After the industrial robot completes the car door welding task, the actual motion trajectory data recording system starts running, reading joint angle sampling values ​​from the controller's memory buffer. Encoder data for each joint is recorded at a frequency of 1000 Hz, including motor rotation angle and torque current values. The tool center point coordinates are calculated using a hand-eye vision system that captures the reflective markers on the welding torch tip. A performance index calculation module runs synchronously, calculating the total welding path time, average speed, and root mean square acceleration. Data timestamps are synchronized with a factory-level network time server, which uses GPS signals to ensure global clock consistency. Recorded files are stored in a segmented format, naturally segmented according to the weld seams. The process of adding actual motion trajectory data to the trajectory database includes a data cleaning step, removing outliers caused by sensor jitter. The trajectory database's table structure includes a task number field, a timestamp field, a joint angle array field, and a tool center point coordinate field. New data records are inserted through database transaction operations, which ensure data integrity constraints. Index update operations optimize query performance; index updates target task type and time range fields. Data archiving strategies retain recent data in high-speed storage while migrating historical data to near-line storage. Database consistency checks validate foreign key relationships and ensure task numbers match those in the main table.

[0052] The training set update of the intelligent optimization algorithm model adopts an incremental learning mode, which retains the original model parameters while incorporating new data. The training set is constructed by extracting the most recent 100 task data from the trajectory database, including successful welding cases and manually corrected trajectories. The feature extraction process recalculates the curvature features of the path point sequence, calculating local curvature values ​​using the three-point method. Velocity distribution features are standardized to the robot's maximum rated speed. Training sample labels are generated by expert system scoring based on welding quality inspection results. A data balancing algorithm addresses class imbalance by repeatedly sampling minority class samples. The learning rate for retraining the intelligent optimization algorithm model uses an annealing strategy, with an initial value of 0.001, decreasing by 10% every 1000 iterations. The training loss function combines trajectory smoothness and time optimality; the trajectory smoothness term calculates the integral of acceleration changes, while the time optimality term evaluates the total motion duration. The gradient clipping threshold is dynamically adjusted, adaptively changing based on the loss function value. The validation set is partitioned chronologically to ensure the model validates future data. Early stopping is used to monitor the validation set loss curve, terminating training after 200 consecutive iterations without improvement. Model checkpoints are saved periodically, including weight blocks and optimizer status. Updated model parameters are deployed using a blue-green deployment strategy, which uses the old model version as a fallback. Parameters are transmitted securely using AES-256 encryption. A version control system records parameter change history, marking the data source and performance metrics for each model version. The model inference service hot-reloads with new parameters without interrupting ongoing welding tasks. A health check interface verifies model availability by sending test data to verify input / output formats.

[0053] The actual motion trajectory data record includes environmental parameters, such as workshop temperature and humidity, and workpiece positioning error. Welding quality inspection results are correlated with the trajectory data; weld quality is assessed using a visual inspection system. A data time alignment algorithm compensates for transmission delays, using interpolation to synchronize data sources with different sampling rates. Anomaly markers record welding spatter events, helping to filter high-quality training data. A data compression algorithm reduces storage space, employing lossy compression to preserve key features. The trajectory database query interface supports complex condition filtering, allowing queries by workpiece type and welding parameters. The data backup strategy implements the 3-2-1 principle, including local backup, off-site backup, and cloud backup. Data anonymization protects production process secrets, hiding key dimensional parameters. Access control restricts data usage, assigning query permissions based on roles. Data lifecycle management automatically cleans up expired data, deleting historical data according to the retention policy.

[0054] The training process of the intelligent optimization algorithm model utilizes a distributed computing framework, which trains multiple model variants in parallel on a GPU cluster. Hyperparameter optimization employs Bayesian optimization, exploring the optimal combination of learning rate and batch size. Training data augmentation techniques apply random perturbations, increasing sample diversity by adding Gaussian noise. Model structure optimization experiments with different network depths, comparing the performance of three-layer and five-layer neural networks. Regularization techniques prevent overfitting, including L2 weight decay and dropout. Training curve visualization monitors the convergence process, displaying changes in the loss function and evaluation metrics. Post-deployment validation testing includes regression testing, using historical data to verify model output consistency. Stress testing evaluates model inference speed, simulating high-concurrency request scenarios. Boundary testing checks the handling of extreme inputs, verifying robustness by inputting abnormal trajectory parameters. Integration testing validates collaboration with the controller, running a complete welding process in a simulation environment. Performance benchmarks compare version differences, recording inference latency and memory usage metrics. Data mining of actual motion trajectory data reveals potential optimization patterns, and data mining algorithms cluster similar motion trajectories. Association rule analysis identifies parameter combination patterns, revealing the correlation between speed parameters and weld quality. Trend analysis predicts equipment performance degradation, establishing a model of accuracy changes over time. Anomaly detection identifies abnormal motion patterns, marking deviations from the norm in trajectory segments. Statistical analysis generates quality reports, calculating statistics for various indicators.

[0055] The update cycle of the intelligent optimization algorithm model is determined by performance monitoring, which tracks the model's performance on the production line. Model decay detection monitors prediction deviations and statistically analyzes the differences between the actual and predicted trajectories. Trigger conditions set performance thresholds, initiating updates when evaluation metrics drop by a certain percentage. A rollback mechanism ensures system reliability, restoring the old version when model updates fail. Version compatibility handles data format changes, maintaining consistency between the old and new model interfaces. Update notifications inform relevant systems via message queues broadcasting version changes. Long-term accumulation of actual motion trajectory data forms a knowledge base containing optimized trajectory patterns for different workpiece models. Transfer learning technology is applied across tasks, transferring welding knowledge to cutting tasks. Meta-learning algorithms extract commonalities in optimization strategies, learning trajectory optimization methods from multiple tasks. A knowledge graph constructs relationships between process parameters, representing constraints between them. Intelligent retrieval supports similar case queries, recommending historical trajectories based on new task characteristics. Continuous learning continuously improves performance, forming a closed loop from data to optimization.

[0056] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0057] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A control method for industrial robot trajectory optimization based on intelligent algorithms, characterized in that, Includes the following steps: Collect motion trajectory data generated by industrial robots when performing various work tasks. The motion trajectory data includes joint position information, tool center point coordinates and motion time series, and store the motion trajectory data in the trajectory database. The motion trajectory data in the trajectory database is cleaned and filtered to extract the feature set of the motion trajectory data, which includes path point sequence, velocity distribution and acceleration profile. An intelligent optimization algorithm model is constructed. The intelligent optimization algorithm model adopts a neural network structure and is trained using a feature set, so that the intelligent optimization algorithm model learns trajectory optimization strategies. Obtain the task description information of the current industrial robot, which includes the target position coordinates and motion constraints. Parse the task description information to obtain the initial trajectory parameters. The initial trajectory parameters are input into the trained intelligent optimization algorithm model, and the intelligent optimization algorithm model outputs the optimized motion trajectory sequence, which includes a list of path points and a velocity curve. Control commands are generated based on the optimized motion trajectory sequence, and the controller of the industrial robot drives the motors of each joint, so that the industrial robot moves according to the optimized motion trajectory sequence.

2. The control method for industrial robot trajectory optimization based on intelligent algorithms as described in claim 1, characterized in that, The steps for collecting motion trajectory data generated by industrial robots when performing various work tasks include: Joint angle data is collected by encoder sensors installed at the joints of industrial robots; Capture the spatial coordinate data of the tool's center point using a vision system; Record the joint angle data and tool center point coordinate data corresponding to each motion timestamp; Associate the joint angle data, tool center point coordinate data and motion timestamp data with the task identifier, which includes the task type and workpiece number. The associated motion trajectory data is stored in the data table of the trajectory database.

3. The control method for industrial robot trajectory optimization based on intelligent algorithms as described in claim 2, characterized in that, The steps for cleaning and filtering motion trajectory data in the trajectory database include: Read motion trajectory data from the data table in the trajectory database; Check the integrity and consistency of motion trajectory data, and remove records containing missing or outlier values; Motion trajectory data is filtered according to a preset accuracy threshold, and records with path errors less than the threshold are retained; Extract path point sequences from the filtered records. The path point sequences are composed of the coordinates of the tool's center point in chronological order. Calculate the velocity and acceleration values ​​between the path point sequences to form the velocity distribution and acceleration profile.

4. The control method for industrial robot trajectory optimization based on intelligent algorithms as described in claim 3, characterized in that, The steps for constructing the intelligent optimization algorithm model include: Initialize the layer structure and parameters of the neural network model, which includes an input layer, hidden layers, and an output layer; The path point sequence, velocity distribution, and acceleration profile in the feature set are used as training input data. Set the optimization objective function, which includes minimizing the path and maximizing the smoothness; Use the gradient descent algorithm to adjust the parameters of the neural network model and minimize the optimization objective function; Verify the performance of the trained intelligent optimization algorithm model on the test dataset and save the model parameters.

5. The control method for industrial robot trajectory optimization based on intelligent algorithms as described in claim 4, characterized in that, The steps for obtaining the current task description information of the industrial robot include: Receive task instructions from user input or the host system; Parse the target location coordinates in the task instructions; the target location coordinates include three-dimensional spatial coordinates. Extract the motion constraints from the task instructions. The motion constraints include the maximum speed limit, acceleration limit, and obstacle avoidance area. The target position coordinates and motion constraints are combined to form the initial trajectory parameters.

6. The control method for industrial robot trajectory optimization based on intelligent algorithms as described in claim 5, characterized in that, The steps for generating the optimized motion trajectory sequence include: Input the initial trajectory parameters into the trained intelligent optimization algorithm model; The intelligent optimization algorithm model calculates the list of path points through forward propagation; Generate a continuous velocity curve by interpolation based on the list of path points; Adjust the velocity curve to meet the motion constraints; Output the optimized motion trajectory sequence, including a list of path points and velocity curves.

7. The control method for industrial robot trajectory optimization based on intelligent algorithms as described in claim 6, characterized in that, The step of generating control commands includes: The optimized motion trajectory sequence is converted into joint space trajectory, and the target angle of each joint is calculated by inverse kinematics. Generate speed control signals for each joint based on the speed curve; Send the target angle and speed control signals to the joint actuators of the industrial robot; Joint actuators control the rotation of motors, enabling the movement of industrial robots.

8. The control method for industrial robot trajectory optimization based on intelligent algorithms as described in claim 7, characterized in that, It also includes the following steps: Real-time monitoring of the actual movement trajectory of industrial robots, and acquisition of actual joint angles and tool center point coordinates through sensors; Compare the actual motion trajectory with the optimized motion trajectory sequence, and calculate the trajectory tracking error; If the trajectory tracking error exceeds the allowable range, adjust the speed control signal in the control command; Repeat the monitoring and adjustment process until the task is completed.

9. The control method for industrial robot trajectory optimization based on intelligent algorithms as described in claim 8, characterized in that, The step of adjusting the speed control signal in the control command includes: Calculate the speed correction amount based on the magnitude and direction of the trajectory tracking error; The speed correction amount is superimposed on the original speed control signal; The adjusted speed control signal is limited to not exceeding the maximum speed limit in the motion constraints. The adjusted speed control signal is sent to the joint actuator.

10. The control method for industrial robot trajectory optimization based on intelligent algorithms as described in claim 9, characterized in that, It also includes the following steps: After the task is completed, record the actual motion trajectory data and its performance indicators; Add the actual motion trajectory data to the trajectory database; Update the training set of the intelligent optimization algorithm model with new data and retrain the model; Save the updated model parameters for subsequent tasks.

Images