Collaborative robot real-time track acquisition and high-precision restoration method

Abstract

The invention discloses a real-time trajectory acquisition and high-precision restoration system and method for a collaborative robot, and the method comprises the steps: collecting robot joint data in real time, and carrying out the preprocessing of the collected data through an improved Ju-Gibbs quaternion; filtering the preprocessed data by using a dynamic adaptive sliding filter according to the change rate and the change amplitude of the real-time data; track reconstruction is carried out according to the filtered data, robot postures in an initial state and an intermediate state are reversely deduced from a final quaternion state by using Ju-Gibbs quaternion inverse operation, and a motion track of the robot is reconstructed; and comparing the reconstructed trajectory with the actual motion trajectory of the robot, calculating an error index to evaluate the accuracy and efficiency of the reconstructed trajectory, and adjusting and optimizing quaternion parameters and a filtering algorithm according to an evaluation result. According to the method, the high-precision track capture and playback capability of the robot in a complex environment is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of robot trajectory collection and restoration, and in particular relates to a method for real-time trajectory collection and high-precision restoration of a collaborative robot. Background Art

[0002] In the field of modern industrial automation, robotics has become one of the key factors in improving production efficiency and ensuring operational safety. The emergence of collaborative robots (Cobots) has greatly promoted the possibility of human-machine collaboration. Collaborative robots can work with human workers without the need for safety fences, thereby significantly improving production flexibility and efficiency. In terms of robot control and path planning, real-time trajectory acquisition and high-precision restoration are the core technologies to ensure that robots can operate efficiently and safely in complex environments. These technologies are not only widely used in the manufacturing industry, but also show great potential in many fields such as medical care, services and entertainment.

[0003] At present, robot trajectory acquisition and restoration technology has made great progress. Typical trajectory acquisition methods include sensor-based data acquisition and vision-based trajectory tracking. The collected trajectory data usually needs to be processed and filtered to reduce the impact of noise on path planning. Existing trajectory restoration technology mainly relies on path planning algorithms, such as Rapidly-exploring Random Tree (RRT) and Artificial Potential Field (APF). These algorithms perform well when dealing with static obstacles, but there are still some shortcomings when facing dynamic obstacles and complex working environments.

[0004] For example, although sensor-based trajectory acquisition methods such as Lidar and depth cameras can provide high-precision location information, there are certain challenges in data processing and real-time performance. On the other hand, vision-based trajectory tracking technologies, such as RGB-D cameras and monocular vision technology, have good environmental adaptability, but their accuracy and stability in complex scenes still need to be improved. The defects and shortcomings of existing technologies mainly include the following two points:

[0005] 1. The trajectory collection accuracy is insufficient.

[0006] Existing trajectory collection methods often have difficulty in balancing collection accuracy and real-time performance when facing complex industrial environments and diverse operating tasks. Especially in high-noise environments, the collected trajectory data is easily disturbed, resulting in reduced path planning accuracy. These problems are mainly reflected in the following aspects:

[0007] Insufficient fast dynamic response: During high-speed motion and complex task execution, traditional posture and position calculation methods are difficult to accurately respond to rapid changes in the robot's state in real time. This lack of response may cause the trajectory execution to deviate from the preset trajectory, affecting the execution quality of the task.

[0008] Data processing lag: When processing large amounts of data generated at high speed, traditional methods are often unable to update the robot's status information in real time due to limited processing capabilities, resulting in data processing lags, which in turn affects the accuracy and efficiency of the overall operation.

[0009] Data noise: When collecting trajectory data, existing sensor technology will be affected by environmental noise, resulting in more noise in the collected data. These noises will cause trajectory deviations in subsequent path planning and affect the robot's operating accuracy.

[0010] 2. Data processing efficiency is low.

[0011] Existing trajectory capture systems usually need to process a large amount of sensor input data, including position, velocity, acceleration and other information. In high-speed robot operations, sensors generate hundreds or even thousands of data points per second. Due to the lack of effective data screening and preprocessing mechanisms, the system often needs to process a large amount of redundant data, which not only increases the consumption of computing resources, but also prolongs the data processing time, thereby reducing the efficiency of the entire system.

[0012] In a dynamically changing environment, especially on an industrial production line, environmental factors (such as vibration, temperature changes, interference between machines, etc.) may change rapidly, which requires the trajectory capture system to be able to process and respond to these changes in real time. The data processing algorithms in the existing technology usually adopt a batch processing method and lack the ability to update and adjust in real time, which is particularly insufficient in application scenarios that require high real-time performance.

[0013] Traditional filtering techniques, such as simple moving average filters or Kalman filters, usually have fixed parameter settings (such as window size or noise estimation), which makes them unable to adapt effectively when faced with rapid changes in data characteristics. When environmental conditions or robot operation modes change, these filters may not accurately reflect these changes, resulting in inaccurate data processing results and increased errors in subsequent trajectory reconstruction.

[0014] 3. The calculation stability is poor.

[0015] Universal lock problem: When using Euler angles as a robot posture description, when the rotation angle is close to 90 degrees, the universal lock problem will occur, causing the posture calculation to lose a degree of freedom, which in turn causes instability in the kinematic calculation. This is unacceptable in precision robot operations because it will lead to incorrect trajectory execution.

[0016] Quaternion normalization problem: When using traditional quaternions for attitude calculation, due to the cumulative error of the quaternion during the calculation process, normalization needs to be performed regularly to prevent the error from increasing. However, in high-speed dynamic operations, frequent normalization operations increase the computational burden and may introduce additional computational errors. Summary of the invention

[0017] Purpose of the invention: In view of the above problems, the present invention proposes a method for real-time trajectory acquisition and high-precision restoration of collaborative robots. The method improves the trajectory acquisition accuracy and trajectory restoration efficiency of collaborative robots in complex industrial environments, and ensures that the robots can operate with high precision and high stability in dynamic obstacle environments.

[0018] Technical solution: To achieve the purpose of the present invention, the technical solution adopted by the present invention is:

[0019] A method for real-time trajectory acquisition and high-precision restoration of a collaborative robot comprises the following steps:

[0020] Step 1: Collect the position and velocity data of the robot joints in real time, including angle, velocity and acceleration;

[0021] Step 2: Use the improved Ju-Gibbs quaternion to preprocess the collected raw data to obtain the preprocessed position and velocity data of the robot joints;

[0022] Step 3: Apply a dynamic adaptive sliding filter to filter the preprocessed data according to the change rate and change amplitude of the real-time data;

[0023] Step 4: Reconstruct the trajectory based on the filtered data. Use the inverse operation of the Ju-Gibbs quaternion to infer the robot postures of the initial state and the intermediate state from the final quaternion state, and reconstruct the robot's motion trajectory.

[0024] Step 5: Compare the reconstructed trajectory with the actual robot motion trajectory, calculate the error index to evaluate the accuracy and efficiency of the reconstructed trajectory, and adjust and optimize the quaternion parameters and filtering algorithm based on the evaluation results.

[0025] Furthermore, the collected raw data is preprocessed using the improved Ju-Gibbs quaternion as follows:

[0026] First, obtain the connection data θ from link l-1 to link l l and n l , where θ l is the angular position about the axis, n l is the axis vector; the expression of the calculated Ju-Gibbs quaternion is:

[0027]

[0028] in, is the Ju-Gibbs quaternion describing the lth link relative to the global reference coordinate system o, o q l represents the unit quaternion of the lth link relative to the reference coordinate system o, where o is the symbol representing the reference coordinate system. n l ,τ l is a vector, is a scalar;

[0029] During the robot joint rotation, the Ju-Gibbs quaternion is written in general form as:

[0030]

[0031] in, is the Ju-Gibbs quaternion describing the attitude from the reference coordinate system o to the l-1th link, is the Ju-Gibbs quaternion describing the relative attitude transformation from the l-1th link to the lth link, is the square of the modulus of the Ju-Gibbs quaternion;

[0032] The Ju-Gibbs quaternion combines the rotation axis and rotation angle of each joint to form a unified mathematical representation to describe the rotation of each joint; the rotation of each joint is continuously combined through quaternion multiplication to express the posture changes of the entire robotic arm.

[0033] Furthermore, a dynamic adaptive sliding filter is applied to filter the preprocessed data as follows:

[0034] First, the difference between each data point in the data stream and the previous data point is calculated to obtain the rate of change and magnitude of change of the data; the data stream refers to the data sequence continuously collected by each sensor of the robot, and the data point includes position, speed or acceleration information;

[0035] Then, according to the rate and magnitude of change, the size of the filter window and the weight coefficient are dynamically adjusted, and the weight coefficient is adjusted according to the time interval between the data point and the current processing point;

[0036] The weight is calculated using the exponential decay model, the formula is w i =e -α*# , where w i is the weight coefficient assigned by the filter to each data point, d is the number of time steps from the data point to the current point, and α is the attenuation coefficient, which controls the speed at which the weight decreases;

[0037] If both the data change rate and the change amplitude exceed the set threshold, the filter window is reduced and the α value is increased; conversely, if both the change rate and the change amplitude do not exceed the set threshold, the filter window is expanded and the α value is reduced.

[0038] Furthermore, the trajectory is reconstructed based on the filtered data as follows:

[0039] After the robot completes a series of actions, the data collected and processed by the sensor system in real time is converted into the final quaternion state. The inverse operation of the Ju-Gibbs quaternion formula is used to infer the robot postures of the initial state and the intermediate state through the final quaternion state, and reproduce the robot's motion trajectory; the mathematical model describing the transition of the robot from one posture to another is as follows:

[0040]

[0041]

[0042] in, is the Ju-Gibbs quaternion of the l-1th joint, representing the pose transformation from the reference coordinate system o to the corrected state d, is the rotation axis vector of the l-1th joint, is the scalar part of the rotation angle of the l-1th joint, d is the correction or target state, and o is the reference coordinate system; is the symmetric rotation axis matrix in the modified state d, is the rotation axis vector of the lth joint, is the symmetric matrix representation of the rotation axis of the lth joint, n l is the rotation axis vector of the lth joint.

[0043] Beneficial effects: Compared with the prior art, the technical solution of the present invention has the following beneficial technical effects:

[0044] The prior art usually uses standard quaternions to process static or slowly changing scenes and describe and calculate the posture of robot joints, but it is easy to cause calculation delay or precision loss in fast dynamic environments. The present invention uses improved Ju-Gibbs quaternions and adds a dynamic adjustment mechanism, so that the quaternions can be adaptively adjusted according to real-time dynamic data, thereby improving the calculation stability and accuracy under fast changing conditions.

[0045] Traditional sliding average filters perform poorly when processing high-frequency data changes because their parameters are fixed and cannot adapt to changes in environmental noise and data fluctuations. The dynamic adaptive sliding filter (DASF) designed in the present invention automatically adjusts the filtering parameters, such as window size and weight factor, according to the characteristics of the real-time data stream, and can adapt to different data fluctuations, more effectively eliminate noise and reduce data redundancy, and improve the accuracy of data processing.

[0046] The method of the invention not only optimizes the data processing flow, but also provides a new solution for the dynamic control of robots, significantly improves the ability of robots to capture and replay high-precision trajectories in complex environments, and has broad industrial application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a flow chart of the method for real-time trajectory collection and high-precision restoration of a collaborative robot of the present invention.

[0048] Figure 2 It is a structural schematic diagram of the collaborative robot real-time trajectory acquisition and high-precision restoration system of the present invention.

[0049] Figure 3 It is the iiwa robot simulation operation.

[0050] Figure 4 This is the data collection process of the real iiwa robot in motion.

[0051] Figure 5 This is the experimental comparison before and after the filter function is introduced at position X.

[0052] Figure 6 This is the experimental comparison before and after the filter function is introduced at position Y.

[0053] Figure 7 This is the experimental comparison before and after the filtering function is introduced at position Z. DETAILED DESCRIPTION

[0054] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and examples.

[0055] The method for real-time trajectory acquisition and high-precision restoration of a collaborative robot described in the present invention has the following process: Figure 1 As shown, the specific implementation steps are as follows:

[0056] Step 1: Data collection: High-precision sensors are used to collect the position and velocity data of the robot joints in real time, including parameters such as angle, velocity, and acceleration, providing raw input for subsequent data processing.

[0057] Step 2: Data preprocessing: The collected raw data is preprocessed using the improved Ju-Gibbs quaternion, mainly for data standardization, laying the foundation for ensuring the accuracy and efficiency of subsequent processing.

[0058] First, get the original data θ l and n l , where θ l is the angular position about the axis, n l are axis vectors, representing the connection from link l-1 to link l. The Ju-Gibbs quaternion can be calculated as:

[0059]

[0060] in, is the Ju-Gibbs quaternion describing the lth link relative to the global reference coordinate system o, o q l represents the unit quaternion of the lth link relative to the reference coordinate system o, where o is the symbol representing the reference coordinate system. n l ,τ l is a vector, is a scalar.

[0061] During the robot joint rotation, the Ju-Gibbs quaternion is written in general form as:

[0062]

[0063] Thus, each joint rotation is obtained in the form of Ju-Gibbs quaternion. is the Ju-Gibbs quaternion describing the attitude from the reference coordinate system o to the l-1th link, is the Ju-Gibbs quaternion describing the relative attitude transformation from the l-1th link to the lth link, is the square of the modulus of the Ju-Gibbs quaternion.

[0064] The Ju-Gibbs quaternion combines the rotation axis and rotation angle of each joint to form a unified mathematical representation to describe the rotation of each joint; the rotation of each joint is continuously combined through quaternion multiplication to express the posture changes of the entire robotic arm.

[0065] Step 3: Data filtering: Dynamic adaptive sliding filter (DASF) is used to further filter the pre-processed data.

[0066] First, the filter automatically adjusts the filter parameters according to the change of data. The specific steps are to first calculate the difference between each data point in the data stream and the previous data point to evaluate the rate and magnitude of change of the data. The data stream refers to the data sequence continuously collected by the robot's sensors, and the data points include position, velocity or acceleration information.

[0067] Then, based on the real-time calculation results of the rate and magnitude of change, the DASF algorithm dynamically adjusts the size of the filter window and the weight coefficient. If the data changes greatly, the window may be reduced and the weight distribution will be more inclined to the nearest data point to quickly respond to the data change; conversely, if the data changes slowly, the window will be expanded and the weight distribution will be more even to smooth out the noise and avoid over-response.

[0068] The weight coefficient is adjusted according to the time interval between the data point and the current processing point. The weight is calculated using an exponential decay model. The weight of recent data points is higher. The formula is w i =e -α*# , where w i is the weight coefficient assigned by the filter to each data point, d is the number of time steps from the data point to the current point, and α is the attenuation coefficient, which controls the speed at which the weight decreases.

[0069] If both the data change rate and magnitude exceed the set thresholds, the filter window will be reduced from the standard size to 50% to enhance the response to new data, and the weight of recent data points will be increased by increasing the α value, so that the weight coefficient is more concentrated on recent data.

[0070] On the contrary, if neither the rate of change nor the magnitude of change exceeds the set threshold, the filter window is expanded to 150% to increase stability and reduce the impact of noise, and the α value is reduced to make the weights more evenly distributed on all data points.

[0071] Step 4: Trajectory reconstruction and playback. Trajectory reconstruction is performed based on the filtered and optimized data, and these data are converted back into motion instructions or position states of the robot joints. In the robot control system, these reconstructed motion instructions are input into the robot's control unit, and the robot's execution system is used to simulate physical motion, thereby reproducing the original or designed motion trajectory.

[0072] After the robot completes a series of actions, the data collected and processed by the sensor system in real time is converted into the final quaternion state. The inverse operation of the Ju-Gibbs quaternion is used to infer the robot posture of the initial state or any intermediate state from the final quaternion state. This calculation involves the composite operation and inverse operation of the quaternion, which will accurately describe the transition of the robot from one posture to another through the following mathematical model:

[0073]

[0074] Therefore, we can get

[0075]

[0076] as well as

[0077]

[0078] in, is the Ju-Gibbs quaternion of the l-1th joint, representing the pose transformation from the reference coordinate system o to the corrected state d, is the rotation axis vector of the l-1th joint, is the scalar part of the rotation angle of the l-1th joint, d is the correction or target state, and o is the reference coordinate system; is the symmetric rotation axis matrix in the modified state d, is the rotation axis vector of the lth joint, is the symmetric matrix representation of the rotation axis of the lth joint, n l is the rotation axis vector of the lth joint.

[0079] Step 5: Performance evaluation and optimization. Finally, the system will evaluate the accuracy and efficiency of trajectory reconstruction based on feedback data from actual applications. According to the evaluation results, necessary adjustments and optimizations are made to the quaternion parameters and filtering algorithms to ensure the best performance of the system in a constantly changing actual environment.

[0080] Specifically, the system compares the reconstructed trajectory with the actual robot motion trajectory and evaluates the accuracy of the reconstructed trajectory by calculating the error indicators, namely the mean square error and the maximum deviation. The system collects feedback data on trajectory execution, including the robot's response time, execution efficiency, and abnormal conditions during operation, for the operator to analyze, process and optimize.

[0081] The present invention also provides a collaborative robot real-time trajectory acquisition and high-precision restoration system, which aims to provide an efficient and accurate robot motion trajectory processing solution by integrating the improved Ju-Gibbs quaternion and the dynamic adaptive sliding filter (DASF). The system includes five main modules, such as Figure 2 As shown, it includes a sensor module, a data processing unit, a control execution unit, a user interface and a communication module, which cooperate with each other through clear data streams and control commands to achieve accurate capture and reproduction of robot motion.

[0082] The sensor module is responsible for real-time acquisition of motion data of the robot's joints and actuators, including but not limited to parameters such as position, velocity, and acceleration. These raw data are sent to the data processing unit, which uses the improved Ju-Gibbs quaternion to perform preliminary data standardization and noise filtering to improve data availability and accuracy. The data processing unit is also equipped with an innovative dynamic adaptive sliding filter (DASF), which can automatically adjust its processing parameters such as window size and filter weight according to changes in real-time data, thereby optimizing data processing effects and reducing errors and delays.

[0083] The processed data is transmitted to the control execution unit, which accurately controls the robot's movements according to the data instructions to achieve high-precision trajectory reproduction. The user interface provides an intuitive platform for operators to enter operating instructions, set system parameters, and monitor the robot's status and system operation in real time. In addition, the communication module supports data exchange with other industrial control systems, realizes external synchronization and remote control of data, and enhances the flexibility and application scope of the system.

[0084] The high-precision trajectory capture and playback system of the present invention has innovative applications in the field of robot control, especially by combining improved Ju-Gibbs quaternion and dynamic adaptive sliding filter (DASF) to improve the accuracy and efficiency of trajectory processing.

[0085] First, the system starts from the "Start" node and enters the "Sensor Data Collection" stage. In this stage, the system uses sensors installed on the robot joints and other key parts to collect various motion data in real time, including but not limited to position, speed and acceleration. These data are the basis of trajectory restoration accuracy and must be collected and transmitted accurately.

[0086] The collected raw sensor data is then sent to the "data preprocessing" module, where the data will undergo preliminary cleaning and formatting to remove possible noise and outliers, providing clean and standardized data input for subsequent advanced processing steps.

[0087] After data preprocessing, the process enters the "Applying Improved Ju-Gibbs Quaternions" stage. This stage is the core of the system, using improved Ju-Gibbs quaternion technology to accurately calculate and model the robot's three-dimensional motion and posture. This technology is superior to traditional quaternion methods and can better handle complex spatial rotations and avoid common mathematical singularity problems.

[0088] The data then flows to the "Data Filtering Processing (DASF)" module. This module uses DASF technology to automatically adjust the filtering parameters according to the real-time changes in the data. This intelligent filtering process greatly improves the availability and accuracy of the data, providing a high-quality data foundation for trajectory reconstruction.

[0089] The filtered data is passed to the "Trajectory Reconstruction Calculation" module, where the system calculates the robot's motion trajectory based on the processed data. This calculation result will be directly used in the "Trajectory Replay" stage to ensure that the robot can accurately reproduce the original motion trajectory.

[0090] After the trajectory playback is completed, the system not only marks the "end" of a complete data processing and trajectory restoration cycle, but also feeds some data back to the "Feedback and Optimization" module. This module analyzes the performance of the trajectory playback, optimizes the system settings, and adjusts the data processing flow to achieve higher accuracy and efficiency.

[0091] In addition, the high-precision trajectory acquisition and restoration system of the present invention has been successfully applied to the KUKA iiwa robot, demonstrating the efficiency and accuracy of the system in actual industrial robot operation. KUKA iiwa is an advanced industrial robot known for its flexibility and precision, and is widely used in automotive manufacturing, electronic assembly, and precision engineering. Figure 3 , Figure 4 ,By integrating this system into KUKA iiwa, the present invention not only verifies ,the functions and performances of the system, but also provides innovative ,technical support for the precise operation of the robot.

[0092] During the implementation process, the system first collects key motion data such as the robot's position, speed, and acceleration in real time through high-precision sensors installed on each joint and key components of KUKA iiwa. These data are captured through the system's "sensor data acquisition" module, ensuring the comprehensiveness and accuracy of data collection.

[0093] The collected data is then transferred to the "data preprocessing" module, which provides clean and standardized data input for subsequent data processing through preliminary data cleaning and formatting. Next, the data is further processed in the "application of improved Ju-Gibbs quaternions" stage. The improved quaternion method in the present invention can effectively handle the spatial rotation and posture adjustment problems encountered by robots in complex industrial environments, thereby improving the accuracy and reliability of data processing.

[0094] During the data filtering stage, the system introduces DASF to automatically adjust the filtering parameters according to the real-time changes in the data. This intelligent processing not only improves the data quality, but also optimizes the system's ability to respond to environmental changes. Through this advanced data processing, the system can accurately calculate the robot's motion trajectory and generate detailed motion instructions in the "Trajectory Reconstruction Calculation" module.

[0095] Finally, these motion instructions are used in the "trajectory playback" stage, and the KUKA iiwa robot is able to accurately reproduce the original motion trajectory, demonstrating highly precise operation capabilities. Figures 5 to 7 The system also includes a "Feedback and Optimization" module, which continuously adjusts and optimizes system settings by analyzing the performance of trajectory playback, further improving the accuracy and efficiency of trajectory restoration.

[0096] Through its successful application on the KUKA iiwa robot, this invention not only demonstrates its technological innovation in the field of high-precision trajectory acquisition and restoration, but also provides valuable practical experience and technical accumulation for the development and application of future robotics technology.

Claims

1. A method for real-time trajectory acquisition and high-precision restoration of a collaborative robot, characterized in that: The steps include: Step 1: Collect the position and velocity data of the robot joints in real time, including angle, velocity and acceleration; Step 2: Use the improved Ju-Gibbs quaternion to preprocess the collected raw data to obtain the preprocessed position and velocity data of the robot joints; Step 3: Apply a dynamic adaptive sliding filter to filter the preprocessed data according to the change rate and change amplitude of the real-time data; Step 4: Reconstruct the trajectory based on the filtered data. Use the inverse operation of the Ju-Gibbs quaternion to infer the robot postures of the initial state and the intermediate state from the final quaternion state, and reconstruct the robot's motion trajectory. Step 5: Compare the reconstructed trajectory with the actual robot motion trajectory, calculate the error index to evaluate the accuracy and efficiency of the reconstructed trajectory, and adjust and optimize the quaternion parameters and filtering algorithm based on the evaluation results.

2. The method for real-time trajectory acquisition and high-precision restoration of a collaborative robot according to claim 1, characterized in that: The collected raw data are preprocessed using the improved Ju-Gibbs quaternion as follows: First, obtain the connection data θ from link l-1 to link l l and n l , where θ l is the angular position about the axis, n l is the axis vector; the expression of the calculated Ju-Gibbs quaternion is: in, is the Ju-Gibbs quaternion describing the lth link relative to the global reference coordinate system o, o q l represents the unit quaternion of the lth link relative to the reference coordinate system o, where o is the symbol representing the reference coordinate system. n l ,τ l is a vector, is a scalar; During the robot joint rotation, the Ju-Gibbs quaternion is written in general form as: in, is the Ju-Gibbs quaternion describing the attitude from the reference coordinate system o to the l-1th link, is the Ju-Gibbs quaternion describing the relative attitude transformation from the l-1th link to the lth link, is the square of the modulus of the Ju-Gibbs quaternion; The Ju-Gibbs quaternion combines the rotation axis and rotation angle of each joint to form a unified mathematical representation to describe the rotation of each joint; the rotation of each joint is continuously combined through quaternion multiplication to express the posture changes of the entire robotic arm.

3. The method for real-time trajectory acquisition and high-precision restoration of a collaborative robot according to claim 1 or 2, characterized in that: Apply a dynamic adaptive sliding filter to filter the preprocessed data as follows: First, the difference between each data point in the data stream and the previous data point is calculated to obtain the rate of change and magnitude of change of the data; the data stream refers to the data sequence continuously collected by each sensor of the robot, and the data point includes position, speed or acceleration information; Then, according to the rate and magnitude of change, the size of the filter window and the weight coefficient are dynamically adjusted, and the weight coefficient is adjusted according to the time interval between the data point and the current processing point; The weight is calculated using the exponential decay model, the formula is w i =e -α*# , where w i is the weight coefficient assigned by the filter to each data point, d is the number of time steps from the data point to the current point, and α is the attenuation coefficient, which controls the speed at which the weight decreases; If both the data change rate and the change amplitude exceed the set threshold, the filter window is reduced and the α value is increased; conversely, if both the change rate and the change amplitude do not exceed the set threshold, the filter window is expanded and the α value is reduced.

4. The method for real-time trajectory acquisition and high-precision restoration of a collaborative robot according to claim 2, characterized in that: The trajectory is reconstructed based on the filtered data as follows: After the robot completes a series of actions, the data collected and processed by the sensor system in real time is converted into the final quaternion state. The inverse operation of the Ju-Gibbs quaternion formula is used to infer the robot postures of the initial state and the intermediate state through the final quaternion state, and reproduce the robot's motion trajectory; the mathematical model describing the transition of the robot from one posture to another is as follows: in, is the Ju-Gibbs quaternion of the l-1th joint, representing the pose transformation from the reference coordinate system o to the corrected state d, is the rotation axis vector of the l-1th joint, is the scalar part of the rotation angle of the l-1th joint, d is the correction or target state, and o is the reference coordinate system; is the symmetric rotation axis matrix in the modified state d, is the rotation axis vector of the lth joint, is the symmetric matrix representation of the rotation axis of the lth joint, n l is the rotation axis vector of the lth joint.

Images