Servo motor-based manipulator cooperative control system and method

Abstract

The application discloses a servo motor-based manipulator cooperative control system and method, and relates to the technical field of manipulator control, and comprises the following steps: injecting a frequency-adjustable reference excitation signal into the ground wire of a servo driver of each servo shaft in a manipulator, and constructing a coupling observation basic sequence; while executing the reference excitation signal injection, performing parallel and synchronous collection on the instantaneous potential of the ground wire of the servo driver of each servo shaft, and based on the coupling observation basic sequence, constructing a complete multi-axis potential response matrix. By introducing the frequency-adjustable excitation signal, combining sparse cross-correlation and adaptive Kalman filtering, the application realizes identification and modeling of the cross-axis potential interference path, dynamically corrects the servo current loop control parameters, thereby effectively suppressing the parasitic current interference, improving the synchronism, stability and anti-interference robustness of multi-axis cooperative control, and overcoming the problems of response lag and out-of-step of the traditional method.

Description

Technical Field

[0001] This invention relates to the field of robotic arm control technology, specifically to a collaborative control system and method for robotic arms based on servo motors. Background Technology

[0002] Servo motor-based collaborative control of robotic arms refers to the use of multiple servo motors to drive and coordinate multiple joints or motion axes of a robotic arm with high precision and real-time synchronization. This enables coordinated and dynamic coupling control of the various parts of the robotic arm in terms of position, speed, and acceleration. This control method typically relies on a distributed servo drive system built using a real-time bus (such as EtherCAT or CANopen). A controller (such as a PLC or motion controller) sends multi-axis coordinated motion commands, and the servo motors execute precisely according to the set trajectory, ensuring that the robotic arm maintains spatial path consistency and dynamic response coordination during complex operations (such as grasping, handling, and assembly). This method is particularly suitable for automation scenarios with high requirements for motion accuracy and dynamic consistency, such as intelligent manufacturing, collaborative robots, and high-precision assembly systems.

[0003] However, existing technologies still have the following shortcomings:

[0004] In existing multi-axis servo motor cooperative control systems, the complex cable wiring structure can lead to unexpected potential coupling between different servo axes, forming a parasitic current closed loop. This parasitic current can cause random transient disturbances to the feedback and control signals of the servo driver, easily causing short-term loss of synchronization or jitter in individual axes. This not only reduces the positioning accuracy of a single axis but also disrupts the synchronous coordination between multiple axes. In severe cases, it can even trigger the system to misinterpret a normal, false cooperative state, leading to unpredictable failures such as overall linkage malfunction.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide a servo motor-based robotic arm collaborative control system and method. By introducing a frequency-adjustable reference excitation signal, the potential coupling behavior between servo axes is actively induced and observed. Combined with sparse cross-correlation analysis and adaptive Kalman filtering, accurate identification and dynamic modeling of cross-axis potential interference paths are achieved. Furthermore, a potential coupling coefficient matrix and an online coupling model are constructed and used for real-time correction and compensation of servo motor current loop control parameters. This scheme effectively suppresses transient disturbances caused by parasitic current coupling, improves the synchronization, response consistency, and disturbance rejection robustness of multi-axis collaborative control, and overcomes the problems of loss of synchronization, jitter, and misjudgment caused by identification lag and parameter rigidity in traditional methods, thus solving the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution: a servo motor-based robotic arm collaborative control method, comprising the following steps:

[0008] S100. Inject frequency-adjustable reference excitation signals into the ground wires of the servo drivers of each servo axis in the robot arm to actively induce observable potential response behavior across servo axes and construct a coupled observation basic sequence.

[0009] S200: While injecting the reference excitation signal, the instantaneous ground potential of each servo axis servo driver is acquired in parallel and synchronously. Based on the coupled observation basic sequence, a complete multi-axis potential response matrix is ​​constructed.

[0010] S300: Perform sparse cross-correlation operation on the multi-axis potential response matrix, extract the cross-axis potential correlation spectrum between each servo axis, and remove the self-excited response components caused by each reference excitation signal source itself to obtain the purified cross-axis potential response relationship.

[0011] S400. Based on the extracted cross-axis potential correlation spectrum, an adaptive Kalman filter structure is constructed, and real-time iteration is performed in combination with historical response data to dynamically infer the potential coupling gain vector between each servo axis.

[0012] S500: Using the ratio of the potential coupling gain vector to the amplitude of the corresponding reference excitation signal as a basis, calculate the potential coupling coefficient matrix between each servo axis, and establish an online coupling model reflecting the current operating state based on the potential coupling coefficient matrix.

[0013] S600 introduces the potential coupling coefficient matrix as a key input parameter into the control structure of the servo controller, divides the redundant control dimension space in the servo controller, and constructs the coupling response matrix. Combining the online coupling model and parasitic current interference characteristics, the current loop proportional-integral control parameters of each servo axis servo driver are dynamically corrected and stability compensated to improve the robustness and anti-disturbance performance of multi-axis collaborative control.

[0014] Preferably, step S100 specifically includes:

[0015] Select a sinusoidal signal source with stable frequency and controllable amplitude, construct multiple reference excitation signal source channels, and connect them to the ground input terminal of the servo driver of each servo axis in the robot hand;

[0016] After the signal connection is established, the excitation control sequence is started based on the preset periodic excitation mode, so that each servo axis ground line receives the reference signal of its own frequency in a specified timing sequence.

[0017] During the excitation operation, the instantaneous potential of the ground wire of each servo axis is acquired in real time and in parallel through a high-resolution data acquisition device, and bandpass filtering and synchronous sampling are performed.

[0018] By combining excitation frequency and time window, the collected data are synchronously analyzed and classified to form a coupled observation basic sequence with time label and frequency attributes.

[0019] Preferably, step S200 specifically includes:

[0020] Configure a multi-channel synchronous sampling device to collect the ground potential of each servo axis in parallel, and set a unified time base to achieve synchronous sampling;

[0021] The acquired potential sequences were aligned according to timestamp frames to construct the original observation data matrix, and the spectrum was divided and the data was labeled according to the excitation frequency and axis number.

[0022] Frequency domain analysis is used to identify the response components of each servo axis to the excitation frequency, and the effective response data is selected. The effective response data is then remapped into a structured multi-axis potential response matrix according to the excitation source axis and the response axis, reflecting the potential coupling strength and distribution characteristics.

[0023] Preferably, step S300 specifically includes:

[0024] For each excitation source axis-response axis combination in the structural multi-axis potential response matrix, the response potential sequence is extracted, and cross-correlation analysis is performed with the corresponding excitation signal as a reference to obtain the initial cross-correlation spectrum.

[0025] A sparsity constraint mechanism is introduced to filter the cross-correlation spectrum, retaining only the effective correlation components that are above a set response intensity threshold, and suppressing irrelevant noise;

[0026] Self-excited response components on the main diagonal are identified and removed, and cross-correlation data containing only cross-axis interference paths are retained. The processed data is then used to construct a purified cross-axis potential response relationship spectrum for subsequent coupling modeling and control input.

[0027] Preferably, step S400 specifically includes:

[0028] Based on the purified transaxis potential response relationship spectrum, a state-space model is established with the potential response value as the observable and the coupling gain vector as the state variable.

[0029] Based on the constructed state-space model, initialize the core parameters in the adaptive Kalman filter structure;

[0030] By combining historical cross-axis potential response data, the state prediction and observation update iteration operation of the Kalman filter is performed to correct the coupling gain estimate in real time.

[0031] After filtering and stabilization, the state output is extracted to form a potential coupling gain vector containing the combination of each excitation source-response axis, which is used as the input for subsequent control modeling.

[0032] Preferably, step S500 specifically includes:

[0033] Extract the potential coupling gain vector and normalize it with the amplitude of the corresponding reference excitation signal to obtain the coupling response value under unit excitation;

[0034] The normalized response values ​​are reorganized into a two-dimensional matrix to construct the potential coupling coefficient matrix between each servo axis.

[0035] Time-domain fusion is performed on the potential coupling coefficient matrix of multiple consecutive cycles to generate an online coupling model that can be updated over time. The online coupling model is then embedded into the servo control process as the input basis for control parameter correction, thereby achieving adaptive adjustment of disturbances.

[0036] Preferably, step S600 specifically includes:

[0037] The potential coupling coefficient matrix is ​​introduced into the servo controller control structure, and the cross-axis coupling influence value corresponding to each servo axis is extracted. Based on the coupling influence value, a redundant control dimension is constructed, a coupling response matrix is ​​generated, and it is mapped to the control space.

[0038] The proportional-integral parameters of the current loop are dynamically corrected based on the coupling response matrix and parasitic disturbance characteristics.

[0039] The correction parameters are loaded into the servo driver, and the update frequency and compensation magnitude are set according to the coupling change trend to achieve adaptive parameter adjustment and stability compensation.

[0040] The servo motor-based robotic arm collaborative control system includes a coupling excitation injection module, a synchronous potential acquisition module, a potential cross-correlation analysis module, a coupling gain estimation module, an online coupling modeling module, and a parameter adaptive compensation module.

[0041] The coupled excitation injection module injects frequency-adjustable reference excitation signals into the ground wires of the servo drivers of each servo axis in the manipulator to construct a coupled observation basic sequence.

[0042] The synchronous potential acquisition module performs parallel synchronous acquisition of the instantaneous ground potential of each servo axis servo driver while injecting the reference excitation signal, and constructs a complete multi-axis potential response matrix based on the coupled observation basic sequence.

[0043] The potential cross-correlation analysis module performs sparse cross-correlation operations on the multi-axis potential response matrix, extracts the cross-axis potential correlation spectrum between each servo axis, and removes the self-excited response components caused by each reference excitation signal source itself.

[0044] The coupling gain estimation module constructs an adaptive Kalman filter structure based on the extracted cross-axis potential correlation spectrum, and performs real-time iteration in combination with historical response data to dynamically infer the potential coupling gain vector between each servo axis.

[0045] The online coupling modeling module uses the ratio of the potential coupling gain vector to the amplitude of the corresponding reference excitation signal as a basis to calculate the potential coupling coefficient matrix between each servo axis, and establishes an online coupling model that reflects the current operating state based on the potential coupling coefficient matrix.

[0046] The parameter adaptive compensation module introduces the potential coupling coefficient matrix as a key input parameter into the control structure of the servo controller. It divides the redundant control dimension space in the servo controller and constructs the coupling response matrix. Combining the online coupling model and parasitic current interference characteristics, it dynamically corrects and compensates for the current loop proportional-integral control parameters of each servo axis servo driver.

[0047] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0048] This invention introduces a frequency-adjustable reference excitation signal during multi-axis servo motor cooperative control to actively induce and observe the potential coupling behavior between servo axes. Combining sparse cross-correlation analysis and an adaptive Kalman filter structure, it achieves accurate identification and dynamic modeling of cross-axis potential interference paths. Based on this, a potential coupling coefficient matrix and an online coupling model are constructed, which are used as control inputs to participate in the real-time correction and compensation of servo motor current loop control parameters. This scheme effectively suppresses the damage to servo control accuracy and stability caused by transient disturbances due to parasitic current coupling, significantly improving synchronization, response consistency, and disturbance rejection robustness during multi-axis cooperative motion. It avoids problems such as loss of synchronization, jitter, and misjudgment caused by the inability to identify coupling paths in real time or fixed control parameters in traditional methods. This provides a highly reliable and adaptable new solution for high-precision multi-axis cooperative control in complex automated equipment. Attached Figure Description

[0049] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0050] Figure 1 This is a flowchart of the servo motor-based robotic arm collaborative control method of the present invention.

[0051] Figure 2 This is a schematic diagram of the module of the servo motor-based robotic arm collaborative control system of the present invention. Detailed Implementation

[0052] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0053] This invention provides, for example Figure 1 The servo motor-based robotic arm collaborative control method shown includes the following steps:

[0054] S100. Inject frequency-adjustable reference excitation signals into the ground wires of the servo drivers of each servo axis in the robot arm to actively induce observable potential response behavior across servo axes and construct a coupled observation basic sequence.

[0055] To address the parasitic current interference that may arise from ground coupling during multi-servo axis coordinated motion of a robotic arm, a method for actively inducing and observing the potential response is proposed, which includes the following steps:

[0056] A stable, amplitude-controllable sinusoidal signal source was selected to construct multiple reference excitation signal source channels, which were then connected to the ground input terminals of the corresponding servo drivers for each servo axis in the robot arm. During the connection process, it was crucial to ensure clear electrical isolation between the signal injection path and the driver ground to prevent the reference excitation signal from causing reverse impact on the drive power supply or control circuit. The frequency range of the reference excitation signal was preferably between 1kHz and 20kHz, which avoids interference from the industrial power supply's main frequency and ensures good spectral separation capability in subsequent observations. The frequencies of each reference excitation signal should be interleaved within the set range to ensure that the excitation carried by each servo axis has a unique identifier in the frequency domain.

[0057] After the aforementioned excitation signal is injected, an excitation control sequence is initiated based on a preset periodic excitation mode, causing each servo axis ground line to sequentially receive its own frequency reference signal in a specified timing order. During the excitation process, it is necessary to ensure that the reference signal of each axis remains stable within two adjacent sampling periods to maintain the repeatability of the potential response behavior. In this way, observable differences in potential response can be effectively induced between servo axes, constructing a cross-axis interference distribution characteristic distinguished by frequency, thereby providing a clear foundation for subsequent potential response analysis.

[0058] During stable operation, the instantaneous potentials on the ground wires of each servo axis servo driver are acquired in real time using a high-resolution data acquisition device, and the acquired data undergoes preliminary preprocessing. Specifically, this includes bandpass filtering of the signal to eliminate main power supply interference and high-frequency noise, and employing a synchronous sampling strategy to ensure that the ground potential data of all servo axes are recorded at the same time reference. The sampling frequency must be significantly higher than five times the highest frequency of the excitation signal to ensure that the acquired signal has sufficient time-domain reconstruction capability. After this step, the original sequence of multi-axis potential responses varying over time can be obtained, providing data support for the construction of the response matrix.

[0059] Based on the corresponding potential responses generated by the excitation signal in the ground lines of different servo axes, the response data are analyzed synchronously in both the time and frequency domains, and the collected data are classified and organized according to the excitation frequency and time window. By mapping the potential response of each axis to its injected specific excitation signal and marking responses that do not belong to the frequency band of the axis, the interference contribution between axes can be effectively separated in the frequency domain. Finally, the potential response behaviors of all servo axes are integrated to form a coupled observation basic sequence with time labels and frequency attributes, laying the foundation for subsequent cross-axis potential correlation spectrum extraction and potential coupling model construction.

[0060] S200: While injecting the reference excitation signal, the instantaneous ground potential of each servo axis servo driver is acquired in parallel and synchronously. Based on the coupled observation basic sequence, a complete multi-axis potential response matrix is ​​constructed.

[0061] While injecting adjustable reference excitation signals into the ground wires of the servo drives of each servo axis of the robot, it is necessary to perform high-precision, time-consistent observation of the ground potential response behavior caused by the excitation in order to construct a complete multi-axis potential response matrix. This process specifically includes the following steps:

[0062] A multi-channel synchronous sampling data acquisition device is configured to connect in parallel to the sampling nodes between the ground wires of each servo axis servo driver and the reference ground point. Each channel employs a high input impedance isolation structure to ensure that no additional interference or pull-down of the original ground potential level is introduced during sampling. For sampling hardware selection, analog-to-digital converters with a resolution of at least 16 bits and a sampling frequency of at least 100kHz are preferred to ensure sufficient time and amplitude resolution for signal components within the excitation frequency range. This step ensures that consistent real-time potential signal inputs are obtained from multiple servo axes, which is a prerequisite for achieving time-aligned observation.

[0063] For the instantaneous potential data acquired from the ground wires of each servo axis, a unified time reference signal is set before sampling begins, and synchronous startup of all channels is achieved through an external clock pulse or fiber optic trigger signal. During sampling, each set of sampled data is timestamped to ensure strict correspondence between data from different axes at the same time during subsequent data fusion. The duration of the sampling window should cover at least several complete excitation signal cycles to provide sufficient resolution spectral data for subsequent frequency domain analysis. Through this step, instantaneous potential observation samples of the multi-axis servo drive ground wires on a completely consistent time axis can be obtained.

[0064] After sampling each channel, the potential sequences corresponding to all servo axis ground lines are frame-aligned according to timestamps to construct the original observation data matrix. Each row of this original observation data matrix corresponds to a sampling time, and each column corresponds to the potential value of a servo axis ground line. Subsequently, combined with the excitation frequency and injection timing information recorded during the aforementioned excitation signal injection process, the original observation data matrix is ​​labeled, and its spectrum is partitioned according to the excitation frequency band. To enhance anti-interference capability, bandpass filtering can be performed on each column of data in the matrix at this stage, retaining only the response signal components near their corresponding excitation frequency, thereby further suppressing the effects of high-frequency noise and power frequency interference.

[0065] After constructing and preprocessing the original observation data matrix, and combining the excitation frequency configuration scheme and servo axis numbering information, the original observation data matrix is ​​remapped into a structured multi-axis potential response matrix with excitation source-response axis as dual indices. The specific steps are as follows: First, extract the time-series potential data from the original observation data matrix. Each column in the original observation data matrix corresponds to the instantaneous potential value of a servo axis ground wire, and each row corresponds to a full-axis synchronous sampling record under a unified timestamp. Second, combining the correspondence between the reference excitation signal frequency injected into each servo driver ground wire and its number in the excitation frequency configuration scheme, establish an index that corresponds one-to-one between the source axis number identifier of the injected excitation signal at each sampling time and the excitation frequency used. Third, based on this index table, frequency domain analysis is performed on each column of the original observation data matrix. Bandpass filters or Fourier transforms are used to identify which columns contain response peaks for specific frequency excitations. Then, combining the excitation frequency and axis number, the "excitation source-response axis" combination to which each response potential trajectory belongs is confirmed. Finally, all frequency domain response data that meet the conditions are extracted, and a new structured multi-axis potential response matrix is ​​constructed using the excitation source number as the row index and the response axis number as the column index. Each element of this structured multi-axis potential response matrix records the observed response value or response function of a certain excitation source axis signal on another servo axis ground line, reflecting the potential coupling strength and distribution characteristics between excitation and response. Through this mapping operation, time-series observation results can be converted into a structured representation of frequency domain coupling relationships, laying a clear and quantifiable foundation for subsequent cross-correlation calculations and the construction of coupling models. This structured multi-axis potential response matrix not only reflects the response of each servo axis to its own excitation, but more importantly, it reveals the coupling response of each excitation source on the ground lines of other non-axis drivers. This provides a solid data foundation for subsequent sparse cross-correlation calculations and extraction of cross-axis potential correlation spectra. The construction process of this structured multi-axis potential response matrix transforms the potential coupling relationship from abstract observation to quantitative description, providing essential prerequisite support for potential coupling modeling and control parameter compensation.

[0066] S300: Perform sparse cross-correlation operation on the multi-axis potential response matrix, extract the cross-axis potential correlation spectrum between each servo axis, and remove the self-excited response components caused by each reference excitation signal source itself to obtain the purified cross-axis potential response relationship.

[0067] To accurately identify parasitic potential conduction paths between multiple servo axes caused by ground coupling, after constructing the structured multi-axis potential response matrix, it is necessary to further perform sparse cross-correlation processing on the matrix to extract the actual cross-axis potential correlation spectrum, and then eliminate the self-excited response interference generated by the reference excitation through an algorithm. The specific steps include the following:

[0068] For each "excitation source axis-response axis" pair in the structured multi-axis potential response matrix, its corresponding response potential time series is extracted, forming several information pairs with clearly defined frequency assignments. For each information pair, using the reference excitation signal corresponding to the excitation source axis as the reference signal and the potential sequence corresponding to the response axis as the response signal, a cross-correlation analysis is performed in the time domain to obtain an initial cross-correlation spectrum. This operation is used to reveal whether there is a linear correlation between the response signal and the excitation signal in frequency and phase, thereby determining whether a cross-axis potential propagation path exists. The peak value and time delay in the cross-correlation spectrum will serve as important parameters for subsequent extraction of coupling features.

[0069] Based on the completion of all "excitation source axis-response axis" cross-correlation analyses, a sparsity constraint mechanism is introduced to avoid interference from low signal-to-noise ratio spurious response data on modeling accuracy. This mechanism, based on L1 norm sparse optimization theory, sets a reasonable response intensity threshold, retaining only the correlated components in the cross-correlation spectrum that are significantly higher than the background noise level. Sparse cross-correlation processing effectively suppresses response artifacts from irrelevant noise channels while highlighting the truly physically coupled cross-axis potential correlation trajectories. Through this operation, the original observation data matrix is ​​compressed into a sparse spectrum containing only strongly correlated paths, providing more focused data input for subsequent modeling.

[0070] To further ensure data purity and modeling relevance, it is necessary to identify and remove self-excited response components induced by the excitation source itself from the aforementioned sparse cross-correlation spectrum. Self-excited responses typically manifest as autocorrelation peaks in the cross-correlation spectrum when the response axis and the excitation source axis share the same index. Since this part of the response does not reflect cross-axis coupling characteristics, it should be removed from the sparse spectrum. Removal methods include: identifying the positions of all main diagonal elements and setting their corresponding cross-correlation values ​​to zero, or performing normalization reduction processing to ensure that only effective coupling information between servo axes with different indices is retained in the spectrum.

[0071] After eliminating self-excited components and performing sparsification, the remaining components are reorganized into a purified cross-axis potential response relationship map. This map, indexed by servo axis numbers, records the measured potential coupling strength and time delay characteristics under excitation, serving as the direct basis for constructing the potential coupling coefficient matrix and establishing online modeling logic. Through this entire process, not only is the cross-axis response path effectively identified and refined, but the quantification accuracy and stability of the potential coupling relationship in subsequent modeling processes are significantly improved, avoiding misleading control strategy judgments due to self-excited interference or spurious signals.

[0072] S400. Based on the extracted cross-axis potential correlation spectrum, an adaptive Kalman filter structure is constructed, and real-time iteration is performed in combination with historical response data to dynamically infer the potential coupling gain vector between each servo axis.

[0073] To further transform the extracted purified cross-axis potential correlation spectrum into coupled quantization parameters that can be used for modeling and dynamic prediction, a gain estimation method based on Kalman filtering theory is proposed. This method constructs an adaptive filter structure and iterates in real time using historical response data to dynamically infer the potential coupling gain vector between servo axes. The process specifically includes the following steps:

[0074] Based on the purified cross-axis potential response spectrum obtained through sparse cross-correlation processing in the previous stage, the initial observation state equation and state transition equation are established. In this step, the potential response values ​​between servo axes are considered as observations, and the coupling gain vector is considered as the state variable to be estimated. To construct the state transition process, the coupling gain vector is assumed to follow a slowly varying model in the short time scale, meaning that its variation amplitude is small and it has a certain degree of stability in continuous time periods. On this basis, the state transition equation is defined as the current state equals the previous state plus a Gaussian white noise perturbation term, while the observation equation is expressed by the weighted product of the current potential response value and the current coupling gain vector, forming a typical linear filtering structure. Through the above modeling, the state space structure and observation input path required for filtering are clarified.

[0075] Based on the constructed state-space model, the core parameters of the adaptive Kalman filter structure, such as the initial state vector, covariance matrix, and noise covariance estimation, are initialized. The initial state vector can be set to a zero vector or a coarse estimate based on the static response calculation; the state covariance matrix is ​​diagonally assigned according to the historical maximum deviation range between different servo axes; the observation noise covariance matrix is ​​initialized based on the sampling accuracy of the response signal acquisition device and the background interference level, and is updated in real time during subsequent iterations. This step ensures that the filter has good convergence initial state and dynamic response capabilities, making the subsequent gain update process more stable and adaptive.

[0076] After initializing the Kalman filter structure, the real-time iterative process of the filter is initiated based on multi-frame cross-axis potential response data acquired historically. Within each iteration cycle, a state prediction step is first executed, predicting the current coupling gain vector and covariance based on the previous estimate and the state transition equation. Subsequently, an observation update step is executed, substituting the actual cross-axis response value acquired at the current moment into the observation equation, comparing it with the predicted value, calculating the innovation (i.e., the observation residual), and using this innovation to correct the current state estimate. This filter update mechanism allows the system to quickly and adaptively adjust to dynamic changes or disturbances in the potential coupling path, obtaining more accurate potential coupling gain values ​​between servo axes in real time.

[0077] After the Kalman filter completes multiple rounds of iterative training using historical data and obtains stable estimation results, the current state estimation output is extracted as the latest potential coupling gain vector. Each component of this coupling gain vector corresponds to the unit potential coupling strength generated by one servo axis to another servo axis under the reference excitation, possessing a clear physical meaning and frequency attribution label. This coupling gain vector can be regarded as a dynamic measure of cross-axis parasitic coupling under the current electrical structure and wiring conditions. It can serve as the input basis for the subsequent online modeling stage and can also directly participate in the adaptive correction of control parameters in the servo control strategy. Through this step, the static spectrum extraction results are successfully transformed into dynamically adjustable physical modeling parameters, establishing a bridge between the observed response and control optimization, and laying the technical foundation for achieving high-precision cooperative control.

[0078] S500: Using the ratio of the potential coupling gain vector to the amplitude of the corresponding reference excitation signal as a basis, calculate the potential coupling coefficient matrix between each servo axis, and establish an online coupling model reflecting the current operating state based on the potential coupling coefficient matrix.

[0079] To achieve quantifiable modeling of the potential coupling relationship between servo motors, after dynamically acquiring the potential coupling gain vector, it is necessary to further introduce the amplitude information of the reference excitation signal, normalize the gain, and thus construct a potential coupling coefficient matrix reflecting the coupling strength. Based on this, an online coupling model that can be updated over time is established to guide the subsequent control parameter correction process. This method not only achieves dynamic modeling of electromagnetic interference but also considers practicality and responsiveness in engineering implementation. Specifically, it includes the following steps:

[0080] The potential coupling gain vector at the current time point output by the Kalman filter structure is extracted, and the amplitude of the reference excitation signal corresponding to each coupling path is determined by referring to the previously recorded reference excitation signal configuration table. In this method, the estimated coupling gain value of each "excitation source axis-response axis" path essentially represents the absolute slope of the response potential change with the excitation amplitude. To eliminate the error caused by the inconsistency of the excitation source amplitude, each gain value needs to be divided by its corresponding excitation amplitude to form a standardized coupling response under a unit excitation amplitude. This normalization operation converts the response intensity on all paths into a dimensionless relative index, facilitating comparative analysis between different paths.

[0081] The normalized response values ​​of each path are reorganized into a two-dimensional matrix, i.e., a potential coupling coefficient matrix is ​​constructed. The row indices of this matrix represent the excitation source axis number, and the column indices represent the response axis number. Each matrix element represents the parasitic coupling potential intensity generated by that excitation source axis to a ground line of a certain response axis under a unit excitation amplitude. This matrix is ​​typically sparse; off-diagonal elements reflect the strength of cross-axis coupling channels, while diagonal elements represent the feedback relationship within the axis (which can be set to zero if necessary). Simultaneously, each value in the matrix retains its time-domain characteristic label, providing a foundation for the next step of constructing a time-dynamic response model. By constructing this potential coupling coefficient matrix, not only is a comprehensive characterization of the number and strength of coupling paths achieved, but a clear basis is also provided for subsequent interference path identification and suppression strategy deployment.

[0082] After obtaining the potential coupling coefficient matrix, time continuity constraints and state recursion logic are further introduced to transform the static coupling description into an online coupling model that can be dynamically updated with the sampling period. To this end, a sliding time window mechanism can be introduced to perform time-domain fusion of the potential coupling coefficient matrices at multiple consecutive time points. The coupling change trend is extracted using methods such as weighted averaging, exponential smoothing, or least squares fitting to obtain the time-smoothed coupling matrix at the current moment. Furthermore, structural correction parameters such as response axis sensitivity factors and inter-axis distance correction factors need to be added to the model to enhance its robustness to local potential shifts under complex wiring structures. This online coupling model can be updated in real time to reflect changes in the coupling mode under the current operating state of the robot, which is a prerequisite for realizing dynamic correction of control parameters.

[0083] This online coupling model is embedded as a queryable dynamic data structure into the servo control flow, allowing it to be called, updated, and evaluated in each control cycle. By comparing it with the output model of the previous cycle, it can detect whether there are abrupt changes, jumps, or drifts in the current potential coupling. If the coupling strength of some coupling channels in the response path exceeds the safety threshold, an early warning mechanism can be triggered, prompting the control loop to enter the parameter adjustment preparation stage. Simultaneously, the coupling strength information output by the model can also serve as an important input for the subsequent adaptive adjustment of the proportional-integral parameters of the current loop, used to precisely control the interference suppression amplitude and corresponding response rate of the servo axis. In this way, the online coupling model not only presents itself as static data but also becomes an indispensable dynamic element in the control logic, realizing a closed-loop response mechanism from "detection-modeling-adjustment".

[0084] S600 introduces the potential coupling coefficient matrix as a key input parameter into the control structure of the servo controller, divides the redundant control dimension space in the servo controller, and constructs the coupling response matrix. Combining the online coupling model and parasitic current interference characteristics, the current loop proportional-integral control parameters of each servo axis servo driver are dynamically corrected and stability compensated to improve the robustness and anti-disturbance performance of multi-axis collaborative control.

[0085] To achieve real-time compensation for parasitic current coupling disturbances in the coordinated control of multi-axis servo motors, based on the successful construction and dynamic updating of the potential coupling coefficient matrix and online coupling model, these are introduced as key inputs into the control structure of the servo controller. By dividing the control redundancy dimension, constructing the coupling response matrix, and dynamically correcting the control parameters, intelligent adjustment of the proportional-integral parameters of the current loop of each servo axis servo driver is ultimately achieved, thereby improving the robustness and disturbance rejection performance of the overall coordinated control. This process specifically includes the following steps:

[0086] The potential coupling coefficient matrix is ​​introduced as a dynamic external input into the servo controller's calculation process and coupled with the current state variables of each axis within the controller in real time. Each off-diagonal element in this matrix reflects the degree of parasitic coupling between different servo axes, while the diagonal elements represent self-feedback effects. This step primarily focuses on cross-axis coupling terms. Before introducing the matrix, its numerical structure needs to be verified and normalized to ensure its amplitude is within the dynamic range acceptable to the controller. After introduction, the controller analyzes the matrix in each control cycle, treating each axis as the response target and all other axes as disturbance sources, extracting their corresponding cross-axis coupling influence values ​​to provide basic data support for subsequent redundant dimension mapping and parameter correction.

[0087] An extended control parameter space is defined within the controller, expanding the original single-axis control structure based on the servo axis closed-loop control strategy to a multi-dimensional coupled control structure. Specifically, for each servo axis, in addition to maintaining the conventional current loop, speed loop, and position loop control channels, a set of redundant control dimensions for disturbance compensation is added. This dimension is used to express the compensation reverse adjustment amount for the external potential coupling effect currently experienced by the servo axis, and its value is derived from the coupling response matrix obtained by transforming the potential coupling coefficient matrix and the state error signals of each axis. In this process, through matrix multiplication, the response quantities of all coupling paths are projected into the control space of each axis, so that each axis obtains a set of coupling-related compensation adjustment signals in addition to the original control signals, forming a multi-dimensional control input combination, thereby suppressing the coupling effect.

[0088] After obtaining the coupling response matrix, the proportional gain and integral time constant of the current loop for each axis are dynamically corrected based on the frequency domain characteristics of parasitic current interference, the transient disturbance model, and the current loop control bandwidth of each axis. Specifically, when a servo axis is identified as being affected by parasitic potential interference from multiple axes in the current cycle, the controller will appropriately reduce its proportional gain to improve the system's tolerance to rapid disturbances; simultaneously, it will increase the integral time constant to avoid excessive amplification of errors by the integral loop, which could cause system oscillations. For servo drives identified as interference source axes, their control stiffness needs to be increased to prevent signal "leakage" to other axes. This process can be combined with amplitude-frequency analysis strategies to compare the coupling response spectrum with a standard interference template, automatically identify the concentrated interference frequency bands, and reconstruct the controller's frequency response characteristics accordingly. In this way, the current loop control parameters of each axis can be customized and distributed according to its coupling relationship, achieving self-balancing of control resources in an interference environment.

[0089] The dynamically corrected set of control parameters is loaded into the servo drive control channel in real time. Taking advantage of the continuous update characteristics of the online coupling model, the parameter adjustment frequency and maximum adjustment amplitude are set to avoid unstable responses caused by frequent jumps. Within each cycle, the controller determines whether to trigger a control parameter update based on the changing trend of the coupling coefficient matrix. If an increasing coupling trend or a new coupling path is detected, gain correction is automatically triggered; if coupling weakens or is in a stable state, the current parameter settings are maintained or a gradual adjustment mode is entered. Furthermore, a parameter recovery mechanism is implemented to ensure that when the servo drive leaves the coupling interference range, the control parameters can smoothly return to their initial settings, maintaining the stability and accuracy consistency of the system over long-term operation. Through these measures, the entire control strategy achieves a comprehensive improvement in dynamic response capability, interference adaptability, and control robustness. It effectively suppresses multi-axis collaborative synchronization loss, sudden malfunctions, or missynchronization problems caused by parasitic potential coupling, significantly enhancing the reliability of multi-axis collaborative control under complex operating conditions.

[0090] This invention introduces a frequency-adjustable reference excitation signal during multi-axis servo motor cooperative control to actively induce and observe the potential coupling behavior between servo axes. Combining sparse cross-correlation analysis and an adaptive Kalman filter structure, it achieves accurate identification and dynamic modeling of cross-axis potential interference paths. Based on this, a potential coupling coefficient matrix and an online coupling model are constructed, which are used as control inputs to participate in the real-time correction and compensation of servo motor current loop control parameters. This scheme effectively suppresses the damage to servo control accuracy and stability caused by transient disturbances due to parasitic current coupling, significantly improving synchronization, response consistency, and disturbance rejection robustness during multi-axis cooperative motion. It avoids problems such as loss of synchronization, jitter, and misjudgment caused by the inability to identify coupling paths in real time or fixed control parameters in traditional methods. This provides a highly reliable and adaptable new solution for high-precision multi-axis cooperative control in complex automated equipment.

[0091] This invention provides, for example Figure 2 The servo motor-based robotic arm collaborative control system shown includes a coupling excitation injection module, a synchronous potential acquisition module, a potential cross-correlation analysis module, a coupling gain estimation module, an online coupling modeling module, and a parameter adaptive compensation module.

[0092] The coupled excitation injection module injects frequency-adjustable reference excitation signals into the ground wires of the servo drivers of each servo axis in the robot, thereby constructing a coupled observation basic sequence.

[0093] The synchronous potential acquisition module performs parallel synchronous acquisition of the instantaneous ground potential of each servo axis servo driver while injecting the reference excitation signal, and constructs a complete multi-axis potential response matrix based on the coupled observation basic sequence.

[0094] The potential cross-correlation analysis module performs sparse cross-correlation operations on the multi-axis potential response matrix, extracts the cross-axis potential correlation spectrum between each servo axis, and removes the self-excited response components caused by each reference excitation signal source itself.

[0095] The coupling gain estimation module constructs an adaptive Kalman filter structure based on the extracted cross-axis potential correlation spectrum, and performs real-time iteration in combination with historical response data to dynamically infer the potential coupling gain vector between each servo axis.

[0096] The online coupling modeling module uses the ratio of the potential coupling gain vector to the amplitude of the corresponding reference excitation signal as a basis to calculate the potential coupling coefficient matrix between each servo axis, and establishes an online coupling model that reflects the current operating state based on the potential coupling coefficient matrix.

[0097] The parameter adaptive compensation module introduces the potential coupling coefficient matrix as a key input parameter into the control structure of the servo controller. It divides the redundant control dimension space in the servo controller and constructs the coupling response matrix. Combining the online coupling model and parasitic current interference characteristics, it dynamically corrects and compensates for the current loop proportional-integral control parameters of each servo axis servo driver.

[0098] The servo motor-based robotic arm collaborative control method provided in this embodiment of the invention is implemented through the aforementioned servo motor-based robotic arm collaborative control system. For details of the specific methods and processes of the servo motor-based robotic arm collaborative control system, please refer to the embodiments of the servo motor-based robotic arm collaborative control method described above, which will not be repeated here.

[0099] The foregoing has only described certain exemplary embodiments of the present invention by way of illustration. Undoubtedly, those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the foregoing drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A servo motor-based robotic arm collaborative control method, characterized in that, Includes the following steps: S100. Inject frequency-adjustable reference excitation signals into the ground wires of the servo drivers of each servo axis in the manipulator to construct a coupled observation basic sequence. S200: While injecting the reference excitation signal, the instantaneous ground potential of each servo axis servo driver is acquired in parallel and synchronously. Based on the coupled observation basic sequence, a complete multi-axis potential response matrix is ​​constructed. S300: Perform sparse cross-correlation operation on the multi-axis potential response matrix, extract the cross-axis potential correlation spectrum between each servo axis, and remove the self-excited response components caused by each reference excitation signal source itself. S400. Based on the extracted cross-axis potential correlation spectrum, an adaptive Kalman filter structure is constructed, and real-time iteration is performed in combination with historical response data to dynamically infer the potential coupling gain vector between each servo axis. S500: Using the ratio of the potential coupling gain vector to the amplitude of the corresponding reference excitation signal as a basis, calculate the potential coupling coefficient matrix between each servo axis, and establish an online coupling model reflecting the current operating state based on the potential coupling coefficient matrix. S600 introduces the potential coupling coefficient matrix as a key input parameter into the control structure of the servo controller, divides the redundant control dimension space in the servo controller, constructs the coupling response matrix, and combines the online coupling model and parasitic current interference characteristics to dynamically correct and compensate the current loop proportional-integral control parameters of each servo axis servo driver.

2. The servo motor-based robotic arm collaborative control method according to claim 1, characterized in that, Step S100 specifically includes: Select a sinusoidal signal source with stable frequency and controllable amplitude, construct multiple reference excitation signal source channels, and connect them to the ground input terminal of the servo driver of each servo axis in the robot hand; After the signal connection is established, the excitation control sequence is started based on the preset periodic excitation mode, so that each servo axis ground line receives the reference signal of its own frequency in a specified timing sequence. During the excitation operation, the instantaneous potential of the ground wire of each servo axis is acquired in real time and in parallel through a high-resolution data acquisition device, and bandpass filtering and synchronous sampling are performed. By combining excitation frequency and time window, the collected data are synchronously analyzed and classified to form a coupled observation basic sequence with time label and frequency attributes.

3. The servo motor-based robotic arm collaborative control method according to claim 2, characterized in that, Step S200 specifically includes: A multi-channel synchronous sampling device is configured to collect ground potential of each servo axis in parallel, and a unified time base is set to achieve synchronous sampling. The collected potential sequences are aligned according to timestamp frames to construct the original observation data matrix, and the spectrum is divided and the data is labeled according to the excitation frequency and axis number. Frequency domain analysis is used to identify the response components of each servo axis to the excitation frequency, and the effective response data is selected. The effective response data is then remapped into a structured multi-axis potential response matrix according to the excitation source axis and the response axis, reflecting the potential coupling strength and distribution characteristics.

4. The servo motor-based robotic arm collaborative control method according to claim 1, characterized in that, Step S300 specifically includes: For each excitation source axis-response axis combination in the structural multi-axis potential response matrix, the response potential sequence is extracted, and cross-correlation analysis is performed with the corresponding excitation signal as a reference to obtain the initial cross-correlation spectrum. A sparsity constraint mechanism is introduced to filter the cross-correlation spectrum, retaining only the valid correlation components that are above a set response intensity threshold; Self-excited response components on the main diagonal are identified and removed, and cross-correlation data containing only cross-axis interference paths are retained. The processed data is then used to construct a purified cross-axis potential response relationship spectrum for subsequent coupling modeling and control input.

5. The servo motor-based robotic arm collaborative control method according to claim 1, characterized in that, Step S400 specifically includes: Based on the purified transaxis potential response relationship spectrum, a state-space model is established with the potential response value as the observable and the coupling gain vector as the state variable. Based on the constructed state-space model, initialize the core parameters in the adaptive Kalman filter structure; By combining historical cross-axis potential response data, the state prediction and observation update iteration operation of the Kalman filter is performed to correct the coupling gain estimate in real time. After filtering and stabilization, the state output is extracted to form a potential coupling gain vector containing the combination of each excitation source-response axis, which is used as the input for subsequent control modeling.

6. The servo motor-based robotic arm collaborative control method according to claim 5, characterized in that, Step S500 specifically includes: Extract the potential coupling gain vector and normalize it with the amplitude of the corresponding reference excitation signal to obtain the coupling response value under unit excitation; The normalized response values ​​are reorganized into a two-dimensional matrix to construct the potential coupling coefficient matrix between each servo axis. Time-domain fusion is performed on the potential coupling coefficient matrix of multiple consecutive cycles to generate an online coupling model that can be updated over time. The online coupling model is then embedded into the servo control process as the input basis for control parameter correction, thereby achieving adaptive adjustment of disturbances.

7. The servo motor-based robotic arm collaborative control method according to claim 6, characterized in that, Step S600 specifically includes: The potential coupling coefficient matrix is ​​introduced into the servo controller control structure, and the cross-axis coupling influence value corresponding to each servo axis is extracted. Based on the coupling influence value, a redundant control dimension is constructed, a coupling response matrix is ​​generated, and it is mapped to the control space. The proportional-integral parameters of the current loop are dynamically corrected based on the coupling response matrix and parasitic disturbance characteristics. The correction parameters are loaded into the servo driver, and the update frequency and compensation magnitude are set according to the coupling change trend to achieve adaptive parameter adjustment and stability compensation.

8. A servo motor-based robotic arm collaborative control system, used to implement the servo motor-based robotic arm collaborative control method according to any one of claims 1-7, characterized in that, It includes a coupling excitation injection module, a synchronous potential acquisition module, a potential cross-correlation analysis module, a coupling gain estimation module, an online coupling modeling module, and a parameter adaptive compensation module; The coupled excitation injection module injects frequency-adjustable reference excitation signals into the ground wires of the servo drivers of each servo axis in the robot, thereby constructing a coupled observation basic sequence. The synchronous potential acquisition module performs parallel synchronous acquisition of the instantaneous ground potential of each servo axis servo driver while injecting the reference excitation signal, and constructs a complete multi-axis potential response matrix based on the coupled observation basic sequence. The potential cross-correlation analysis module performs sparse cross-correlation operations on the multi-axis potential response matrix, extracts the cross-axis potential correlation spectrum between each servo axis, and removes the self-excited response components caused by each reference excitation signal source itself. The coupling gain estimation module constructs an adaptive Kalman filter structure based on the extracted cross-axis potential correlation spectrum, and performs real-time iteration in combination with historical response data to dynamically infer the potential coupling gain vector between each servo axis. The online coupling modeling module uses the ratio of the potential coupling gain vector to the amplitude of the corresponding reference excitation signal as a basis to calculate the potential coupling coefficient matrix between each servo axis, and establishes an online coupling model that reflects the current operating state based on the potential coupling coefficient matrix. The parameter adaptive compensation module introduces the potential coupling coefficient matrix as a key input parameter into the control structure of the servo controller. It divides the redundant control dimension space in the servo controller and constructs the coupling response matrix. Combining the online coupling model and parasitic current interference characteristics, it dynamically corrects and compensates for the current loop proportional-integral control parameters of each servo axis servo driver.

Images