A multi-motor synchronous positioning system and method

Abstract

The application relates to the technical field of industrial control, and discloses a multi-motor synchronous positioning system and method, which comprises the following steps: establishing a synchronous positioning reference relationship between a reference motor and a driven motor; determining a dynamic threshold to generate an event trigger flag when a displacement deviation exceeds a boundary; calculating a buffer correction amount by using visual observation data and mechanical clearance to correct a position instruction to be executed and generate a local control amount; determining a deceleration starting moment by fusing vision and motion state; generating a feedforward compensation amount by an updated feedforward network supervised by an actual residual error; generating a nonlinear enhanced error signal based on a position error after the system execution time reaches the deceleration starting moment; and uniformly limiting the local control amount, the event trigger gated position correction amount, the feedforward compensation amount and the enhanced error signal to generate a unique comprehensive control input. The application effectively reduces communication burden, improves end fine-tuning precision, and realizes high-precision multi-motor synchronous positioning.

Description

Technical Field

[0001] This invention relates to the field of industrial control technology, and in particular to a multi-motor synchronous positioning system and method. Background Technology

[0002] Multi-motor synchronous positioning is used in mobile platforms, synchronous conveyor platforms, gantry-type moving objects, and multi-axis positioning equipment. These applications require multiple motors to maintain position synchronization along associated trajectories and to dock at the target position after operation. Existing solutions typically employ master-slave axis following, cross-coupling control, periodic synchronous communication, industrial Ethernet synchronous interpolation, or communication delay compensation to achieve multi-axis synchronization.

[0003] Existing multi-motor synchronization and multi-axis motion control can improve trajectory consistency by utilizing synchronization error, disturbance estimation, and trajectory interpolation. However, when the outer loop consistently sends correction messages at a fixed high-frequency cycle, it increases the burden on the main controller and communication link; if the visual results are only used for post-docking verification, they cannot correct position commands that have not yet been issued in advance; if the feedforward model does not have actual residual supervision and freeze conditions, it may reuse unverified prediction results for parameter updates; and when multiple processing stages output control commands separately, the final action may not be unique.

[0004] In the prior art, Chinese patent document CN110705034B discloses an event-triggered position tracking control method for permanent magnet synchronous motors. This method introduces an event-triggered mechanism into single-motor position tracking control to save network resources, but it does not address visual prediction buffer correction, actual residual supervision feedforward compensation, or unified arbitration of multiple control quantities in multi-motor synchronous positioning. Chinese patent document CN107229291B discloses a synchronization method for an online visual positioning system, which achieves time synchronization between the servo system and the visual positioning system through camera trigger signals, but it does not address event-triggered correction or the synthesis of a unique integrated control input. Chinese patent document CN106773683B discloses a synchronization control method and device for dual-side drive motors, which plans a position-given curve using a virtual spindle and calculates a speed / torque feedforward curve, but it does not address visual buffer correction, model updates for actual residual supervision, or anomaly freezing strategies. None of the above solutions organize synchronization reference, event triggering, visual buffer correction, feedforward compensation, end-effector deceleration, unified arbitration, and anomaly freezing into a continuous chain of methods with a defined input-output order.

[0005] Therefore, a synchronous positioning method that does not depend on a specific motor mounting location, connection structure, or motion mechanism construction is needed. Summary of the Invention

[0006] The present invention aims to at least solve one of the technical problems existing in related technologies. To this end, the present invention provides a multi-motor synchronous positioning system and method.

[0007] The first technical solution provided by this invention is: A multi-motor synchronous positioning method includes: The system acquires the positioning object identifier, target position, and position command to be executed for the multi-motor system, and establishes a synchronous positioning reference relationship between the reference motor and the driven motor based on the position reference system and timestamp. The current load status, reference motor speed, and operating condition stability are obtained to determine a dynamic threshold. The displacement of the driven motor in the synchronous positioning reference relationship is compared with the displacement of the reference motor to generate a displacement deviation. When the absolute value of the displacement deviation is greater than the dynamic threshold, a true event trigger flag is generated. The measured and predicted positions within the effective visual time window are obtained. The buffer correction amount is calculated by combining the mechanical clearance estimate. The buffer correction amount is used to correct the position command to be executed to generate the input error. The input error is then input to the local closed-loop controller to generate the local control quantity. The position and velocity estimates are obtained by fusing the visually measured position and the motion state of the previous stage, and the deceleration start time is determined by combining the allowable deceleration and the safety distance. The operating characteristics, including at least the dynamic threshold, the buffer correction amount, and the reference motor speed, are input into the feedforward network model to generate the feedforward compensation amount. When the system execution time reaches the deceleration start time, the position error between the target position and the real-time position is obtained; based on the position error, the preset enhancement coefficient and the nonlinear exponential parameter, an enhancement error signal is generated; The displacement deviation is gated based on the event trigger flag to generate a position correction amount; the local control amount, the position correction amount, the feedforward compensation amount, and the enhanced error signal are superimposed to form a candidate synthetic control amount; the candidate synthetic control amount is uniformly limited to generate a unique comprehensive control input, which is then output to the motor control execution carrier.

[0008] Furthermore, the step of establishing the synchronous positioning reference relationship between the reference motor and the driven motor includes: The displacement of each motor, the visual positioning results, and the position command to be executed are converted to the same displacement reference and aligned according to the timestamp. Determine whether the communication delay is within a preset communication compensation window; when the communication delay exceeds the communication compensation window, it is determined to be a communication anomaly, triggering control freeze and stopping the execution of subsequent calculation steps.

[0009] Further, the step of determining the dynamic threshold includes: based on an initial threshold benchmark, superimposing the load state response term, the velocity change response term, and the stability response term to obtain the dynamic threshold; the calculation formula is as follows: In the formula, For dynamic thresholds; Used as the initial threshold benchmark; As the load threshold weight; This refers to the load status quantity; Weights are used for speed thresholds; The reference motor speed; As a stability threshold weight; For operating condition stability.

[0010] Further, the step of correcting the input error generated by the position instruction to be executed using the buffer correction amount includes: The difference between the predicted position and the visually measured position, minus the estimated mechanical clearance, and then multiplied by the buffer compensation coefficient, yields the buffer correction amount; the calculation formula is as follows: In the formula, This is the buffer correction amount; This is the buffer compensation coefficient; Predicting location based on visual perception; The location is determined by visual measurement. For estimating mechanical clearance; The input error is generated by subtracting the actual feedback position of the driven motor from the position command to be executed after being corrected by the buffer correction amount.

[0011] Further, the step of determining the deceleration start time includes: traversing all candidate times corresponding to the current position, and selecting the earliest time when the difference between the target position and the position estimate first meets the braking distance requirement as the deceleration start time; the braking distance determination formula is as follows: In the formula, This is the moment when deceleration begins; Candidate moments; For the target location; For location estimation; For speed estimation; To allow for deceleration; For a safe distance.

[0012] Further, the step of inputting the operating characteristics, including at least the dynamic threshold, the buffer correction amount, and the reference motor speed, into the feedforward network model includes: A three-layer feedforward network is constructed as the feedforward network model; Collect the reference motor motion parameters, current load state, historical synchronization deviation of the driven motor, the buffer correction amount, the dynamic threshold, and the operating condition stability of the current control cycle, and splice them to form the feature vector of the operating characteristics; The feature vector is input into the three-layer feedforward network, and after hidden layer activation and output layer weight mapping, the feedforward compensation amount is output. The feedforward network model uses actual residual supervision for weight updates; the model update steps include: Obtain the actual synchronization deviation at the end of the current control cycle; The final steady-state error after the motor stops is obtained as the actual positioning residual; Obtain the feedforward change; The supervision loss is constructed using the square of the actual synchronization deviation, the square of the actual positioning residual, and the square of the feedforward change, and the weight gradient is calculated based on the supervision loss to update the weight parameters of the feedforward network model.

[0013] Furthermore, when calculating the weight gradient based on the supervised loss to update the weight parameters of the feedforward network model, the following boundary constraints are applied: Calculate the candidate weights obtained from gradient descent; A symmetric saturation constraint is applied to the candidate weight values: when the magnitude of the candidate weight value is greater than the preset upper limit of the weight magnitude, the corresponding positive or negative sign is maintained and the upper limit of the weight magnitude is taken as the updated weight; When the actual positioning residual exceeds the limit or communication is abnormal, the previous valid weight parameter is frozen.

[0014] Furthermore, the rule for generating the enhanced error signal is as follows: retain the sign direction of the position error, and add the absolute value of the position error to the nonlinear enhancement term; the nonlinear enhancement term is obtained by multiplying the enhancement coefficient by the nonlinear exponent parameter power of the absolute value of the position error; the calculation formula is as follows: In the formula, To enhance the error signal; This refers to the positional error; For enhancement coefficient; For nonlinear exponential parameters; sgn( ) is a symbolic function.

[0015] Further, the steps of gating the displacement deviation based on the event trigger flag to generate a position correction quantity, and superimposing them to form a candidate synthetic control quantity, include: When the event trigger flag is true, the product of the preset event correction gain and the displacement deviation is taken as the position correction amount; when the event trigger flag is false, the position correction amount is zero. The candidate synthetic control quantity is obtained by superimposing the product of the local control quantity, the position correction quantity, the feedforward compensation quantity, and the preset enhancement error gain with the enhancement error signal. The candidate synthesized control variables are subjected to symmetric saturation processing based on the upper limit of the amplitude, and the output is the unique synthesized control input that is restricted within the safety boundary; the calculation formula is as follows: In the formula, This is the integrated control input for the i-th driven motor; and These are local control variables and feedforward compensation variables, respectively. This serves as a flag for the event to be triggered. and These are event correction gain and error enhancement gain, respectively. and These are displacement deviation and enhanced error signals, respectively. To comprehensively control the upper limit of the input amplitude; This is a symmetric saturation processing function.

[0016] The second technical solution provided by this invention is as follows: A multi-motor synchronous positioning system for performing a multi-motor synchronous positioning method includes: The reference relationship establishment module is used to obtain the positioning object identifier, target position, and position command to be executed of the multi-motor system, and establish a synchronous positioning reference relationship between the reference motor and the driven motor based on the position reference system and timestamp; The threshold and deviation processing module is used to obtain the current load state quantity, the reference motor speed and the operating condition stability to determine the dynamic threshold, compare the displacement of the driven motor with the displacement of the reference motor in the synchronous positioning reference relationship to generate a displacement deviation, and generate a true event trigger flag when the absolute value of the displacement deviation is greater than the dynamic threshold. The vision pre-correction module is used to obtain the visual measured position and predicted position within the effective visual time window, calculate the buffer correction amount in combination with the mechanical clearance estimate, use the buffer correction amount to correct the position command to be executed to generate input error, and input the input error to the local closed-loop controller to generate local control quantity; The deceleration time planning module is used to fuse the visually measured position and the motion state of the previous stage to obtain position and velocity estimates, and to determine the deceleration start time by combining the allowable deceleration and the safety distance. The feedforward adaptive compensation module is used to input operating characteristics, including at least the dynamic threshold, the buffer correction amount, and the reference motor speed, into the feedforward network model to generate the feedforward compensation amount. The nonlinear enhancement module is used to obtain the position error between the target position and the real-time position when the system execution time reaches the deceleration start time, and generate an enhancement error signal based on the position error, a preset enhancement coefficient and a nonlinear exponent parameter. The integrated arbitration control module is used to perform gating processing on the displacement deviation based on the event trigger flag to generate a position correction amount; to superimpose the local control amount, the position correction amount, the feedforward compensation amount, and the enhanced error signal to form a candidate synthetic control amount; to perform unified amplitude limiting on the candidate synthetic control amount to generate a unique integrated control input; and to output it to the motor control execution carrier.

[0017] The above-described one or more technical solutions in the embodiments of the present invention have at least one of the following technical effects: 1. By introducing an event triggering mechanism based on dynamic thresholds, it is possible to eliminate the need to send correction messages when the outer loop deviation is small, which greatly saves network resources and reduces the communication burden on the main controller. At the same time, a nonlinear enhanced error signal is introduced at the end to ensure that the system can still generate sufficient correction instructions within a small error range, prevent the signal attenuation at the end, and achieve high-precision fine-tuning.

[0018] 2. Unlike traditional vision results which are only used for post-docking verification, this solution generates buffer correction amounts through vision data and mechanical clearance prediction results. Before the control commands are issued, the position commands that have not yet been issued in the buffer are corrected in advance and digested in the underlying closed loop, avoiding the following error caused by time delay.

[0019] 3. The feedforward compensation network forces the use of the actual synchronization deviation and the final positioning residual after operation for supervised updates, avoiding the repeated use of unverified prediction results for model training; and implements a strict model freezing strategy when communication is abnormal or residuals go out of bounds, effectively suppressing the amplification of accumulated errors.

[0020] 4. Faced with the outputs of multiple control dimensions such as local closed-loop control, event-triggered deviation compensation, feedforward compensation, and nonlinear enhancement, this solution performs safety arbitration through a unified amplitude limiting equation, ensuring that the motor actuator can receive the "unique integrated control input" within the safety boundary, thus eliminating the hidden dangers of collision and non-unique action quantities when multiple loops are jointly controlled.

[0021] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0023] Figure 1 This is a flowchart of a multi-motor synchronous positioning method according to an embodiment of the present invention; Figure 2 This is a comparison diagram of the evolution of enhanced error signals in an embodiment of the present invention. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention. The following embodiments are used to illustrate this invention but cannot be used to limit the scope of this invention.

[0025] The execution sequence of the method of the present invention is illustrated by an embodiment. The position data acquisition carrier, vision data acquisition carrier, synchronous communication carrier, controller, and motor control execution carrier are only used to provide input, perform calculations, transmit data, and implement control outputs; this embodiment does not consider the installation position, connection method, or mechanical structure of these carriers as technical improvements.

[0026] In this embodiment, the position data acquisition carrier is, for example, a servo motor's built-in encoder or an external grating ruler, used to output the displacement data of each motor; the vision data acquisition carrier is, for example, an industrial camera, used to capture images of the positioning object within an effective visual time window; the synchronous communication carrier is, for example, an industrial Ethernet bus supporting timestamps, used to transmit timestamped data and control commands; the controller is, for example, a programmable motion controller or an industrial PC, used to execute the various calculation steps of this embodiment; and the motor control execution carrier is, for example, a servo driver, used to execute comprehensive control inputs. The above hardware implementation is merely an example and does not constitute a limitation on the method of this invention.

[0027] like Figure 1 The diagram illustrates the workflow of a multi-motor synchronous positioning method according to the present invention. The implementation process of this method will be described in detail below.

[0028] Before synchronous positioning begins, the controller reads the positioning object identifier, target position, allowable synchronization deviation, effective visual time window, communication compensation window, drive safety input boundary, allowable deceleration range, and safety distance range to determine one reference motor and at least one driven motor. The controller converts the displacement of each motor, the visual positioning results, and the position commands to be executed in the buffer to the same displacement reference, and aligns them according to the object identifier and timestamp to form a synchronous positioning reference relationship. If the object identifier is inconsistent, the displacement reference is not converted, the difference in sampling timestamps exceeds the allowable alignment error, or the necessary control boundary is missing, the formal deviation calculation will not begin, the previous effective control state will be maintained, and data will be re-acquired or re-aligned.

[0029] The communication compensation window is used to define the upper limit of acceptable communication delay. When the communication delay is within the communication compensation window, the controller completes data alignment based on the timestamp and allows participation in subsequent synchronization and positioning calculations. When the communication delay exceeds the communication compensation window, it is determined to be a communication anomaly and a freeze is triggered.

[0030] After the synchronous positioning reference relationship is confirmed to be valid, the local control carrier of each motor executes current closed-loop and speed closed-loop. The controller reads the displacement of the reference motor and the displacement of the i-th driven motor, and calculates the displacement deviation and trigger flag according to formula (1).

[0031] (1) in: For the slave motor index, For controlling the periodic index; Let be the displacement deviation of the i-th driven motor relative to the reference motor, with the dimension of mm; Let be the displacement of the i-th driven motor, with dimensions in mm; The displacement of the reference motor is measured in mm. This is a dimensionless event trigger flag. The threshold value is a dynamic threshold with units of mm. Obtained by sampling the position of the driven motor; Obtained by sampling the position of the reference motor; This is an indicator function. It takes the value 1 when the condition in parentheses is true (i.e., the absolute value of the displacement deviation is greater than the dynamic threshold), and 0 otherwise.

[0032] The trigger flag in equation (1) is used for gating non-periodic position compensation and the event correction term in equation (8).

[0033] When the trigger flag is zero, the position correction amount is zero, and no new outer loop position compensation command is sent; when the trigger flag is one, a position compensation command with the target execution timestamp is generated. If the sampling timestamps do not match or the communication delay exceeds the communication compensation window, the result of this trigger is not used for model update, the controller maintains local feedback and freezes the event correction item.

[0034] Before comparing displacement deviations, the controller reads the current load, reference motor speed and operating condition stability, and calculates the dynamic threshold according to formula (2).

[0035] (2) In the formula: The threshold value is a dynamic threshold with units of mm. The initial threshold reference is in mm. The load threshold weight is measured in mm. This is a load state quantity, dimensionless; The velocity threshold weight has the dimension s; The reference motor speed is measured in mm / s. The stability threshold weight is in mm. The stability under operating conditions is dimensionless.

[0036] It is obtained by calibrating the end-point allowable synchronization error and the high-speed safe synchronization error; Obtained by calibration under no-load and rated load conditions; It is determined by the current effective load relative to the rated load calibration range; Obtained from the calibration of speed variation and safety synchronization error; Estimated from the speed of the reference motor; Obtained from the operating condition stability calibration; It is determined by the effective state variables of operating resistance, temperature, vibration, and historical synchronization deviation.

[0037] Load state parameters are obtained from motor current observation, torque estimation, or external force sensors; operating stability is determined by the statistical variance of running resistance, temperature, vibration sensors, and historical synchronization deviation.

[0038] When the load or operating condition is invalid, the synchronization deviation changes abruptly, or the communication is abnormal, the previous effective threshold is maintained.

[0039] The controller maintains a timestamped position command buffer. The visual data acquisition carrier acquires an image of the positioning object within the effective visual time window. The controller calculates the visually measured position and real-time velocity from the image, and predicts the position of the target at the sampling time based on the current position, velocity, and processing delay. The controller calculates the buffer correction amount for the i-th slave motor according to equation (3).

[0040] (3) In the formula: This is the buffer correction amount, with units of mm; This is the buffer compensation coefficient, which is dimensionless. The location is predicted visually, with units in mm; The position is determined visually, with units of mm. This is an estimate of mechanical clearance, with the dimension mm.

[0041] It is obtained by calibration of visual delay, communication delay, and mechanical clearance; It is predicted based on the current visual position, velocity, and processing latency; Obtained by visual positioning calculation; It is identified by commutation error and low-speed positioning residual.

[0042] Used to correct position commands that have not yet been issued.

[0043] The output of equation (3) is only used to correct position commands that have not yet been issued and whose timestamps are later than the current execution time. The corrected target position command is subtracted from the current actual feedback position of the motor to generate the local position error. This error is input to the local position, speed, and current cascaded closed-loop controller, and its final calculation result is reflected in the local control quantity of equation (8). In this context, instructions that have already been executed are not modified.

[0044] The mechanical clearance estimate is identified from historical commutation errors, low-speed positioning residuals, or current effective following errors, and is updated after hysteresis confirmation following commutation. In cases of visual frame drops, insufficient visual confidence, or visual processing delays exceeding permissible limits, the previous stable buffer correction is maintained, and the current visual window is not used to update the mechanical clearance estimate.

[0045] The controller uses a Kalman filter or equivalent state estimator to fuse visual position observations and the motion state of the previous stage to obtain position and velocity estimates, and determines the deceleration start time according to equation (4). The common displacement reference of the reference motor and the driven motor, the time-stamped deviation calculation and trigger gating are respectively connected to the retrieved multi-motor synchronization, visual position correction and unified time reference path; this embodiment further limits these inputs to the continuous consumption relationship of subsequent buffer correction, deceleration determination, feedforward compensation and unified amplitude limiting.

[0046] (4) In the formula: The moment when deceleration begins, with dimensions in seconds; Let be a candidate time point, with dimensions s; The target location is in mm. This is a location estimate, with units in mm; For velocity estimation, the dimension is mm / s; To allow for deceleration, the dimension is mm / s. 2 ; For safety distance, the unit is mm.

[0047] The candidate motion time corresponding to the current position is discretely traversed according to the control cycle; Obtained from the target input of the location task; It is obtained by fusing visual observation with the motion state of the previous stage; It is obtained by fusing visual observation with the motion state of the previous stage; It is determined by the load, driving capability, and friction state; It is determined by processing delay, communication delay, and mechanical backlash margin.

[0048] The allowable deceleration is determined within the safe deceleration range based on the current effective load, driving capability, and operating friction state; the safe distance covers the remaining distance error caused by visual processing delay, communication delay, and mechanical clearance. Equation (4) selects the earliest moment that meets the braking distance condition and outputs the deceleration trigger to the subsequent end control steps. When the visual state is invalid, the new prediction result is frozen, and the previous effective deceleration trigger or conservative early deceleration is adopted.

[0049] The controller calls the lightweight feedforward model. This model is a three-layer feedforward network. The input features include the reference motor position, speed and acceleration, current load, historical synchronization deviation of the driven motor, buffer correction state, dynamic threshold state and operating condition stability, and the feedforward compensation amount of the i-th driven motor is obtained according to Equation (5).

[0050] (5) In the formula: It represents the feedforward compensation amount of the i-th driven motor in the k-th control cycle, with the dimension being the control input unit; This is the output layer weight matrix, with dimensions matching the dimensions of the adaptive control input. It is an activation function with dimensionless dimensions; is the input layer weight matrix, with dimensions matching the feature dimensions; Let be the input feature vector of the i-th driven motor in the k-th control cycle, with dimensions of mixed feature quantity; For hidden layer bias, the dimensions are adapted to the hidden layer dimensions; This is the output layer bias, and the unit is the control input unit.

[0051] Obtained by updating the actual residuals; It is obtained from a three-layer feedforward network structure; Obtained by updating the actual residuals; It is obtained by splicing together the reference motion parameters, load, historical deviation, buffer correction, threshold and stability of the current control cycle; Obtained by updating the actual residuals; It is obtained by updating the actual residuals.

[0052] The feedforward compensation amount in equation (5) is not issued independently, but is entered into the comprehensive control input of equation (8).

[0053] The model parameters are updated only with the supervision loss consisting of the actual synchronization deviation, the actual positioning residual, and the feedforward change, and are updated according to Equation (6).

[0054] (6) In the formula: The updated weights for the l-th layer are represented by the corresponding weight units. For the first Layer Weights (gradient terms) before periodic update In That is, the current weight), and the units are all the corresponding weight units; The learning rate is expressed in units of the corresponding weights. 2 / mm 2 ; For monitoring losses, the dimension is mm. 2 ; This represents the upper limit of the magnitude of the candidate weight values, with the dimension being the corresponding weight dimension. and These are the actual synchronization deviation and the actual positioning residual, respectively, both in mm. The actual synchronization deviation used for monitoring updates is directly taken from the displacement deviation at the end of the control cycle; the actual positioning residual is defined as the final steady-state error after the motor stops, which is different from the transient position error during operation. To determine the residual weights, they are dimensionless; The weights are for feedforward smoothing and have the dimension in mm. 2 / (Control input unit) 2 ; It represents the feedforward change, with dimensions in the control input unit.

[0055] It is formed by actual synchronization deviation, positioning residual, and feedforward variation. and Calibrated by end-positioning error and drive smoothness; Calibration is performed based on actual residual monitoring training. It is calibrated by model stability and upper limit of feedforward mutation.

[0056] sat(v;B) represents symmetric saturation of the candidate quantity v: take B when v is greater than B, take v when -B is not greater than v and v is not greater than B, take -B when v is less than -B; B is positive and has the same dimension as v.

[0057] In equation (6), each loss term is mm. 2 gradient term multiplied by It was later restored to the weight unit.

[0058] After saturation, the candidate weight values ​​do not exceed the positive and negative amplitude boundaries, and the corresponding signs are maintained when the boundaries are reached. As online model parameters of Equation (5); the weights after saturation constraint are used to prevent the weight magnitude from continuing to expand in the abnormal window.

[0059] The supervision loss in Equation (6) consists of the square of the actual synchronization deviation, the square of the actual positioning residual, and the square of the feedforward change after dimensional conversion. The weight update thread has a lower priority than the local inner loop and only uses the model online at the control boundary. The candidate weights obtained from the gradient update are used as model parameters in Equation (5) after being constrained by symmetry. The model's own output cannot be used as training labels. When the actual synchronization deviation or positioning residual exceeds the limit, the visual confidence is insufficient, communication is abnormal, the comprehensive control quantity is saturated, or the supervision loss increases continuously, the previous effective model parameters are frozen and the feedforward compensation change is limited.

[0060] When approaching the target position (i.e., when the system execution time has passed the deceleration start time and the absolute value of the position error is less than the preset end fine-tuning threshold), the controller obtains the position error based on the difference between the target position and the real-time position, and generates an enhanced error signal according to formula (7).

[0061] (7) in: To enhance the error signal, the dimension is mm; The positional error is expressed in mm. To enhance the coefficient, the dimension is... ; This is a nonlinear exponential parameter, dimensionless; sgn( ) is a symbolic function, dimensionless.

[0062] The difference between the target position and the real-time position is used to characterize the trajectory tracking error of a single axis, distinguishing it from the displacement deviation between multiple axes. ; The end-positioning error response and oscillation boundary calibration are used to determine the position. It is determined by the end response sensitivity and oscillation risk.

[0063] sgn( )exist The value is 1 when it is greater than zero, 0 when it is equal to zero, and -1 when it is less than zero. Therefore, the error direction is preserved and the output is zero when there is zero error.

[0064] The enhanced error term and end fine-tuning of Equation (8) are used; the sign function makes the control direction of this term consistent with the position error direction.

[0065] In equation (7), the position error is the signed difference between the target position and the real-time position. The enhancement coefficient is determined by the end-positioning error response, and the nonlinear exponent is greater than zero and does not exceed one. The sign function takes a positive value when the position error is positive, a zero value when it is zero, and a negative value when it is negative, thus preserving the position error direction and making the enhancement error signal zero when the error is zero. The enhancement error signal and the deceleration trigger in equation (4) are used together to form the end-positioning fine-tuning candidate quantity; when end-positioning oscillation or reverse overshoot occurs, the enhancement coefficient is reduced, and when the oscillation boundary is reached, the enhancement term is frozen and the local feedback and verified event correction are maintained.

[0066] Under the same target position, real-time position, nonlinear exponent, and driving safety input boundary, the relationship between position error and enhancement error signal is calculated according to equation (7). The method of this invention retains the enhancement coefficient and nonlinear term; the nonlinear term corresponding to the enhancement coefficient is removed by comparing the path, so that the enhancement error signal is equal to the position error, while keeping other inputs and boundaries consistent.

[0067] When the system execution time reaches the deceleration start time When the local control quantity is triggered to enter the deceleration mode, the controller will form a candidate synthetic control quantity from the local control quantity, the position correction quantity gated by the trigger flag, the feedforward compensation quantity and the enhanced error signal, and then perform unified arbitration and amplitude limiting according to formula (8).

[0068] (8) In the formula: This is the integrated control input for the i-th driven motor, with the dimension being the control input unit; and These are the local control quantity and the feedforward compensation quantity, respectively, both in units of control input. The local control quantity... The input error of the closed loop, obtained from the local current-speed closed loop, is corrected by the buffer. The corrected target position command is generated by subtracting the current actual feedback position of the motor. This is the event trigger flag, which takes the value 0 or 1 and is dimensionless. and These are the event correction gain and the enhanced error gain, respectively, both measured in control input units of mm. and All dimensions are in mm; The upper limit of the comprehensive control input is defined in units of control input.

[0069] The threshold is obtained from equation (1). Difference is calculated from the location samples. and Certified by the corresponding response, Determined by the drive safety input boundary; This is a symmetric saturation processing function.

[0070] sat(v; Symmetrical saturation is applied to the candidate integrated control input v: v is greater than Time to take , - Not greater than v and v is not greater than v When v is less than - Time take- .

[0071] If positive, all terms in the candidate quantities before saturation are control input units; after saturation... lie in- to Between these points, the sign of the candidate quantity is preserved and the amplitude is limited when the boundary is reached.

[0072] As the sole external integrated control input for motor control execution; When the event correction term is zero, the event correction term is zero after saturation. It is directly sent to the driver execution stage.

[0073] The output of equation (8) is the only external integrated control input received by the i-th driven motor in the current control phase. No position correction is introduced when the trigger flag is zero; the previous effective feedforward compensation or zero feedforward is used when the feedforward model is frozen; no new enhancement is introduced when the enhancement term is frozen; the positive limit is used when the candidate synthetic control value exceeds the positive limit, the negative limit is used when it is below the negative limit, and the candidate value is maintained when it is between the two. The displacement, synchronization deviation, positioning residual, and control saturation state after the integrated control input is issued are written into the subsequent state record.

[0074] When the reference motor needs to be replaced during the motion phase, the controller first freezes the event trigger correction, buffer correction and model update, redetermines the new set of reference motors and driven motors, and re-executes the alignment of object identification, displacement reference and sampling timestamp; only after the new synchronous positioning reference relationship is confirmed to be valid will the calculation and control output be restored.

[0075] The above method is carried out by a position data acquisition carrier, a vision data acquisition carrier, a synchronous communication carrier, a controller, and a motor control execution carrier. Each hardware carrier only provides the displacement, image, timestamp, and operating status required by this method, or performs a unique integrated control input. Its specific installation location, connection structure, and mechanical structure do not change the processing order, formula relationships, or abnormal branches of this method.

[0076] A multi-motor synchronous positioning system for performing a multi-motor synchronous positioning method includes: The reference relationship establishment module is used to obtain the positioning object identifier, target position, and position command to be executed of the multi-motor system, and establish a synchronous positioning reference relationship between the reference motor and the driven motor based on the position reference system and timestamp; The threshold and deviation processing module is used to obtain the current load state quantity, the reference motor speed and the operating condition stability to determine the dynamic threshold, compare the displacement of the driven motor with the displacement of the reference motor in the synchronous positioning reference relationship to generate a displacement deviation, and generate a true event trigger flag when the absolute value of the displacement deviation is greater than the dynamic threshold. The vision pre-correction module is used to obtain the visual measured position and predicted position within the effective visual time window, calculate the buffer correction amount in combination with the mechanical clearance estimate, use the buffer correction amount to correct the position command to be executed to generate input error, and input the input error to the local closed-loop controller to generate local control quantity; The deceleration time planning module is used to fuse the visually measured position and the motion state of the previous stage to obtain position and velocity estimates, and to determine the deceleration start time by combining the allowable deceleration and the safety distance. The feedforward adaptive compensation module is used to input operating characteristics, including at least the dynamic threshold, the buffer correction amount, and the reference motor speed, into the feedforward network model to generate the feedforward compensation amount. The nonlinear enhancement module is used to obtain the position error between the target position and the real-time position when the system execution time reaches the deceleration start time, and generate an enhancement error signal based on the position error, a preset enhancement coefficient and a nonlinear exponent parameter. The integrated arbitration control module is used to perform gating processing on the displacement deviation based on the event trigger flag to generate a position correction amount; to superimpose the local control amount, the position correction amount, the feedforward compensation amount, and the enhanced error signal to form a candidate synthetic control amount; to perform unified amplitude limiting on the candidate synthetic control amount to generate a unique integrated control input; and to output it to the motor control execution carrier.

[0077] The table below shows the comparison data of the enhanced error signal output of the method of the present invention and the comparative scheme under different positional error conditions. The common constraint is: enhancement coefficient. Nonlinear exponential parameters .

[0078] Table 1 Comparison of Enhanced Error Signal Evolution Data

[0079] Note: Both sets of data are verified under the same common input conditions and physical boundaries.

[0080] The difference between this invention and the comparative scheme in terms of method lies in that... The method of this invention (with retained enhancement term): This invention retains the nonlinear enhancement term based on the enhancement coefficient and the nonlinear exponent. This allows the multi-motor positioning system to still generate a larger, controllable end-point correction signal within a small position error range (typically constrained by oscillation boundaries and integrated control input amplitude limits). This mechanism ensures that the system still has a high response gain to eliminate small errors when approaching the target position.

[0081] Existing technology (enhancement term removal): The corresponding nonlinear enhancement term was removed from the comparison solution, leaving the system with only a linear error response. As can be seen from the comparison graphs and data, under this setting, the end-correction gain is significantly smaller, and the control command is prone to attenuation when the error is small, affecting the efficiency and accuracy of final docking and fine-tuning.

[0082] The methodological differences are directly reflected in the composition of the calculation formulas. The comparison of the calculation formulas for the two schemes is as follows: The method formula of this invention: The formula retains the nonlinear term. .in This is a sign function used to ensure that the control direction of this term is consistent with the direction of the position error.

[0083] The formula for the comparison scheme: The nonlinear term based on the enhancement coefficient was removed, so that the final output enhancement error signal is completely equivalent to the initial position error.

[0084] Figure 2 The enhancement error signal is given under the same input conditions and physical boundaries. With position error The comparison curves show the changes; the comparison scheme curves are those after removing the nonlinear enhancement terms based on the enhancement coefficient and nonlinear exponent, according to... Calculated.

[0085] After employing the method of this invention, an enhanced error signal is generated by combining the position error with an enhancement coefficient and a nonlinear exponent, and then frozen during oscillation or reverse overshoot. With position error The enhancement is increased and improved, while remaining controllable within the common physical boundary; whereas in the comparison scheme, this enhanced error signal... Only the linear response is retained, and the end-correction gain is small.

[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-motor synchronous positioning method, characterized in that, include: The system acquires the positioning object identifier, target position, and position command to be executed for the multi-motor system, and establishes a synchronous positioning reference relationship between the reference motor and the driven motor based on the position reference system and timestamp. The current load status, reference motor speed, and operating condition stability are obtained to determine the dynamic threshold, and the displacement of the driven motor in the synchronous positioning reference relationship is compared with the displacement of the reference motor to generate the displacement deviation; When the absolute value of the displacement deviation is greater than the dynamic threshold, a true event trigger flag is generated; The measured and predicted positions within the effective visual time window are obtained. The buffer correction amount is calculated by combining the mechanical clearance estimate. The buffer correction amount is used to correct the position command to be executed to generate the input error. The input error is then input to the local closed-loop controller to generate the local control quantity. The position and velocity estimates are obtained by fusing the visually measured position and the motion state of the previous stage, and the deceleration start time is determined by combining the allowable deceleration and the safety distance. The operating characteristics, including at least the dynamic threshold, the buffer correction amount, and the reference motor speed, are input into the feedforward network model to generate the feedforward compensation amount. When the system execution time reaches the deceleration start time, the position error between the target position and the real-time position is obtained; based on the position error, the preset enhancement coefficient and the nonlinear exponential parameter, an enhancement error signal is generated; The displacement deviation is gated based on the event trigger flag to generate a position correction amount; the local control amount, the position correction amount, the feedforward compensation amount, and the enhanced error signal are superimposed to form a candidate synthetic control amount; the candidate synthetic control amount is uniformly limited to generate a unique comprehensive control input, which is then output to the motor control execution carrier.

2. The multi-motor synchronous positioning method according to claim 1, characterized in that, The step of establishing the synchronous positioning reference relationship between the reference motor and the driven motor includes: The displacement of each motor, the visual positioning results, and the position command to be executed are converted to the same displacement reference and aligned according to the timestamp. Determine whether the communication delay is within a preset communication compensation window; when the communication delay exceeds the communication compensation window, it is determined to be a communication anomaly, triggering control freeze and stopping the execution of subsequent calculation steps.

3. The multi-motor synchronous positioning method according to claim 1, characterized in that, The step of determining the dynamic threshold includes: based on an initial threshold benchmark, superimposing the load state response term, velocity change response term, and stability response term to obtain the dynamic threshold; the calculation formula is as follows: In the formula, For dynamic thresholds; Used as the initial threshold benchmark; As the load threshold weight; This refers to the load status quantity; Weights are used for speed thresholds; The reference motor speed; As a stability threshold weight; For operating condition stability.

4. The multi-motor synchronous positioning method according to claim 1, characterized in that, The step of correcting the input error generated by the instruction to be executed at the specified position using the buffer correction amount includes: The difference between the predicted position and the visually measured position, minus the estimated mechanical clearance, and then multiplied by the buffer compensation coefficient, yields the buffer correction amount; the calculation formula is as follows: In the formula, This is the buffer correction amount; This is the buffer compensation coefficient; Predicting location based on visual perception; The location is determined by visual measurement. For estimating mechanical clearance; The input error is generated by subtracting the actual feedback position of the driven motor from the position command to be executed after being corrected by the buffer correction amount.

5. The multi-motor synchronous positioning method according to claim 1, characterized in that, The step of determining the deceleration start time includes: traversing all candidate times corresponding to the current position, and selecting the earliest time when the difference between the target position and the position estimate first meets the braking distance requirement as the deceleration start time; the braking distance determination formula is as follows: In the formula, This is the moment when deceleration begins; Candidate moments; For the target location; For location estimation; For speed estimation; To allow for deceleration; For a safe distance.

6. The multi-motor synchronous positioning method according to claim 1, characterized in that, The step of inputting operating characteristics, including at least the dynamic threshold, the buffer correction amount, and the reference motor speed, into the feedforward network model includes: A three-layer feedforward network is constructed as the feedforward network model; Collect the reference motor motion parameters, current load state, historical synchronization deviation of the driven motor, the buffer correction amount, the dynamic threshold, and the operating condition stability of the current control cycle, and splice them to form the feature vector of the operating characteristics; The feature vector is input into the three-layer feedforward network, and after hidden layer activation and output layer weight mapping, the feedforward compensation amount is output. The feedforward network model uses actual residual supervision for weight updates; the model update steps include: Obtain the actual synchronization deviation at the end of the current control cycle; The final steady-state error after the motor stops is obtained as the actual positioning residual; Obtain the feedforward change; The supervision loss is constructed using the square of the actual synchronization deviation, the square of the actual positioning residual, and the square of the feedforward change, and the weight gradient is calculated based on the supervision loss to update the weight parameters of the feedforward network model.

7. The multi-motor synchronous positioning method according to claim 6, characterized in that, When calculating the weight gradient based on the supervised loss to update the weight parameters of the feedforward network model, the following boundary constraints are applied: Calculate the candidate weights obtained from gradient descent; A symmetric saturation constraint is applied to the candidate weight values: when the magnitude of the candidate weight value is greater than the preset upper limit of the weight magnitude, the corresponding positive or negative sign is maintained and the upper limit of the weight magnitude is taken as the updated weight; When the actual positioning residual exceeds the limit or communication is abnormal, the previous valid weight parameter is frozen.

8. The multi-motor synchronous positioning method according to claim 1, characterized in that, The rule for generating the enhanced error signal is as follows: retain the sign direction of the position error, and add the absolute value of the position error to the nonlinear enhancement term; the nonlinear enhancement term is obtained by multiplying the enhancement coefficient by the nonlinear exponent parameter raised to the power of the absolute value of the position error; the calculation formula is as follows: In the formula, To enhance the error signal; This refers to the positional error; For enhancement coefficient; For nonlinear exponential parameters; sgn( ) is a symbolic function.

9. The multi-motor synchronous positioning method according to claim 1, characterized in that, The steps of gating the displacement deviation based on the event trigger flag to generate a position correction quantity, and superimposing them to form a candidate synthetic control quantity, include: When the event trigger flag is true, the product of the preset event correction gain and the displacement deviation is taken as the position correction amount; when the event trigger flag is false, the position correction amount is zero. The candidate synthetic control quantity is obtained by superimposing the product of the local control quantity, the position correction quantity, the feedforward compensation quantity, and the preset enhancement error gain with the enhancement error signal. The candidate synthesized control variables are subjected to symmetric saturation processing based on the upper limit of the amplitude, and the output is the unique synthesized control input that is restricted within the safety boundary; the calculation formula is as follows: In the formula, This is the integrated control input for the i-th driven motor; and These are local control variables and feedforward compensation variables, respectively. This serves as a flag for the event to be triggered. and These are event correction gain and error enhancement gain, respectively. and These are displacement deviation and enhanced error signals, respectively. To comprehensively control the upper limit of the input amplitude; This is a symmetric saturation processing function.

10. A multi-motor synchronous positioning system for performing the multi-motor synchronous positioning method as described in any one of claims 1-9, characterized in that, include: The reference relationship establishment module is used to obtain the positioning object identifier, target position, and position command to be executed of the multi-motor system, and establish a synchronous positioning reference relationship between the reference motor and the driven motor based on the position reference system and timestamp; The threshold and deviation processing module is used to obtain the current load state quantity, the reference motor speed and the operating condition stability to determine the dynamic threshold, compare the displacement of the driven motor with the displacement of the reference motor in the synchronous positioning reference relationship to generate a displacement deviation, and generate a true event trigger flag when the absolute value of the displacement deviation is greater than the dynamic threshold. The vision pre-correction module is used to obtain the visual measured position and predicted position within the effective visual time window, calculate the buffer correction amount in combination with the mechanical clearance estimate, use the buffer correction amount to correct the position command to be executed to generate input error, and input the input error to the local closed-loop controller to generate local control quantity; The deceleration time planning module is used to fuse the visually measured position and the motion state of the previous stage to obtain position and velocity estimates, and to determine the deceleration start time by combining the allowable deceleration and the safety distance. The feedforward adaptive compensation module is used to input operating characteristics, including at least the dynamic threshold, the buffer correction amount, and the reference motor speed, into the feedforward network model to generate the feedforward compensation amount. The nonlinear enhancement module is used to obtain the position error between the target position and the real-time position when the system execution time reaches the deceleration start time, and generate an enhancement error signal based on the position error, a preset enhancement coefficient and a nonlinear exponent parameter. The integrated arbitration control module is used to perform gating processing on the displacement deviation based on the event trigger flag to generate a position correction amount; to superimpose the local control amount, the position correction amount, the feedforward compensation amount, and the enhanced error signal to form a candidate synthetic control amount; to perform unified amplitude limiting on the candidate synthetic control amount to generate a unique integrated control input; and to output it to the motor control execution carrier.

Images