A peristaltic pump control method for a pneumatic mechanical gripper

Abstract

The application provides a peristaltic pump control method for a pneumatic mechanical gripper, and relates to the field of intelligent control, and comprises the following steps: constructing a pneumatic conversion module, a peristaltic pump PID control module, a driving execution module, a bending perception module and an error feedback processing module; calculating a bending angle difference of the mechanical gripper and obtaining a bending error through pneumatic conversion; outputting a position type PID control signal, and online setting Kp, Ki and Kd through an adaptive migration optimization algorithm; updating an individual by using a direction perception consensus index, a virtual adsorption point set and an adsorption step length; generating a time offset factor correction time sequence based on a historical response memory function; and driving a motor by using a correction signal to realize accurate control of the mechanical gripper. The method can significantly reduce overshoot and steady-state error, shorten response time, suppress oscillation, reduce manual parameter adjustment and improve robustness.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of intelligent control, and particularly relates to a peristaltic pump control method for a pneumatic mechanical gripper. BACKGROUND

[0002] As an important execution component of soft robot, the pneumatic mechanical gripper usually realizes bending, closing and unfolding actions by relying on the inflation and deflation process of the flexible pneumatic cavity; the soft material has significant nonlinear deformation characteristics, and the change of the bending angle is highly dependent on the dynamic change of the driving air pressure; as a miniature driving device with continuous flow output, fast reaction and adjustable output pressure, the peristaltic pump is selected to provide air source driving for the pneumatic mechanical gripper, and the peristaltic pump periodically extrudes the hose by rotating the pressure roller to form a negative pressure and a positive pressure area inside the pump cavity, so as to realize the suction and discharge of the gas.

[0003] Migration Optimization Algorithm (MOA) is an intelligent optimization method based on the idea of swarm intelligence, and its core idea is derived from the migration behavior of biological individuals in the process of adapting to environmental changes; the algorithm is widely used in solving mathematical function optimization, parameter setting, path planning and control system problems. SUMMARY

[0004] Based on the problems existing in the prior art, the application provides a peristaltic pump control method for a pneumatic mechanical gripper, which improves the standard migration optimization algorithm, and performs adaptive setting and dynamic adjustment of the PID controller parameters of the peristaltic pump by designing an adaptive migration optimization algorithm.

[0005] A peristaltic pump control method for a pneumatic mechanical gripper, the specific method is:

[0006] S1, a closed-loop control system model of the peristaltic pump in the pneumatic mechanical gripper is constructed, and the model comprises a pneumatic conversion module, a peristaltic pump PID control module, a driving execution module, a bending perception module and an error feedback processing module;

[0007] S2, the target bending angle of the mechanical gripper is calculated according to the error feedback processing module, and the angle difference between the real-time bending angle of the mechanical gripper and the target bending angle is input into the pneumatic conversion module, and the corresponding peristaltic pump speed error value is output;

[0008] S3, a position type PID calculation control signal is obtained by using the peristaltic pump PID control module according to the peristaltic pump speed error value; the control signal is calculated by adaptively optimizing the parameters of the peristaltic pump PID through the adaptive migration optimization algorithm to obtain the best Kp, Ki and Kd parameter values;

[0009] The adaptive migration optimization algorithm fuses the synthetic migration driving force vector by constructing the direction perception consensus index of the target individual relative to the candidate individuals in its local neighborhood, and introducing an amplitude regulation factor for position update of the target individual;

[0010] According to the current individual position, a set of virtual adsorption points is constructed, the adsorption strength weight is calculated according to the spatial distance of each adsorption point relative to the current position, and the adsorption point offset vector and the adsorption step coefficient are introduced for individual position update;

[0011] S4, the dynamic derivative of the historical response memory function is used to calculate the response lag offset, a time offset factor is constructed to correct the control signal in time sequence, and a corrected control signal is obtained;

[0012] S5, the corrected control signal is input to the driving execution module, the actual rotating speed of the peristaltic pump motor is adjusted, and the precise control of the peristaltic pump is realized.

[0013] Preferably, the pneumatic conversion module linearly proportionally converts the bending angle of the mechanical gripper and the preset pneumatic conversion time coefficient, proportionally operates the angle error according to the preset pneumatic conversion time coefficient, obtains an initial conversion amount, calculates a modulation parameter according to the ratio of the real-time bending angle to the maximum allowed bending angle, and operates the modulation parameter with the initial conversion amount to construct a corresponding peristaltic pump speed; wherein the pneumatic conversion time coefficient is a proportional factor between the bending angle and the peristaltic pump speed, and is preset as a fixed numerical value;

[0014] Specifically, the error signal between the set target bending angle of the mechanical gripper and the actual bending angle is received, and the error is converted into a target rotating speed error signal according to the built-in bending coefficient of the system.

[0015] Preferably, the pneumatic mechanical gripper peristaltic pump PID control module proportionally, integrally and differentially adjusts the converted rotating speed error signal and outputs a corrected control signal; after receiving the corrected control signal, the driving execution module constructs a second-order response model based on voltage driving to dynamically adjust the driving voltage of the peristaltic pump motor and output the corresponding motor actual rotating speed; the bending perception module obtains the current rotating speed of the motor in real time and feeds back the data to the error feedback processing module; the error feedback processing module compares the actual rotating speed with the target rotating speed to generate an error signal, which is fed back to the pneumatic conversion module in a closed loop to realize dynamic regulation of the peristaltic pump.

[0016] Preferably, the Kp, Ki and Kd parameters of the peristaltic pump PID are adaptively optimized, each individual position is taken as a three-dimensional vector, and the proportion, integral and differential coefficients of the adjusting temperature are mapped , the specific method is as follows:

[0017] S401. Based on the optimization range [lb,ub] set by the adaptive migration optimization algorithm, generate multiple initial individual positions. Each individual contains the three-dimensional values ​​of the proportional, integral and derivative coefficients of the peristaltic pump PID controller, and construct the starting point of the peristaltic pump PID controller parameter space search for the closed-loop control system model of the peristaltic pump.

[0018] S402. In the global search phase, the individual position is updated through a migration strategy guided by multiple consensuses; in the local development phase, an individual position update strategy driven by adsorption structure response is constructed to achieve adaptive update of the target individual's position in the parameter space.

[0019] S403. The updated individual positions are parsed into peristaltic pump PID controller parameters and applied to the closed-loop control system model of the peristaltic pump. The objective function is defined based on the absolute value of the error between the actual peristaltic pump speed value and the target peristaltic pump speed value. The feedback is used for subsequent selection and iteration. The smaller the objective function value, the more accurate the peristaltic pump PID controller parameters are.

[0020] S404. Repeat steps S402 to S403 until the set maximum number of iterations is reached, and output the optimal peristaltic pump PID controller parameter results for the closed-loop control system of the peristaltic pump.

[0021] Preferably, the adaptive migration optimization algorithm constructs a direction-aware consensus index of the target individual relative to candidate individuals in its local neighborhood, specifically including:

[0022] For any two individuals and Based on the spatial position in the current iteration and Construct the Euclidean distance matrix It is used to measure the degree of difference between individuals in spatial distribution; for each target individual At the same time, it should be considered that it ranks higher in fitness. A collection of named individuals And its nearest predecessor in Euclidean distance A collection of individuals The local neighborhood is defined by the intersection operation. .

[0023] Based on the local neighborhood of the target individual Within the target individual, multiple neighboring individuals are sequentially selected as a reference group. For each reference individual, the spatial distance and fitness difference between the reference individual and the candidate individuals are calculated. The spatial distance is input into a first exponential adjustment function for transformation, and the fitness difference is input into a second exponential adjustment function for transformation. The two transformed adjustment results are multiplied and fused, and an average reduction operation is performed over all local neighborhoods to construct the direction perception consensus index of the target individual towards the candidate individuals. The candidate individuals are randomly selected individuals within a local neighborhood.

[0024] Preferably, a synthetic migration driving force vector is formed by fusion, and an amplitude regulation factor is introduced to update the target individual's position, specifically including:

[0025] Based on the aforementioned direction-aware consensus index, state-oriented information from multiple candidate individuals is fused to construct a migration driving force vector that integrates multi-source information. The state guidance information is to extract the position offset vector of each candidate individual relative to the target individual, and to normalize the magnitude of the offset vector according to a preset norm constraint.

[0026] For the current target individual, an amplitude regulation factor related to the degree of local guidance consistency is constructed based on the direction-aware consensus index associated with the target individual. The amplitude control factor is combined with the migration driving force vector. The scaling combination is performed, and a preset learning rate parameter is introduced to perform proportional adjustment. A perturbation vector of a standard normal distribution is superimposed, and a control weight coefficient is introduced to fine-tune the perturbation amplitude. Based on the above method and the current individual position, the individual position is accumulated and updated.

[0027] Preferably, a set of virtual adsorption points is constructed based on the current individual location, and the adsorption intensity weight is calculated based on the spatial distance of each adsorption point relative to the current location, specifically as follows:

[0028] Obtain multiple adsorption reference positions for the i-th individual in the current iteration state. For the current individual's position at a given time, construct a point set containing multiple virtual adsorption reference positions. These adsorption reference positions are generated by introducing multi-level perturbation amplitudes and nonlinear anti-motion functions, specifically: around the position of the i-th individual. An adsorption point set is constructed by generating a set of virtual adsorption points at different scales and with different perturbation methods. ;

[0029] For each adsorption point constructed, by its relative current position Calculate the corresponding adsorption strength weight based on the spatial distance. .

[0030] Preferably, an adsorption point offset vector and an adsorption step size coefficient are introduced to update the individual position. Based on the adsorption intensity weight corresponding to each adsorption reference position, a normalized weighted aggregation strategy is used to fuse the position information of each adsorption point. Specifically, this includes: for each adsorption point, calculating its offset vector relative to the current position of the target individual, and using its adsorption intensity weight as a weighting factor to construct a corresponding displacement sub-item; normalizing and fusing the weighted displacement sub-items of all adsorption points proportionally to form the final aggregated guiding vector; then introducing the adsorption step size coefficient to scale the guiding vector and superimposing it on the current target individual position to update the target individual position.

[0031] Preferably, the adsorption step length coefficient is constructed as follows: based on each target individual, according to the set of adsorption points of the target individual. The spatial distance between each adsorption point and the current position is used to construct a first structural response factor to measure the overall offset level of the adsorption points in the spatial distribution. Two adsorption points are randomly selected from the set of adsorption points, and a second structural response factor is constructed based on their directional angle relationship with respect to the target individual to characterize the directional consistency between adsorption directions. The two types of structural response factors are jointly modeled according to a set fusion function to construct an adsorption step size coefficient for adjusting the individual update amplitude.

[0032] Preferably, obtaining the corrected control signal includes:

[0033] Obtain the peristaltic pump in the previous fixed time window The rotational speed deviation data within the time frame is combined with memory weights for weighted integration to form a historical response memory function. The mathematical model is as follows:

[0034] ;

[0035] in, This is the target speed value for the peristaltic pump. The memory weight decays over time, and Δ represents the deposition window length;

[0036] Based on the derivative of the historical response memory function, the time offset that the current control signal of the system needs to be advanced or delayed is dynamically estimated.

[0037] Based on the current controller output value u(t) and the time offset, a control signal is generated after time reconstruction and correction.

[0038] Compared with existing technologies, it has the following technical effects and advantages:

[0039] This invention employs a dual position update mechanism of multiple consensus guidance and adsorption structure response, which can maintain the stable control response of the peristaltic pump control system under multi-source disturbances and dynamic environments, and significantly enhances the system's robustness to nonlinear disturbances.

[0040] Among them, a multi-consensus-guided migration strategy is introduced, and the adaptability index and spatial distribution characteristics of candidate solutions are comprehensively analyzed to construct a direction-aware consensus index, which effectively avoids the local optimum problem caused by the reliance on a single target for guidance in traditional tuning methods; by dynamically adjusting the migration step size through a consistency control factor, high-precision adaptive adjustment of the PID parameters of the peristaltic pump controller under different operating stages is achieved, thereby significantly improving the stability and accuracy of the pneumatic mechanical gripper control.

[0041] By combining the adsorption structure response driving mechanism, the controller update amplitude can be flexibly adjusted according to the structural differences between nodes and the adsorption behavior, so that the peristaltic pump control system has a stronger ability to cope with dynamic load fluctuations and boundary disturbances, and ensures the bending angle accuracy and response speed of the pneumatic manipulator.

[0042] This method dynamically tunes the parameters of the peristaltic pump PID controller using an adaptive migration optimization algorithm, achieving precise control of the motor drive. Compared to the traditional fixed PID parameter method, it can effectively reduce the bending angle error of the robotic gripper and improve the accuracy of micro-angle movements. Simultaneously, a response hysteresis correction module is introduced, constructing a time offset factor through a historical response memory function. This allows for automatic adjustment of the timing characteristics of the control signal based on system hysteresis behavior, enhancing the adaptability of the peristaltic pump control system of the pneumatic robotic gripper to time-varying disturbances. Addressing the actual needs of pneumatic robotic grippers, a triple mapping model of bending angle, rotational speed, and voltage is constructed, forming a complete closed-loop control link in conjunction with a sensing and feedback module. This approach is suitable for precise micro-angle operation scenarios of pneumatic robotic grippers. Attached Figure Description

[0043] Figure 1 This is a flowchart of the peristaltic pump control method for a pneumatic robotic gripper.

[0044] Figure 2 Flowchart of an adaptive optimization method for the Kp, Ki, and Kd parameters of a peristaltic pump PID controller.

[0045] Figure 3 This is a flowchart of the control signal correction method.

[0046] Figure 4 This is a comparison chart showing the fitness values ​​of the improved method of this invention and the standard migration optimization algorithm for adaptive tuning of the parameters of a peristaltic pump PID controller.

[0047] Figure 5 This is a graph showing the parameter optimization results of the PID controller for the peristaltic pump of a pneumatic robotic gripper.

[0048] Figure 6 The diagram shows the effect of the improved method of this invention on the control signal correction during the peristaltic pump speed regulation process, compared with the standard migration optimization algorithm. Detailed Implementation

[0049] This application provides a peristaltic pump control method for pneumatic manipulators, solving the problems of large control fluctuations, slow response, reliance on manual experience for parameter tuning, and difficulty in adapting to complex operating scenarios in existing pneumatic manipulator control processes. This method introduces an adaptive migration optimization algorithm to dynamically optimize the proportional (Kp), integral (Ki), and derivative (Kd) parameters of the peristaltic pump PID controller. Combined with pneumatic conversion modeling and historical response offset correction mechanisms, it achieves precise adjustment and timing reconstruction of the control signal, thereby improving the pneumatic manipulator's response sensitivity and control stability to changes in bending angle, and enhancing the automation and accuracy of the pneumatic manipulator's motion process.

[0050] Please see Figure 1 As shown in the embodiment of this application, a peristaltic pump control method for a pneumatic robotic gripper controls the bending angle of the gripper by controlling the gas pressure output by the peristaltic pump. The specific steps are as follows.

[0051] S1. Construct a closed-loop control system model for the peristaltic pump in a pneumatic manipulator gripper. The model includes a pneumatic conversion module, a peristaltic pump PID control module, a drive execution module, a bending sensing module, and error feedback.

[0052] In this embodiment, a peristaltic pump PID control module, a drive execution module, an error feedback processing module, and a bending sensing module are constructed in Siumulink. The bending sensing module uses the absolute value of the error between the actual peristaltic pump speed and the target peristaltic pump speed as the objective function, and uses the feedback peristaltic pump speed control effect as the Kp, Ki, and Kd parameters of the peristaltic pump PID to measure the accuracy of adaptive optimization. The peristaltic pump PID control module adopts a positional PID algorithm. The error feedback processing module calculates the error value between the set speed and the real-time actual speed of the peristaltic pump through a pneumatic conversion module. The drive execution module realizes the control signal to control the peristaltic pump speed through the controlled model.

[0053] Furthermore, in this embodiment, the control time of the closed-loop control system of the peristaltic pump is set to 100 seconds.

[0054] S2. The target bending angle of the robotic arm is calculated and set according to the error feedback processing module, and the angle difference between the target bending angle of the robotic arm and the real-time bending angle of the robotic arm is input into the pneumatic conversion module and the corresponding peristaltic pump speed error value is output.

[0055] Furthermore, in step S2, the pneumatic conversion module quickly adjusts the error between the target bending angle required by the pneumatic manipulator and the real-time bending angle. The peristaltic pump speed error value that can be handled by the closed-loop control system of the peristaltic pump. ;

[0056] In actual control systems, as the bending angle increases, the deformation response of soft materials exhibits nonlinear characteristics. Within the designed calibration angle range, the relationship between its input and output can be experimentally fitted to be nonlinear. Therefore, the use of nonlinear approximation in the modeling process affects the control accuracy.

[0057] In this embodiment, the bending angle of the pneumatic manipulator and the rotational speed of the peristaltic pump have a non-linear relationship. Based on the error between the target bending angle and the real-time bending angle, this angle error is used as an input parameter. An initial conversion value is obtained by proportionally calculating the angle difference according to a preset pneumatic conversion time coefficient. Based on this, a modulation parameter is calculated according to the ratio of the real-time bending angle to the maximum allowable bending angle. This modulation parameter is then processed with the initial conversion value to form a conversion result. The method for constructing the pneumatic conversion module is as follows:

[0058] ;

[0059] The unit of the peristaltic pump speed error value is rad / s. For the maximum bending angle, The pneumatic conversion time coefficient is a proportional factor between the bending angle and the peristaltic pump speed. In this embodiment, the pneumatic coefficient adopts a fixed value of 0.5, and the maximum bending angle of the pneumatic mechanical gripper is 160 degrees.

[0060] S3. Based on the peristaltic pump speed error value, the positional PID of the peristaltic pump PID control module is used to calculate the control signal; during the calculation of the control signal, the parameters of the peristaltic pump PID are adaptively optimized through an adaptive migration optimization algorithm to obtain the optimal Kp, Ki and Kd parameter values.

[0061] The adaptive migration optimization algorithm constructs a direction-aware consensus index of the target individual relative to candidate individuals in its local neighborhood, fuses them to form a synthetic migration driving force vector, and introduces an amplitude adjustment factor to update the position of the target individual.

[0062] A set of virtual adsorption points is constructed based on the current individual position. The adsorption intensity weight is calculated based on the spatial distance of each adsorption point relative to the current position. The adsorption point offset vector and the adsorption step size coefficient are introduced to update the individual position.

[0063] In this embodiment, the adaptive migration optimization algorithm includes two stages: global search and local development. The global search stage is a migration strategy guided by multiple consensuses. It constructs a direction-aware consensus index of the target individual relative to candidate individuals in its local neighborhood, fuses them to form a synthetic migration driving force vector, and introduces an amplitude regulation factor to update the position of the target individual. The local development stage is an individual position update strategy driven by adsorption structure response. It constructs a set of virtual adsorption points based on the current individual position, calculates the adsorption intensity weight based on the spatial distance of each adsorption point relative to the current position, and introduces an adsorption point offset vector and an adsorption step size coefficient to update the individual position.

[0064] Furthermore, based on the peristaltic pump speed error value, the position-type PID of the peristaltic pump PID control module is used to calculate the control signal;

[0065] Furthermore, the peristaltic pump speed error value is utilized. The proportional coefficient Kp, integral coefficient Ki, and derivative coefficient Kd of the positional PID controller are used to calculate the output control signal. The mathematical model is as follows:

[0066] ;

[0067] In this embodiment, The system time sampling interval is set to 5 seconds.

[0068] Furthermore, during the calculation of the control signal, the Kp, Ki, and Kd parameters of the peristaltic pump PID are adaptively optimized using an adaptive migration optimization algorithm, the process of which is as follows: Figure 2 As shown, the adaptive migration optimization algorithm is constructed by selecting and migrating to the target migration stage and the adaptation stage of the new environment at the destination. The parameters of the peristaltic pump PID controller are adaptively tuned, and the proportional coefficient Kp, integral coefficient Ki, and derivative coefficient Kd of the optimal peristaltic pump PID controller are output and input to the peristaltic pump PID control module.

[0069] In this embodiment, the specific steps are as follows:

[0070] S401. Based on the optimization range [lb,ub] set by the adaptive migration optimization algorithm, generate multiple initial individual positions. Each individual contains three-dimensional values ​​of proportional, integral and derivative coefficients used to adjust the temperature, and construct the starting point for the peristaltic pump PID controller parameter space search of the closed-loop control system model of the peristaltic pump.

[0071] S402. In the global search phase, the individual position is updated through a migration strategy guided by multiple consensuses; in the local development phase, an individual position update strategy driven by adsorption structure response is constructed to achieve adaptive update of the target individual's position in the parameter space.

[0072] S403. The updated individual positions are parsed into peristaltic pump PID controller parameters and applied to the closed-loop control system model of the peristaltic pump. The objective function is defined based on the absolute value of the error between the actual peristaltic pump speed value and the target peristaltic pump speed value. The feedback is used for subsequent selection and iteration. The smaller the objective function value, the more accurate the peristaltic pump PID controller parameters are.

[0073] S404. Repeat steps S402 to S403 until the set maximum number of iterations is reached, and output the optimal peristaltic pump PID controller parameter results for the closed-loop control system of the peristaltic pump.

[0074] Furthermore, during implementation, multiple initial individual positions are generated; more specifically, Nmax initial individual positions are generated. The specific mathematical model for implementation is as follows:

[0075] ;

[0076] In the formula, Let j be the value of the position of the i-th individual in the j-th dimension. To find the optimal value for the j-th dimension of the lower bound, To optimize the value of the j-th dimension of the upper bound, the lower bound lb and the upper bound ub are set during the initialization phase.

[0077] In this embodiment, the problem dimension D=3 of the adaptive migration optimization algorithm, that is, the individual position vector is a 3-dimensional value, the population size is Nmax=60, and lb=[0.1,0,0.1]; ub=[20,80,30]; The values ​​are random numbers between 0 and 1; the position of each individual is mapped to the proportional, integral, and derivative coefficients of the peristaltic pump PID. More specifically: , , .

[0078] Furthermore, a multi-consensus-guided migration strategy is constructed and used in the global search phase of the adaptive migration optimization algorithm. The individual position is updated using the multi-consensus-guided migration strategy, and the top 10 individuals in terms of objective function value and the top 10 individuals with the closest Euclidean distance are taken to form a local neighborhood set.

[0079] In this embodiment, the specific implementation steps are as follows:

[0080] S301, For any two individuals and Based on the spatial position in the current iteration and Construct the Euclidean distance matrix It is used to measure the degree of difference between individuals in spatial distribution; for each target individual At the same time, it should be considered that it ranks higher in fitness. A collection of named individuals And its nearest predecessor in Euclidean distance A collection of individuals The local neighborhood is defined by the intersection operation. .

[0081] S302, in the local neighborhood of the target individual Within the group, multiple neighboring individuals are selected sequentially as a reference individual group; for each reference individual, the spatial distance and fitness difference between the reference individual and the candidate individuals are calculated; the fitness is calculated using an objective function.

[0082] The spatial distance is input into the first exponential adjustment function for transformation processing, and the fitness difference is input into the second exponential adjustment function for transformation processing. The two transformed adjustment results are multiplied and fused, and an average reduction operation is performed in all local neighborhoods to construct the direction perception consensus index of the target individual to the candidate individual.

[0083] The candidate individuals are randomly selected individuals within a local neighborhood. The selection is based on whether the candidate individual is in the same direction and has a similar position to other individuals in the neighborhood. If a candidate individual is close to other members in its neighborhood in both spatial location and fitness, it is considered to have stronger consistency and guidance in that direction.

[0084] The specific implementation model is as follows:

[0085] ;

[0086] In the formula, Let be the consensus index of the k-th candidate individual to the i-th individual, used to measure the importance of the target individual's update direction reference. For candidate individual positions, For local neighborhood The positions of individuals other than candidate individuals. and They are respectively , The fitness value represents The parameter is used to control the degree of influence of spatial distance on the degree of consensus. The larger the value, the more sensitive it is to spatial distribution. In practice, the value is set to 0.5.

[0087] S303. Based on the direction-aware consensus index, the state guidance information of multiple candidate individuals is fused to construct a migration driving force vector that integrates multi-source information; the state guidance information is the position offset vector of each candidate individual relative to the target individual, and the offset vector is normalized according to the preset norm constraint form. The migration driving force vector is used to construct a position vector direction that guides the target individual to update.

[0088] The specific implementation model is as follows:

[0089] ;

[0090] In the formula, Let be the composite driving direction vector obtained by the i-th individual at its current position. A very small constant, set to prevent division by zero errors, is used, with a value of 0.0001.

[0091] S304. Introduce a consensus mapping mechanism to transform the direction-aware consensus index into a migration amplitude control factor. The individual position is updated based on the amplitude factor, the driving direction information, and the perturbation mechanism. The specific implementation model is as follows:

[0092] ;

[0093] In the formula, This represents the value of the j-th dimension of the individual position in the (iter+1)-th iteration; Let be the value of the j-th dimension of the individual position in the iter-th iteration. This is the initial learning rate, initially set to 0.7. The introduced normal distribution perturbation is used to enhance the exploration capability of the adaptive migration optimization algorithm. It follows the pattern N(0,I), where N(0,I) is a vector sampled from a three-dimensional normal distribution.

[0094] Furthermore, an individual position update strategy driven by the adsorption structure response is constructed. In this embodiment, the specific implementation model is as follows:

[0095] S201. Obtain multiple adsorption reference positions of the i-th individual in the current iteration state. Specifically, for the position of the current individual at a given time, construct a point set containing multiple virtual adsorption reference positions. The adsorption reference positions are generated by introducing multi-level perturbation amplitudes and nonlinear anti-motion functions to simulate the spatial perturbation behavior under fractal structures, thereby enhancing the target individual's response capability to multi-scale local environments. The specific mathematical model is as follows: around the position of the i-th individual... A set of virtual adsorption points is generated by different scales and perturbation methods. , ;

[0096] In the formula, The c-th scaling factor controls the amplitude of the perturbation and is set as a decreasing sequence. As c increases, the perturbation amplitude decreases. C is the largest scale factor, which is set to 5 during implementation. To control the oscillation frequency of the nonlinear function, the value is set to 2;

[0097] Furthermore, for each adsorption point constructed, its relative position to the current location is considered. Calculate the corresponding adsorption strength weight based on the spatial distance. ;

[0098] ;

[0099] In the formula, This is the c-th adsorption site;

[0100] S202. Based on the adsorption intensity weight corresponding to each adsorption reference position, a normalized weighted aggregation strategy is used to fuse the position information of each adsorption point. Specifically, this includes: for each adsorption point, calculating its offset vector relative to the current position of the target individual, and using its adsorption intensity weight as a weighting factor to construct the corresponding displacement sub-item; normalizing and fusing the weighted displacement sub-items of all adsorption points proportionally to form the final aggregated guiding vector; then introducing an adsorption step size coefficient to scale the guiding vector and superimposing it on the current position of the target individual to update the target individual's position. The specific mathematical model is as follows:

[0101] ;

[0102] In the formula, This represents the position of the i-th individual in the (iter+1)-th iteration. Let i be the position of the i-th individual in the iter-th iteration. This represents the adsorption step size coefficient.

[0103] Furthermore, in this implementation, the specific method for constructing the adsorption step size coefficient is as follows: based on the positional differences between each adsorption point in the adsorption point set of the i-th individual and the individual itself, a first structural response factor is calculated to characterize the degree of aggregation and dispersion of the adsorption point set. ;

[0104] Based on the angular relationship between any two adsorption points in the set of adsorption points relative to the current individual, a second structural response factor is calculated to characterize the consistency of adsorption orientation. ;

[0105] By jointly modeling the first structural response factor and the second structural response factor, an adsorption step size coefficient is generated for dynamically adjusting the update amplitude of the i-th individual. ;

[0106] in, The initial value of the adsorption step size coefficient is set to 2 in this embodiment.

[0107] Furthermore, an error-time weighted integral is constructed as the objective function to evaluate and optimize the speed control effect of the peristaltic pump control process. In this embodiment, the system time sampling interval is 5 seconds, therefore, the discrete form of the integral time absolute error is adopted in the implementation process. The mathematical model is as follows:

[0108] ;

[0109] In the formula, In this implementation, the closed-loop control system for the peristaltic pump controls the maximum operating time. The value is consistent with the control time of the closed-loop control system of the peristaltic pump in the Simulink simulation model.

[0110] S4. Calculate the response hysteresis offset based on the dynamic derivative of the historical response memory function, construct a time offset factor to perform timing reconstruction correction on the control signal, and obtain the corrected control signal. .

[0111] In this embodiment, a historical response memory mechanism is introduced to dynamically correct the timing of control signal action, thereby enabling the construction of more predictive control commands. The process is as follows: Figure 3 As shown;

[0112] S101, Obtain the peristaltic pump in the previous fixed time window The internal rotational speed deviation data is combined with memory weights for weighted integration to form a historical response memory function. The mathematical model is as follows:

[0113] ;

[0114] in, This is the target speed value for the peristaltic pump. The memory weight decays over time, and Δ is the deposition window length, set to 5 seconds;

[0115] S102. Based on the derivative of the historical response memory function, dynamically estimate the time offset that the current control signal of the system needs to be advanced or delayed.

[0116] In this embodiment, the time offset is dynamically adjusted by combining the current peristaltic pump speed error, the historical response derivative trend, and output stability. The offset is proportional to the speed deviation, and the time correction amplitude is constructed through nonlinear scaling of the error trend and an exponential suppression mechanism for the bending angle fluctuation of the pneumatic gripper. The mathematical model is as follows:

[0117] ;

[0118] in, This is the time offset. The offset scaling exponent converts the angular velocity difference into a time offset in s·rps⁻¹, and M(t) represents the cumulative effect of the peristaltic pump speed error within the time window in rad·s⁻¹. This is the target speed value for the peristaltic pump. This refers to the real-time rotational speed of the peristaltic pump. The standard deviation of the bending angle within a fixed time window, in rad. The volatility suppression index is expressed in rad⁻¹.

[0119] In this embodiment, the target peristaltic pump speed is obtained by calculating the target bending angle using a pneumatic conversion module. Based on the real-time peristaltic pump speed recorded by the peristaltic pump encoder, the pneumatic conversion module performs an inverse transformation to calculate the speed within a fixed time window. The real-time bending angle within the time frame is calculated, and its standard deviation is determined. Finally, the time offset is dynamically adjusted. The derivative of the historical response memory function represents the trend of the peristaltic pump speed error, with units of rad·s⁻. 2 Preferably, the offset scaling index Set to 0.1 s·rps⁻¹, and set the fluctuation suppression index to 1.0 rad⁻¹.

[0120] S103. Generate a time-reconstructed corrected control signal based on the current controller output value u(t) and the time offset. .

[0121] In this embodiment, the reconstructed control signal is fused using a linear weighting method to the current moment. With forward-looking future moments To control the output and improve the stability and predictability of peristaltic pump control, the specific mathematical model is as follows: ;

[0122] in, These are weighting coefficients; preferably, This indicates that the current control value accounts for 40% and the offset time control value accounts for 60%.

[0123] S5. Input the corrected control signal to the drive execution module to adjust the actual speed of the peristaltic pump motor and achieve precise control of the peristaltic pump.

[0124] In this embodiment, the control signal corrected for time offset is input to the drive execution module. After receiving the control signal, the drive execution module adjusts the drive voltage of the peristaltic pump motor according to its magnitude, thereby dynamically adjusting the actual speed of the motor and achieving precise control of the pressing rhythm of the peristaltic pump roller. Furthermore, in this embodiment, the drive execution module is a second-order response control model of the peristaltic pump constructed using Kirchhoff's laws, where the input is the corrected control signal, the value of the corrected control signal is used as the drive voltage input to the drive execution module, and the output is the rotational speed of the peristaltic pump.

[0125] The second-order response control model of a peristaltic pump is as follows:

[0126] ;

[0127] in, For the stator winding inductance of the peristaltic pump motor, For the stator winding resistance of the peristaltic pump motor, The moment of inertia of the peristaltic pump motor. This is the torque constant of the peristaltic pump motor. The electromotive force constant is . Where is the real-time rotational speed of the peristaltic pump in rad / s, and B is the viscous damping coefficient;

[0128] In this embodiment, the stator winding inductance of the peristaltic pump motor is 1.5 mH, the stator winding resistance is 0.5 Ω, and the moment of inertia is 1.5 × 10⁻⁶. −4 The peristaltic pump has a torque constant of 0.2 N·m / A, a viscous damping coefficient of 0.01 N·m·s / rad, and an electromotive force constant of 0.058 V·s / rad; the second-order response control model of the peristaltic pump is converted into a complex frequency domain transfer function in Siumulink.

[0129] The complex frequency domain transfer function is input into the driver execution module of Siumulink.

[0130] Furthermore, in this embodiment, to further verify the accuracy of the micro-angle movement of the pneumatic manipulator, the bending angle of the manipulator was initially set to 60 rad. Through the pneumatic conversion model, the peristaltic pump needs to operate at 120 rad / s. Through the second-order response control model of the peristaltic pump, the theoretically required control signal value is 10. To better verify the control effect, the micro-angle movement experiment was conducted with a target bending angle of 12 rad. The steady-state speed of the peristaltic pump was approximately 24 rad / s, and the required control signal value was 2. The Siumulink simulation system settings were then completed.

[0131] The entire system program was run in MATLAB, and the Siumulink simulation system was started simultaneously. The system was trained through 60 iterations, and the simulation lasted 100 seconds. The output of this embodiment demonstrates the effect of the method of the present invention on the peristaltic pump control of the pneumatic mechanical gripper. Figure 4 As shown, experimental results indicate that, under the same initial conditions and parameter range, the existing standard method (standard migration optimization algorithm) gets stuck in a local optimum after the 28th iteration, with a minimum fitness value of 15.0603; the method proposed in this invention converges to the global optimum in the 48th generation, with a minimum fitness value of 10.4655. Compared with the existing method, the method of this invention has a faster convergence speed and better control effect.

[0132] The fitness graph corresponding to the 48th generation converges to the global optimum is used to optimize the optimal Kp, Ki and Kd parameter values ​​obtained in the 48th iteration. The optimal Kp = 2.2547, the optimal Ki = 8.4359 and the optimal Kd = 0.9729. The peristaltic pump PID control module is reconstructed using the optimal Kp, Ki and Kd parameter values.

[0133] Further comparison of the changes in the system control signal over time before and after optimization reveals that in the known experiment, the initial bending angle of the mechanical gripper was 60 rad, reaching the target bending angle of 24 rps after 20 seconds. Based on the above calculations, the corrected control signal target value first needs to reach 10, and after 20 seconds, the control signal becomes 2. The experimental results are as follows... Figure 6 In existing methods, the control signal exhibits periodic oscillations within the range of t=10 to t=60, which is difficult to adapt to pneumatic control and results in significant overshoot and repeated oscillations. The method of the present invention reaches a peak value in the early stage of response and then quickly stabilizes, eventually settling near the target set value with almost no oscillations. This indicates that, under the same conditions, the method of the present invention can effectively suppress control signal overshoot.

Claims

1. A peristaltic pump control method for a pneumatic robotic gripper, characterized in that, A closed-loop control system model for a peristaltic pump in a pneumatic manipulator is constructed. The model includes a pneumatic conversion module, a peristaltic pump PID control module, a drive execution module, a bending sensing module, and an error feedback processing module. The error feedback processing module calculates the target bending angle of the robotic arm and the angle difference between the target bending angle and the real-time bending angle of the robotic arm. The angle difference is input into the pneumatic conversion module and outputs the corresponding peristaltic pump speed error value. Based on the peristaltic pump speed error value, the positional PID of the peristaltic pump PID control module is used to calculate the control signal; during the calculation of the control signal, the parameters of the peristaltic pump PID are adaptively optimized through an adaptive migration optimization algorithm to obtain the optimal Kp, Ki and Kd parameter values; The adaptive migration optimization algorithm constructs a direction-aware consensus index of the target individual relative to candidate individuals in its local neighborhood, fuses them to form a migration driving force vector, and introduces an amplitude adjustment factor to update the position of the target individual. A set of virtual adsorption points is constructed based on the current individual position. The adsorption intensity weight is calculated based on the spatial distance of each adsorption point relative to the current position. The individual position is updated by introducing the adsorption point offset vector and the adsorption step size coefficient. The response lag offset is calculated based on the dynamic derivative of the historical response memory function. A time offset factor is constructed to perform timing reconstruction correction on the control signal, resulting in a corrected control signal. The corrected control signal is input to the drive execution module to adjust the actual speed of the peristaltic pump motor, thereby achieving precise control of the peristaltic pump.

2. The peristaltic pump control method for a pneumatic robotic gripper according to claim 1, characterized in that, The pneumatic conversion module linearly converts the bending angle of the mechanical gripper to a preset pneumatic conversion time coefficient. Based on the preset pneumatic conversion time coefficient, it performs proportional calculation on the angle error to obtain an initial conversion amount. It then calculates the modulation parameter based on the ratio of the real-time bending angle to the maximum allowable bending angle, and performs calculations on the modulation parameter and the initial conversion amount to construct the corresponding peristaltic pump speed.

3. The peristaltic pump control method for a pneumatic robotic gripper according to claim 2, characterized in that, The Kp, Ki, and Kd parameters of the peristaltic pump PID are adaptively optimized, with each individual position treated as a three-dimensional vector and mapped to the proportional, integral, and derivative coefficients of the temperature regulation. The specific method is as follows: S401. Based on the optimization range [lb,ub] set by the adaptive migration optimization algorithm, generate multiple initial individual positions. Each individual contains the three-dimensional values ​​of the proportional, integral and derivative coefficients of the peristaltic pump PID controller, and construct the starting point of the peristaltic pump PID controller parameter space search for the closed-loop control system model of the peristaltic pump. S402. During the global search phase, the individual location is updated through a migration strategy guided by multiple consensus mechanisms. In the local development phase, an individual position update strategy driven by adsorption structure response is constructed to achieve adaptive update of the target individual's position in the parameter space; S403. The updated individual positions are parsed into peristaltic pump PID controller parameters and applied to the closed-loop control system model of the peristaltic pump. The objective function is defined based on the absolute value of the error between the actual peristaltic pump speed value and the target peristaltic pump speed value. The feedback is used for subsequent selection and iteration. The smaller the objective function value, the more accurate the peristaltic pump PID controller parameters are. S404. Repeat steps S402 to S403 until the set maximum number of iterations is reached, and output the optimal peristaltic pump PID controller parameter results for the closed-loop control system of the peristaltic pump.

4. The peristaltic pump control method for a pneumatic robotic gripper according to claim 3, characterized in that, The direction-aware consensus index specifically includes: For any two individuals and Based on the spatial position in the current iteration and Construct the Euclidean distance matrix It is used to measure the degree of difference between individuals in spatial distribution; for each target individual At the same time, it should be considered that it ranks higher in fitness. A collection of named individuals And its nearest predecessor in Euclidean distance A collection of individuals The local neighborhood is defined by the intersection operation. ; Based on the local neighborhood of the target individual Within the target individual, multiple neighboring individuals are sequentially selected as a reference group. For each reference individual, the spatial distance and fitness difference between the reference individual and the candidate individuals are calculated. The spatial distance is input into a first exponential adjustment function for transformation, and the fitness difference is input into a second exponential adjustment function for transformation. The two transformed adjustment results are multiplied and fused, and an average reduction operation is performed over all local neighborhoods to construct the direction perception consensus index of the target individual towards the candidate individuals. The candidate individuals are randomly selected individuals within a local neighborhood.

5. A peristaltic pump control method for a pneumatic robotic gripper according to claim 4, characterized in that, Introducing the amplitude regulation factor to achieve adaptive updating of individual positions specifically includes: Based on the aforementioned direction-aware consensus index, state-oriented information from multiple candidate individuals is fused to construct a migration driving force vector that integrates multi-source information. The state guidance information is to extract the position offset vector of each candidate individual relative to the target individual, and to normalize the magnitude of the offset vector according to a preset norm constraint. For the target individual, an amplitude regulation factor related to the degree of local guidance consistency is constructed based on the direction-perceived consensus index associated with the target individual. The amplitude control factor is combined with the migration driving force vector. The scaling combination is performed, and a preset learning rate parameter is introduced to perform proportional adjustment. A perturbation vector of a standard normal distribution is superimposed, and a control weight coefficient is introduced to fine-tune the perturbation amplitude. Based on the above method and the current individual position, the individual position is accumulated and updated.

6. A peristaltic pump control method for a pneumatic robotic gripper according to claim 5, characterized in that, Obtain multiple adsorption reference positions for the i-th individual in the current iteration state. For the current individual's position at a given time, construct a point set containing multiple virtual adsorption reference positions. These adsorption reference positions are generated by introducing multi-level perturbation amplitudes and nonlinear anti-motion functions, specifically: around the position of the i-th individual. An adsorption point set is constructed by generating a set of virtual adsorption points at different scales and with different perturbation methods. ; For each adsorption point constructed, by its relative current position Calculate the corresponding adsorption strength weight based on the spatial distance. ; Based on the adsorption intensity weight corresponding to each adsorption reference position, a normalized weighted aggregation strategy is adopted to fuse the position information of each adsorption point. Specifically, this includes: for each adsorption point, calculating its offset vector relative to the current position of the target individual, and using its adsorption intensity weight as a weighting factor to construct the corresponding displacement sub-item; normalizing and fusing the weighted displacement sub-items of all adsorption points proportionally to form the final aggregated guiding vector; then introducing an adsorption step size adjustment factor to scale the guiding vector and superimposing it on the current position of the target individual to update the target individual's position.

7. A peristaltic pump control method for a pneumatic robotic gripper according to claim 6, characterized in that, Based on each target individual, according to the set of adsorption points of the target individual. The spatial distance between each adsorption point and the current position is used to construct a first structural response factor. Two adsorption points are randomly selected from the set of adsorption points, and a second structural response factor is constructed based on their directional angle relationship with respect to the target individual. The two types of structural response factors are jointly modeled according to a set fusion function to construct an adsorption step size coefficient for adjusting the individual update amplitude.

8. A peristaltic pump control method for a pneumatic robotic gripper according to claim 7, characterized in that, The corrected control signal is obtained by: Obtain the peristaltic pump in the previous fixed time window The rotational speed deviation data within the time frame is combined with memory weights for weighted integration to form a historical response memory function. The mathematical model is as follows: ； in, This is the target speed value for the peristaltic pump. The memory weight decays over time, and Δ represents the deposition window length; Based on the derivative of the historical response memory function, the time offset that the current control signal of the system needs to be advanced or delayed is dynamically estimated. Based on the current controller output value u(t) and the time offset, a control signal is generated after time reconstruction and correction.

Images