Three-fingered dexterous hand adaptive PID control method and system

Abstract

The invention provides a three-fingered dexterous hand self-adaptive PID control method, which comprises the following steps: initializing a system, and setting PID parameter initial values of all joints of a three-fingered dexterous hand and motion trails of all the joints; all joints of the three-finger dexterous hand are driven to move according to the movement track, and position feedback signals of all the joints are collected in real time; generating a motion response curve of each joint based on the position feedback signal, and extracting overshoot, response speed and steady-state error from the motion response curve as key performance indexes; according to the key performance indexes, PID parameters are adjusted on line according to a preset rule; according to the adjusted PID parameters, the PID parameters are repeatedly applied to the same motion trail for multiple times, through multiple closed-loop iteration adjustment, the PID parameters are automatically converged to the optimal value, and the three-fingered dexterous hand is controlled in a self-adaptive mode. Based on the key performance indexes and the preset rules, the method does not need to depend on any accurate mathematical model, and thoroughly overcomes the inherent defect of poor robustness of a conventional model dependent control method.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of robot dexterous hand control, in particular to a three-fingered dexterous hand adaptive PID control method and system. BACKGROUND

[0002] Currently, the PID control methods of three-fingered dexterous hand mainly include traditional independent joint PID control, model-based PID control and neural network PID control. The traditional independent joint PID designs an independent controller for each joint, which is simple in structure but completely ignores the dynamic coupling between finger joints, resulting in poor coordination and weak anti-interference ability. The model-based PID (such as the calculated torque method) performs feedforward compensation through a dynamic model, which solves the coupling problem, but its performance is highly dependent on the accuracy of the model, and the complex model has large calculation amount and difficult real-time performance guarantee. The neural network PID adjusts parameters using the self-learning ability of the network, which has strong adaptability, but requires a large amount of data for training, and has problems such as complex calculation, difficult real-time deployment and lack of stability theoretical proof. The above methods are difficult to maintain optimal grasping performance under model uncertainty, variable load and external disturbance. SUMMARY

[0003] In view of the defects in the prior art, the purpose of the present application is to provide a three-fingered dexterous hand adaptive PID control method.

[0004] In a first aspect of the present application, a three-fingered dexterous hand adaptive PID control method is provided, comprising: System initialization, setting the initial value of the PID parameters of each joint of the three-fingered dexterous hand and the motion trajectory of each joint; Driving each joint of the three-fingered dexterous hand to move according to the motion trajectory and collecting the position feedback signal of each joint in real time; Generating a motion response curve of each joint based on the position feedback signal, and extracting overshoot, response speed and steady-state error from the motion response curve as key performance indicators; According to the key performance indicators, adjusting the PID parameters online according to a preset rule; According to the adjusted PID parameters, repeat the application to the same motion trajectory for multiple times, and through multiple closed-loop iteration adjustments, the PID parameters automatically converge to the optimal value, and the three-fingered dexterous hand is adaptively controlled.

[0005] Optionally, the initial value of the PID parameters includes proportional gain Kp, integral gain Ki and derivative gain Kd. The initial value of the PID parameters is based on the given value in the initial state control to achieve the expected device adjusted data; and the motion trajectory is the joint angle change curve of the three-fingered dexterous hand when grasping the target object.

[0006] Optionally, the driving each joint of the three-fingered hand to move according to the motion trajectory, and collecting position feedback signals of each joint in real time, comprises: acquiring the motion trajectory set for the three-fingered hand; driving the three-fingered hand to move according to the set motion trajectory, and acquiring signals of each joint from the initial position to the target position of the motion trajectory; converting the position signals of each joint into electric signals, and feeding back the positions of all joints.

[0007] Optionally, the generating a motion response curve of each joint based on the position feedback signals, and extracting overshoot, response speed and steady-state error from the motion response curve as key performance indicators, comprises: acquiring the feedback signals of each joint, and constructing a motion response curve; extracting overshoot, response speed and steady-state error from the motion response curve as key performance indicators; wherein the overshoot feedbacks the amount by which the maximum position value actually reached by the joint of the three-fingered hand during the process of following the motion trajectory exceeds the target position value of the trajectory; the response speed feedbacks the time consumed by the joint of the three-fingered hand from starting at the initial position of the motion trajectory to actually reaching the target position for the first time; and the steady-state error feedbacks the difference between the position value actually maintained by the joint of the three-fingered hand after reaching a stable state and the target position value of the motion trajectory.

[0008] Optionally, the acquiring the feedback signals of each joint, and constructing a motion response curve, comprises: acquiring the feedback signals of the positions of each joint, recording the time stamp, current position and target position corresponding to each data point of the feedback signals, and calculating the control deviation between the target position and the current position; inputting the control deviation into a PID control algorithm, and performing weighted summation on the proportional term, integral term and differential term of the control deviation to output a control signal; generating a motion response curve of each joint according to the position feedback signals of each joint and the data points of each feedback signal.

[0009] Optionally, the preset rule comprises: setting overshoot threshold, response speed threshold and steady-state error threshold; adjusting PID parameters by acquiring overshoot, response speed and steady-state error extracted from the motion response curve; if the overshoot exceeds the threshold, reducing the proportional gain Kp; if the response speed is lower than the threshold, increasing the proportional gain Kp; If there is a steady-state error, increase the integral gain Ki until the system is stable and the performance meets the requirements, output the PID parameters in the current stable state.

[0010] Optionally, the PID parameters adjusted are repeatedly applied to the same motion trajectory multiple times through multiple closed-loop iterative adjustments to automatically converge the PID parameters to optimal values for adaptive control of the three-fingered hand, comprising: Obtain the PID parameters in the current stable state and apply the PID parameters to the same motion trajectory in the next time; Drive each joint of the three-fingered hand to move along the same motion trajectory and construct a motion response curve of the PID parameters in the current stable state; According to the motion response curve of the PID parameters in the current stable state, check whether the set rules are effective; if the set rules are not ideal, adjust the threshold or rule logic for iterative optimization until the system performance meets the requirements.

[0011] In a second aspect, the application provides a three-fingered hand adaptive PID control method system, comprising: An initialization module for system initialization, setting the initial values of the PID parameters of each joint of the three-fingered hand and the motion trajectory of each joint; A motion control module for driving each joint of the three-fingered hand to move along the motion trajectory and collecting position feedback signals of each joint in real time; A performance evaluation module for generating a motion response curve of each joint based on the position feedback signals and extracting key performance indicators from the motion response curve; A parameter adjustment module for adjusting the PID parameters online according to the key performance indicators and a pre-set rule; An iterative optimization module for repeatedly applying the adjusted PID parameters to the same motion trajectory multiple times through multiple closed-loop iterative adjustments to automatically converge the PID parameters to optimal values.

[0012] Optionally, the parameter adjustment module includes an expert rule base, and the rule base stores multiple adjustment rules in the form of "if-then"; Optionally, the multiple adjustment rules in the form of "if-then" include setting overshoot threshold, response speed threshold and steady-state error threshold; and adjusting the PID parameters by obtaining overshoot, response speed and steady-state error from the motion response curve; If the overshoot exceeds the threshold, decrease the proportional gain Kp; If the response speed is lower than the threshold, increase the proportional gain Kp; If there is a steady-state error, increase the integral gain Ki.

[0013] The adaptive PID control method of the three-fingered hand provided in the application adopts real-time collection of joint position feedback signals of the three-fingered hand to generate a motion response curve, extracts overshoot, response speed and steady-state error as key performance indicators, and then adjusts PID parameters online according to a preset rule and automatically converges the parameters to the optimal through multiple closed-loop iterations. Through performance evaluation based on real-time motion response (such as overshoot and response speed) and preset rules, PID parameters are automatically adjusted online, and the parameters are automatically converged to the optimal through the closed-loop iteration of "motion-evaluation-adjustment", thereby realizing high adaptability control without relying on accurate mathematical models, overcoming the inherent defects of traditional model-dependent control methods, such as poor robustness, complex intelligent algorithm calculation and the need for a large amount of training, reducing the dependence on manual parameter setting, and enabling the three-fingered hand to maintain excellent grasping performance and control stability when facing different objects, loads and external disturbances.

[0014] Other technical effects brought by the additional features will be further described in the corresponding embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0015] Other features, objects and advantages of the application will become more apparent from the following detailed description of non-limiting embodiments with reference to the attached drawings: Figure 1 A flowchart of an adaptive PID control method of a three-fingered hand according to an exemplary embodiment. DETAILED DESCRIPTION

[0016] The application will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the application, but do not limit the application in any form. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the application, and these all belong to the protection scope of the application. The parts not described in detail in the following embodiments can be implemented by using existing technologies.

[0017] In the prior art, the traditional independent joint PID control ignores the strong coupling characteristics between multiple joints, resulting in poor grasping coordination and weak anti-interference ability; the model-based control method (such as the calculated torque method) considers the coupling relationship, but it is severely dependent on accurate dynamic models, and in actual application, the control performance is easily degraded due to model mismatch, and the calculation is complex; and the neural network and other intelligent control methods have problems such as complex design, large amount of calculation and the need for a large amount of prior data. Based on the above problems, the adaptive PID control method of the three-fingered hand is provided to solve the above problems.

[0018] Referring toFigure 1 As shown in the embodiment of the present application, a self-adaptive PID control method for a three-fingered hand includes: S1, system initialization, setting the initial value of the PID parameters of each joint of the three-fingered hand and the motion trajectory of each joint; S2, driving each joint of the three-fingered hand to move according to the motion trajectory, and collecting the position feedback signal of each joint in real time; S3, generating the motion response curve of each joint based on the position feedback signal, and extracting the overshoot, response speed, and steady-state error from the motion response curve as key performance indicators; S4, adjusting the PID parameters online according to the key performance indicators and the preset rules; S5, applying the adjusted PID parameters to the same motion trajectory repeatedly, and adjusting through multiple closed-loop iterations to make the PID parameters automatically converge to the optimal value, thereby self-adaptively controlling the three-fingered hand.

[0019] Specifically, first, the system is initialized, the initial value of the PID parameters of each joint of the three-fingered hand is set, and the preset motion trajectory of each joint is set; then the joints of the three-fingered hand are driven to move according to the set motion trajectory, while the position feedback signal of each joint is collected in real time; based on the collected position feedback signal, the motion response curve corresponding to each joint is generated, and the overshoot, response speed, and steady-state error are extracted from the motion response curve as key indicators for evaluating control performance; then, according to the above key performance indicators, the PID parameters are dynamically adjusted online according to the preset parameter adjustment rules; finally, the adjusted PID parameters are applied to the same motion trajectory, and through the closed-loop adjustment mechanism of multiple loop iterations, the PID parameters automatically converge to the optimal value, thereby realizing self-adaptive control of the three-fingered hand.

[0020] It should be noted that the overshoot refers to the amount by which the actual maximum position value of the joint of the three-fingered hand exceeds the target position value of the trajectory during following the motion trajectory, which is usually expressed as a percentage of the excess amount and the target position value. The response speed refers to the time consumed from the start of the joint of the three-fingered hand at the initial position of the motion trajectory to the first time the actual position reaches the target position of the trajectory. The steady-state error refers to the difference between the actual position value maintained by the joint of the three-fingered hand after the motion reaches a stable state and the target position value of the motion trajectory.

[0021] The above embodiments of the present application solve the problem that the traditional fixed PID parameters are difficult to adapt to the complex motion scene of the three-fingered hand by collecting joint position feedback signals in real time, extracting key performance indicators, dynamically adjusting PID parameters online according to the indicators, combining a closed-loop iteration mechanism to promote the parameters to converge to the optimal value, reducing the overshoot of joint motion, improving the response speed, and reducing the steady-state error. At the same time, the adaptive optimization of the PID parameters does not require human intervention, enhances the versatility and flexibility of the control method, can adapt to control requirements under different motion trajectories, and improves the precision, stability and dynamic response performance of the three-fingered hand motion control.

[0022] In some specific embodiments of the present application, the PID parameter initial value includes a proportional gain Kp, an integral gain Ki, and a differential gain Kd. The given value of the PID parameter initial value is data after the initial state control reaches the expected state, and the motion trajectory is the joint angle change curve when the three-fingered hand grasps the target object.

[0023] The joint motion trajectory is targeted at the target position, which is controlled by communicating the target angle position. The final measurement standard is the overshoot range, response speed, and position convergence of the motion state. The initial position is obtained by the angle sensor at the joint of the three-fingered hand, and the target angle is a position that the user hopes the finger to reach, so the target angle is set by the user.

[0024] It should be noted that the initial value of the PID parameter is determined by debugging. Specifically, first, set the integral and differential to zero, and only adjust the proportional gain (P) until the system output starts to work in a sustained oscillation state. After starting to work, adjust the Ki parameter to accelerate the control of the three-fingered hand to reach the expected state. The initial value is the approximate data obtained after debugging the given value.

[0025] Specifically, in the control process, the PID parameter initial value is first determined, the proportional gain Kp, the integral gain Ki, and the differential gain Kd are selected as the key parameters, and the initial value given is based on the data from the initial state of the three-fingered hand to the control reaching the expected state. At the same time, the motion trajectory is set as the joint angle change curve when grasping the target object, and the target position is controlled by the target angle position communicated. The final measurement standard is the overshoot range, response speed, and position convergence of the motion state. The initial position is obtained by the angle sensor at the joint of the three-fingered hand, and the target angle is set by the user. For the debugging of the PID parameter initial value, first set the integral and differential to zero, and only adjust the proportional gain (P) to the system output sustained oscillation, then adjust the Ki parameter to accelerate the control to reach the expected state, and finally obtain the approximate initial value data.

[0026] The above-mentioned embodiments of the present application can reduce the blindness of parameter adjustment, accelerate system stabilization, and make motion control more accurate and flexible by using angle sensors to obtain initial positions and self-defined target angles, combining with communication to send target angle positions to control target positions, and combining with multiple indicators for measurement. Through a specific PID parameter debugging method, appropriate initial values are determined to improve control effect, which can directly serve the three-fingered hand to complete the grasping task, effectively improving the precision and success rate in grasping operation.

[0027] In some specific embodiments of the present application, each joint of the three-fingered hand is driven to move according to a motion trajectory; position feedback of each joint of the three-fingered hand is collected, and signals of each joint from the initial position to the target position of the motion trajectory are obtained in real time through the angle sensor at each joint; the position signals of each joint are converted into electrical signals to feedback the position of each joint.

[0028] It should be noted that during the motion control process, the joint motion may appear overshoot or hysteresis phenomenon. The last few frames of data before reaching the target are analyzed to calculate and determine whether the current phenomenon is overshoot or hysteresis. If overshoot occurs, Kp is reduced, and Ki is increased to adjust. If the target cannot be reached and oscillates near the target position, Ki is reduced, and Kd is increased to gradually achieve complete control.

[0029] After the position signal is converted into an electrical signal, when the angle sensor fails, an error and abnormality feedback will occur, including: a) the angle sensor data does not change; b) the sensor data exceeds the preset range; c) during the control process, the angle sensor data does not change for a long time.

[0030] In the control process of the three-fingered hand, the above-mentioned embodiments of the present application first drive each joint of the three-fingered hand to start moving according to a pre-set motion trajectory; during the joint motion, the angle sensor installed at each joint is used to collect position feedback information of each joint in real time and continuously from the initial position to the target position of the motion trajectory; then, the collected signal representing the joint position is converted into an electrical signal form to complete the feedback of the position of each joint.

[0031] The above-mentioned embodiments of the present application can accurately grasp the motion state of the joint by using the angle sensor to obtain the signal of each joint from the initial position to the target position, ensuring that the motion process is controllable. At the same time, the position signal is converted into an electrical signal for feedback, so that the control system can timely and accurately understand the deviation between the actual position and the expected position of each joint, realize rapid adjustment of the control strategy, improve the precision and stability of the motion of the three-fingered hand, reduce the motion error, enhance its adaptability to different tasks and environments, and better complete various fine operations.

[0032] In some embodiments of the present application, a motion response curve of each joint is generated based on the position feedback signal, and overshoot, response speed and steady-state error are extracted from the motion response curve as key performance indicators, including: Obtaining the feedback signal of each joint to construct a motion response curve; Extracting overshoot, response speed and steady-state error from the motion response curve as key performance indicators.

[0033] Wherein, the overshoot of the three-fingered hand joint in the process of following the motion trajectory is the amount that the maximum actual position value exceeds the target position value of the trajectory; the response speed is the time consumed from the start of the motion trajectory at the initial position to the actual position first reaching the target position; and the steady-state error is the difference between the actual position value maintained after the motion of the three-fingered hand joint reaches a stable state and the target position value of the motion trajectory.

[0034] The above embodiments of the present application can comprehensively and accurately quantify the dynamic characteristics and control accuracy of each joint of the three-fingered hand in the motion process by obtaining the feedback signal of each joint to construct a motion response curve and extracting overshoot, response speed and steady-state error as key performance indicators. The overshoot can reflect the degree of system response exceeding the target value, avoiding excessive motion; the response speed reflects the rapid reaction ability of the joint to the control command, ensuring timely action; and the steady-state error measures the deviation between the final stable state of the system and the target value, ensuring motion accuracy. The above key performance indicators provide clear and specific basis for subsequent online adjustment of PID parameters, which helps to realize more accurate, stable and efficient motion control of the three-fingered hand.

[0035] In some embodiments of the present application, the feedback signal of each joint is obtained to construct a motion response curve, including: Obtaining the feedback signal of each joint position, recording the time stamp, current position and target position corresponding to each data point of the feedback signal, and calculating the control deviation of the target position and the current position; Inputting the control deviation into the pid control algorithm, and performing weighted summation on the proportional term, integral term and differential term of the control deviation to output a control signal; Generating a motion response curve of each joint according to the position feedback signal of each joint and the data points of each feedback signal.

[0036] Specifically, in the joint motion control process of the three-fingered hand, the key information corresponding to each data point is recorded in real time, including the timestamp of the data point, the current position of the joint, the preset target position, and the control deviation error between the current position and the target position; secondly, in each control period, a unified time reference is first obtained to ensure timing consistency, then the joint target position at that time is calculated according to the preset motion trajectory, and then the feedback signal is sampled to obtain the current actual position of the joint. Based on the target position and the current position, the control deviation is calculated, the deviation is input into the PID control algorithm, and the corresponding control signal output is output to drive the joint motion by weighted sum operation on the proportional term, integral term and differential term of the deviation; finally, by continuously collecting and integrating the timestamp, position data, deviation data and other information recorded in each control period, a complete motion response curve is generated.

[0037] The above embodiments of the present application record the timestamps and position information of the feedback signal data points in detail, calculate the deviation, comprehensively grasp the gap between the joint motion state and the target, ensure the normativity and stability of the control, make the PID control algorithm accurately play a role, output reasonable control signals, and thus draw a motion response curve through the obtained data, intuitively present the joint motion characteristics, realize in-depth analysis of the system performance, and provide support for optimizing the control strategy and improving the motion accuracy and stability of the three-fingered hand.

[0038] It should be noted that the Pid control algorithm is prior art, which applies the pid parameters adjusted by engineers to the pid control algorithm. After the parameters are set, if the control environment changes, the pid parameters need to be adjusted again. In the present application, the three-fingered hand self-adaptive PID control method provides a pid parameter that has not yet converged but is very close as an initial value at the beginning. When slight changes occur during system operation, the important parameter pid parameter in the pid control algorithm can be adjusted automatically.

[0039] In the present application, the overshoot, response speed and steady-state error are extracted from the motion response curve as key performance indicators, specifically: A tolerance for steady-state error is set, and a time window is observed. For example, within a certain control period, within this time window, all position deviations error are less than the set tolerance, indicating that the system is no longer oscillating around the target position, and the motion tends to be stable; the slope of the time-position curve is calculated. In the stable state, the slope should tend to zero. Within this time window, the actual position change rate is continuously lower than a very small threshold, indicating that the joint has stopped moving.

[0040] Key indicators: The overshoot is calculated by comparing the peak value in the response curve with the average value after the movement stabilizes, and the overshoot = ((peak value - steady state value) / steady state value) 100%.

[0041] The response speed is determined by the time interval between the first 10% and 90% of the determined position curve crossing the steady state value. The shorter the time interval, the faster the system responds.

[0042] The steady state error is the difference between the expected target position and the actual average steady state position after the system is determined to be in a stable state, i.e. steady state error = target value - actual steady state value, which measures the final control accuracy of the system.

[0043] In some embodiments of the present application, the preset rules include setting overshoot threshold, response speed threshold and steady state error threshold; adjusting PID parameters by obtaining overshoot, response speed and steady state error from the motion response curve; if the overshoot exceeds the threshold, reducing the proportional gain Kp; if the response speed is lower than the threshold, increasing the proportional gain Kp; if there is a steady state error, increasing the integral gain Ki until the system is stable and the performance meets the standard, and outputting the current PID parameters in the stable state.

[0044] It should be noted that the preset rules are expert rule library, which is based on classic PID control theory and practical engineering experience. There is a clear physical relationship between PID parameters (Kp, Ki, Kd) and system performance indicators (such as overshoot, response speed, steady state error): Among them, the proportional gain Kp: mainly affects the response speed and overshoot. Increasing Kp will improve the response speed, but may cause the overshoot to increase; reducing Kp will reduce the overshoot, but may reduce the response speed. Integral gain Ki: mainly eliminates steady state error. Increasing Ki can accelerate the elimination of steady state error, but may introduce oscillation. The differential gain Kd: mainly improves stability and suppresses overshoot. Increasing Kd can reduce overshoot, but may be sensitive to noise. The rule library is designed based on the analysis of system response characteristics, and a reasonable adjustment direction can be made.

[0045] The establishment of the preset rules is usually a systematic process involving the following steps: First, the threshold values of the key performance indicators, overshoot threshold, response speed threshold and steady state error threshold, need to be defined. The threshold values are set based on the control requirements of the three-fingered hand, such as: the overshoot threshold is set to 10%~20%; the response speed threshold is set to the time requirement to reach the target position; the steady state error threshold is set to the allowable position error range. The threshold values are determined through simulation, experiment and historical data to ensure that the system meets the stability and accuracy requirements.

[0046] Based on the relationship between performance indicators and PID parameters, a series of "if-then" rules are formulated. The set rules include: If overshoot > threshold, decrease Kp; If response speed < threshold, increase Kp; If steady-state error ≠ 0, increase Ki; If the system is stable and the performance meets the requirements, keep the current parameters.

[0047] After the rule base is initially established, it needs to be verified through simulation and actual experiments.

[0048] Run the typical motion trajectory on the three-fingered hand and observe the system response after parameter adjustment.

[0049] According to the response curve, check if the rules are effective. If the rules are not ideal, adjust the threshold or rule logic for iterative optimization until the system performance meets the requirements.

[0050] The above embodiments of the present application adjust the PID parameters by setting an expert rule base containing overshoot, response speed and steady-state error threshold, providing control specifications for online adjustment of PID parameters. The key performance indicators accurately extracted from the motion response curve can be compared with the pre-set threshold to quickly determine the system running state. When the overshoot exceeds the threshold, the proportional gain Kp is decreased to effectively suppress the system's excessive response. When the response speed is lower than the threshold, Kp is increased to improve the system's reaction speed. When there is a steady-state error, the integral gain Ki is increased to eliminate the deviation, achieving fast and accurate adjustment of PID parameters to the appropriate value, making the system stable and meeting the performance requirements, and improving the precision, stability and adaptability of the three-fingered hand motion control.

[0051] In some specific embodiments of the present application, for the adjusted PID parameters, the same motion trajectory is repeatedly applied multiple times through multiple closed-loop iterative adjustments to automatically converge the PID parameters to the optimal value, and the three-fingered hand is adaptively controlled, including: Get the PID parameters in the current stable state, apply the PID parameters to the next same motion trajectory, drive the joints of the three-fingered hand to move along the same motion trajectory, and construct the motion response curve of the PID parameters in the current stable state. According to the motion response curve of the PID parameters in the current stable state, check if the set rules are effective. If the set rules are not ideal, adjust the threshold or rule logic for iterative optimization until the system performance meets the requirements.

[0052] Specifically, the PID parameters obtained by adjustment in the current stable state are applied to the next same motion control of the three-fingered hand. After the motion control using the new parameters, the system response is observed, and a response curve is generated according to the observation. Then, whether the previously set rules are effective is checked according to the generated response curve. If the rules are found to be not ideal, that is, the system performance does not meet the expectation, the preset overshoot, response speed, steady-state error threshold, or the logic of the rules is adjusted, and then the above process is performed again. Through multiple iterations of the above process, the system performance meets the preset requirements.

[0053] In the above embodiment, the process from the initial motion to the target motion (for example, the process of the finger from the upright state to the bent state) includes multiple bending angles. In the PID adjustment process, the PID parameters of different bending angles are adjusted simultaneously in the motion trajectory from the upright state to the bent state. The PID parameters after multiple iterations in the entire motion trajectory can be applied to different bending angles in the process of the finger from the upright state to the bent state.

[0054] The above embodiment of the present application forms a complete closed-loop feedback optimization mechanism by applying the adjusted PID parameters to the next motion control and generating a response curve to check the effectiveness of the rules. The control strategy can be continuously adjusted and optimized according to the actual system response, avoiding the blindness of parameter adjustment. When the rules are not ideal, the threshold or the rule logic is adjusted in time for iterative optimization, so that the PID parameters can automatically converge to the optimal value, thereby realizing adaptive control of the three-fingered hand and improving the precision and stability of the motion of the three-fingered hand under different working conditions.

[0055] It should be noted that when the converged and adjusted PID parameters are obtained, the next motion can be repeated the last motion or controlled to reach a new position, which is applicable.

[0056] The logic is that a PID parameter suitable for the current system is obtained after the first few motion controls, and this parameter is suitable for the next motion control. If the parameter needs to be adjusted after the next motion control, the adjusted PID parameter will be used in the next motion after the end of this control. If the parameter does not need to be adjusted, the next motion will still use the same parameter for motion control. The parameter is adjusted after each motion control and is used in the next motion, which is dynamically changed.

[0057] Based on the same technical concept, in another embodiment of the present application, a three-fingered hand adaptive PID control system is also provided, which includes: The initialization module is configured to initialize the system, set initial values of PID parameters of each joint of the three-fingered hand, and set motion trajectories of each joint. The motion control module is configured to drive each joint of the three-fingered hand to move according to the motion trajectories, and collect position feedback signals of each joint in real time. The performance evaluation module is configured to generate motion response curves of each joint based on the position feedback signals, and extract key performance indicators from the motion response curves. The parameter adjustment module is configured to adjust the PID parameters on line according to the key performance indicators and a preset rule. The iterative optimization module is configured to apply the adjusted PID parameters to the same motion trajectories repeatedly, and automatically converge the PID parameters to optimal values through multiple closed-loop iterative adjustments.

[0058] Specifically, the initialization module is first used to perform system initialization, set initial values of PID parameters of each joint of the three-fingered hand, and set motion trajectories corresponding to each joint. Then, the motion control module is started to drive each joint of the three-fingered hand to move according to the preset motion trajectories, and collect position feedback signals of each joint in real time. Subsequently, the performance evaluation module generates motion response curves of each joint based on the collected position feedback signals, and extracts key performance indicators such as overshoot, response speed, and steady-state error from the curves. Then, the parameter adjustment module adjusts the PID parameters on line according to the extracted key performance indicators and a preset rule. Finally, the iterative optimization module applies the adjusted PID parameters to the next motion control of the three-fingered hand, and automatically converges the PID parameters to optimal values through multiple closed-loop iterative adjustments.

[0059] The above embodiments of the present application realize adaptive control of the three-fingered hand through the cooperative work of the modules. The initialization module sets initial parameters and motion trajectories, reducing the difficulty and time of subsequent adjustment. The motion control module ensures that the joints can move accurately according to the predetermined trajectories and feedback position information in real time. The performance evaluation module generates motion response curves and extracts key performance indicators, comprehensively and accurately reflecting the system running state. The parameter adjustment module adjusts the PID parameters on line according to a preset rule, quickly responds to system changes, and improves control effect. The iterative optimization module automatically converges the PID parameters to optimal values through multiple closed-loop iterative adjustments, improving the stability of the motion of the three-fingered hand.

[0060] In some specific embodiments of the present application, the parameter adjustment module includes an expert rule base, and the rule base stores a plurality of adjustment rules in the form of "if-then".

[0061] In some embodiments of the present application, the plurality of adjustment rules in the form of "if-then" include setting overshoot threshold, response speed threshold and steady-state error threshold; adjusting PID parameters by extracting overshoot, response speed and steady-state error from the motion response curve; if the overshoot exceeds the threshold, reducing the proportional gain Kp; if the response speed is lower than the threshold, increasing the proportional gain Kp; if there is a steady-state error, increasing the integral gain Ki.

[0062] In the above embodiments of the present application, by constructing an expert rule base storing a plurality of "if-then" form rules, the quantitative thresholds of overshoot, response speed and steady-state error are explicitly set, and specific adjustment strategies are developed for abnormal situations of different key performance indicators, specifically: reducing the proportional gain Kp when the overshoot is excessive, increasing the proportional gain Kp when the response speed is insufficient, and increasing the integral gain Ki when there is a steady-state error. The PID parameter adjustment has clear logical basis and quantifiable judgment criteria, avoiding subjectivity and blindness in parameter adjustment. At the same time, the "if-then" rule form is simple and intuitive, with high execution efficiency, which can quickly respond to abnormal feedback of each performance indicator, realize precise and dynamic adaptation of PID parameters, and enhance the robustness and practicality of the control method.

[0063] In another embodiment of the present application, a terminal is provided, including a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor, when executing the program, can be used to execute the self-adaptive PID control method of the three-fingered hand.

[0064] The present application will be further described below in conjunction with specific application examples / comparative examples, so as to better understand the above technical solutions of the present application. It should be understood that the following are only some examples and do not limit the present application.

[0065] Application Example 1: The complete working process of the technical solution of the present application from initialization to parameter self-adaptation is described in detail through a specific example as follows: Example Background: Task: Control the three-fingered hand to grasp a cylindrical object with a diameter of 5 cm Control object: index finger tip joint of the dexterous hand Target position: bend from straight state (0°) to grasping position (45°) First step: system initialization -PID parameter initial value setting: Proportional gain Kp = 1.2 (provide basic response speed); Integral gain Ki = 0.05 (initially eliminate steady-state error); Differential gain `Kd = 0.08` (provides preliminary damping); Motion Trajectory Planning: Generate an S-shaped smooth trajectory from 0° to 45°; total motion time set to 1.5 seconds; trajectory includes acceleration, constant velocity, and deceleration phases.

[0066] Second Step: First Motion Execution and Data Collection Motion Execution: Controller drives motor according to planned trajectory; real-time reading of actual position feedback from encoder; Data Recording (Constructing the First Motion Response Curve): Record timestamp sequence: `[0, 1, 2,..., 1500]` ms Record actual position sequence: `[0, 0.3, 1.1,..., 45.2, 44.9, 45.0]`° Third Step: Performance Evaluation and Index Calculation Stability Determination: Check last 200ms data: position fluctuates between 44.9°-45.1° Fluctuation range <0.5°, determine as having entered stable state Key Index Extraction: 1. Overshoot Analysis: Find position peak: `47.8°` (appears at 1.2 seconds) Calculate overshoot: `(47.8 - 45.0) / 45.0 × 100% = 6.2%` 2. Response Speed Analysis: Calculate 10% position point (4.5°): arrival time `T1 = 180ms` Calculate 90% position point (40.5°): arrival time `T2 = 980ms` Rise time: `980 - 180 = 800ms` 3. Steady-state Error Analysis: Target position: `45.0°` Actual steady-state position: `44.95°` (average of last 100 sampling points) Steady-state error: `45.0 - 44.95 = 0.05°` Fourth Step: Rule-based PID Parameter Adjustment Apply preset adjustment rules: Rule 1: Overshoot 6.2% > Threshold 5% → Perform: Reduce Kp or Increase Kd → Select: `Kp = 1.2 × 0.9 = 1.08`, `Kd = 0.08 × 1.1 = 0.088` Rule 2: Rise time 800ms > Threshold 700ms → Execute: Increase Kp → Adjust: `Kp = 1.08 × 1.1 = 1.188` Rule 3: Steady-state error 0.05° ≈ 0, within acceptable range → Execute: Keep Ki unchanged → `Ki = 0.05` Adjusted PID parameters: - `Kp = 1.188`, `Ki = 0.05`, `Kd = 0.088` Step 5: Iterative Optimization and Convergence Second motion execution: Using adjusted PID parameters Repeat steps 2-4 Performance improvement effects: Overshoot: from 6.2% to 4.1% (satisfies <5% requirement) Rise time: from 800ms to 720ms (close to 700ms target) Steady-state error: remains at 0.05° (satisfies accuracy requirement) Third iteration: Further fine-tune parameters All performance indicators meet preset requirements System determines that parameters have converged, stops adaptive adjustment Final results After three "motion-evaluation-adjustment" closed-loop iterations: PID parameters are automatically optimized from initial values `(1.2, 0.05, 0.08)` to optimal values `(1.25, 0.05, 0.095)` Dexterous hand achieves fast, smooth, and accurate grasping movements Overshoot <5%, response time <700ms, steady-state error <0.1° System has adaptive ability to cope with different objects and load changes The above describes some specific embodiments of the present application. It should be understood that the present application is not limited to the above specific embodiments, and those skilled in the art can make various modifications or modifications within the scope of the claims, which does not affect the essential content of the present application. The above preferred features can be combined in any way without conflict.

Claims

1. A self-adaptive PID control method for a three-fingered dexterous hand, characterized in that, The system initialization sets the initial values of the PID parameters of each joint of the three-fingered hand and the motion trajectory of each joint. Drive each joint of the three-fingered hand to move according to the motion trajectory, and collect the position feedback signals of each joint in real time. Generate the motion response curve of each joint based on the position feedback signals, and extract the overshoot, response speed and steady-state error from the motion response curve as key performance indicators. According to the key performance indicators, the PID parameters are adjusted online according to the preset rules. According to the adjusted PID parameters, the same motion trajectory is repeatedly applied multiple times, and the PID parameters are automatically converged to the optimal value through multiple closed-loop iteration adjustments, so as to adaptively control the three-fingered hand. The initial values of the PID parameters include proportional gain Kp, integral gain Ki and derivative gain Kd.

2. The self-adaptive PID control method of a three-fingered dexterous hand according to claim 1, characterized in that, The initial values of the PID parameters are based on the data after the initial state control reaches the expected state based on the given value; and the motion trajectory is the joint angle change curve of the three-fingered hand when grabbing the target object. The driving of each joint of the three-fingered hand according to the motion trajectory and the real-time collection of the position feedback signals of each joint include:

3. The self-adaptive PID control method of a three-fingered dexterous hand according to claim 1, characterized in that, Obtain the motion trajectory set for the three-fingered hand, and drive the three-fingered hand to move according to the set motion trajectory; Collect the position feedback of each joint of the three-fingered hand, and obtain the signal of each joint from the initial position to the target position of the motion trajectory in real time through the angle sensor at each joint; Convert the position signals of each joint into electrical signals to feedback the positions of all joints. The generation of the motion response curve of each joint based on the position feedback signals and the extraction of the overshoot, response speed and steady-state error from the motion response curve as key performance indicators include:

4. The self-adaptive PID control method of a three-fingered dexterous hand according to claim 3, characterized in that, Obtain the feedback signals of each joint, and construct the motion response curve; Extract the overshoot, response speed and steady-state error from the motion response curve as key performance indicators; Wherein, the overshoot feedbacks the maximum position value actually reached by the joint of the three-fingered hand during the process of following the motion trajectory, which exceeds the target position value of the trajectory; the response speed feedbacks the time consumed from the start of the joint of the three-fingered hand at the initial position of the motion trajectory to the actual position first reaching the target position; and the steady-state error feedbacks the difference between the actual position value maintained by the joint of the three-fingered hand after reaching a stable state and the target position value of the motion trajectory. The acquisition of the feedback signals of each joint and the construction of the motion response curve include:

5. The self-adaptive PID control method of a three-fingered dexterous hand according to claim 4, characterized in that, Obtain the feedback signals of the positions of each joint, record the time stamp, current position and target position corresponding to each data point of the feedback signals, and calculate the control deviation between the target position and the current position; Input the control deviation into the pid control algorithm, and perform weighted summation on the proportional term, integral term and differential term of the control deviation to output a control signal; Generate the motion response curve of each joint according to the position feedback signals of each joint and the data points of each feedback signal. The preset rules include:

6. The self-adaptive PID control method of a three-fingered dexterous hand according to claim 4, characterized in that, ​ setting overshoot threshold, response speed threshold and steady-state error threshold; adjusting PID parameters by extracting overshoot, response speed and steady-state error from the motion response curve; decreasing proportional gain Kp if overshoot exceeds threshold; increasing proportional gain Kp if response speed is below threshold; increasing integral gain Ki if there is steady-state error until system is stable and performance meets requirements, outputting current stable state PID parameters.

7. The self-adaptive PID control method of a three-fingered dexterous hand according to claim 6, characterized in that, repeatedly applying the adjusted PID parameters to the same motion trajectory multiple times through multiple closed-loop iterations to automatically converge the PID parameters to optimal values for adaptive control of the three-fingered hand, comprising: applying the current stable state PID parameters to the same motion trajectory next time; driving each joint of the three-fingered hand to move along the same motion trajectory and constructing a motion response curve of the current stable state PID parameters; checking if the set rules are effective according to the motion response curve of the current stable state PID parameters; if the set rules are not ideal, adjusting threshold or rule logic for iterative optimization until system performance meets requirements.

8. A system of self-adaptive PID control method of a three-fingered dexterous hand according to any one of claims 1-7, characterized in that, comprising: an initialization module for system initialization, setting PID parameter initial values of each joint of the three-fingered hand and motion trajectory of each joint; a motion control module for driving each joint of the three-fingered hand to move along the motion trajectory and collecting position feedback signals of each joint in real time; a performance evaluation module for generating a motion response curve of each joint based on the position feedback signals and extracting key performance indicators from the motion response curve; a parameter adjustment module for adjusting the PID parameters online according to the key performance indicators and preset rules; an iterative optimization module for repeatedly applying the adjusted PID parameters to the same motion trajectory multiple times through multiple closed-loop iterations to automatically converge the PID parameters to optimal values.

9. The control system of claim 8, wherein, The parameter adjustment module is provided with an expert rule base, and the rule base stores multiple "if-then" form adjustment rules.

10. The control system of claim 9, wherein, The multiple "if-then" form adjustment rules include setting overshoot threshold, response speed threshold and steady-state error threshold; adjusting PID parameters by extracting overshoot, response speed and steady-state error from the motion response curve; decreasing proportional gain Kp if overshoot exceeds threshold; increasing proportional gain Kp if response speed is below threshold; increasing integral gain Ki if there is steady-state error.

Images