PSO fuzzy variable lead-lag control system and method for six-axis parallel robot

Abstract

The application discloses a PSO fuzzy variable inductance control system and method for a six-axis parallel robot, and relates to the field of robot compliance control technology. The system comprises a data storage module, a PSO optimization module, an error calculation module, a fuzzy controller, a damping calculation module and an inductance control module. The data storage module stores a force and parameter mapping table; the PSO optimization module optimizes the basic damping and gain coefficient online and updates the mapping table with the minimum overshoot and return-to-steady time as the target when the overshoot exceeds the limit; the error calculation module calculates the position and speed errors; the fuzzy controller outputs a damping adjustment factor; the damping calculation module queries target parameters according to the contact force and calculates dynamic damping; and the inductance control module corrects the robot pose. The application solves the boundary mismatch problem caused by traditional parameter static setting, realizes adaptive optimization and fine matching, significantly reduces the overshoot and shortens the return-to-steady time, and improves the compliance and disturbance rejection performance.

Description

Technical Field

[0001] This invention relates to the field of robot compliant control technology, and in particular to a PSO fuzzy variable admittance control system and method for a six-axis parallel robot. Background Technology

[0002] Six-axis parallel robots (commonly known as six-degree-of-freedom parallel robots) are high-precision spatial motion mechanisms possessing high stiffness, high load-to-weight ratio, and complex spatial trajectory planning capabilities. Among them, Stewart-type parallel robots, due to their compact structure and fast dynamic response, are widely used in contact operations such as aerospace precision assembly, complex surface grinding and polishing. In these physical interaction tasks, contact forces inevitably occur between the robot's end effector and environmental objects or workpieces. Especially when performing high-precision assembly or surface machining, they are highly susceptible to sudden pulse force disturbances (such as rigid tool contact, assembly jamming, etc.). If the robot itself lacks an effective compliance mechanism, rigid impacts will occur, potentially damaging the workpiece or mechanism. Therefore, introducing active compliance control strategies, such as variable admittance control, to absorb disturbance energy and adapt to environmental constraints has become a key technology for ensuring operational safety and quality.

[0003] However, existing variable admittance control technology has the following technical shortcomings when dealing with the complex dynamic characteristics of six-axis parallel robots and the harsh execution conditions in industrial settings: On the one hand, in traditional fuzzy variable admittance control strategies, the setting of key control parameters (such as the basic damping coefficient and gain coefficient) usually relies on the manual experience of control engineers for offline tuning and fixation. For six-axis parallel robots such as Stewart, their dynamic characteristics are highly nonlinear, and the inertia, stiffness, and coupling characteristics of the mechanism vary significantly under different poses. At the same time, the magnitude and direction of external disturbance pulse forces in actual industrial contact operations are highly random. This statically fixed parameter setting method is difficult to adapt to dynamically changing force conditions and the attitude changes of the mechanism itself. The limitations of manual experience tuning lead to boundary mismatches between control parameters and actual working conditions when the system faces complex and variable external disturbances, which in turn causes the damping adjustment mechanism to fail. This directly results in a large position overshoot at the moment of contact and prolongs the system's settling time, making it difficult to provide consistent and optimal compliance disturbance rejection performance across the entire workspace.

[0004] On the other hand, existing admittance control models often employ fixed parameter structures and lack refined matching mechanisms for different contact force amplitudes. When the external contact force varies over a large range, a single model or piecewise coarse parameter adjustments cannot simultaneously ensure sensitivity in the small force range and safety in the large force range. This results in sluggish robot response under light contact conditions and insufficient disturbance rejection capability under heavy impact conditions, further restricting the improvement of work efficiency and process quality. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a PSO fuzzy variable admittance control system and method for a six-axis parallel robot.

[0006] A PSO fuzzy variable admittance control system for a six-axis parallel robot includes a data storage module, a damping calculation module, an admittance controller, an inverse kinematics module, a position closed-loop regulator, an absolute encoder, a forward kinematics module, a summing node, a differentiation module, a fuzzy controller, a pose recording module, and a PSO optimization module. The data storage module is used to store the parameter mapping table; The damping calculation module is used to calculate based on external contact force. Query the parameter mapping table to obtain the current base damping. and current gain coefficient And combined with damping adjustment factor Calculate dynamic damping ; The admittance controller is used to adjust based on dynamic damping. Solve the desired pose ; The inverse kinematics module is used to determine the desired pose. Convert to target cylinder length command ; The position closed-loop adjuster is used to adjust according to the target cylinder length command. Control the movement of the telescopic cylinder; The absolute encoder is used to acquire the actual physical length. ; The kinematics forward solving module is used to determine the actual physical length. Calculate the actual pose ; The summation node is used to calculate the initial pose. With actual pose The difference is used to obtain the position error. ; The differentiation module is used to calculate the position error. Determine the velocity error by taking the derivative ; The fuzzy controller is used to determine the position error. and speed error Output damping adjustment factor ; The pose recording module is used to monitor the actual pose. Overshoot; The PSO optimization module is used to perform parameter optimization when the overshoot is greater than a preset threshold, and write the optimization results into the data storage module.

[0007] Furthermore, the parameter mapping table includes a type a mapping table and a type b mapping table; The type a mapping table records the calibration force of the mapping relationship. Optimize basic damping and optimize gain coefficient ; The type b mapping table records the calibration range of the mapping relationships. Basic damping and gain coefficient .

[0008] Furthermore, the damping calculation module executes the following logic: Obtaining external contact force The amplitude is rounded to the nearest integer to obtain the integer force value; First, look up the calibration force that is equal to the integer force value in the class A mapping table. If a match exists, it is considered an exact match, and the corresponding optimized basic damping is applied. and optimize gain coefficient Assign a value to the current base damping and current gain coefficient ; If it does not exist, then search the class b mapping table for information containing external contact forces. The rated force range If it exists, it is determined to be an interval match, and the corresponding basic damping is adjusted. and gain coefficient Assign a value to the current base damping and current gain coefficient .

[0009] Furthermore, when the PSO optimization module performs parameter optimization, it specifically executes the following initialization and testing steps: Initialization steps: Stop acquiring data from the data storage module, and check the current external contact force. The amplitude is rounded to the nearest integer and written into the Class A mapping table as the calibration force. ; based on current basic damping and current gain coefficient Using the origin as the reference point, generate a system containing... A population of particles, each containing the fundamental damping to be measured. and the gain coefficient to be measured ; Parameter update and testing steps: Extract the first... The fundamental damping of the individual particles to be measured and the gain coefficient to be measured Assign a value to the current basic damping and current gain coefficient Control the corresponding telescopic cylinder movements of the six-axis parallel robot and maintain the preset test duration; After the test, record the robot's actual pose. As the final stable position; Extracting overshoot and recovery time Among them, overshoot The actual pose during the test Maximum deviation from the final stable position, stabilization time actual pose The time required to enter and remain within the preset error band; according to and Calculate the first Fitness score of each particle .

[0010] Furthermore, when the PSO optimization module performs parameter optimization, it also performs the following update and termination steps: Optimal solution update steps: Based on fitness score Update the individual historical best position data for each particle. Extract this round fitness score The minimum value among them is taken as the lowest score in this round. The lowest score in this round Compared to the lowest score in the previous round Perform a comparison, if Then the global minimum score will be updated to And update the corresponding particle positions to the globally optimal position data. ;according to and Generate the next generation particle coordinate matrix; Termination determination steps: based on and Calculate the numerical improvement rate; determine whether the following conditions are met simultaneously: globally optimal location data. The corresponding overshoot is less than a preset threshold, and the numerical improvement rate is less than the convergence tolerance threshold or the number of iterations. ; If satisfied, Analysis for optimizing basic damping and optimize gain coefficient Write to the mapping table of class A, then exit parameter optimization; if not satisfied, let... Reset particle sequence number Return to the execution parameter update and test steps.

[0011] A PSO fuzzy variable admittance control method for a six-axis parallel robot, including normal operation mode and parameter calibration mode; Normal operating modes include: Parameter matching steps: Obtain external contact force The current basic damping is obtained by querying the mapping relationship. and current gain coefficient ; Dynamic damping calculation steps: Based on the current foundation damping and current gain coefficient and damping adjustment factor Calculate dynamic damping ; Admittance control calculation steps: Based on dynamic damping Solve the desired pose And drive the telescopic cylinder to move; Closed-loop feedback adjustment steps: Acquire actual pose Calculate position error and speed error Fuzzy inference output damping adjustment factor ; Mode switching determination steps: Monitor actual pose If the overshoot exceeds the preset threshold, switch to parameter calibration mode. Parameter calibration modes include: Initialization steps: Generate a collection of... A population of particles, each containing the parameter to be measured; Iterative testing steps: Assign the parameter to be tested to the current parameter to control the robot's actions, and test to obtain the overshoot. and recovery time And calculate the fitness score; Optimal update step: Update the global optimal position data based on the fitness score. ; Termination determination step: Determine whether the termination condition is met. If it is met, save the optimization results and switch back to normal operation mode.

[0012] Furthermore, the parameter matching step specifically includes: Obtaining external contact force The amplitude is rounded to the nearest integer to obtain the integer force value; First, look up the calibration force that is equal to the integer force value in the class A mapping table. If it exists, then the corresponding optimized basic damping will be applied. and optimize gain coefficient Assign a value to the current base damping and current gain coefficient ; If it does not exist, then search the class b mapping table for information containing external contact forces. The rated force range If it exists, then the corresponding basic damping will be... and gain coefficient Assign a value to the current base damping and current gain coefficient .

[0013] Furthermore, the iterative testing steps specifically include: Extract the first The fundamental damping of the individual particles to be measured and the gain coefficient to be measured Assign a value to the current basic damping and current gain coefficient Control the corresponding telescopic cylinder movements of the six-axis parallel robot and maintain the preset test duration; After the test, record the robot's actual pose. As the final stable position; Extracting overshoot and recovery time Among them, overshoot The actual pose during the test Maximum deviation from the final stable position, stabilization time actual pose The time required to enter and remain within the preset error band; according to and Calculate the first h Fitness score of each particle .

[0014] Furthermore, the optimal update step specifically includes: Based on fitness score Update the individual historical best position data for each particle. ; Extract this round fitness score The minimum value among them is taken as the lowest score in this round. ; The lowest score in this round Compared to the lowest score in the previous round Perform a comparison, if Then the global minimum score will be updated to And update the corresponding particle positions to the globally optimal position data. ; according to and Generate the next generation of particle coordinate matrices.

[0015] Furthermore, the termination determination step specifically includes: according to and Calculate the numerical improvement rate; Determine if the following conditions are met simultaneously: globally optimal location data The corresponding overshoot is less than a preset threshold, and the numerical improvement rate is less than the convergence tolerance threshold or the number of iterations. ; If satisfied, the globally optimal location data will be used. Analysis for optimizing basic damping and optimize gain coefficient Write to the class A mapping table, exit parameter calibration mode, and switch back to normal operation mode; If not satisfied, let Return to the execution of the iterative test steps.

[0016] Beneficial technical effects of the present invention: First, this invention constructs a hierarchical and refined parameter mapping mechanism, which combines a-type and b-type mapping tables to balance the system's sensitive response to light contact with its safe resistance to disturbances under heavy impact.

[0017] Then, the present invention designs a micro population PSO algorithm, which balances optimization accuracy and engineering real-time performance, and is particularly suitable for industrial low-level controller deployment.

[0018] Secondly, this invention constructs a closed-loop evaluation system based on dynamic performance indicators. The system and control method use overshoot and settling time as evaluation indicators, directly transforming the control objective into an optimization objective, which significantly improves the stability of contact operations.

[0019] Finally, this invention achieves adaptive optimization and precise matching of control parameters, breaking the traditional model of static parameter tuning based on manual experience, solving the boundary mismatch problem caused by parameter solidification, and ensuring the consistency of compliant control performance throughout the entire workspace. Attached Figure Description

[0020] Figure 1 This is a structural block diagram of the PSO fuzzy variable admittance control system applicable to a six-axis parallel robot in this embodiment of the invention; Figure 2 It is the nonlinear mapped mesh surface of the fuzzy controller in this embodiment of the invention; Figure 3 This is the physical experimental platform used for experimental verification in the embodiments of the present invention; Figure 4 These are the Z-axis displacement stabilization curves for three control strategies—ordinary admittance, empirical fuzzy, and PSO-Fuzzy optimization—in embodiments of this invention.

[0021] In the picture: 1. AUBO-i5 robot, 2. Standard weight, 3. Stewart parallel robot. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0023] Example 1 A PSO fuzzy variable admittance control system for a six-axis parallel robot includes a data storage module, a damping calculation module, an admittance controller, an inverse kinematics module, a position closed-loop regulator, an absolute encoder, a forward kinematics module, a summing node, a differentiation module, a fuzzy controller, a pose recording module, and a PSO optimization module. These modules together constitute a dynamic feedback control closed loop.

[0024] The connection relationships and data processing logic of each module are as follows: Data storage module: Stores type A mapping tables and type B mapping tables. Type A mapping tables record calibration values. Optimize basic damping and optimize gain coefficient Class B mapping tables record the calibration range. Basic damping and gain coefficient .

[0025] Damping calculation module: includes parameter matching unit and dynamic calculation unit.

[0026] Parameter matching unit: Acquires external contact force The amplitude is rounded to the nearest integer to obtain an integer force value; the calibration force equal to this integer force value is first searched in the type a mapping table. If a match exists, it is considered an exact match, and the corresponding optimized basic damping is applied. and optimize gain coefficient Assign a value to the current base damping and current gain coefficient If it does not exist, then search the class b mapping table for information containing external contact forces. The rated force range If it exists, it is determined to be an interval match, and the corresponding basic damping is adjusted. and gain coefficient Assign a value to the current base damping and current gain coefficient .

[0027] Dynamic calculation unit: Receives current foundation damping Current gain coefficient and damping adjustment factor Calculate and obtain dynamic damping .

[0028] Admittance controller: based on external contact force To obtain dynamic damping Combined with the initial pose Construct a second-order admittance dynamic model and solve for the desired pose. .

[0029] Second-order admittance dynamics model: In the formula, The inertia matrix representing the admittance model; Real-time dynamic damping representing the admittance model The stiffness matrix representing the admittance model; Indicates the acceleration of the desired pose; The velocity representing the desired pose; This represents the desired pose generated by the solution; This indicates the initial position that is manually set for writing.

[0030] Inverse kinematics module: Determines the desired pose Decoupling into multi-channel target cylinder length commands .

[0031] Position closed-loop regulator: based on target cylinder length command Multiple drive signals are generated to control the corresponding telescopic cylinder movements of the six-axis parallel robot.

[0032] Absolute encoder: Collects the actual physical length of each telescopic cylinder. .

[0033] Kinematics forward solution module: based on actual physical length Calculate the actual pose .

[0034] Summation node: Calculates the initial pose With actual pose The difference is used to obtain the position error. .

[0035] Differentiation module: for position error Differentiate to generate velocity error .

[0036] Fuzzy controller: based on position error and speed error Perform fuzzy inference and output the damping adjustment factor. To the damping calculation module; when external contact force When it is 0, the damping adjustment factor will be... Set to 0.

[0037] After receiving the two error signals, the fuzzy controller sequentially performs three calculation steps: fuzzification, rule reasoning, and defuzzification.

[0038] First, a fuzzification step is performed. The system will then perform a position error... Speed ​​error And the damping adjustment factor to be output They are all uniformly divided into seven specific fuzzy subsets: negative large (NB), negative medium (NM), negative small (NS), zero (ZO), positive small (PS), positive medium (PM), and positive large (PB).

[0039] Table 1 Damping Adjustment Factors Fuzzy control rule table In Table 1, the columns are explained as follows: Position error : Calculated from the summation nodes ( This reflects the distance and direction of the robot's current pose from its initial pose.

[0040] Positive values ​​(PS, PM, PB): Indicate that the robot's end effector deflects in the positive direction.

[0041] Negative values ​​(NS, NM, NB): represent the robot end effector shifting in the negative direction.

[0042] Zero value (ZO): Indicates that it is at the initial equilibrium position with no positional deviation.

[0043] In Table 1, the row descriptions are as follows: speed error : Obtained by differentiating the position error using the differentiation module, it reflects the instantaneous motion trend of the robot's end effector deviating from the initial pose.

[0044] Positive values ​​(PS, PM, PB): Indicate that the current instantaneous motion direction is in the positive direction.

[0045] Negative values ​​(NS, NM, NB): indicate that the current instantaneous direction of motion is in the negative direction.

[0046] Zero value (ZO): Characterizes a state where the velocity of motion is zero, that is, a state of complete stillness, or a final stable position where the direction of motion is about to reverse.

[0047] Table 1 shows the damping adjustment factors. Explanation: Damping adjustment factor The intersection of the row and column is the output of the fuzzy controller, which is obtained by directly adjusting the dynamic damping of the system. This achieves adaptive control, with the following specific effects: Increase damping (positive output PB / PM / PS): When the system is moving in the direction of the external contact force, the command instantly increases the dynamic damping to forcibly absorb and dissipate the impact kinetic energy. The core effect is to suppress overshoot.

[0048] Reduce damping (output negative values ​​NB / NM / NS): When the system is in the stage of force withdrawal and returning to the initial posture, the command actively reduces the damping to eliminate motion drag resistance. The core effect is to reduce the stabilization time.

[0049] Maintain damping (output zero value ZO): When the system is in its initial pose, with minimal deviation, or at a motion reversal point, the command maintains the basic damping unchanged, with the core effect being to maintain a stable state.

[0050] For position errors, the fuzzy controller uses the Gaussian membership formula for calculation: In the formula, It represents the membership degree of the positional error to a specific fuzzy subset; This represents the expected value of the center of this subset; The standard deviation parameter represents the width of the subset distribution.

[0051] For velocity error, the Gaussian membership formula is also used for calculation: In the formula, This represents the membership degree of the velocity error to a specific fuzzy subset; This represents the expected value of the center of this subset; The standard deviation parameter represents the width of the subset distribution.

[0052] Then, the rule inference step is executed. The fuzzy controller calls its internally stored two-dimensional fuzzy rule base (containing 49 control rules, as shown in Table 1, and their three-dimensional mapping relationship is as follows). Figure 2 (As shown). During the inference process, the controller performs a logical AND operation between the position error membership degree and the velocity error membership degree to generate the first... Trigger weight of the rule It acts as a computational bridge connecting input and output.

[0053] Finally, the defuzzification step is performed. The fuzzy controller uses the centroid method, and the specific defuzzification calculation formula is as follows: In the formula, Indicates the rule sequence number; Indicates the first The trigger weight value of the rule; Indicates the first Each control rule corresponds to a constant centroid value for the output fuzzy subset.

[0054] The fuzzy controller, by sequentially performing the fuzzification, rule inference, and defuzzification described above, ultimately outputs the damping adjustment factor precisely. .

[0055] Pose recording module: monitors actual pose When the overshoot exceeds the preset maximum allowable overshoot threshold, an optimization trigger command is generated, and the system enters parameter calibration mode.

[0056] PSO optimization module: In parameter calibration mode, the following logic is executed: 1. Initialization steps: The damping calculation module stops acquiring data from the data storage module; the current external contact force is... The amplitude is rounded to the nearest integer, and the integer force value is written into the Class A mapping table as the calibration force. ; based on current basic damping and current gain coefficient Using the origin as the reference point, generate a system containing [various parameters] within the preset constraint range. A miniature population data matrix of individual particles, each particle containing the underlying damping to be measured. and the gain coefficient to be measured Let the iteration rounds be... Particle serial number The initial global minimum score Set to the maximum floating-point value.

[0057] 2. Parameter Update and Testing Steps: Extract the first... The fundamental damping of the individual particles to be measured and the gain coefficient to be measured Assign the value to the current basic damping in the damping calculation module. and current gain coefficient Control the corresponding telescopic cylinder movements of the six-axis parallel robot and maintain the preset test duration; after the test, record the robot's actual pose. As the final stable position; the pose recording module extracts the overshoot. and recovery time Among them, overshoot The actual pose during the test Maximum deviation from the final stable position, stabilization time actual pose The time required to enter and remain within the preset error band; the PSO optimization module based on the overshoot. and recovery time Calculate the first Fitness score of each particle The specific calculation formula is as follows: In the formula, This is the preset overshoot weighting coefficient; This is the preset stabilization time weighting coefficient.

[0058] 3. Traversal and judgment steps: Let ,judge Check if the condition is met; if it is met, return to step 2; if it is not met, proceed to step 4.

[0059] 4. Optimal solution update steps: First, based on the fitness score... Update the individual historical best position data for each particle. Subsequently, extract this round fitness score The minimum value among them is taken as the lowest score in this round. The lowest score in this round Compared to the lowest score in the previous round Compare: If Then the global minimum score will be updated to And update the corresponding particle positions to the globally optimal position data. ;like ≥ Then the global minimum score and Remain unchanged; finally, and Substitute the velocity and position update formulas to generate the next generation particle coordinate matrix.

[0060] Velocity and position update formulas: In the formula, Indicates the first The search velocity vector updated for each particle; Indicates inertia weight; Indicates the first The search velocity vector of each particle in the current iteration; Indicates self-awareness learning factors; Represents a random number between 0 and 1; Indicates the first The parameter coordinate vector of each particle in the current iteration; Represents social cognitive learning factors; This represents the next-generation parameter coordinate vector.

[0061] 5. Termination Judgment Step: Based on and Calculate the numerical improvement rate; Determine if the following conditions are met simultaneously: globally optimal location data The corresponding overshoot is less than a preset threshold, and the numerical improvement rate is less than the convergence tolerance threshold or the number of iterations. If satisfied, Analysis for optimizing basic damping and optimize gain coefficient Write to the class A mapping table, exit parameter calibration mode, and restore the damping calculation module's permission to retrieve data from the data storage module; if not satisfied, set... Reset particle sequence number Return to step 2.

[0062] Example 2 A PSO fuzzy variable admittance control method for a six-axis parallel robot includes a normal operation mode and a parameter calibration mode, which are switched according to the overshoot state. The specific steps are as follows: I. Normal Operating Mode S1: Parameter Matching Obtaining external contact force The amplitude is rounded to the nearest integer to obtain an integer force value. The calibration force equal to this integer force value is first retrieved from the type A mapping table. If a match exists, it is considered an exact match, and the corresponding optimized basic damping is applied. and optimize gain coefficient Assign a value to the current base damping and current gain coefficient If it does not exist, then search the class b mapping table for information containing external contact forces. The rated force range If it exists, it is determined to be an interval match, and the corresponding basic damping is adjusted. and gain coefficient Assign a value to the current base damping and current gain coefficient .

[0063] S2: Dynamic Damping Calculation Receive current basic damping and current gain coefficient and damping adjustment factor Calculate and obtain dynamic damping When external contact force When it is 0, the damping adjustment factor Set to 0.

[0064] S3: Admittance control solution With external contact force To provide excitation, dynamic damping is incorporated. and initial pose Construct a second-order admittance dynamic model and solve for the desired pose. ; Desired pose Decoupling to target cylinder length command And drive the telescopic cylinder to move.

[0065] S4: Closed-loop feedback regulation Collect actual physical length Calculate the actual pose ; Calculate the initial pose With actual pose The difference is used to obtain the position error. The derivative yields the velocity error. Based on positional error and speed error Fuzzy inference output damping adjustment factor To correct dynamic damping .

[0066] S5: Mode Switching Detection Monitoring actual pose The overshoot is determined, and it is determined whether the overshoot is greater than the preset maximum allowable overshoot threshold. If it is greater, the parameter calibration mode is switched and step S6 is executed. If it is not greater, the normal operation mode is continued and the process returns to step S1.

[0067] II. Parameter Calibration Mode S6: Initialization Stop acquiring data from the data storage module and apply the current external contact force. The amplitude is rounded to the nearest integer and written into the Class A mapping table as the calibration force. With current basic damping and current gain coefficient Generate within the constraint range, using the origin as the reference point. Each particle contains the underlying damping to be measured. and the gain coefficient to be measured Let the iteration rounds be... Particle serial number The initial global minimum score Set to the maximum floating-point value.

[0068] S7: Parameter Update and Testing Extract the first The fundamental damping of the individual particles to be measured and the gain coefficient to be measured Assign a value to the current basic damping and current gain coefficient Control the corresponding telescopic cylinder movements of the six-axis parallel robot and maintain the preset test duration; after the test, record the robot's actual pose. As the final stable position; extract the overshoot. and recovery time Overshoot The actual pose during the test Maximum deviation from the final stable position, stabilization time actual pose The time required to enter and remain within the preset error band; based on the overshoot. and recovery time Calculate the first Fitness score of each particle .

[0069] S8: Traversal Determination make ,judge Check if the condition is met; if it is met, return to S7; if it is not met, execute S9.

[0070] S9: Optimal Update First, based on fitness score Update the individual historical best position data for each particle. Subsequently, extract this round fitness score The minimum value among them is taken as the lowest score in this round. The lowest score in this round Compared to the lowest score in the previous round Compare: If Then the global minimum score will be updated to And update the corresponding particle positions to the globally optimal position data. ;like ≥ Then the global minimum score and Remain unchanged; finally, according to and Generate the next generation of particle coordinate matrices.

[0071] S10: Termination Decision and Mode Switching according to and Calculate the numerical improvement rate; determine whether the following conditions are met simultaneously: globally optimal location data. The corresponding overshoot is less than a preset threshold, and the numerical improvement rate is less than the convergence tolerance threshold or the number of iterations. , This represents the maximum number of iterations. If satisfied, the globally optimal position data will be used. Analysis for optimizing basic damping and optimize gain coefficient Write to the class A mapping table, exit parameter calibration mode, switch back to normal operation mode, and return to step S1; if not satisfied, let... Reset particle sequence number Return to S7.

[0072] Experimental verification To verify the effectiveness of the system proposed in this invention in a real physical system, a physical experimental platform based on the Stewart parallel robot 3 was built, as follows: Figure 3 As shown, the platform mainly consists of a Stewart parallel robot 3, a Heberson FT060S six-axis force / torque sensor, and an AUBO-i5 robot 1. The Stewart parallel robot 3 platform uses servo motors to drive electric cylinders, with a single electric cylinder achieving a repeatability of ±0.02mm, and the system control cycle is set to 10ms. The six-axis force sensor is installed at the end of the platform, with a range of 600N (Z-axis), a sampling frequency of 1000Hz, and transmits force feedback signals in real time via an RS-485 bus with a baud rate of 115200bps. The AUBO-i5 robot 1 serves as a disturbance actuator, with an end-effector repeatability of ±0.02mm, used to simulate pulse force disturbances under assembly conditions.

[0073] The experiment used an AUBO-i5 robot 1 to suspend a 20N standard weight 2, applying an external force along the Z-axis of the Stewart parallel platform at a fixed speed and target position to simulate pulse disturbance. The experiment consisted of two phases: loading and sinking, and force removal and stabilization, monitoring the dynamic response process of the system. Three comparative experiments were conducted: 1. Ordinary admittance group: fixed damping coefficient B =150; 2. Empirical Fuzzy Group: Employs empirical fuzzy parameters that have not been optimized by PSO; 3. PSO-Fuzzy Optimization Group: The fuzzy strategy for optimizing basic damping and gain proposed in this invention using PSO is adopted.

[0074] Response Feature Comparison Analysis Figure 4The Z-axis displacement stabilization curves for three control strategies are shown.

[0075] During the loading and sinking phase: The ordinary admittance group exhibits significant underdamped oscillations when subjected to load, with a settling time of 2.7s and an overshoot of 35.5%.

[0076] The empirical fuzzy group dynamically adjusted the damping through fuzzy logic, which played a certain role in buffering the sinking impact. However, there was still a slight overshoot at the bottom, with a stabilization time of 1.5s and an overshoot of 13.2%.

[0077] The PSO-Fuzzy optimization group exhibited a smooth falling trajectory, with the settling time reduced to 0.8 s and the overshoot decreased to 2.4%.

[0078] During the recovery and stabilization phase: The ordinary admittance group showed significant oscillations, with a settling time of 2.3s and an overshoot of 30.7%.

[0079] The oscillations of the empirical fuzzy group were somewhat suppressed, but still significant, with a settling time of 1.3s and an overshoot of 9.7%.

[0080] The PSO-Fuzzy optimization group exhibited excellent stabilization curves, reducing the stabilization time to 1 s and the overshoot to 3.5%.

[0081] Experimental results show that the PSO fuzzy variable admittance control system proposed in this invention can significantly reduce position overshoot and shorten settling time in contact operations compared with traditional fixed parameter admittance control and ordinary fuzzy admittance control, effectively improving the compliant disturbance rejection performance of the six-axis parallel robot.

[0082] It should be understood that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. All equivalent modifications or substitutions made by those skilled in the art to the technical solutions of the present invention are included within the protection scope of the claims of the present invention.

Claims

1. A PSO fuzzy variable inductance control system for a six-axis parallel robot, characterized by, It includes a data storage module, a damping calculation module, an admittance controller, an inverse kinematics module, a position closed-loop regulator, an absolute encoder, a forward kinematics module, a summing node, a differentiation module, a fuzzy controller, a pose recording module, and a PSO optimization module; The data storage module is used to store the parameter mapping table; The damping calculation module is configured to calculate a dynamic damping according to an external contact force and a current gain coefficient and a damping adjustment factor and a current gain coefficient and a damping adjustment factor ​ The admittance controller is configured to control the dynamic damping solving the desired pose ; The inverse kinematics module is used to determine the desired pose. Convert to target cylinder length command ; The position closed-loop adjuster is used to adjust according to the target cylinder length command. Control the movement of the telescopic cylinder; The absolute encoder is used to acquire the actual physical length. ; The kinematics forward solving module is used to determine the actual physical length. Calculate the actual pose ; The summation node is used to calculate the initial pose. With actual pose The difference is used to obtain the position error. ; The differentiation module is used to calculate the position error. Determine the velocity error by taking the derivative ; The fuzzy controller is used to determine the position error. and speed error Output damping adjustment factor ; The pose recording module is used to monitor the actual pose. Overshoot; The PSO optimization module is used to perform parameter optimization when the overshoot is greater than a preset threshold, and write the optimization result into the data storage module. The parameter mapping table includes a type a mapping table and a type b mapping table; The type a mapping table records the calibration force of the mapping relationship. Optimize basic damping and optimize gain coefficient ; The type b mapping table records the calibration range of the mapping relationships. Basic damping and gain coefficient ; The damping calculation module executes the following logic: Obtaining external contact force The amplitude is rounded to the nearest integer to obtain the integer force value; First, look up the calibration force that is equal to the integer force value in the class A mapping table. If a match exists, it is considered an exact match, and the corresponding optimized basic damping is applied. and optimize gain coefficient Assign a value to the current base damping and current gain coefficient ; If it does not exist, then search the class b mapping table for information containing external contact forces. The calibration force range If it exists, it is determined to be an interval match, and the corresponding basic damping is adjusted. and gain coefficient Assign a value to the current base damping and current gain coefficient .

2. The PSO fuzzy variable admittance control system for a six-axis parallel robot according to claim 1, characterized in that, When the PSO optimization module performs parameter optimization, it specifically executes the following initialization and testing steps: Initialization steps: Stop acquiring data from the data storage module, and check the current external contact force. The amplitude is rounded to the nearest integer and written into the Class A mapping table as the calibration force. ; based on current basic damping and current gain coefficient Using the origin as the reference point, generate a system containing... A population of particles, each containing the fundamental damping to be measured. and the gain coefficient to be measured ; Parameter update and testing steps: Extract the first... The fundamental damping of the individual particles to be measured and the gain coefficient to be measured Assign a value to the current basic damping and current gain coefficient Control the corresponding telescopic cylinder movements of the six-axis parallel robot and maintain the preset test duration; After the test, record the robot's actual pose. As the final stable position; Extracting overshoot and recovery time Among them, overshoot The actual pose during the test Maximum deviation from the final stable position, stabilization time actual pose The time required to enter and remain within the preset error band; according to and Calculate the first Fitness score of each particle .

3. The PSO fuzzy variable admittance control system for a six-axis parallel robot according to claim 2, characterized in that, When the PSO optimization module performs parameter optimization, it also performs the following update and termination steps: Optimal solution update steps: Based on fitness score Update the individual historical best position data for each particle. Extract this round fitness score The minimum value among them is taken as the lowest score in this round. ; The lowest score in this round Compared to the lowest score in the previous round Perform a comparison, if Then the global minimum score will be updated to And update the corresponding particle positions to the globally optimal position data. ;according to and Generate the next generation particle coordinate matrix; Termination determination steps: based on and Calculate the numerical improvement rate; determine whether the following conditions are met simultaneously: globally optimal location data. The corresponding overshoot is less than a preset threshold, and the numerical improvement rate is less than the convergence tolerance threshold or the number of iterations. , This represents the maximum number of iterations. If satisfied, Analysis for optimizing basic damping and optimize gain coefficient Write to the class A mapping table and exit parameter optimization; If not satisfied, let Reset particle sequence number Return to the execution parameter update and test steps.

4. A PSO fuzzy variable admittance control method for a six-axis parallel robot, characterized in that, The method is applied to the PSO fuzzy variable admittance control system for a six-axis parallel robot as described in claim 1, and includes a normal operation mode and a parameter calibration mode. Normal operating modes include: Parameter matching steps: Obtain external contact force The current basic damping is obtained by querying the mapping relationship. and current gain coefficient ; Dynamic damping calculation steps: Based on the current foundation damping and current gain coefficient and damping adjustment factor Calculate dynamic damping ; Admittance control calculation steps: Based on dynamic damping Solve the desired pose And drive the telescopic cylinder to move; Closed-loop feedback adjustment steps: Acquire actual pose Calculate position error and speed error Fuzzy inference output damping adjustment factor ; Mode switching determination steps: Monitor actual pose If the overshoot exceeds the preset threshold, switch to parameter calibration mode. Parameter calibration modes include: Initialization steps: Generate a collection of... A population of particles, each containing the parameter to be measured; Iterative testing steps: Assign the parameter to be tested to the current parameter to control the robot's actions, and test to obtain the overshoot. and recovery time And calculate the fitness score; Optimal update step: Update the global optimal position data based on the fitness score. ; Termination determination step: Determine whether the termination condition is met. If it is met, save the optimization results and switch back to normal operation mode.

5. The PSO fuzzy variable admittance control method for a six-axis parallel robot according to claim 4, characterized in that, The parameter matching step specifically includes: Obtaining external contact force The amplitude is rounded to the nearest integer to obtain the integer force value; First, look up the calibration force that is equal to the integer force value in the class A mapping table. If it exists, then the corresponding optimized basic damping will be applied. and optimize gain coefficient Assign a value to the current base damping and current gain coefficient ; If it does not exist, then search the class b mapping table for information containing external contact forces. The calibration force range If it exists, then the corresponding basic damping will be... and gain coefficient Assign a value to the current base damping and current gain coefficient .

6. The PSO fuzzy variable admittance control method for a six-axis parallel robot according to claim 4, characterized in that, The iterative testing steps specifically include: Extract the first The fundamental damping of the individual particles to be measured and the gain coefficient to be measured Assign a value to the current basic damping and current gain coefficient Control the corresponding telescopic cylinder movements of the six-axis parallel robot and maintain the preset test duration; After the test, record the robot's actual pose. As the final stable position; Extracting overshoot and recovery time Among them, overshoot The actual pose during the test Maximum deviation from the final stable position, stabilization time actual pose The time required to enter and remain within the preset error band; according to and Calculate the first h Fitness score of each particle .

7. The PSO fuzzy variable admittance control method for a six-axis parallel robot according to claim 6, characterized in that, The optimal update steps specifically include: Based on fitness score Update the individual historical best position data for each particle. ; Extract this round fitness score The minimum value among them is taken as the lowest score in this round. ; The lowest score in this round Compared to the lowest score in the previous round Perform a comparison, if Then the global minimum score will be updated to And update the corresponding particle positions to the globally optimal position data. ; according to and Generate the next generation of particle coordinate matrices.

8. The PSO fuzzy variable admittance control method for a six-axis parallel robot according to claim 7, characterized in that, The termination determination step specifically includes: according to and Calculate the numerical improvement rate; Determine if the following conditions are met simultaneously: globally optimal location data The corresponding overshoot is less than a preset threshold, and the numerical improvement rate is less than the convergence tolerance threshold or the number of iterations. ; If satisfied, the globally optimal location data will be used. Analysis for optimizing basic damping and optimize gain coefficient Write to the class A mapping table, exit parameter calibration mode, and switch back to normal operation mode; If not satisfied, let Return to the execution of the iterative test steps.

Images