A novel non-calibration visual servo control method for fruit sorting parallel robot

Abstract

The present invention discloses a novel non-calibrated visual servo control method for a fruit sorting parallel robot, and belongs to the field of visual servo for fruit sorting parallel robots. Firstly, for the fruit sorting parallel robot, a visual mapping model of the fruit sorting parallel robot is established with image information as input and robot motion information as output, so that the non-calibrated visual servo control of the parallel robot can be realized by using the visual mapping model; then, an integral sliding mode controller based on an extended state observer is designed; secondly, a new adaptive law is designed for the switching gain of the integral sliding mode controller based on the extended state observer; finally, the non-calibrated visual servo adaptive integral sliding mode control of the fruit sorting parallel robot is realized through software programming. The visual servo adaptive integral sliding mode control method based on the extended state observer of the present invention can improve the control accuracy of the non-calibrated visual servo system of the fruit sorting parallel robot.

Description

Technical Field

[0001] The invention belongs to the field of robot visual servo, and relates to an uncalibrated visual servo adaptive integral sliding mode control method of a fruit sorting parallel robot, aiming to improve the control accuracy of a visual servo system of the fruit sorting parallel robot. Background Art

[0002] Parallel robots have many advantages such as low cost, high precision, high speed and high rigidity, and they perform well in fruit sorting operations. With the increasing maturity of machine vision technology, the flexibility and accuracy of the robot system can be significantly improved by using visual servo technology to control the fruit sorting parallel robot. In the traditional calibration-based fruit sorting parallel robot visual servo system, it is necessary to obtain information such as camera parameters, target model and current robot posture in advance, and obtain the image Jacobian matrix through calculation iteration to obtain the robot's control instructions. The calibration of these system parameters is an extremely complex and time-consuming process. In addition, there are problems such as the limited camera shooting position and the image is easily doped with noise during image acquisition and processing during the calibration process. Therefore, there are often errors in the detection of image feature points. In addition, the fruit sorting parallel robot visual servo system is a highly nonlinear system. Without calibration, the robot visual servo control system faces the problem that the image Jacobian matrix is ​​difficult to calculate accurately in real time, and there are uncertainties such as unmodeled dynamics and external interference. These problems can easily lead to the inaccuracy of the fruit sorting parallel robot visual servo system, and in severe cases, even lead to the instability of the entire visual servo system.

[0003] In the article "Research on Tracking Control Strategy of Uncalibrated Visual Servo Robot" (Chen Mei, Che Shangyue, Control Engineering, 2019, Vol. 26, No. 6, pp. 1055-1059), an image-based uncalibrated visual servo fuzzy control method is designed for a robot visual servo system. The image captured by the camera is used for target detection, and the image noise is reduced by the frame difference method. A fuzzy controller is designed to perform visual servo control on the robot system. However, this method requires taking multiple frames of images in advance for image processing and target recognition, and fails to achieve real-time visual servo control. It also does not consider the uncertainty of the actual robot system, such as unmodeled dynamics and external interference.

[0004] In the article "Design of Industrial Robot Hand-Eye Visual Servo Control System" (Wang Hongqi, Wu Haibo, Dong Hao, etc., Sensors and Microsystems, 2019, Vol. 38, No. 4, pp. 111-113), a composite image Jacobian matrix estimation method based on Kalman filter is proposed to solve the problem of difficulty in accurately obtaining image Jacobian matrix and uncertainty of target motion in industrial robot visual servo system. Finally, a proportional controller is designed to track the target image features. However, the Kalman filter algorithm is very sensitive to noise interference of the system. In actual robot visual servo tasks, image noise, robot noise, etc. are usually difficult to determine, and unknown noise will cause the filter accuracy to decrease or even diverge. Summary of the invention

[0005] In order to overcome the shortcomings of the prior art, the present invention proposes a novel uncalibrated visual servo adaptive integral sliding mode control method for a fruit sorting parallel robot. Firstly, for the fruit sorting parallel robot, a visual mapping model of the fruit sorting parallel robot is established with image information as input and robot motion information as output, and the visual signal obtained by the camera is used as the feedback information of the visual servo system of the parallel robot. The image Jacobian matrix representing the real-time changes of the mapping relationship between the robot image space and the task space is regarded as the unmodeled dynamics of the system, and the visual mapping model is rewritten into a form suitable for the design of the extended state observer. The extended state observer is used to perform lumped estimation of the unmodeled dynamics and unknown external disturbances of the visual mapping model of the sorting parallel robot. At the same time, the target posture of the parallel robot is determined by setting four feature points, and the image coordinates of the actual posture of the parallel robot are obtained by image processing, so that the visual mapping model can be used to realize the uncalibrated visual servo control of the parallel robot. Then, an integral sliding mode controller based on the extended state observer is designed to improve the robustness of the visual servo system to overcome the image feature point detection error and external disturbance, thereby improving the control accuracy of the visual servo system. Secondly, a new adaptive law is designed for the switching gain of the integral sliding mode controller based on the extended state observer to achieve the effect of suppressing the chattering of the sliding mode control.

[0006] The technical solution of the present invention is: a novel non-calibrated visual servo adaptive integral sliding mode control method for fruit sorting parallel robot, comprising the following steps:

[0007] 1) The kinematics of the fruit sorting parallel robot are analyzed by analytical method and its inverse solution is obtained, and the robot kinematics forward solution and the robot Jacobian matrix are further calculated;

[0008] 2) Based on the kinematic model of the sorting robot established in step 1), combined with the visual imaging principle of the fruit sorting parallel robot, a visual mapping model of the fruit sorting parallel robot is established, and the image Jacobian matrix of the fruit sorting parallel robot is further obtained;

[0009] 3) The image Jacobian matrix representing the real-time changes in the mapping relationship between the robot image space and the task space is regarded as the unmodeled dynamics of the system, and the visual mapping model is rewritten into a form suitable for the design of an extended state observer. The extended state observer is introduced to perform lumped estimation of the unmodeled dynamics and unknown external disturbances of the fruit sorting parallel robot visual servo system;

[0010] 4) Based on the extended state observer in step 3), a visual servo integral sliding mode controller is designed to realize robot visual servo control;

[0011] 5) Design a new adaptive law for the switching gain of the visual servo integral sliding mode controller based on the extended state observer in step 4) to reduce the sliding mode chattering;

[0012] 6) Through software programming, the uncalibrated visual servo adaptive integral sliding mode control of the fruit sorting parallel robot is realized;

[0013] The present invention proposes a method for adaptive integral sliding mode control of a fruit sorting parallel robot without calibration visual servoing. Further, in the step 2), according to the definition of robot visual imaging principle, a point P in the task space of the fruit sorting parallel robot is expressed in image space coordinates as:

[0014]

[0015] Where f is the focal length of the camera (unit: mm), (x, y) represents the coordinates of the image space of the point (unit: mm), C = (x c ,y c ,z c ) represents the position of the point in the camera coordinate system. To facilitate subsequent computer image recognition and processing, the physical unit of the image space is converted into pixel units:

[0016]

[0017] Where (u0, v0) is the center pixel of the image space, (u, v) represents the coordinates of the image space at that point (unit: pixel), and a x is the number of pixels per unit on the X axis of the camera, a y The number of pixels per unit on the Y axis of the camera.

[0018] Furthermore, the state expression of the extended state observer in the fruit sorting parallel robot visual servo system in step 3) is:

[0019]

[0020]

[0021] In the formula, is a predicted value of the image Jacobian matrix of the fruit sorting parallel robot system, defined as Z1 represents the observation value of the extended state observer on the position of the feature point in the image space, Z2 represents the total amount of unmodeled dynamics and external disturbances estimated by the extended state observer for the visual servo system, ε i ,α i ,β i (i=1,…,4) and b1, b2 are non-negative adjustable design parameters of the extended state observer, P * is the expected position of the feature point in the image space, P is the actual position of the feature point in the image space, and ε is the image feedback error of the visual servo controller.

[0022] Further, in step 4), according to the definition of the visual servo system, the visual servo control error variable is designed as follows:

[0023] e=pp * (5)

[0024] p is the actual position of the feature point in the image space, p * For the expected position of the feature point in the image space, the visual servo integral sliding surface is designed as:

[0025] s=e+λ∫e (6)

[0026] Where λ is the controller parameter and the first-order derivative of s is:

[0027]

[0028] In step 5), the control law of the designed visual servo adaptive sliding mode controller is:

[0029]

[0030] In the formula, is the derivative of the switching gain, k m is an adjustable parameter of the adaptive law, whose value is greater than 0, where

[0031]

[0032] Where μ>0 is an adjustable parameter in the adaptive law.

[0033] The present invention proposes for the first time a novel calibration-free visual servo adaptive integral sliding mode control method for a fruit sorting parallel robot, which is applied to realize visual servo motion control of the fruit sorting parallel robot. The characteristics and beneficial effects are as follows:

[0034] 1. Combining the principle of robot vision imaging, a visual servo system is designed that does not require advance camera and robot position calibration. The image features are used as the feedback signal of the system. During the robot movement, the controller is designed by calculating the relationship between the image features and the robot movement speed. The visual controller uses this deviation as feedback to drive the robot to move until the feature point in the image space moves from the current position to the desired position.

[0035] 2. The real-time changing image Jacobian matrix that represents the mapping relationship between the robot's image space and task space is regarded as the unmodeled dynamics of the system, and the visual mapping model is rewritten into a form suitable for the design of an extended state observer. The extended state observer is used to perform a lumped estimate of the unmodeled dynamics and unknown external disturbances of the visual mapping model of the sorting parallel robot, thereby solving the problem that it is difficult for an uncalibrated visual servo system to obtain an accurate image Jacobian matrix in real time and reducing the large amount of parameter calculations required to obtain the image Jacobian matrix.

[0036] 3. A visual servo integral sliding mode controller based on an extended state observer is designed for the uncalibrated visual servo of a fruit sorting parallel robot to improve the robustness of the visual servo system in overcoming image feature point detection errors, visual mapping model modeling errors and external interference, thereby improving the control accuracy of the visual servo system. Furthermore, a new adaptive law is designed for the switching gain of the visual servo integral sliding mode controller based on the extended state observer to achieve the effect of suppressing the sliding mode control jitter. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 This is a prototype diagram of a parallel robot for fruit sorting. In the diagram: 1. Parallel mechanism 2. Calibration board 3. End clamping mechanism 4. Camera 5. Fruit in a string 6. Display screen 7. Control cabinet 8. Detection cabinet 9. Light source;

[0038] Figure 2 This is the structure diagram of the fruit sorting parallel robot (the sorting parallel robot is mainly composed of a static platform, a main (auxiliary) dynamic platform and four sets of branch chains connecting the static and dynamic platforms);

[0039] Figure 3 This is the overall structure diagram of the visual servo control system for a fruit sorting parallel robot;

[0040] Figure 4 It is the x-axis image feature tracking simulation diagram;

[0041] Figure 5It is the y-axis image feature tracking simulation diagram;

[0042] Figure 6 It is a simulation diagram of z-axis image feature tracking;

[0043] Figure 7 It is the trajectory diagram of the robot's spatial end effector;

[0044] Figure 8 is the image space target tracking map;

[0045] Fig. 9 A graph showing the tracking effect of feature points in the image space obtained by the camera. DETAILED DESCRIPTION

[0046] The specific implementation modes of the present invention will be further described below with reference to the accompanying drawings.

[0047] The technical solution adopted by the present invention is to adopt the following steps:

[0048] 1) The kinematics of the fruit sorting parallel robot is analyzed by analytical method to obtain its inverse solution, and the robot kinematics forward solution and the robot Jacobian matrix are further calculated;

[0049] 2) Based on the kinematic model of the fruit sorting parallel robot established in step 1), combined with the visual imaging principle of the fruit sorting parallel robot, a visual mapping model of the fruit sorting parallel robot is established, and the image Jacobian matrix of the fruit sorting parallel robot is further obtained;

[0050] 3) The image Jacobian matrix that represents the real-time changes in the mapping relationship between the robot image space and the task space is regarded as the unmodeled dynamics of the system, and the visual mapping model is rewritten into a form suitable for the design of an extended state observer. The extended state observer is introduced to perform lumped estimation of the unmodeled dynamics and unknown external disturbances of the visual servo system of the sorting parallel robot;

[0051] 4) Based on the extended state observer in step 3), a visual servo integral sliding mode controller is designed to realize the calibration-free visual servo control of the fruit sorting parallel robot;

[0052] 5) Using the switching gain of the visual servo integral sliding mode controller based on the extended state observer in step 4), a new adaptive law is designed to reduce the sliding mode chattering;

[0053] 6) Through software programming, the uncalibrated visual servo adaptive integral sliding mode control of the fruit sorting parallel robot is realized;

[0054] Firstly, the kinematic analysis of the fruit sorting robot is carried out. A fruit sorting parallel robot visual mapping model with image information as input and robot motion information as output is established, and the visual signal obtained by the camera is used as the feedback information of the parallel robot visual servo system. The image Jacobian matrix representing the real-time changes of the mapping relationship between the robot image space and the task space is regarded as the unmodeled dynamics of the system, and the visual mapping model is rewritten into a form suitable for the design of the extended state observer. The unmodeled dynamics and unknown external disturbances of the visual mapping model of the sorting parallel robot are lumped and estimated by the extended state observer. At the same time, the target posture of the parallel robot is determined by setting four feature points, and the image coordinates of the actual posture of the parallel robot are obtained by image processing, so that the visual servo control of the parallel robot can be realized by using the visual mapping model. Then, an integral sliding mode controller based on the extended state observer is designed to improve the robustness of the visual servo system to overcome the image feature point detection error, the modeling error of the visual mapping model and the external disturbance, thereby improving the control accuracy of the visual servo system. Secondly, a new adaptive law is designed for the switching gain of the sliding mode controller based on the extended state observer to achieve the effect of suppressing the chattering of the sliding mode control. Finally, through software programming, the uncalibrated visual servo adaptive integral sliding mode control of the fruit sorting parallel robot is realized. The specific method is as follows:

[0055] 1. Use analytical method to perform inverse kinematics solution of the fruit sorting parallel robot, and further obtain the kinematics forward solution and robot Jacobian matrix. i is the rotation angle of the active arm in the i-th (i=1,2,3,4) branch chain (unit: rad), the main platform mass point is regarded as P1, the auxiliary platform is regarded as mass point and P2, in the robot coordinate system, the position vector of P1(P2) is r=[x,y,z] T for:

[0056]

[0057] Where s is the displacement of point P2 to P1 along the positive direction of the z-axis (m), e i =e(cosγ i sinγ i 0) T , e is the distance from the center of mass of the static platform to the rotation center of the i-th (i=1,2,3,4) branch chain, γ i =(i-1)π / 2(i=1,2,3,4),l1,l2,u i ,w i They represent the length and unit vector of the active arm and the driven arm of the i-th (i=1, 2, 3, 4) branch of the fruit sorting parallel robot, respectively.

[0058] The analytical method is used to perform the inverse kinematics solution of the fruit sorting parallel robot. Furthermore, the derivatives of the inverse kinematics equations are taken to obtain:

[0059]

[0060] Where J is the robot Jacobian matrix, Robot direct Jacobian matrix, is the robot indirect Jacobian matrix, where ω i is the velocity vector of the i-th (i=1,2,3,4) slave arm, where is the velocity vector of the active arm, v i =(-sinγ i cosγ i 0) T , where γ i =(i-1)π / 2(i=1,2,3,4), where is the state variable of the system.

[0061] 2. Through the kinematic model of the fruit sorting parallel robot and combined with the visual imaging principle of the fruit sorting parallel robot, a visual mapping model of the sorting robot is established.

[0062] The fruit sorting parallel robot visual servo system controls the robot motion through the robot visual information and uses the image information for feedback. Let f = (f1 f2 f3 f4) T is the 2D vector in the robot image space, and the robot terminal velocity is According to the robot Jacobian matrix definition:

[0063]

[0064] Where J represents the mapping relationship between the robot joint space and the end Cartesian space, and the robot image feature velocity is available:

[0065]

[0066] Among them J f Image Jacobian matrix for the fruit sorting parallel robot.

[0067] 3. The idea of ​​extended state observer is introduced into the design of uncalibrated visual servo system of fruit sorting parallel robot. The image Jacobian matrix representing the real-time changes of the mapping relationship between the robot image space and the task space is regarded as the unmodeled dynamics of the system. The visual mapping model is rewritten into a form suitable for the design of extended state observer. The extended state observer is introduced to perform lumped estimation of the unmodeled dynamics and unknown external disturbances of the visual servo system of the sorting parallel robot:

[0068]

[0069]

[0070] In the formula, is a predicted value of the image Jacobian matrix of the fruit sorting parallel robot system, defined as Z1 represents the observation value of the extended state observer on the position of the feature point in the image space, Z2 represents the total amount of unmodeled dynamics and external disturbances estimated by the extended state observer for the visual servo system, ε i ,α i ,β i (i=1,…,4) and b1, b2 are non-negative adjustable design parameters of the extended state observer, P * is the expected position of the feature point in the image space, P is the actual position of the feature point in the image space, and ε is the image feedback error of the visual servo controller.

[0071] 4. The fruit sorting parallel robot is fixed in the camera mode of the eye in the hand. Since the sorting parallel robot has 4 degrees of freedom, in order to ensure that the visual servo has a unique solution, 4 feature points are set to perform visual servo control on the robot. The servo expression is:

[0072]

[0073] Where U is the motion vector of the robot in the robot control coordinate system, J f is the image Jacobian matrix, which represents the mapping relationship between the end effector in the robot coordinate system and the image coordinate system, and P is the image feature.

[0074] Considering the errors in the image detection process and the external disturbances in the visual servo system, let the lumped disturbance effect on the visual mapping model be ω(t), then the state expression of the visual mapping model considering the lumped disturbance effect is:

[0075]

[0076] Where u is the motion control variable of the robot. Assume:

[0077]

[0078] Then we have:

[0079]

[0080] According to the fruit sorting parallel robot visual servo system designed in step 2, the visual servo control error variables are designed as follows:

[0081] e=pp *(20)

[0082] Since the variable structure sliding mode control system has a simple algorithm, fast response speed, and robustness to external noise interference and parameter perturbations, it has been widely used in the field of robot control. The present invention designs an integral sliding mode algorithm that can overcome the uncertainty of the system and has strong robustness to interference and unmodeled dynamics, especially for nonlinear visual servo systems. The designed visual servo integral sliding mode surface is:

[0083] s=e+λ∫e (21)

[0084] Where λ is the controller parameter and the first-order derivative of s is:

[0085]

[0086] Substituting formula (22) into the robot visual servo model, we get:

[0087]

[0088] In the formula, Prediction of the image Jacobian matrix for a parallel robot system for fruit sorting.

[0089] In order to ensure the tracking performance of the visual servo system under model and disturbance uncertainty, the following visual servo integral sliding mode controller is designed:

[0090]

[0091] In the formula, k * >0 is the ideal controller gain, satisfying k * ≥λ|e|.

[0092] 5. For the switching gain of the visual servo integral sliding mode controller based on the extended state observer in step 4, a new adaptive law is further designed as follows:

[0093]

[0094] In the formula, is the derivative of the switching gain, k m is an adjustable parameter of the adaptive law, whose value is greater than 0, where

[0095]

[0096] Where μ>0 is an adjustable boundary parameter. Finally, the visual servo adaptive integral sliding mode controller is designed as:

[0097]

[0098] 6. Through software programming, the "host computer + slave computer" structure is adopted to establish a distributed fruit sorting parallel robot control system. The visual servo control quantity of the fruit sorting parallel robot is calculated by equation (27) in step 5. Figure 1 The robot control cabinet shown sends control commands to the motor drivers of each active joint of the fruit sorting parallel robot to drive the robot to complete the visual servo movement.

[0099] The following provides embodiments of the present invention:

[0100] Example 1

[0101] like Figure 1 As shown, the structure of the fruit sorting parallel robot is: 1. parallel mechanism 2. calibration plate 3. end clamping mechanism 4. camera 5. string fruit 6. display screen 7. control cabinet 8. detection cabinet 9. light source.

[0102] The present invention proposes a novel non-calibrated visual servo adaptive integral sliding mode control method for a fruit sorting parallel robot to realize visual servo motion control of the fruit sorting parallel robot. The specific implementation of the control method is as follows:

[0103] 1. The kinematics of the fruit sorting parallel robot are analyzed by analytical method to obtain its inverse solution, and the robot kinematics forward solution and the robot Jacobian matrix are further calculated.

[0104] like Figure 2 As shown in the figure, the fruit sorting robot consists of 4 sets of branched chains connecting the dynamic platform and the static platform. The camera is used to acquire the features of the target, and the motor tends the robot end effector to move in the x, y, and z axes (in millimeters). The end effector can rotate along the z axis with a rotation angle of θ (in rad), as shown in the figure. Figure 3 As shown: the main platform mass point is regarded as P1, the auxiliary platform is regarded as mass point and P2, and its kinematic equation is as follows:

[0105]

[0106] where e i =e(cosγ i sinγ i 0) T Represents the vector value of the radius difference of the inscribed circle from the moving platform to the static platform. In the above formula, γ i =(i-1)π / 2 is the structural angle of the static platform robot, l1 represents the length of the active arm of the branch chain, l2 represents the length of the driven arm (unit: mm), s=p(θ / 2π), p is the pitch of the robot screw.

[0107] 2. By establishing the kinematic model of the sorting robot and combining it with the visual imaging principle of the fruit sorting parallel robot, a visual mapping model of the fruit sorting parallel robot is established.

[0108] The fruit sorting parallel robot visual servo system controls the robot movement through the robot visual information and uses the image information for feedback. The control block diagram is as follows: Figure 4 As shown. Since the sorting parallel robot has 4 degrees of freedom, in order to determine the position and posture of the parallel robot, 4 feature points are set to perform visual servo control on the robot, and f = (f1 f2 f3 f4) T is the 4-dimensional vector in the robot image space, and the robot terminal velocity is According to the robot Jacobian matrix definition:

[0109]

[0110] Where J represents the mapping relationship between the robot joint space and the end Cartesian space. The moving speed of the feature point in the image space is available:

[0111]

[0112] Where J f Image Jacobian matrix for the fruit sorting parallel robot.

[0113] 3. The image Jacobian matrix representing the real-time changes in the mapping relationship between the robot image space and the task space is regarded as the unmodeled dynamics of the system, and the visual mapping model is rewritten into a form suitable for the design of an extended state observer. The extended state observer is introduced to perform a lumped estimate of the unmodeled dynamics and unknown external disturbances of the fruit sorting parallel robot visual servo system. Its state expression is:

[0114]

[0115]

[0116] In the formula, is a predicted value of the image Jacobian matrix of the fruit sorting parallel robot system, defined as Z1 represents the observation value of the extended state observer on the position of the feature point in the image space, Z2 represents the total amount of unmodeled dynamics and external disturbances estimated by the extended state observer for the visual servo system, ε i ,α i ,β i (i=1,…,4) and b1, b2 are non-negative adjustable design parameters of the extended state observer, P *is the expected position of the feature point in the image space, P is the actual position of the feature point in the image space, and ε is the image feedback error of the visual servo controller.

[0117] 4. The fruit sorting parallel robot uses the eye-in-hand camera fixed mode, and its servo expression is:

[0118]

[0119] Where U is the motion vector of the robot in the robot control coordinate system, J f is the image Jacobian matrix, which represents the mapping relationship between the end effector in the robot coordinate system and the image coordinate system, and P is the image feature.

[0120] Considering the errors in the image detection process and the external disturbances in the visual servo system, let the lumped disturbance effect on the visual mapping model be ω(t), then the state expression of the visual mapping model considering the lumped disturbance effect is:

[0121]

[0122] In the formula, u is the motion control quantity of the robot. Assume:

[0123]

[0124] Then we have:

[0125]

[0126] Design visual servo control error variable e=pp * , p is the actual position of the feature point in the image space, p * For the expected position of the feature point in the image space, the visual servo integral sliding surface is designed as:

[0127] s=e+λ∫e (36)

[0128] Where λ is the controller parameter and the first-order derivative of s is:

[0129]

[0130] Substituting formula (37) into the designed visual servo system model, we can obtain:

[0131]

[0132] In order to ensure the tracking performance of the visual servo system under unmodeled dynamics and external disturbances, the following visual servo integral sliding mode controller is designed:

[0133]

[0134] Where k * >0 is the ideal controller gain, satisfying k * ≥λ|e|.

[0135] 5. For the switching gain of the visual servo integral sliding mode controller based on the extended state observer in step 4, a new adaptive law is designed as follows:

[0136]

[0137] In the formula, is the derivative of the switching gain, k m is an adjustable parameter of the adaptive law, whose value is greater than 0, where

[0138]

[0139] Where μ>0 is an adjustable boundary parameter. Finally, the visual servo adaptive sliding mode controller is designed as:

[0140]

[0141] The fruit sorting parallel robot consists of three parts: the upper computer, the lower computer and the servo drive system. The upper computer is configured as follows: operating system Windows 7, processor Intel (R) Core (TM) i7-3770, main frequency 3.40GHz; the lower computer controller uses the UMAC programmable multi-axis motion controller, which can control the high-precision motion of up to 32 motors at the same time, and together with the upper computer, it forms a distributed control system. The UMAC controller consists of four parts: CPU board, I / O board, axis board and power board. The CPU motherboard uses the TURBOPMAC2OPT-5C0 PMAC-CPU board, the I / O board uses the ACC-11E expansion board with 24 ports on the input and output ends, and the axis board uses two AC The C-24E2A board has one four-channel board for controlling the parallel mechanism and one two-channel board for controlling the clamping mechanism. The power board is the ACC-E1 power board. The servo drive system is mainly composed of a servo driver and a servo motor. The parallel mechanism uses four sets of Panasonic MSMF102L1H6M servo motors and matching MDDLT55SF servo drivers. The clamping mechanism uses a set of Panasonic MSMF5AZL1U2M servo motors and matching MDDLT05SF servo drivers. In addition, to ensure the stability of the servo system, an ABR090-020-S1-P1 reducer with a reduction ratio of 9 is selected, and a 23-bit absolute encoder is used for position detection.

[0142] The visual system of the fruit sorting parallel robot mainly includes cameras, lenses, and image acquisition cards: the camera used is an area scan camera developed by Teledyne DALSA of Canada, model Genie Nano-M2420, which uses Sony IMX264 monochrome sensor, with a resolution of 2448*2048, a pixel depth of 8bit or 12bit, and a pixel size of 3.45um*3.45um. When TurboDriver is turned on, it can achieve an image acquisition rate of 35fps at a resolution of 5.1M pixels, and is directly connected to the computer through the GigE Vision standard; the lens used is a lens developed by VS Technology of Japan, model SV-0614H, with a focal length of 6mm, an aperture range of 1.4~16, and a maximum CCD size of 2 / 3 inches, which is installed on the camera through the C-mount interface; the image acquisition card uses the Intel-EXPI9402PT Gigabit Ethernet card, which can realize the image acquisition function, and its bus type is PIC-E 4x, has 2 RJ-45 network cable interfaces, and uses Intel 82571GB controller.

[0143] The visual servo control quantity of the fruit sorting parallel robot is calculated by formula (42), and the control quantity is programmed by the host computer and Figure 1 The robot control cabinet shown sends control commands to the motor drivers of each active joint of the fruit sorting parallel robot to drive the robot to complete the visual servo movement.

[0144] The static target tracking visual servo simulation experiment of the fruit sorting parallel robot is carried out through MATLAB. During the simulation, it is assumed that there is Gaussian white noise in the image detection process with a power of 10 -4 w, and at the same time add the external disturbance 1.5sin(t) of the visual servo system, Figure 5 , Figure 6 and Figure 7 The visual servo integral sliding mode control method (SMC) without an extended state observer, the visual servo integral sliding mode control method based on an extended state observer without adaptive action (ESO-SMC) and the visual servo adaptive integral sliding mode control method based on an extended state observer (ESO-ASMC) of the present invention are compared respectively. When tracking the same static target (Target) under interference conditions, the tracking error results in the X-axis, Y-axis, and Z-axis directions are shown as follows: Figure 5 , Figure 6 and Figure 7 , Figure 5 , Figure 6 and Figure 7The results show that the visual servo adaptive sliding mode control method based on the extended state observer has smaller errors in the X, Y, and Z axes, reaches the stable stage faster, and can maintain a stable state, indicating that the method proposed in the present invention has superior anti-interference performance and better visual servo control effect. Figure 8 It represents the running path of the robot end effector in the robot space under the uncalibrated visual servo adaptive integral sliding mode control method of the fruit sorting parallel robot proposed in the present invention, Fig. 9 Represents the tracking effect of feature points in the image space obtained by the camera, and analyzes Fig. 9 It can be seen that the maximum pixel absolute error of target tracking is 1.069 pixels, and the pixel root mean square error is 0.717 pixels, which shows that the uncalibrated visual servo adaptive integral sliding mode control method of the fruit sorting parallel robot proposed in the present invention has high accuracy.

[0145] In summary, the present invention provides a novel calibration-free visual servo adaptive integral sliding mode control method for a fruit sorting parallel robot. First, for the fruit sorting parallel robot, a fruit sorting parallel robot visual mapping model with image information as input and robot motion information as output is established, and the visual signal acquired by the camera is used as feedback information of the parallel robot visual servo system; the image Jacobian matrix representing the real-time changes of the mapping relationship between the robot image space and the task space is regarded as the unmodeled dynamics of the system, and the visual mapping model is rewritten into a form suitable for the design of an extended state observer, and the unmodeled dynamics and unknown external disturbances of the visual mapping model of the sorting parallel robot are lumped and estimated by the extended state observer. The target posture of the parallel robot is determined by setting four feature points, and the image coordinates of the actual posture of the parallel robot are obtained by image processing, so that the visual mapping model can be used to realize the calibration-free visual servo control of the parallel robot; then, an integral sliding mode controller based on an extended state observer is designed to improve the robustness of the visual servo system to overcome image feature point detection errors and external interference, thereby improving the control accuracy of the visual servo system; secondly, a new adaptive law is designed for the switching gain of the integral sliding mode controller based on the extended state observer to achieve the effect of suppressing the chattering of the sliding mode control; finally, the calibration-free visual servo adaptive integral sliding mode control of the fruit sorting parallel robot is realized through software programming. The visual servo adaptive integral sliding mode control method based on the extended state observer of the present invention can improve the control accuracy of the calibration-free visual servo system of the fruit sorting parallel robot.

[0146] It should be understood that the above embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent modifications of the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

Claims

1. A novel non-calibrated visual servo adaptive integral sliding mode control method for fruit sorting parallel robot, characterized in that: The steps include: 1) Taking the fruit sorting parallel robot as the controlled object, the analytical method is used to perform the inverse kinematics solution of the fruit sorting parallel robot, and the kinematics forward solution and the robot Jacobian matrix are obtained; 2) Based on the kinematic model of the sorting robot established in step 1), combined with the robot visual imaging principle, a fruit sorting parallel robot visual mapping model is established; 3) The image Jacobian matrix representing the real-time changes in the mapping relationship between the robot image space and the task space is regarded as the unmodeled dynamics of the system, and the visual mapping model is rewritten into a form suitable for the design of an extended state observer. The extended state observer is introduced to perform lumped estimation of the unmodeled dynamics and unknown external disturbances of the fruit sorting parallel robot visual servo system; 4) Design a visual servo integral sliding mode controller based on the extended state observer in step 3); 5) Design a new adaptive law for the switching gain of the visual servo integral sliding mode controller based on the extended state observer in step 4); 6) Through software programming and the use of a distributed fruit sorting parallel robot control system with a "host computer + slave computer" structure, the fruit sorting parallel robot can achieve uncalibrated visual servo adaptive integral sliding mode control; In the step 2), according to the robot visual imaging principle, a point P in the task space of the fruit sorting parallel robot is expressed in image space coordinates as: Where f is the focal length of the camera (unit: mm), (x, y) represents the coordinates of the image space of the point (unit: mm), C = (x c ,y c ,z c ) represents the position of the point in the camera coordinate system. The physical unit of the image space is converted to the pixel unit: Where (u0, v0) is the center pixel of the image space, (u, v) represents the coordinates of the image space at that point (unit: pixel), and a x is the number of pixels per unit on the X axis of the camera, a y The number of pixels per unit on the Y axis of the camera; It is characterized in that: the state expression of the expanded state observer in the fruit sorting parallel robot visual servo system in step 3) is: In the formula, u is the controller output; is a predicted value of the image Jacobian matrix of the fruit sorting parallel robot system, defined as Z1 represents the observation value of the extended state observer on the position of the feature point in the image space, Z2 represents the total amount of the unmodeled dynamics and external disturbances estimated by the extended state observer for the visual servo system, and ε i ,α i ,β i (i=1,…,4) and b1, b2 are non-negative adjustable design parameters of the extended state observer, P * is the expected position of the feature point in the image space, P is the actual position of the feature point in the image space, and ε is the image feedback error of the visual servo controller.

2. The method according to claim 1, characterized in that: In step 4), according to the definition of the visual servo system, the visual servo control error variable e is designed as follows: e=p-p * (5) Where p is the actual position of the feature point in the image space, p * is the expected position of the feature point in the image space, Design the visual servo integral sliding surface s as: s=e+λ∫e (6) Where λ is the controller parameter and the first-order derivative of s is: .

3. The method according to claim 1, characterized in that In step 5), the adaptive law of the designed visual servo adaptive integral sliding mode controller is: In the formula, is the derivative of the switching gain, k m is an adjustable parameter of the adaptive law, whose value is greater than 0, where Where μ>0 is an adjustable boundary parameter.

Images