A robot vibration suppression method based on impedance compensation predictive control
By establishing a dynamic model of robot-pneumatic end effector, predicting and optimizing impedance compensation force, real-time vibration suppression during robot processing is achieved, solving the complexity and universality of vibration suppression in the existing technology, and improving the Processing quality.
Patent Information
- Application Number
- CN202211195506.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-29
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2042-09-29
AI Technical Summary
The prior art is difficult to effectively suppress vibrations during robot processing, especially when the end of the robot is in flexible contact with the workpiece, resulting in a decrease in processing quality, and common vibration suppression methods are complex and not universal.
Using an impedance compensation prediction control method, the impedance compensation force is predicted and optimized by establishing a dynamic model of the robot-pneumatic end effector, and real-time force compensation is used to perform real-time force compensation, so as to suppress the robot vibration.
Real-time vibration suppression during the robot processing process is achieved, processing quality is improved, force fluctuations and force feedback time lags are reduced, control system is simplified, and overall vibration damping effect is improved.
Smart Images

Figure CN115533900B_ABST
Abstract
Description
Technical field:
[0001] The invention relates to the field of robot manufacturing control, and in particular to a robot vibration suppression method using a force-controlled end effector based on impedance compensation prediction. Background technology:
[0002] Robotic processing is flexible and has high production efficiency. It is gradually replacing manual processing, which has unstable processing quality and difficulty in meeting production volume needs. Various studies have been carried out on it around the world and it has been applied in many fields.
[0003] Robotic processing methods can be divided into two types according to the different end effectors used: robot direct processing with tool and robot with force-controlled end effector. The processing accuracy of the former is affected by the positioning accuracy of the robot, which makes it difficult for the blade processing quality to meet the technical requirements. The latter constitutes a macro-micro structure processing system. Since its end effector has independent degrees of freedom, it can further achieve precise control based on the robot's position positioning.
[0004] The lightweight design of the robot and its complex transmission system result in the low rigidity of the robot, which leads to vibration during the process of the robot end contacting the rigid body. During the processing, the vibration of the robot will be transmitted to the end, resulting in path deviation and contact force fluctuation, which seriously affects the processing quality. The force-controlled end effector includes a support part, a feed part, a guide part and a rotating part. The connecting flange is connected to the robot wrist, and the upper and lower base plates are used to install the rest. The air intake of the cylinder is controlled by applying voltage to the electrical proportional valve, and the cylinder rod is extended and retracted to drive the pneumatic spindle to complete the processing under the constraint of the guide rod. Therefore, the end effector is a pneumatic end effector. The rotational motion of the tool is provided by the pneumatic spindle.
[0005] At present, in order to suppress the vibration of the robot processing process, the robot is often modeled and analyzed and the joints are controlled. For example, the patent with the publication number CN109202884A discloses a robot vibration suppression method. The method includes the following steps: real-time acquisition of motion data of each joint of the flexible robot; based on the acquired motion data of each joint, the dynamic equations of the robot rigid body, joint deformation and the dynamic equations of the desired trajectory of the robot end are established, and the required joint torque value is obtained by solving the inverse dynamics calculation; according to the expected damping matrix of the robot body flexibility, the joint drive torque is optimized to obtain the optimized joint torque value; according to the optimized joint drive torque value, the joint torque control is performed to achieve robot vibration suppression. However, since the patent controls the robot body, it is necessary to modify the robot control algorithm to suppress the robot vibration. Since most robots are not open source, it is difficult to achieve. In addition, there are many joints and degrees of freedom that need to be controlled, and dynamic modeling is difficult, which leads to a very complex robot control problem. When the environment is disturbed, the control effect will be greatly reduced.
[0006] Some researchers at home and abroad have suppressed vibration by adding shock absorbers or vibration-damping materials. However, since the shock absorber has its own control system and additional input energy, the vibration-damping system may become a secondary vibration source, which in turn reduces system performance. The vibration-damping method of designing fixtures and auxiliary supports for specific workpieces depends on the specific workpiece materials and structures and is not universal. In addition, the method of designing the vibration-damping structure of the end effector cannot achieve real-time control of vibration. Summary of the invention:
[0007] In view of the deficiencies in the prior art, the technical problem that the present invention intends to solve is to provide a robot vibration suppression method based on impedance compensation predictive control, which can suppress the vibration of the robot processing process in real time, make the robot end contact with the workpiece flexibly, and improve the processing quality.
[0008] In order to achieve the above object, the present invention comprises the following steps:
[0009] A robot vibration suppression method based on impedance compensation predictive control, characterized in that it comprises the following steps:
[0010] The impedance compensation relationship is established based on the robot-pneumatic end effector machining dynamics model. The impedance compensation relationship is:
[0011] y S (s)=-(B r s+K r )×x(s)
[0012] Among them, y S is the impedance compensation force, B ris the damping of the pneumatic end effector, s is the complex variable in the inverse Laplace transform, K r is the stiffness of the pneumatic end effector, x is the displacement of the cylinder rod of the pneumatic end effector;
[0013] Build an impedance-compensated predictive controller:
[0014] (2-1) Given a step input to the system, the input is the control voltage of the pneumatic end effector cylinder, and the system response is a=[a 1 a 2 …a N ] T Indicates that N is the modeling time domain, establishes a prediction model, assumes that M is the control time domain, P is the optimization time domain, and the response in the P optimization time domain constitutes a dynamic matrix A, which is expressed as:
[0015]
[0016] The prediction model is:
[0017] in, is the contact force prediction sequence, is the initial prediction sequence of contact forces, is the initial prediction sequence of impedance compensation force, which is the prediction value of the future P moments at time k and is a P-dimensional column vector; Δu M (k) is the control voltage increment of the contact force at time k in the M control domain, which is an M-dimensional column vector, and A is Δu M (k) The corresponding dynamic matrix;
[0018] (2-2) The ratio of the output contact force and displacement of the pneumatic end effector is defined as the contact stiffness, and the average value of the contact stiffness at time k-2 and the contact stiffness at time k-1 is used as the predicted contact stiffness value at time k:
[0019]
[0020] In the formula, K e ′(k) is the contact stiffness at time k, y(k-1) is the output value of the force sensor of the pneumatic end effector at time k-1, and x(k-1) is the output value of the displacement sensor of the pneumatic end effector at time k-1; y(k-2) is the output value of the force sensor of the pneumatic end effector at time k-2, and x(k-2) is the output value of the displacement sensor of the pneumatic end effector at time k-2;
[0021] The contact stiffness K at time k e ′(k) and the initial prediction sequence of the contact force according to Get the initial prediction sequence of displacement The initial prediction sequence of displacement Substitute into the impedance compensation relationship to obtain the initial prediction sequence of impedance compensation force expression:
[0022]
[0023] Among them, Z m is the compensation impedance, Z m =-(B r s+K r ),
[0024] (2-3) The initial prediction sequence of impedance compensation force obtained in step 2-2 Substitute the expression into the prediction model in step 2-1 to obtain a new prediction model;
[0025] (2-4) Contact force prediction sequence including compensation force information Perform rolling optimization: set the quadratic performance index J(k) to be the minimum as the optimization goal:
[0026] Among them, ω P (k) is the expected force at time k, which is also a P-dimensional column vector;
[0027] Combined with the necessary conditions for extreme values The control law ψ(k) under the optimal performance index condition is obtained:
[0028]
[0029] Among them, Q is the P-dimensional diagonal error weight matrix, and Z is the M-dimensional diagonal control weight matrix;
[0030] Considering that there are some time lags and non-minimum phase dynamic characteristics in the dynamic response process of the system, and it is impossible to achieve stable tracking of the control target output to the expected value within this time domain P, in order to realize the real dynamic optimization process of the control system, the optimization time domain P should be greater than the time lag and non-minimum phase dynamic characteristics time domain, and the corresponding error weight coefficient is taken as 0, which can be obtained:
[0031]
[0032] q i is the element on row i and column i of the error weight matrix Q;
[0033] Control voltage increment at time k Δu(k) = d T ψ(k), where d T =[1 0…0], d T is a P-dimensional vector;
[0034] The rolling optimization is as follows: at time k, the output at time P in the future is obtained according to the system model, and the system control rate is optimized by minimizing the quadratic performance index, thereby obtaining the control voltage increment Δu of M optimal contact forces M (k), but at time k, only the first element Δu(k) is taken to act on the pneumatic end effector, and the optimization is performed again at time k+1;
[0035] (2-5) The feedback correction process is as follows:
[0036] Contact force prediction sequence and impedance compensation force initial prediction sequence Perform feedback correction: Compare the actual value y(k+1) of the output force obtained at time k+1 with the predicted value at time k at the future time P. (that is, the output value of the prediction model), and the error value e(k+1) is obtained: Subscript 1 indicates that in Δu M (k) Under the action of the first quantity;
[0037] Introduce the correction vector h = [h 1 h 2 …h P ] T Get P predicted output values after correction at time k, and the predicted output value of contact force after error correction
[0038]
[0039] By shifting the matrix Get the initial prediction sequence of contact force at time k+1 Continue to make predictions for the next moment and input them into the new prediction model;
[0040] At this time, the prediction system is based on the prediction of the dynamic response of the end contact force. After completing the prediction of the displacement of the pneumatic end effector and the impedance compensation force, the initial prediction sequence of the contact force and the initial prediction sequence of the compensation force are used as the input of the control law and the input of the feedback correction, so that the compensation force participates in the rolling optimization and prediction correction links;
[0041] The pneumatic end effector is controlled by taking the first term of the optimal control law as the control voltage increment, thus obtaining an impedance compensation predictive controller.
[0042] The impedance compensation predictive controller is applied to the robot machining process control and is used to compensate the output force of the pneumatic end effector to suppress the robot vibration.
[0043] All elements in the control weight matrix Z are 0.
[0044] The pneumatic end effector contacts the workpiece, and the robot generates vibrations to input the displacement signals and force signals obtained by the corresponding sensors into the impedance compensation prediction controller. The force signals and displacement signals are directly acquired by the corresponding sensors of the pneumatic end effector.
[0045] The control quantity output by the impedance compensation predictor is used as the voltage input quantity of the cylinder of the pneumatic end effector.
[0046] Compared with the prior art, the present invention has the following beneficial effects:
[0047] The present invention introduces the compensation impedance force into the dynamic matrix predictive control, uses the prediction function of the predictive control to solve the time lag problem of impedance compensation, and uses the feedback correction function to enable the output force of the pneumatic end effector to better track the expected force under the action of compensation control and unknown interference. The impedance compensation force is taken as part of the prediction object, and is predicted through the relationship between the end output contact force and the displacement, and the initial prediction sequence of the contact force and the initial prediction sequence of the compensation force are used as the input of the control law and the input of the prediction correction, so that the compensation force participates in the rolling optimization and prediction correction links, so that the sum of the compensation force and the contact force is close to the expected force (set value), the force fluctuation is reduced, and the vibration is reduced, so as to obtain a contact force closer to the expected force and a compensation force value with the best vibration suppression effect on the robot during the entire processing process.
[0048] Compared with the existing technology, the robot vibration suppression method based on impedance compensation predictive control provided by the present invention has the following advantages over other vibration suppression methods: the present application focuses on the control of the end effector, not the robot joint control, does not need to modify the robot dynamics interface, and does not change the robot's dynamic characteristics; it does not make the control system and mechanical system more complicated; it reduces the time lag in the force feedback and force compensation process; it simultaneously performs rolling optimization on the contact force and the compensation force (the cylinder simultaneously outputs the end contact force and the cylinder rod extension (force sensor and displacement sensor), so it is a simultaneous rolling optimization) to improve the vibration reduction effect. The present application can accurately compensate for the vibration caused by the low stiffness of the robot during the entire processing process, including the entire processing process of the contact state and the stable contact stage, and corrects the predicted value at every moment to improve the processing quality.
[0049] The present invention solves the vibration problem of the industrial robot during the entire processing process in which the robot carries a force-controlled end effector in rigid contact with the workpiece, and solves the problems of force fluctuation, force overshoot and reduced processing quality caused by vibration. Description of the drawings:
[0050] The specific embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings:
[0051] Figure 1It is a flow chart of a robot vibration suppression method based on impedance compensation predictive control in an embodiment of the present invention;
[0052] Figure 2 It is a schematic diagram of the structure of the pneumatic end effector;
[0053] Figure 3 is a diagram of the robot-pneumatic end effector processing dynamics model in an embodiment of the present invention;
[0054] Figure 4 is a time lag analysis diagram of impedance compensation in an embodiment of the present invention;
[0055] Figure 5 4 is a control block diagram of an impedance compensation prediction controller in an embodiment of the present invention.
[0056] Among them: robot 1, pneumatic end effector 2, mounting plate 21, cylinder 22, force sensor 3, support plate 23, air spindle 24, connecting flange 25, guide rod 26, displacement sensor 27, inclination sensor 28, tool 29, workpiece 4. Specific implementation method:
[0057] like Figure 1 As shown, the robot vibration suppression method based on impedance compensation predictive control includes the following steps:
[0058] (1) Establish impedance compensation relationship based on robot-pneumatic end effector machining dynamics model;
[0059] (2) designing an impedance compensation prediction controller based on the impedance compensation relationship obtained in step (1) and a dynamic matrix prediction algorithm;
[0060] (3) The pneumatic end effector contacts the workpiece, the robot generates vibration, and the displacement signal and force signal obtained by the corresponding sensor are input into the impedance compensation predictive controller;
[0061] (4) The impedance compensation predictive controller is applied to the pneumatic end effector for use in robot machining process control. The controller is used to compensate the output force of the pneumatic end effector, complete real-time force feedback and force compensation, and suppress robot vibration.
[0062] The pneumatic end effector 2 is a prior art, and its structure is as follows: Figure 2As shown, the connection flange 25 of the pneumatic end effector 2 is fixed to the robot wrist with bolts at the top, and is fixed to the mounting plate 21 at the bottom. The feed part and the sensor are between the mounting plate and the support plate 23. The feed part is a cylinder 22, which is bolted to the mounting plate. The cylinder rod is extended and retracted under the restriction of the guide rod 26. The guide rod is welded between the mounting plate and the support plate. The sensors are a force sensor 3, a displacement sensor 27, and an inclination sensor 28. In the middle of the support plate are a threaded air spindle 24 and a tool 29.
[0063] Robot-pneumatic end effector machining dynamics model Figure 3 As shown in the figure, it is an equivalent spring-mass-damper model. The dynamic equation established by the robot-pneumatic end effector processing dynamics model is expressed as:
[0064]
[0065]
[0066]
[0067] Where, M is the mass of robot 1, B is the damping of robot 1, K is the stiffness of robot 1, m is the mass of pneumatic end effector 2, and B r is the damping of the pneumatic end effector 2, K r is the stiffness of the pneumatic end effector 2, B s is the damping of the force sensor 3 between the pneumatic end effector 2 and the environment, K s is the stiffness of the force sensor 3; X 1 is the vibration displacement of robot 1, X 2 is the vibration displacement of the pneumatic end effector 2, X e is the force sensor vibration displacement, F n is the axial reaction force of the workpiece 4 on the tool 29, F r is the force between robot 1 and pneumatic end effector 2, X 2 -X 1 is the displacement of the cylinder rod. and They are the vibration speed of robot 1, the vibration speed of pneumatic end effector 2, and the vibration speed of force sensor 3; and They are the vibration acceleration of robot 1, the vibration acceleration of pneumatic end effector 2 and the vibration acceleration of force sensor 3.
[0068] Assume that the force sensor vibration displacement X e After Laplace inverse transformation, the end contact force F of the system can be obtained. n and the disturbance force F rThe transfer function of , s is the complex variable in the transformation:
[0069]
[0070] Z r (s)=B r s+K r
[0071] Among them, impedance is expressed as the ratio of force to displacement, and admittance is expressed as the ratio of displacement to force. 1 is the admittance of robot 1, Y 2 is the admittance of the pneumatic end effector 2, Z r is the coupling impedance between robot 1 and pneumatic end effector 2, Z s is the impedance between the pneumatic end effector 2 and the force sensor 3.
[0072] Admittance Y 1 Reflects the resonant frequency of the vibration of robot 1, the pole is At the resonance frequency of robot 1, a large vibration will be generated and coupled to the impedance Z r Transmitted to the pneumatic end effector 2.
[0073] In direct force feedback, a compensation impedance Zm is introduced, and the sum of the compensation impedance and the coupling impedance is set to zero, that is, Z r and Z m Numerically equal, both are |B r s+K r |, can offset Y 1 (s) brings about the disturbance, and the contact force F is obtained n and the expected force F d The relationship between:
[0074]
[0075] in,
[0076] According to the ratio of force to displacement, the impedance compensation force expression is obtained:
[0077] y S (s)=-(B r s+K r )×x(s),
[0078] Among them, y S is the impedance compensation force, and x is the displacement of the cylinder rod, that is, the value obtained by the displacement sensor 27.
[0079] Figure 4It is an impedance compensation time-delay diagram. The control voltage is input to the cylinder, and the cylinder outputs the contact force and displacement. The impedance compensation process is: the impedance compensation force is calculated from the output displacement. The impedance compensation force is obtained by inversely calculating the control voltage increment from the cylinder transfer function. At the same time, after force feedback, the force error is input into the PID controller to obtain the input control voltage of the cylinder. The PID controller makes the sum of the output contact force and the impedance compensation force track the desired force. The time lag of the impedance compensation process is that the impedance compensation force including the displacement at the current moment is obtained at the k moment, and the voltage compensation amount obtained by the cylinder transfer function can only act on the cylinder at the k+1 moment. The calculation of the impedance compensation force is based on the displacement output by the displacement sensor 27 at the current moment, but the force compensation can only be achieved by applying the control voltage to the cylinder 22 at the next moment, thus generating the time lag phenomenon of impedance compensation.
[0080] Figure 5 The control flow chart of the impedance compensation prediction controller is as follows: Through contact stiffness K e ′(k) gets the initial prediction sequence of displacement Then the compensation impedance Z m (k) Obtain the initial prediction sequence of impedance compensation force The control voltage increment Δu is obtained by minimizing the quadratic performance index M (k), Δu M The first term Δu(k) of (k) acts on the cylinder and outputs the actual contact force y(k+1) and displacement x(k+1) at time k+1. According to the system dynamic response matrix A, considering Δu 1 (k) Output force prediction under control That is, the future change output generated by the control voltage increment, the actual output contact force y(k+1) of the cylinder and For comparison, the error value obtained is corrected by the correction vector h to output the contact force prediction value after error correction Then, the initial prediction sequence of contact force at the next moment is obtained through the shift matrix S Continue the cycle to participate in feedback correction at the next moment.
[0081] like Figure 5 As shown in the figure, the specific steps of impedance compensation prediction controller design are as follows:
[0082] Given a step input to the system, the input is the control voltage of the cylinder 22, and the step response of the pneumatic end effector 2 is recorded, using a=[a 1 a 2 …a N ] T Indicates that N is the modeling time domain, and P is the step response in the optimization time domain to form the dynamic matrix A; let M be the control time domain, P be the optimization time domain, and A can be expressed as:
[0083]
[0084] A prediction model is established, wherein the prediction model is:
[0085]
[0086] in, is the contact force prediction sequence, is the initial prediction sequence of contact forces, is the initial prediction sequence of the impedance compensation force, which is the prediction value of the future P moments at time k, so it is a P-dimensional column vector, Δu M (k) is the control voltage increment of the contact force at time k in the M control time domain, which includes the compensation voltage increment Δu corresponding to the impedance compensation contact force at time k S (k), are M-dimensional column vectors, A is Δu M (k) The corresponding dynamic matrix; the actual voltage-controlled increment of the contact force is Δu(k), Δu(k) = u(k) - u(k-1).
[0087] The ratio of the output contact force of the pneumatic end effector 2 to the displacement is the contact stiffness, which is determined by factors such as the material of the workpiece 4 and the material of the tool 29. In addition, the dynamic characteristics of the pneumatic end effector 2 and the workpiece 4 will continue to change during the machining process, resulting in a small range of fluctuations in this value. In order to ensure the accuracy of vibration compensation, the contact stiffness is replaced by the average value of the contact stiffness at time k-2 and the contact stiffness at time k-1 instead of the predicted contact stiffness value at time k:
[0088]
[0089] K e ′(k) is the contact stiffness at moment k, y(k-1) is the output value of force sensor 3 at moment k-1, x(k-1) is the output value of displacement sensor 27 at moment k-1; y(k-2) is the output value of force sensor 3 at moment k-2, x(k-2) is the output value of displacement sensor 27 at moment k-2.
[0090] Initial prediction sequence of the contact stiffness and the contact force Get the initial prediction sequence of displacement and the compensation force initial prediction sequence expression:
[0091]
[0092]
[0093] According to the transfer function of the cylinder, The compensation voltage increment Δu can be obtained S (k).
[0094] Initial prediction sequence of compensation force The expression is substituted into the prediction model to obtain a new prediction model.
[0095] The rolling optimization process is as follows:
[0096] In predictive control, the prediction error can be maintained within a bounded range by performing rolling optimization on the control target, but it still cannot guarantee that the control increment will not fluctuate greatly, thus causing the controlled system to become unstable. Therefore, the minimum quadratic performance index J(k) is used as the optimization target:
[0097]
[0098] Among them, ω P (k) is the expected force at time k, which is also a P-dimensional column vector;
[0099] Combined with the necessary conditions for extreme values The control law ψ(k) under the optimal performance index condition can be obtained:
[0100]
[0101] Among them, Q is the P-dimensional diagonal error weight matrix, and Z is the M-dimensional diagonal control weight matrix.
[0102] Considering that there are some time lags and non-minimum phase dynamic characteristics in the dynamic response process of the system, and it is impossible to achieve stable tracking of the control target output to the expected value within this time domain P, in order to realize the true dynamic optimization process of the control system, the optimization time domain P should be larger than the time lag and non-minimum phase dynamic characteristics time domain, then the corresponding error weight coefficient q i It can be taken as 0, and we can get:
[0103]
[0104] qi is the element in row i and column i of the error weight matrix Q,
[0105] When the error weight coefficient is determined, the control weight coefficient becomes an adjustable factor, and all elements in the control weight matrix Z are set to 0.
[0106] The control voltage increment at time k is Δu(k)=d T ψ(k), where d T =[1 0…0], which is a P-dimensional vector.
[0107] The rolling optimization is as follows: at time k, the output at time P in the future is obtained according to the system model, and the system control rate is optimized using a quadratic performance index, thereby obtaining M optimal contact force voltage control increments Δu M (k), but at time k, only the first element Δu(k) is taken to act on the pneumatic end effector 2, and the optimization is performed again at time k+1.
[0108] The feedback correction process is as follows:
[0109] Contact force prediction sequence and impedance compensation force initial prediction sequence Perform feedback correction: Compare the actual value y(k+1) of the output force obtained at time k+1 with the predicted value of the future P moments predicted at time k Get the error value: Subscript 1 indicates that in Δu M (k) Under the action of the first quantity.
[0110] Introduce the correction vector h = [h 1 h 2 …h P ] T Get P predicted output values after correction at time k, and the predicted output value of contact force after error correction
[0111]
[0112] By shifting the matrix Get the initial prediction sequence of contact force at time k+1 Continue to make predictions for the next moment and input them into the new prediction model; the shift matrix S is a P×P matrix consisting of a unit diagonal matrix with the first row deleted and the last row added, which is the same as the previous row;
[0113] The pneumatic end effector is controlled by taking the first term of the optimal control law as the control voltage increment, thus obtaining an impedance compensation predictive controller.
[0114] The force signal and displacement signal are directly acquired by the force sensor 3 and displacement sensor 27 of the pneumatic end effector 2. The control quantity output by the impedance compensation predictor is used as the voltage input quantity of the cylinder 22.
[0115] The displacement output at time k contains the robot vibration information at time k, and the compensation force can only be calculated at time k+1. This is a time lag phenomenon. The present application obtains the compensation force value at time k+1 at time k, and can perform force compensation at time k+1, eliminating time lag, and combines the dynamic matrix prediction algorithm to continuously perform rolling optimization of the contact force and impedance compensation force throughout the contact process, continuously predict the impedance compensation force value and feedback correction prediction value, and can perform real-time and accurate force compensation for robot vibration.
[0116] The method of the present application is applied to a pneumatic end effector, which can achieve vibration suppression of the robot, improve processing quality, reduce vibration, and improve force tracking performance. By controlling the cylinder in the end effector, the purpose of reducing robot vibration is achieved.
[0117] The present invention solves the vibration problem of the industrial robot during the entire processing process in which the robot carries a force-controlled end effector in rigid contact with the workpiece, and solves the problems of force fluctuation, force overshoot and reduced processing quality caused by vibration.
[0118] Any matters not described in the present invention are applicable to the prior art.
Claims
1. A robot vibration suppression method based on impedance compensation predictive control, characterized in that: The following steps are involved: The impedance compensation relationship is established based on the robot-pneumatic end effector machining dynamics model. The impedance compensation relationship is: y S (s)=-(B r s+K r )×x(s) Among them, y S is the impedance compensation force, B r is the damping of the pneumatic end effector, s is the complex variable in the inverse Laplace transform, K r is the stiffness of the pneumatic end effector, x is the displacement of the cylinder rod of the pneumatic end effector; Build an impedance-compensated predictive controller: (2-1) Given a step input to the system, the input is the control voltage of the cylinder of the pneumatic end effector, and the system response is a = [a1 a2…a N ] T Indicates that N is the modeling time domain, establishes a prediction model, assumes that M is the control time domain, P is the optimization time domain, and the response in the P optimization time domain constitutes a dynamic matrix A, which is expressed as: The prediction model is: in, is the contact force prediction sequence, is the initial prediction sequence of contact forces, is the initial prediction sequence of impedance compensation force, which is the prediction value of the future P moments at time k and is a P-dimensional column vector; Δu M (k) is the control voltage increment of the contact force at time k in the M control domain, which is an M-dimensional column vector, and A is Δu M (k) The corresponding dynamic matrix; (2-2) The ratio of the output contact force and displacement of the pneumatic end effector is defined as the contact stiffness, and the average value of the contact stiffness at time k-2 and the contact stiffness at time k-1 is used as the predicted contact stiffness value at time k: In the formula, K e '(k) is the contact stiffness at time k, y(k-1) is the output value of the force sensor of the pneumatic end effector at time k-1, and x(k-1) is the output value of the displacement sensor of the pneumatic end effector at time k-1; y(k-2) is the output value of the force sensor of the pneumatic end effector at time k-2, and x(k-2) is the output value of the displacement sensor of the pneumatic end effector at time k-2; The contact stiffness K at time k e '(k) and the initial prediction sequence of the contact force according to Get the initial prediction sequence of displacement The initial prediction sequence of displacement Substitute into the impedance compensation relationship to obtain the initial prediction sequence of impedance compensation force expression: Among them, Z m is the compensation impedance, Z m =-(B r s+K r ), (2-3) The initial prediction sequence of impedance compensation force obtained in step 2-2 Substitute the expression into the prediction model in step 2-1 to obtain a new prediction model; (2-4) Contact force prediction sequence including compensation force information Perform rolling optimization: set the minimum quadratic performance index J(k) as the optimization goal: Among them, ω P (k) is the expected force at time k, which is also a P-dimensional column vector; Combined with the necessary conditions for extreme values The control law ψ(k) under the optimal performance index condition is obtained: Among them, Q is the P-dimensional diagonal error weight matrix, and Z is the M-dimensional diagonal control weight matrix; Considering that there are some time lags and non-minimum phase dynamic characteristics in the dynamic response process of the system, and it is impossible to achieve stable tracking of the control target output to the expected value within this time domain P, in order to realize the real dynamic optimization process of the control system, the optimization time domain P should be greater than the time lag and non-minimum phase dynamic characteristics time domain, and the corresponding error weight coefficient is taken as 0, which can be obtained: q i is the element on row i and column i of the error weight matrix Q; Control voltage increment at time k Δu(k) = d T ψ(k), where d T =[1 0…0], d T is a P-dimensional vector; The rolling optimization is as follows: at time k, the output at time P in the future is obtained according to the system model, and the system control rate is optimized by minimizing the quadratic performance index, thereby obtaining the control voltage increment Δu of M optimal contact forces M (k), but at time k, only the first element Δu(k) is taken to act on the pneumatic end effector, and the optimization is performed again at time k+1; (2-5) The feedback correction process is as follows: Contact force prediction sequence and impedance compensation force initial prediction sequence Perform feedback correction: Compare the actual value y(k+1) of the output force obtained at time k+1 with the predicted value at time k at the future time P. Get the error value e(k+1): Subscript 1 indicates that in Δu M (k) Under the action of the first quantity; Introduce the correction vector h = [h1 h2…h P ] T Get P predicted output values after correction at time k, and the predicted output value of contact force after error correction By shifting the matrix Get the initial prediction sequence of contact force at time k+1 Continue to make predictions for the next moment and input them into the new prediction model; At this time, the prediction system is based on the prediction of the dynamic response of the end contact force. After completing the prediction of the displacement of the pneumatic end effector and the impedance compensation force, the initial prediction sequence of the contact force and the initial prediction sequence of the compensation force are used as the input of the control law and the input of the feedback correction, so that the compensation force participates in the rolling optimization and prediction correction links; The pneumatic end effector is controlled by taking the first term of the optimal control law as the control voltage increment, thus obtaining an impedance compensation predictive controller. The impedance compensation predictive controller is applied to the robot machining process control and is used to compensate the output force of the pneumatic end effector to suppress the robot vibration.
2. The robot vibration suppression method based on impedance compensation predictive control according to claim 1, characterized in that: All elements in the control weight matrix Z are 0.
3. The robot vibration suppression method based on impedance compensation predictive control according to claim 1, characterized in that: The pneumatic end effector contacts the workpiece, and the robot generates vibrations to input the displacement signals and force signals obtained by the corresponding sensors into the impedance compensation prediction controller. The force signals and displacement signals are directly acquired by the corresponding sensors of the pneumatic end effector.
4. The robot vibration suppression method based on impedance compensation predictive control according to claim 1, characterized in that: The control quantity output by the impedance compensation predictor is used as the voltage input quantity of the cylinder of the pneumatic end effector.
Citation Information
Patent Citations
Vibration suppression method of flexible multi-joint robot and vibration control system of flexible multi-joint robot
CN109202884A
Force control end effector vibration suppression method based on impedance control
CN112936291A
Force control method and device for reducing grinding and polishing vibration
CN112959217A
Impedance control method, and controller and robot using the same
US20220009095A1
Vibration suppression and dynamic balancing for retargeting motions onto robotic systems
US20220134552A1
